Biomedical Research in Space Flight

NASA SP-5078
t4'71 - = 77 I_I
_ i_ i¸¸¸
BIOMEDICAL RESEARCH AND

COMPUTER APPLICATION IN
MANNED SPACE FLIGHT
A REPORT
=|
NOTICE. This document was prepared under the sponsorship of the National
Aeronautics and Space Administration. Neither the United States Government nor
any person acting on behalf of the United States Government assumes any liability
resulting from the use of the information contained in this document, or warrants
that such use will be free from privately owned rights.
For sale by the Superintendent of Documents,

U.S. Government Printing Office, Washington, D. C. 20402
Price
Library of Congress Catalog Card Number 72-608485
FOREWORD
The National Aeronautics and Space Administration has established a Tech-

nology Utilization Program designed to transfer technological developments that
may have useful commercial applications. From NASA laboratories and contractors,
aeronautics and space-related technology is gathered and evaluated. Items which
have potential industrial use are made generally available. This survey of computer
uses in the field of medicine is one of a series of NASA publications that presents
information of direct or indirect interest to the nonaerospace community.
During the past twenty-five years, computers have grown from slow-speed
vacuum-tube machines with very small storage capacity and very limited processing
capability to very fast, very large, extremely versatile devices that use micro-
miniature electronic components. During the same period of time, there has been
an exponential growth of knowledge in the medical sciences. To a considerable
degree, America's twelve-year old program of space exploration has stimulated
development in both areas.
With the increase of medical knowledge has come a multiplication in the quantity
and kinds of medical data. It seems logical to expect that the computer would be
applied to aid in the reduction and interpretation of this data, but the full potential
of computers in medicine has not yet been realized.
This report summarizes the areas of medicine in which computers can be employed
and examines in detail several cases where computers have been applied in connec-
tion with the medical aspects of NASA's manned space flight program. Treated
are such problems as those of automated medical data storage and retrieval systems,
continuous monitoring and interpretation of electrocardiograms, and computer-
aided medical diagnosis. The approach is cautious throughout, with the emphasis
almost constantly on ways to permit the computer to perform various clerical
functions while leaving critical decisions to a human monitor.
RONALD J. PHILIPS, Director

Technology Utilization Office
III
PREFACE
Research into the biomedical variables surrounding safe and efficient manned
space flight has revealed the need for the advanced _pplication of computers in the
solution of various problems. This book presents manned spaceflight-oriented re-
search that emphasizes the computerized approach both in terms of a general over-
view and as utilized in the solution of specific problem area&
Since the articles presented here have not previously been made available to the
general scientific community, it is believed that this book will preserve and dissemi-
nate much knowledge concerning manned space flight research and the general
and specifically related application of computers.
This organized collection of articles resulted from activities encouraged, and in
most cases supported, by the National Aeronautics and Space Administration.
The material presented here spans the tenures of George M. Knauf, M.D.; William
Randolph Lovelace, II, M.D.; and Brigadier General Jack Bollcrud, USAF, _iC,
in the Office of the Director of Space Medicine, NASA Headquarters. Concurrently,
under each of the foregoing Directors, Dr. Jefferson F. Lindsey, Jr., served as Head
of the Biomathematical Staff.
The efforts of Charles A. Berry, M.D., and A. Duane Catterson, M.D., of the
Manned Spacecraft Center, Houston, Texas, in furthering the research efforts re-
ported here are to be particularly noted. Likewise, appreciation is expres_d to
Mr. Waiter B. Sullivan, Jr., NASA Headquarters, who was directly concerned as
monitor on several of these projects, and to Dr. Mae _f. Link, NASA Headquarters,
for her coordination efforts in expediting the publication of this book.
J. W. HUMPHREYS, JR.
M[ajor General, USAF, AIC

Director, Space Medicine
.Manned Space Flight
CONTRIBUTORS AND EDITORS
•_braham, Sidney: StatJsffcian, Medical Sy3tems Develop- McAllisler, James W.- Engineer, Medical Systems
' ment Laboratory, Heart Disease Control Program, Development Laboratory
National Center for Chronic Disease Control, Public Minckler, Tare M., M.D.: Chief, Section of Medical
, Health Service, U.8. Department of Health, Education, Information Management Systems, Department of
and Welfare, Washington, D.C. Biomathematics, The University of Texas, M. D.
Butch, Neii R., M.D.: Head, Psychophysiology Division, Anderson Hospital and Tumor Institute, Houston,
Baylor University College of Medicine, Texas Medical Texas
Center, Houston, Texas Moseley, Edward C., Ph.D.: Biomedical Research Office,
Caceres, Cesar A., M.D.: Chief, Instrumentation Field National Aeronautics and Space Administration,
Station, Heart Disease Control Program; and Associate Manned Spacecraft Center, ttouston, Texas
Professor of M_licine, George Washington University, Paine, Stephen T., Ph.D.: Life Sciences Operation, North
Washington, D.C. American Aviation, Inc., El Segundo, California
Ca|atayud, Juan B., M.D. : Assistant Professor of Medicine, Proctor, Lorne D., M.D.: Chairman, Department of
George Washington University, Washington, D.C. Neurology and Psychiatry, Henry Ford Hospital,
Dorsett, RonMd G.: Texas Research Institute o/' Mental Detroit, _fichigan
Sciences, Baylor University College of Medicine, Texas Rosner, Stuart W., M.D.: Medical Officer, Medical
Medical Center, Houston, Texas Systems Development Laboratory
Hochberg, Howard' M., M,D.: Medical Officer, Medical Sashin, Jerome, M.D. : Medical Officer, Medical Systems
Systems Development Laboratory Development Laboratory
Horton, Caroline L., B.S.: Senior Computer Programmer, Sells, S. B., Ph.D.: Research Professor (Psychology) and
Section of Medical Information Management Systems, Director, Institute of Behavioral Research, Texas
Department of Biomathematics, The University of Christian University, Fort Worth, Texas
Texas, M.D. Anderson Hospital and Tumor Institute, Stein, Mervyn R., M.D.: Clinical Laboratory Service,
Houston, Texas Clinical Center, National Institutes of Health, Bethesda,
Jackson, Larry K., M.D. : Staff Physician, Medical Systems Maryland
Development Laboratory Surgent, Louis V., Ph.D.: Research Staff Specialist,
]ones, Robert L., Ph.D.: Associate Professor, Baylor Research Department, Electronics Division, General
University College of Medicine; formerly with the Dynamics, Inc., Rochester, New York
Biomedieal Research Office, National Aeronautics Townsend, John C., Ph.D.: Professor and Director of
and Space Administration, Manned Spacecraft Center, the Human Factors Program, Department of Psy-
Houston, Texas chology, The Catholic University of America, Wash-
Knoblock, Edward C., Ph.D., Colonel, USA: Director, ington, D.C.; and Consultant to the Office of Space
Division of Biochemistry, Walter Reed Army Institute Medicine
of Research, Washington, D.C. Vordeman, Ahbie L., M.D.: Texas Research Institute
Lester, Boyd K., M.D.: University of Oklahoma Medical of Mental Sciences, Baylor University, College of
Center, Oklahoma City, Oklahoma Medicine, Texas Medical Center, Houston, Texas
Linflsey, Jefferson F., Jr., Ed.D.: Assistant to the Presi- Weihrer, Anna Lea: Mathematician, Medical Systems
den[., Southern Illinois University, Carbondale; formerly Development Laboratory
witfi the Office of Space Medicine, National Aero- Winer, David E.: Engineer, Medical Systems Development,
nau'tics and Space Administration, Washington, D.C. Laboratory
VI
CONTENTS
Cha pter Page

1 Medical Usage of Computer Sdence--John C. Townsend ............................ 1
Utilization of Computers in MedlcaJ Diagnosis ................................. 1

Models and Simulation of Biological Systems.................................. 6
Coding of Physiologic Data for Analysis by Computers .......................... 8
Analysis and Transmission of Physiological Data ............................... 10
Physiological Human Centrifuge Studies ...................................... 11
Radiation Treatment ...................................................... 1!
Medical Libraries ........................................................ 12
Patients' Medical Records ................................................. 12
Medical Schools' Computing Facilities ........................................ 13
References ............................................................. 16
Computer Applications in the Behavioral Sclences--S. B. Sells ........................ 19
New Research Capability ................................................. 19

Psychological Research Applications ......................................... 20
How Far Can Computers Simulate Behavior? ................................. 29
References ............................................................. 32
Medical Data from Flight in Space: Objectives and Methods of AnalysTs--]e_Terson F.

Lindsey, Jr.................................................................. 35
Preparation of Time-Line Medical Data ...................................... 37

Types of Analyses ....................................................... 42
Integrated-Computer-System Concept ........................................ 47
References ............................................................. 47
Automated Medlcai-Monitorlng Aids for Support of Operational Flight--Robert L. Jones

and Edward C. Maseley ....................................................... 49
General MedlcaI-Monltoring Systems........................................ 51

MSC's MedicaI-Monltorlng and Research System ............................... ,51
Early Apollo Monitoring System ............................................ 57
Improved Physlological Monitoring System for Apol/o ........................... 58
References ............................................................. 60
MEDATA: A Medlcal-lnformatlon-Management System--Tare M. Minckler and Caroline L

Horton ..................................................................... 61
Medical Information ...................................................... 61

Information Management .................................................. 63
The Future of Medata .................................................... 73
References ............................................................. 73
Development of a Computer Program for Automated Recovery of Laboratory Data--

Edward C. Knoblock, Mervyn R. Stein, and Stephen T. Paine .......................... 75
Preprocessor and Processor................................................ 76

Modes of Operation ..................................................... 76
Multiple files ........................................................... 77
VIII CONTENTS
Reports ................................................................ 77
Summary ............................................................... 78
Appendix ................................................................. 81
Continuous Monitoring and Interpretation of Electrocardiograms from Space--Cesar A.
Caceres, Anna Lea Weihrer, Sidney Abraham, David E. Winer, Larry K. Jackson, Jerome Sash/n,
Stuart W. Rosner, Juan B. Calatayud, .lames W. McAIltster, and Howard M. Hochberg ...... 91
On-Line Analysis of Limited-Time Signals ..................................... 91
Display Methods ......................................................... 93

Template Investigations ................................................... 97
Noise .................................................................. 1 O0
Predictive Values ........................................................ 104

Automatic Care and Evaluation Units ........................................ 112
Commentary ............................................................ I 16
Summary and Conclusions ................................................. 116
Period Analysis of an Electroencephalogram from an Orbiting Command Pilot--Nell R.
Burch, Ronald G. Dossett, Abbie L. Vorderman, and Bayd K. Lester ..................... 1 t7
Method ................................................................ | 18
Results ................................................................. 121
Discussion .............................................................. 138
Acknowledgment ........................................................ 139

References ............................................................. 139
Anafysls of E|ectroencephafographlc Data from Orbff by Three Di_'erent Computer-Oriented

Methods--L. D. Proctor ....................................................... 141
Zero-Crossings Technique .................................................. 141
Parametric Analysis of the Space-Fllght EEG's ................................. 142

Z/C Method of Data-Reduction ............................................. 142
Digital Smoothing and Peak-Countlng Technique ............................... 146
Analyses by Digital Smoothing and Peak-Countlng Technique ..................... 148

Welbull Statistic ......................................................... 155
Welbull Analysis of Flight EEG's ............................................ 157

General Discussion ....................................................... 161
Acknowledgment ........................................................ 162
References ............................................................. 162
10 Tracklng of an Astronaut's State by Physical Measurements of Speech--Louis V. Surgent 163
Summary ............................................................... 163

Introduction ............................................................. 163
Measurement of Speech Parameters ......................................... 164

Results ................................................................ 168
Recommendations ........................................................ 182
Acknowledgments ........................................................ 185
Appendix A Assessment of Astronauts' Changes of State for Speech Research .......... 187
Probable-State Anafysls ........................................... 187
Alternatives to Complete Probable-State Analysis ...................... 190

Appendix B Probable-State Definitions ......................................... 193
References .................................................................. 197

CHAPTER 1
MEDICAL USAGE OF COMPUTER SCIENCE

John C. Townsend
Since the early 1940's, the development of UTILIZATION OF COMPUTERS IN MEDICAL

computers has been exponential. Unfortunately DIAGNOSIS
their utilization in certain fields of endeavor has
lagged far behind their availability. The medical Since 1959 there have been several concerted
field affords an outstanding example of this lag, efforts in research aimed at use of digital com-
one reason apparently being the unfamiliarity of puters for medical diagno_s. The approach has
medical personnel with the availability of spe- been cautious. Even the most recent literature is
cific computer applications to the medical field. peppered with such phrases as "of course the
Because there is a grouting body of literature re- computers will only serve as an aid to the physi-
lating computers to medicine, which could close cian when he makes his diagnosis," and "the
the gap, reporting portions of the literature will computer is to be used to make only presumptive
contribute heavily to this chapter. diagnoses." However, the results of certain ap-
Certain observations made during collection of proaches seem to permit a more optimistic
material for this chapter succinctly point up the viewpoint. Let us look at the completed research
status of computer utilization in the medical field. and see just what can and what cannot be done
The observations were based on information in the area of medical diagnosis.
gathered from 84 medical schools, many civilian Use of the computer for medical diagnoses is
and military installations, and recognized ex- really an attempt at simulation of the reasoning
perts; the biomedical journals were surveyed process and the memory, of the physician in
also. In too many cases the working medical re- dealing with the patient's symptoms. The inlbut
spondents stated that they were just then looking data to the system are the symptoms. The litera-
into computers and hoped to be able to use them ture reports several ways in which symptoms
in the near future. Often the apparent local inter- have been gathered from patients, ms well as
est in computer use was mostly held by nonmedi- several ways in which the appropriate cluster of
cal personnel working in medical settings. On the symptoms has been established for a particular
whole the response presented a picture of ex- disease.
tremely spotty use by medical departments.
Often where computers were available, and a Obtaining Input Data on Symptoms
list of the uses of computers by resident medical Most commonly used in this regard is the self-
personnel was provided, there appeared, by in- _dministered printed form of the CornelI 5[edical
ference from examination of the scope and meth- Index - Health Questionnaire (CMI); it solicits
ods of usages, to be lack of knowledge of the full yes-or-no answers to 195 questions covering the
potential of computers in a particular area of patient's medical history and complaint. The
medicine. On the other hand, computers were patient's age and sex are added as input in-
being used very efficiently at some institutions. formation. The yes-no type of answer mates
The approach in this chapter will be separate well with the binary input requirements of the
treatment of the various categories of use of computer. It is obvious, however, that the input
computers in the medical field. Although some of system information is mainly subjective.
categories overlap, a good deal of compart- In an attempt to evaluate the use of subjective
mentalization was not too difficult. symptoms in diagnosis of disease by computer,
2 BIOMEDICAL RESEARCH AND COMPISTER APPLICATION
Rinaldo et al. (ref. 1) chose epigastric pain be- textbook on hemotology. By transfer of the data
cause it embodied the classic problems in symp- to cards and their processing in a computer, the
tom diagnosis. On the first interview, the patient symptoms as identified in the patient can be cor-
was asked to answer "yes" or "no" to the follow- related with the diagnostic criteria. Lipkin and
ing eight questions: Right-upper-quadrant pain? Hardy (ref. 2) used this approach and, on the
Clusters? Brief, irregular? Food relief? Food ag- basis of the correlation coefficients obtained,
gravation? Positional aggravation? Weight loss, sorted the cases into three groups. Group-I was
20 lb? Persistent? The patient's age also was identical with one disease category; Group-II
obtained. These symptoms were solicited as pre- with several disease categories; and Group-III
dictors of six diagnostic categories: hiatus hernia, not with any disease category. For Group-I[ the
duodenal ulcer, gastric ulcer, gallstones, func- indication was that additional information was
tional, and cancer. The six diagnoses were used needed to establish the diagnosis of any one
because they were most comprehensive and made disease. Addition of further information resulted
a list short enough for handling by the IBM-650 in the symptoms being correlated with only one
computing facilities. disease. Further observation of Group-III re-
Other investigators have dealt with more ob- vealed that more than one hematologic disease
jective symptom input. Lipkin and Hardy (ref. was present.
2) attempted to use computers in the differential Each item of information previously coded in
diagnosis of hemotological diseases; they coded each of the diseases was assigned a numerical
patients' clinical and laboratory data for appli- value. On the basis of its contribution to estab-
cation to punch cards. The data contained in- lishment of a diagnosis, the item was assigned a
formation from the case history and physical positive, negative, or zero weight. Each item
examination, and from peripheral-blood, bone- therefore might have a different weight in pre-
marrow, and other laboratory examinations. diction of each disease. The ratio of the weight
of the hospital data to the sum of all weights of
Logical Concepts in Computer Diagnosis the data of the disease was determined. The true
The next step in prediction of the disease disease was considered to be the one that scored
category to which the patient belongs is to at- highest in terms of weighted averages.
tempt to program the computer in a manner Since this pioneering study in 1957, researchers
that will relate the symptoms to the disease. The have become much more sophisticated. Perhaps
physician, in his ordinary attempt to do this, the most complete and elegant description of
dwells on the patient's symptoms, sex, and age what was to become the common approach to
and tries to assign a diagnostic value to each the "reasoning process" of the computer, in diag-
symptom for each disease possibility. He then nosis of disease, is by Ledley and Lusted (ref. 3).
attempts to relate for each disease the total Under the title "Reasoning Foundations of
value of the symptoms for each disease to the Medical Diagnosis," they presented the logical,
totals held in his memory, which were obtained probabilistic, and value-theory concepts, and the
from other patients of the same sex and age conditional probability or learning device that
range who are known to have the disease. He supports the decision of the computer and the
draws his conclusions and makes his presumptive interpreter. Their work was aimed at suggestion
diagnosis from this information. of a basis from which pr._ctical procedures could
In attempts to simulate the physician's thought be (and have been) worked out.
process by computers, certain common procedures Basic to the reasoning behind medical diagnosis
have been followed and certain interesting varia- is the concept of probability, since few diagnoses
tions have been attempted. Since the physician are made with absolute certainty. These logical
asks his memory for reference data, the computer concepts by Lcdley and Lusted start with two
likewise must have a memory. One way to estab- sources of information: the symptoms presented
lish such a memory would be, for instance in the by the patient and medical knowledge concern-
ease of hemotological-disease prediction, to list ing the symptom relation to certain disease
the symptoms of the disease from a standard entities. By use of the symbolism common to
MEDICAL 1JSAGE OF COMPUTER SCIENCE 3
the language of the propositional calculus of on utility theory. Here a mathematical model
symbolic logic as the basis for communication of that fits the operations of the physician, as he
t_he concepts involved in the logical process, makes his decision, is most useful. A good decision
patients can easily be classified according to their is the choice among alternatives that maximizes
attributes. Boolean functions are used to classify the probability of achievement of the correct
patients into more than four classes, which would diagnosis. Such strategy is mathematical in form
result from dealing with more than two attributes; and can be computerized with ease. If one is
they are therefore used to express the contents of satisfied with an expected value approach based
the logical system which includes symptoms, on a utility theory, the model for decision-making
medical knowledge, and diagnosis. The logical discussed by Von Neumann and Morgenstern
problem is determination of the diseases f. If (ref. 4) is available.
medical knowledge E is known and if the patient Crumb and Rupe (ref. 5) suggest a general
has symptoms G, the prediction is that he has plan for a logical sequence in development of
disease f. Ledley and Lusted present this idea in techniques using computers as diagnostic aids;
such a way that one need only determine Boolean the following steps are derived from their method:
function f that satisfies this formula:
(1) Test-area selection: One selects a group of
E-+ (G--*.f) diseases that are difficult to diagnose because of
similarity of the symptoms.
They call this the fundamental formula of medical
(2) Test-data compilation: Data relating symp-
diagnosis.
toms to disease are compiled in such a way that
Application of the logical basis of this diagnostic
statistical calculation of correlation constants is
system requires the display of all combinations
possible.
of a patient's symptoms. A given patient must fit
(3) Development of the correlation technique:
into one of the mutually exclusive categories of
Trial solutions are attempted and established for
conceivable disease complexes at a time. Each
the original data. The teclmique is developed
such complex is a possible medical diagnosis.
from them rather than from irregularities that
However, in the absence of certainty, one must
would be occasioned by introduction of new
quantify the probability of the patient having a
variables. The nonlinearities in preparation and
particular disease with the medium of conditional use of the correlation-constant table are tried
probabilities. The conditional probability is the
and accepted as they prove helpful in dealing
probability of a patient having a particular symp-
with further data collected.
tom or symptom complex when selected from a
(4) Adaptation of data format and computer
population having a particular disease; the
programs: One selects the appropriate form for
mathematics of this approach is a product of
input to the computer and writes the program to
Baye's theorem.
be used in the computations.
Probability enters into the diagnosis as a prob-
lem in evaluation of the conditional probabilities The system yields what are called probability-
for a single patient. Since these probabilities index numbers. While these indexes do not rep-
change constantly as further diagnoses are made, resent quantitative probabilities, they are used
the system must be corrected constantly by use in ranking of the diseases from "most likely" to
of the diagnosed cases as input data for further "least likely" as the true diagnosis for the patient.
probability calculations. In this manner tile It has been pointed out (ref. 5) that the big weak-
system is constantly improved and kept current ness of the technique lies in the fact that the
with local conditions. Such updating is a rela- disease list must contain the correct diagnosis.
tively simple matter, easily accomplished by a Crumb and Rupe reported certain computer
computer-trained person. considerations of importance. The storage re-
In connection with this approach to making quirements of the diagnostic undertaking might
computer-aided decisions, value must be con- be satisfied by a magnetic tape unit, a large-
sidered. Most often mentioned in connection capacity, random-access memory unit such as
with this problem is the one-person game based the IBM Ramac. The total operational time for
4 BIOMEDICAL RESEARCH AND COMPUTER APPLICATION
the computer is approximately 2 min to handle statistical measures upon which to base the
completely a 48X48 matrix. They believed that interpretation.
an inexpensive, medium-size, general-purpose
Computers and Programs in Medical Diagnosis
computer with magnetic-tape facility, such as
the Bendix G-15, would be satisfactory. The above discussion shows two approaches to
Brodman et al. (ref. 6) did not provide the solution of the relation between symptoms and
computer with sets of symptoms for particular diagnoses: (1) simple and multiple correlational
diseases; rather they required the machine techniques and (2) conditional probabilities.
system to develop empirically its own diagnostic Either technique requires use of computers for
criteria. The computer was provided with a priori efficient accomplishment. Various programs exist
information based on the accumulated experience for performance of such correlational procedures
and knowledge of the medical profession and was on all sizes of computers. A computer as small and
instructed how to determine diagnostic syndromes as standard as the IBM-1401 or the IB_I-1620
from each patient's sex, age, and medical com- is perfectly adequate in performing such compu-
plaint, correlated with the hospital diagnosis; no tations where the size of the matrix is not more
additional information or assistance was given. than 30X30. Larger matrices require more passes
This appears to be both a reasonable and an un- of the data through the computer and thus more
biased approach to evaluation of diagnosis by time, so that the efficiency of the process is
machine. lessened when a small computer is used.
Because the method of application of condi- Brodman (ref. S) describes the operation of the
tional probabilities works most easily when one computer in relating the input information to the
particular disease is dealt with, a rather compli- diagnosis, and van Woerkom (ref. 9) provides an
cated situation exists when the patient has more actual computer program for performance of the
than one disease. It is possible to study disease Brodman computations. In general the machine
complexes rather than single diseases, neverthe- has stored in its memory tile symptoms as
less, by dealing only with the symptoms that gathered from tile CMI and tlle patient's age and
have higher incidence within the disease category sex. The computer assigns to every complaint a
than in the total population. But, on the whole, diagnostic value for each disease and corrects
few real difficulties have been encountered in this sum for the patient's age. The computer
application of conditional probabilities in de- then relates the patient's corrected sum in each
termination of a diagnosis. disease to the mean of the corrected sums ob-
Tolles et al. (ref. 7) attempted diagnosis of tained from other patients for whom a valid
disease ill the cardiovascular system where diagnosis has been obtained. This yields L which
quantitative data are quite readily available. is the likelihood of the patient having any given
Their input data to the computer were obtained disease. Brodman used the formula
mainly from the electrocardiogram (ECG, which
provided electrical information about the heart), L = Npso/Nm5
from the bMlistocardiogram (for data concerning
the mechanical characteristics of the heart), tlle where L is the likelihood, Np is the patient's
phonocardiogram (for data concerning the valvu- corrected sum in e,_ch disease, .V,_ is the mean of
lar action of the heart), and the arterial pulse the corrected sums obtained from other patients
wave (for a reflection of the arterial system). known to have the disease, and the lower limit of
"The pathological condition chosen for study, in 5 in the denominator prevents incorrect high
regard to predictability of its diagnosis, was values from being a_igned to this factor when
left-ventricular hypertrophy. Complete measure- the mean sum for a disease is low.
ments were made on five heart beats from each Thus the computer output is tile likelihood of
of 15 subjects, and the average value of each a p_tient having each of the diseases (60 diseases
point was entered in punched cards to serve as a in Brodman's study). An L of 35 or more identi-
source of input data to the computer. Means, ties the disease in the patient. An IBM-704 used
variances, and correlations were computed as by Brodman in his computations proved to be
quite satisfactory; with it likelihood indexes can validity, the results indicated applicability of the
be calculated in less than 1 sec. technique to a sample different from the one
Tolles et al. (ref. 7, diagnosis of Ieft-ventricular used in its establishment.
hypertrophy) calculated means, variances, and When the information input to the computer
correlation coefficients; calculations were made is systematically biased, the decisions made by
from 45 subjects. Because the correlation coeffi- the system are incorrect but in a predictable
cients were calculated from a 45)<45 matrix, the manner. When the input is bimsed randomly, the
labor would have been enormous without the aid computer yields incorrect (but logical) decisions
of a computer. that are unpredictable.
In Rinaldo's study of epigastric pain, the
Valldatioa of Computer Diagnosis
accuracy of the computer in selecting a correct
The question arises, of course, of just how well X-ray diagnosis, based on an eight-item question-
the computer diagnosis correlates with a fnal naire covering subjective pain symptoms alone,
diagnosis made by well-established methods hav- was ascertained by loading of the computer with
ing the highest validity. All reports in the litera- the subjective symptoms and the X-ray diag-
ture express a great deal of concern in this area noses of 204 patients. The data were then used
and describe in detail the validation procedures by the computer for prediction of the radiological
and results utilized.
diagnoses of the next 96 patients. The computer
In this Brodman study the diagnosis predicted correctly identified 73 percent of the patients
by computer was compared with the diagnosis having hiatus hernia, 69 percent having duodenal
made by hospital physicians after el;citing a ulcers, 27 percent having gastric ulcers, 75 per-
history and performing physical and laboratory cent having gall stones, 3S percent functional,
examinations. There was wide variation among and 33 percent having gastric carcinomas. Func-
diseases in the percentage of cases correctly tional disorders caused confusion among the other
diagnosed by the computer. Where the CMI major diagnostic categories; variability of the
elicited many symptoms pertinent to identifica- data from the patients harmed the validity of the
tion of a disease, such as ulcers of the duodenum, technique. It was suggested that symptom diag-
as many as 68 percent of the computer's diagnosis could be improved by different weighting of
noses were correct. However, where the CMI the significance of symptoms and by alteration of
elicited no pertinent information concerning the one's attitude toward interview data.
symptoms of some diseases, such as benign neo- The study by Tolles et al. (ref. 7) yielded 69
plasm of the skin, no such cases were identified. significant correlations from the 990 calculated;
The machine and physician experienced in use 1 percent (1O) of the 990 could be expected to
of the CSII were compared in their ability to occur by chance alone at the 1-percent level of
diagnose 60 diseases from the CSII data ex- confidence--the level accepted as significant in
clusively. The physician correctly identified 43 this study. Therefore, some of the correlations
percent of the cases (other than psychoneurosis) that could not be explained on a rational basis
while the machine correctly identified 48 percent., by the authors should be attributed solely to
but the difference in percentage was not statis- chance. In order to permit the computer to
tically significant. The physician was clearly classify a given individual as normal or as one
superior to the machine in diagnosing psycho- having a given pathology, Tolles et al. used the
neurosis. In some cases the physician could means, variance, and correlation coefficients to
diagnose correctly from the CMI data when the construct a multidimensional probability model;
machine could not; however, the machine made thus the probability that a given patient would
few incorrect diagnoses. There was no significant belong to a particular disease or control group
difference between machine and physician in could be calculated. The results showed that the
incidence of incorrect diagnoses: 4.9 and 2.0 ECG, the ballistocardiogram, and the arterial
percent, respectively. pulse wave could separate the normals from the
In a cross-validation study in which samples non-normals, but that the phonocardiogram
from 1948-49 and 1956 were compared for could not. The ballistocardiogram and the ECG
could separate the hypertensive and the valvulitis effective. Presumptive diagnoses by computer,
groups from the others, but the arterial pulse based on both subjective and objective data in
wave could not. The use of such variables im- sufficient quantities, would of course be com-
proved separation of the normals from the pletely satisfactory.
non-normals by 29 orders of magnitude. The Schwichtenberg (ref. 10) has made the point
authors feel that the method warrants general that extension of the use of computers into the
application and have no doubt of the value of life sciences other than medicine, and indeed
computers for diagnoses. their use within medicine in such activities as
The validity of the approach used by van computer diagnosis, depends on development of
Woerkom and Brodman is found in their discus- a standard means of communication; he feels
sion of their conditionM-probability approach. that promulgation of a "current medical termi-
They found that a modified conditional-proba- nology" would expedite such communication. The
work of the American Medical Association in de-
bility approach, when tested with their sample,
produced certain expected inconsistencies; almost veloping and editing Current Medical Terminology
always categories were favored having the largest for 1963 and 1964, as well as Current Surgical
number of attributes with relatively high fre- Terminology and Current Medical Surgical Ab-
quencies. They failed to get probabilities high breviations, is either complete or well under way.
enough for consideration from categories having Current Medical Terminology for 1963 (revised in
small sample sizes and small numbers of at- 1964) has already been coded for computers.
tributes of relatively high frequency. They re-
marked that the latter often would have been
MODELS AND SIMULATION OF BIOLOGICAL
identified by a physician, while the machine SYSTEMS
failed to do so. The validity of their approach
rests on the fact that, if a patient's cluster of Since the introduction of computers, work on
symptoms resembles the cluster of the average development of mathematical models of biological
patient, he can be a_qsigned to the correct cate- processes has progressed at a tremendous rate.
gory and thus validly diagnosed. Complicated systems have been simulated, with
The literature makes it clear that the com- the effects of variables within and upon the
puter's advantages for medical diagnosis are not systems tested, in attempts at more complete
questioned. Among these advantages are the understanding of their workings. Although need
facts that the complete memory of the computer-- for such efforts was recognized quite long ago, it
but not of the man--is available for making a was not until the advent of speedy and easy
diagnosis; that the diagnostic time is practically dealing with such complicated models by com-
instantaneous once the symptoms input is in- puters that real progress was made.
serted in the machine; and, through use of an Of particular interest in the development of
instrument such as the CMI, a presumptive models of biological systems is the analog com-
diagnosis mny be made even before the physician puter; in general it is smaller and less expensive
sees the patient. On the other hand, the most than its digital counterpart. However, the real
severe limitations of the technique are the scope reason for use of analog computers in modeling
and quality of the symptom input. If the input and simulation is that one is essentially able to
is from only the subjective complaints of the deal with a direct electrical analog of a physical
patient, the diagnosis varies with the accuracy phenomenon. Taskett (ref. 11) reports that the
of that set of data. It is the old story: You get analog computer permits the research worker to
no more out of an anMysis than you put into it. get the feel of a physical system because the re-
A patient has more significant data relevant to suits- are displayed graphically and faster than
his diagnosis than is contained in his medical is customary with a digital computer.
history. To the extent that these other items, The analog computer is usually better at per-
such as laboratory and radiological data, are im- forming mathematical operations on measured
portant to the diagnosis, their exclusion as variables such as are encountered in electro-
information renders the diagnosis so much less encephalographie and electrocardiographic work.
If the researcher is interested in curve-fitting, determination of volume and rate constants from
density-discriminations, power spectra, and auto experimental data. The model was applied to two
or cross correlations, the analog approach is systems: The first dealt with use of the analog
preferable. Thus, for the purposes of model as a device for analysis of data, the specific
building and simulation of biological systems, situation chosen being flow and diffusion of the
the analog computer is preferable. However, cerebrospinal system; the second dealt with
when it comes to handling large ma_ses of data metabolism of iodine by the human thyroid
from a compilation and statisticM-processing gland. It was found that the voltage curves pro-
point Of view, the analog is far inferior to the duced by the compartments bore a scale relation
digital computer. Likewise, when one is not to the activity curves of the biological model.
interested in understanding the basic operations A good example of how simulation of a system
of a system but only wants, in a practical sense, yields new hypotheses, concerning the explana-
to make decisions concerning the operation of a tion of phenomena taking place within the system,
particular subject on a real-time basis and on is reported by Morse (ref. 14). He evaluated the
line, it is better to use an analog-to-digital con- significance of various physiological parameters
verter of the biological data and to let the digital that contribute to the human ballistocardiogram.
computer take over. Input information to the model consisted of an
We now review current efforts at modeling and arterial pulse wave and an arterial ballistogenic
simulation, with an eye to understanding of the function, which is a mathematical representation
scope of such modeling efforts, and demonstration of the arterial system in which the pulse wave
of the role played in them by computers. travels. A digital computer performed all the
DeLand (ref. 12) reports an attempt to simulate necessary mathematical calculations. Morse found
a large biological system by use of an analog that, although the digital computer is slower than
computer--a method based on Gibbs' free-energy the analog computer in performing such calcula-
hypothesis. A mathematical format was em- tions, it has greater flexibility with regard to the
ployed, with the actual computations being input data and the mathematicM processes that
accomplished by the method of steepest descent. follow. This investigation showed that the clinical
Tile respiratory function of blood in the human ballistocardiographic abnormalities of I- and
lung was chosen as the subject for the model. The J-wave diminution may be explained as due to
method was based on the postulate that chemical changes both in the slope of the rising pulse
mixtures tend toward a reaction equilibrium that pressure and in the elasticity of the great vessels.
minimizes the potential, or free energy, of the More exact determinations would allow further
system. The solution of the equilibrium problem elaboration of the ballistocardiographic model.
consisted of a set of mole numbers that minimized Of particular interest in the area of modeling
the free-energy function. The time-dependent and simulation is the research of Crosbie et al.
aspects of the system were also simulated. With (ref. 15). An electrical analog was developed to
use of a digital rather than an analog computer simulate the physiological responses of man to
to achieve the same results, it was discovered heat and cold; it was based on the fundamental
that the digital approach gave more accurate and equation for heat balance, developed to account
reproducible results but had a longer solution for heat-loss by radiation, convection, and
time. Since it was desired to simulate the dy- evaporation. Since physiologic temperature in-
namic system in real time, the analog computer volves at least three types of control mode
was considered better because of its characteristic (proportional, rate, and "on-off"), one must be-
parallel computation and its fast solution. come involved in nonlinear differential equations.
An apparatus for simulation of compartmental The simulator solved such equations in its at-
biological systems was developed by Brownell tempt to predict steady-state situations of rectal
et al. (ref. 13). The analysis was accomplished by temperature, skin temperature, metabolic rate,
use of an electrical analog. The apparatus was vasomotor state, and evaporative heat-loss during
designed for rapid simulation of linear compart- both rest and exercise. Equations based on the
mental biological systems and for numerical controls mentioned permit simulation of dynamic
responses to sudden change in environmental from a simulated system as hypotheses to be

temperature, air velocity, relative humidity, and verified in the real system that is being simu-
metabolic rate. lated, he stands only to gain from the simulation
Warner and Cox (ref. 16) investigated the experience.
effect on heat rate of stimulation of the sympa- Great strides have been made in construction
thetic and vagus efferent nerves to the heart; of models of biological systems. Equally well
they hoped for information concerning the nature known are the methods of setting up of models
of the physical and chemical events involved in relating man to his environmental forces. Results
this transformation and the dynamic and steady- yielded concern about (1) the understanding of
state parameters of this link of the heart-rate reactions within a human physiological system,
control system. Computers were used to increase (2) the relation of the action of one physiological
the accuracy of voltage-to-frequency conversion system upon another within the human, and (3)
as well as to perform other important functions the reaction of the total human system to in-
in handling of the situation. Differentiations took ternal and external environmental forces. Besides
place on an analog computer. The entire pro- the nicety of knowing what happens in each of
cedure leading to the results was tied to com- these instances, knowledge can be gained from
puters at every step, so that the accuracy in manipulation of certain variables and combina-
simulation and calculation far exceeded what tions of variables to determine their effects on
was feasible without computer aid. the total-system operation. If the model holds
Many investigators have made similar use of in known instances, and if then its components
computerized models. Prominent examples are are experimentally manipulated iT_ previously un-
the work of Clynes (ref. 17) on laws of respiratory known relations, the effects of these changes in
sinus arrhythmia, as derived from computer the situational variables upon system performance
simulation; Pace (ref. 1S) on analog-computer can be predicted.
simulation of a neural element; and Wood (ref.
19) on mathematical analysis of indicator dilu- CODING OF PHYSIOLOGIC DATA FOR
tion techniques, where electrical analogs of ANALYSIS BY COMPUTERS
mathematical models were used to express the
distortion of an indicator dilution curve during Physiological phenomena, picked up directly
traverse of a segment of the circulation. from subjects by an appropriate lead system, can
These studies point up the tremendous possi- be recorded on magnetic tape and processed di-
bilities of models developed for research purposes. rectly by a digital computer. One such system is
Modeling and simulation are not feasible if the reported by Taback et al. (ref. 20); they used a
computations are done by hand, nor can one corrected orthogonaI three-lead system. When a
construct a real-time model that does not depend significant cardiac cycle is selected by a tech-
on rapid calculation for its maximum efficiency. nician from an electrocardiographic magnetic
Use of the analog computer even in following a tape record, it is automatically sampled at
simple function is highly desirable. Use of both 1-msec intervals, and the numerical v,qlues are
types of computers jointly for making real-time, stored on a magnetic tape in a form acceptable
on-line analyses of complex functions is indis- by a digital computer. The researcher is then
pensable. free to establish various objective analyses for
Biological systems of interest to medicine are the data and to proceed under each of the pro-
seldom simple. 5[ultivariate situations are routine grams. The authors offer several reasons for
if a large percentage of all relations within a utilizing an automated approach to the study of
system are to be understood. It must be remem- ECG's. The gist of the reasons follows:
bered, however, that correlation is not proof; (1) The method provides a completely objective
while the end product of the syslem being simulated analysis that, of course, does not reflect error due
may be the same as that of the simulation system, to different interpretations by different readers.
it may have been produced by different means. (2) Fidelity of data is maintained in accuracy
However, when the researcher treats his results and in frequency range since no use is made of
MEDICAL LrSAGE OF COMPUTER SCIENCE 9
direct-writing instruments with their restricted tracings, and ballistocardiograms, can be analyzed
frequency response. on the same equipment, provided that certain
(3) The convenience of magnetic tape for stor- minor modifications are made to the equipment
age and retrieval of large amounts of data, in a (ref. 22).
form suitable for comparison, is well known. Pipberger et al. (ref. 23) have been concerned
(4) Because of the high speed of the digital with the advantages and disadvantages of scalar
computer, large masses of data can be analyzed lead recordings, vector-loop displays, curves of
statistically with great efficiency; this computer spatial magnitude, orientation and velocity, polar
provides an easy and flexible method for in- vectors, and eigenvectors. A new method of
vestigation of many new and involved approaches differential electrocardiography described was
to analysis of the data. based on computer ranges of various diagnostic
Using an IB51-704 a central processing facility entities. It was pointed out that the leads that
accepts the original analog tape recordings and, discriminate best between diagnostic groups are
under the guidance of a medical technician, pro- obtained by resolution of orthogonal leads. The
duces a digital magnetic tape from the informa- informational content of three corrected orthog-
tion in the form of IBSI words. The data are then onal leads is comparable to that of the standard
processed on the 704 according to the program 12-lead ECG. Transformation of data appears to
selected by the researcher for the particular be necessary for recovery of the clinical informa-
analysis. tion contained in the 12-lead system.
Tile speed and efficiency of this objective ap- An automatic method is reported (ref. 24) for
proach are quite obvious. 5_Iore research work processing mass data in clinical medicine, known
seems to be needed in determination of criteria as FOSDIC (Film Optical Sensing Device for
for the analyses. Furthermore one should note Input to Computers). An analog scanner, acti-
that, by coupling the results of the analysis with vated by a digital computer, automatically reads,
a program for diagnosis of cardiac diseases by recodes, and transcribes the physiological data
computers, a presumptive diagnosis can be on magnetic tape in binary language. Research
achieved. The techniques described here for the on the system is aimed at its evaluation for
cardiac analysis are applicable to other biomedical reduction of clinical information and correlation
data. of analog records, such as are provided by the
A great deal of concentrated work on this ap- ECG, with other clinical data.
proach has been supported by the Veterans Ad- Many more biomedical data are produced and
ministration; it is summarized by Berson et al. normally collected than are required for analysis.
(ref. 21). Analog-computer methods for analysis As a result some researchers have expressed con-
were first considered but then rejected in favor cern regarding the sampling of data in excess of
of the digital approach, primarily because of the their needs (ref. 25). This is an important prob-
latter's flexibility. An analog-to-digital conversion lem that requires solution. Emph_is on this
of the ECG was made from the original magnetic difficulty has been increased by its enhancement
tape recordings. Specific complexes of the ECG, of the problem of transmission of medical in-
free of disturbing artifacts, are chosen by an formation over limited-capacity telemetry chan-
operator, who then presses a button setting the nels. Bayevskiy (ref. 26) suggests that information
automatic process into motion. The ECG on each theory may solve the problem. Only between 10
channel is converted into digital form and re- and 1 percent of the data collected in some situa-
corded on digital tape at the rate of one con- tions may be needed for satisfactory analyses.
version per millisecond. The tape is then analyzed With an ECG one is dealing with data containing
by use of the IB.Sf-704. The results of tests of approximately 400 binary units per second. Re-
many different programs showed that completely duction of this enormous number of bits, by
automatic computer analysis of the P-QRS-T elimination of parts having no practical use,
complex is highly accurate and of value in large- requires special medical devices. By means of such
scale statistical and epidemiological studies. Other devices the electrocardiographic information can
analog data, such as phonocardiograms, pulse be reduced to a symbolic code.
Transposing Bayevskiy's work on the telemetry ware section, reduced by a computer program,
system to the problem of coding ECG's for input and presented in tabular form for analysis. Cady's
to a computer, we see that we can apply his work (ref. 32) on the development of a computer
principle quite well. He states that the principle program for measurement of ECG-wave charac-
of coding an ECG is based on identification of teristics is notable in this area.
the alpha, beta, gamma, and delta constituents; Glaser's (ref. 33) automatic system for proc-
determination of the integral values of the bio- essing microelectrode data is of interest in con-
currents at the output of each filter for a period nection with the uses of computers for analysis of
of 2 sec; and the general integral values of the biomedical data. He reports that the data-
biopotential in each lead. Thus 20 pulses are processor designed and built is useful for trans-
available each second, with reduction of the data ferring the single-unit microelectrode recordings
by 99 percent. Although there is no set of mathe- from analog magnetic tape to punched paper
matical principles for production of statistical tape while preserving the exact time sequences
codes for medical data, there is little doubt that of the neurological events. A Flexowriter can
the computer itself will be the vehicle for doing yield a hard copy from the paper tape or produce
the coding as well as the processing of data. IBSf cards for processing by a computer. Thus
Schmitt and Caceres (ref. 27) voice the hope one can utilize the digital computer in studying
that computers directly accepting biologic data such phenomena as spontaneous activity and
in symbolic or pattern form will eventually evolve. adaptive cell processes during and after periods
More research is needed in the area of coding of stimulation.
for acquisition, anMysis, and presentation of Such studies a_s these demonstrate the scope of
brain-wave data. Computers have been used the digital computer in study of bioelectrical
extensively in development of techniques for phenomena. In most cases both procedure and
dealing with measurements derived from electro- apparatus are available, so that maximization of
encephalographic recordings (refs. 28 and 29, the use of computers in coding and processing of
and subsequent work by the same authors). such data is possible.
It is most handy for a physician to have a
digital readout of physiological data when his
ANALYSIS AND TRANSMISSION OF
task is to monitor the physiological status of an PHYSIOLOGICAL DATA
individual. Siahaya et al. (ref. 30) report a tech-
nique that yields digital readout of systolic and Computers have been used extensively for
diastolic blood pressure, heart rate, and respira- various mathematical and statistical procedures
tory minute volume, applicable to wireless during the collection and analyses of physiologic
telemetry from aerospace vehicles; blood-pressure data. The neurophysiologist is particularly fortu-
data obtained indirectly in analog fashion are nate in having such a tool; let us cite a specific
converted to digital information. instance. The computer technique for autocorre-
Heart-rate QSR complex of the ECG, after pass- lation is of great value; it is a method for compar-
ing through wave-shape-recognition and noise-re- ing a one-time series, such as an ECG, with itself
jection circuitry, is totaled by a digital computer. displaced in time. With a lag of 1, 2..., many
A voltage-to-frequency converter feeds into a comparisons are made. If there is a signal in the
digital counter, which in turn converts analog tracing that is obscured by noise, and if its re-
information to digital and integrates to yield currence is phase locked, the correlation of the
respirato.ry minute volume. The data are re- tracing with itself is highly positive when the
corded in a predetermined sequence once each lag introduced equals the repeating period of the
minute, and the process is controlled by a pro- signal and whole multiples of the repeating period.
grammer. The autocorrelation is smaller in value when other
Hovey et al. (ref. 31) report design of an auto- lag periods are used. Thus one can pull informa-
matic data-acquisition system that is very useful tion from the data that is otherwise obscured.
for minimizing large amounts of biomedical data. Cross correlations axe of equal value. With this
The data are digitaIized and recorded by a hard- method of correlation, two different tracings are
MEDICAL USAGE OF COMPUTER SCIENCE 11
compared for determination of whether or not immediately analyzed by a digital computer, with
they contain common characteristics. verbal delivery of the resultant diagnosis over the
Vinograd (ref. 34) suggested use of computers same dial-telephone system. Approximately 8
for investigation of new physiologic parameters min elapsed between the patient's entry into the
such as rate of change, and rate of rate of change. ECG room and delivery of a complete P-QRS-T
From this suggestion Townsend (refs. 35, 36) de- analysis to the transmitting physician.
veloped the necessary methodology. Another physiological recording was trans-
In detection of evoked responses, one can de- mitted by a relay satellite. On April 23, 1965, an
tect most efficiently with a computer responses ECG was transmitted from England via the
evoked by stimuli (ref. 37). The procedures, British transmission system to the relay satellite
including the programs for performance of auto and thence to the receiving station in New
and cross correlations, are now readily available Jersey. It then went by land line to the Mayo
to the researcher and can be routinely employed. Clinic at Minneapolis where it was fed into a
Other procedures that are quite helpful and computer. From the printout, a diagnosis was
equally available are the methods of phase detec- made and the results were back in England
tion and use of averaging techniques on the within 1 min. There appears to be no obstacle to
computer for analysis of physiological tracings. such long-range diagnosis of brain disorders, with
Adey and Walter (ref. 38) elaborate on these proper use of satellite transmission and computer
approaches, pointing up the advantages of the analysis. It is hoped that coupling of computer
computerized techniques for performance of these diagnosis with this system will replace the physi-
procedures. In the area of space-flight medical cisn until after a presumptive diagnosis is made.
data, Lindsey (ref. 39) has introduced a valuable
concept linking statistical procedures with a
PHYSIOLOGICAL HUMAN CENTRIFUGE
computerized time-line approach.
STUDIES
Analog-computer techniques have been used
in many physiological analyses, as have digital Wood et al. (ref. 44) at the 5Iayo clinic have
approaches. Typical applications of the analog been studying the reactions of a physiological
techniques are those by Murphy and Crane (ref. system to transient reproducible degrees of stress,
40), involving the analog computation of respira- as a useful means of discovering the mechanisms
tory-response curves; in Randolph's research of action of the system. Through acceleration
(ref. 41) into application of analog computers to they produced reactions in the cardiovascular
ESR spectroscopy; and in the studies of Chance system that resulted in sudden decrease in
et al., with the electric analog computer, of the arterial pressure at head level, stagnant ano×ia of
mechanism of catalase action (ref. 42). the retina and brain, and hydrostatic effects of
There has been considerable interest in trans- acceleration, which alter the ventilation-perfusion
mission of physiological recordings on an on-line ratios in the lungs. A multiplicity of variables
basis. Of course the techniques of telemetry are had to be dealt with, presenting a task that
quite well known and largely standardized. Of was ideally suited for computer solution. The
more general interest to the medical profession is hardware used is well described, including the
the effort to use telephone transmission of physio- 5IAYDAC (Mayo analog-to-digital conversion
logical tracings, such as the ECG, for on-line system; ref. 44).
computer diagnosis. Berson et al. (ref. 43) care-
fully considered this problem and report success;
RADIATION TREATMENT
their transmission system included a regular dial-
telephone network for sending and receiving One obstacle to application of automatic
electrocardiographic information. In accuracy the calculation of multifield dose distribution is the
received data were superior to those transmitted lack of a workable mathematical description of
by analog systems, since a pulse-code-modulation the percentage-depth dose distribution within a
system, with parity bit transmission and check- single beam. A formula was found by Sterling
ing, was employed. The data received were et al. (ref. 45) for approximating the percentage-
]2 BIOMEDICAL RESEARCH AND COMPUTER APPLICATION
depth distribution resulting from a _°Co beam, of (subset) are more closely related to each other
any portal size, at 80 cm SSD. Their equation than to the rest of the universe to which they
can be used to yield computer printouts of com- belong. The method (ref. 48) involves a statistical
bined multifield distributions. With a different technique that can be used for computation of
program it can also print out the combined the word associations of use in retrieval of docu-
isodose curves directly to scale, with small error. ments from a collection in which the documents
Thus one can plan and replan treatments ac- are described by index words occurring in the
cording to each individual patient's needs (ref. 46). text or in abstracts of the documents. The com-
puterized technique shows evidence of being
useful in large-scale associated-document retrieval.
MEDICAL LIBRARIES
A good many scientific libraries are computer- PATIENTS' MEDICAL RECORDS

izing their card catalogs; some produce catalog
cards and others have gone to book catalogs. The need for automatic information-processing
Kilgour et al. (ref. 47) further report that some of hospital records has been made vividly clear
have gone to an information-retrieval system for (ref. 49). The population explosion will steadily
catalog-card information that utilizes a large, increase demands on the modern hospital for
high-speed computer. Little, however, is being better facilities, staff, and medical services. The
done to computerize the retrieval of catalog and combination of more patients, with better medi-
index information except for the MEDLAI_S cal coverage, and more medical knowledge will
system of the National Library of Medicine and result in much more data that will have to reach
the American Society of Metals' information- quickly the men maldng the decisions. Further-
retrieval system. Both these systems are based more, all medical data on one patient or on all
on use of magnetic tape. Kilgour et al. also discuss patients--past and present--should be available
the Columbia-Harvard-Yale Medical Libraries' for different multiple-item correlations to be used
Computerization Project; its goal is increase in for treatment or research. By proper handling of
the speed and completeness with which a user patients' medical records, collected data will be
obtains catalog and index information in a disseminable throughout the hospital as needed;
library. It is hoped that other systems like errors will be eliminated, because the information
MEDLARS will be included in the library's will be transmitted electronically; statistical in-
facilities. The complete system, including both formation regarding the inpatient population will
hardware and software, is hoped to be sufficiently be made available routinely or on demand; the
generalizable for use by most conventional li- machine will ask for an instruction if one does
braries. An on-line computer catalog, located at not come when it is due; it will be possible to
one institution, will allow the processing of re- select cases as a group for studies, surveys, or
quests from various information stations, located research; problems associated with the location,
at other libraries, by means of a telecommunica- filing, and retrieval of patients' records will be
tions system. A random-access memory unit will virtually eliminated; and it will be possible to
hold the catalog files. It is planned that the com- determine trends, and so anticipate conditions
puter will hold the catalog files. It is also planned developing within a patient or within a hospital.
that the computer programs will be designed for A list of problems has been published (ref. 49)
the IBM-1401, 4K-core, two-tape-drive com- with which the medical librarian may expect to
puter; the 1401 is in most general use, being used deal. Among them are definition and determina-
at five times as many instMlations as its nearest tion of satisfactory solutions to medicolegal
competitor, the IBM-1620. aspects of automated record systems, establish-
There has been some interesting research in the ment of a set of recognition rules to limit access
area of "clumping" for associated-document to persons entitled to address the system (for a
retrieval. A clump, in terms of word association, given purpose or for particular data); and decision
is a group of objects, from some universe of on the type of output copy for certain demands.
objects, such that the members of the group Of course the first step in conversion from a
manual method of handling medical data to art One large-scale effort toward establishment of
automatic one involving use of computers is a computerized system for handling of extensive
performance of a system analysis. Slavin (ref. 50) data and records was made by a large northeast-
says that this step entails analysis of the com- ern general hospital. A Hospital Computer
ponents comprising the existing patient-data Project examined the feasibility of use of a com-
system; the clinical folder of the individual patient puter to improve patient care and to provide
is the most significant component. Slavin further new techniques for research on information in the
points out that there are two major types of data medical record. The goals of the project were as
contained in the medical record: hard--numerical follows:
or identifying data such as for personnel-identifi- (1) Use of a time-shared computer, with remote
cation, laboratory-test results, etc.; and soft-- input-output devices, to increase the rapidity trod
evaluative information expressed in narrative accuracy of collection, recording, transmission,
form, consisting mainly of physicians' comments. retrieval, and summary of information
Entering of these data in a computerized system (2) Decrease in the amount of routine paper
presents a problem, two aspects of which are (1) work required of the nursing staff
recording of the essential data and (2) their (3) Arrangement and consolidation of informa-
collection from various locations in the hospital. tion for effective utilization by the medical staff
The problem of simplification of the collection (4) Development of a system that would store
and recording of data in a standard way has large amounts of complex medical information for
been coped with very well in connection with rapid and easy search and retrieval for facilitation
medical records of astronaut candidates (refs. 51 of clinical research
and 52). Flight surgeons recorded the results of A status report indicated that one of the biggest
physical examination of the candidates by check- factors in successful operation of the system is the
ing marks on mark-sense cards for reading by a education of the hospital staff at all levels. A
computer; the medical histories and results of questionnaire, seeking personal reactions of the
laboratory tests could be recorded similarly. Ad- staff, led to the conclusion that the greater the
vantages and disadvantages of more than one knowledge and involvement of the participant
approach to the card system are discussed (refs. the more favorable and helpful his attitude.
51 and 52), the conclusion being that despite However, there appeared to be a deeper reaction
disadvantages the mark-card system is superior against the total picture of automation. An ef-
to the usual coded-IBM-card method using work fective educational program must accompany such
sheets and punch-card operators. an installation if it is to be efficient.
Lindberg (ref. 53) reports on the program sur-
rounding installation of an IBM-1410 computer
MEDICAL SCHOOLS' COMPUTING
at a university's medical center. The initial
FACILITIES
project involved development of a system for
reporting all clinical-laboratory determinations Considerable materials were received from the
through a computer. The data were transmitted medical schools contacted during this study; some
to the ward by a preliminary system based on mailed boxes of publications, descriptions of their
an IB51-1912 card-reader and Teletype Corpora- facilities, and lists of projects completed by
tion transmitting equipment. Thus the data were personnel using the facilities. The response in
already on punched cards and ready for processing terms of facilities available or planned at the
by the computer. In a new system under develop- medical schools was so overwhelming that the
ment, the data pass through the computer, before topic cannot be treated on a school-by-school
being transmitted, for correction of errors and for basis; moreover, some schools placed restrictions
general maximization of quality control. Lindberg on publication of some of their materials. An
points out the need for suitable coding procedures overall review of the materials received will be
for the interrogative medical history and the presented with the aim of establishing the nature
physical examination. Later he deals further with of the average facility now either planned or in
the technique (ref. 54). operation.
14 BIOMEDICAL :RESEARCH AND COMPUTER APPLICATION
The title of this section, indicating that it deals virtually all operations are routine at present. As
with computing facilities of medical schools, may yet there are no tie-ins with other computing
give the reader a wrong impression. In some in- facilities, but a teleprocessing link with a larger
stances the dean of a university's medical school computing center is contemplated. There are no
sent materials that reflected the availability of special data-handling or data-reduction tech-
its computing facilities on a university-wide niques for physiological data. Analog-to-digital
basis, although in fret the university's facilities conversion is anticipated in the near future. No
were available to the medical school. computer switching techniques are utilized.
Generally the medicM school lagged behind the Below are described the computing facilities
rest of the university's schools in the magnitude considered among the best available at larger
of its computing facilities. Some medical schools medical centers where medical research is ex-
had no computing equipment that they could tensive. The material is taken partly from the
call their own. The average medical school, if it publication of a facility that shall not be identified.
owned a computing facility independently of the The current basic equipment is an IB_I-1401
university's other schools and centers, had a computer with 16 000-character core memory; it
small computer such as the IBM-1401 and an includes a 1402 card reader and punch, a 1403
odd assortment of analog computers often home- printer, four 729-IV tape units, two 1311 disk
made and hybrid in type. At medical schools at drives, and an IBM-1231 Optical _Iark Page
which individuals had special interests in use of Reader. Auxiliary card-handling equipment in-
computers for modeling or simulation of bio- cludes a sorter, reproducer, and interpreter and
medical process, there was an abnormally high four 026 key punches.
concentration of computer activity, supported by Access is available to a computer center operat-
either an exceptional computing facility at the ing a directly coupled system consisting of an
school or a tie-in with a large facility in the IBM-7040 connected to an IB31-7094 computer.
vicinity. The 7040 handles all input/output and buffering,
while the more powerful computer, the 7094,
Software and Hardware at Medical Schools
compiles, _sembles, and executes jobs. The 7094
Below are listed the computing facilities of one is a high-speed binary machine with 32 768 ad-
respondent in the study; the school is average in dressable locations of 36-bit word size. A large
that its hardware is typical of that available in variety of programming l'mguages may be used,
and/or planned for most medical schools: and an extensive library of available programs
may make it unnecessary for the user to write
Available : new programs.
Two source languages are available as part of
1 IB.5[-1620, 20K
the operating system of the 1401: FORTRAN
1 IB_I-1622 card reader-punch
and AUTOCODER. Source-language programs
1 IB_[-1443 printer
in FORTRAN are treated as input data for the
2 IBM-1311 disk drives
FORTRAN compiler program; the compiler is
20 IBM 1316 disk packs
maintained on magnetic tape and is available for
Supporting unit-record equipment
use whenever a job is processed. The compiler
checks for grammatical errors in the source
On order:
language and translates the program into object
1 IB.SI 360 model 30E language for the 1401 computer. FORTRAN
1 1403 printer compilers are available for most computers in
2 1403 disk drives use today, so that a program written in this
1 1402 card reader-punch language can be translated and run on another
1 1050 remote terminal machine with only minor changes.
When a program is to be used repetitively and
The computer at this medical school is used for has been thoroughly checked for ensurance that
business and accounting, research, and teaching; it operates properly, it can be made available as
an object deck, punched in machine code that IMP029 Bartlett test for homogeneity of
operates the computer directly. Attempts are variance
made to collect such programs of general useful- IMP030 Adjustment for heterogeneous vari-
ness and make them available to others. Some ance
programs are in such general use that they are ISIP031 Correlation coefficients, with miss-
kept on file on a magnetic-disk unit, and with ing data
control cards they can be called for u_e; the ISIP033T Test-paired and unpaired data
FORTRAN compiler and the AUTOCODER IMP034 Probability chart
assembler are programs kept available in this IMP035A Frequency distribution
fashion. I5:IP036 5Iax-min
These general-purpose programs are currently IMP037 Z-test
available, all written in FORTRAN: IMP039 Exponential fit
ISR002 LogF
16 × 16 Correlation
Direct-difference t-test When one finds specifc laboratories as a func-
Polynomial curve fitting tional unit within a medical center, he also finds a
t-Test/F-test computing facility that reflects its special needs.
ChP If the laboratory is one that makes great use of
Regression lines biomedical data collected in analog form, a com-
Analysis of variance puter complex is established to deal with specific
Life table and survival rates as well as general needs. Table 1 is the published
Mantel/Haentzcl chP analysis list of hardware and software available at or
under study for one typical laboratory dealing
This computer center has also obtained the
with biomedical data, and established in connec-
following MEDCOMP programs:
tion with a large midwestern medical center.
IMP001 Means and standard deviations The general picture of computer use in and
IhIP004 Linear fit around the medical schools of this country is one
IMP005 Analysis of variance, one-way of great variation. Some schools reported no use
IMP006 Analysis of variance, two-way; no of computers for biomedical data. :_[ost schools
replication have either their own computing equipment or
IMP007 Latin square ready access to some equipment on the campus;
IMP008 Analysis of variance, two-way; no in these cases the amount of research conducted
missing data in conjunction with limited and unspecialized
I_,IP009 Analysis of variance, three-way; no computing equipment was often surprising. At
missing data some institutions--invariably the larger medical
IMP010 Scattergram and grapher schools near large urban areas and well supported
IMP012 Multiple regression by research grants and other outside sources--
ISIP013 Polynomial fit one finds highly sophisticated use of computers by
IMP014 Frequency table generator workers in the medical area; not in all such cases
ISIP015 5Iarshall test is there a large computing facility at the school,
IMP018 Analysis of covariance but usually there is a tie-in with a large nearby
IMP021 Matrix inverter computer complex. Such lie-ins bring the best-
IMP022A Expanded histogram available computing facilities to the researcher
I5IP024 Biserial correlation coefficients
on a time-sharing basis. Smaller schools seem to
IMP025 Generalized accumulator
give more consideration to having a basic com-
IMP026 Numerical integrator
puter, such as the IB._[-1401 or the IB._I-1620,
IMP027 Two-way analysis of variance,
with missing data plus a tie-in with a facility having an IB._:I-7090
IMP028 Analysis of variance, three-way, or the equivalent; in this way, maximum facilities
with missing data are available at minimum cost.
TABLE 1.--Equipmenl at One Typical Laboratory REFERENCES
Quantity Description Remarks 1. RINALDO, J. A.; SCHEINOK, P.; AND RUPE, C. E.: A
Mathematical Analysis of Epigastric Pain. Symptom
Hardware available Diagnosis, vol. 59, no. 2, 1963, pp. 145-154.
2. LIPKIN, M.; AND'HARDY, J. D.: Differential Diagnosis
IBM-1620 model-II 1/0 typewriter; auto- of lIematologic Diseases Aided by Mechanical
CPU; memory, 20K matte floating pt.; Correlation of Data. Science, vol. 125, March
auto. divide; in- 1957, pp. 551-552.
direct addressing 3. LEDLEY, R. S. ; AND LUSTED, L. B. : Reasoning Founda-
1 Additional 20K core Model 1625 tions of Medical Diagnosis. Science, vol. 130, no.
storage 3366, July 1959, pp. 9-21.
1 Card reader-punch Model 1622; input 4. VON NEU_ANN, J.; AND _:IORGENSTERN, 0.: Theory
speed, 500 cpm; of Games and Economic Behavior. Princeton
output speed, 250 University Press, 1944.
cpm 5. CRUMB, C. B., JR.; AND R UPE, C. E.: The Automatic
1 Magnetic-tape unit Model 7330 and 1921; Digital Computer as an Aid in Medical Diagnosis.
with control high-density Proc. Eastern Joint Computer Conf., vol. 16,
3 Card printer-punch Keypunch machine, 1959, pp. 174-180.
model 026 6. BRODMAN, K.; VAN WOERKOM, J.; ERDMANN, A. J.,
JR.; AND GOLDSTEIN, L. S.: Interpretation of Symp-
Hardware under consideration • toms with a Data Processing 5Iachine. A. M. A.
4 b Disk storage unit Arch. Internal Meal., vol. 103, May 1959, pp.
776-782.
6 b Magnetlc-tape unit
Paper-tape reader- On-line use 7. TOLLES, W. E.; STEINBERG, C. A.; CARBERRY, W. J.;
punch AND FREIMAN, A. lI.:Experimental Techniques and
High-speed printer On-line use Results of a Study Using a Computer as a Diagnostic
Digital X-Y plotter On-line use Aid. Trans. N.Y. Acad. Sci., vol. 23, no. 3, Dec.
100K b Additional core stor- 20K per unit 1961, pp. 246-258.
8. BRODMAN, K.: Diagnostic Decisions by Machine. IRE
age in a system
Trans. Med. Electron., vol. ME-7, July 1960, pp.
Optical input/output On-line use
216-219.
subsystem
Several Card printer-punch Model-026, off-line 9. VAN WOERKOM, A. J.: Program for a Diagnostic
use Model. IRE Trans. Med. Electron., vol. ME-7, 1960.
Accounting machine, Off-line use 10. SCHWICHTENBERG, A. II.: What about the Future?
IBM-407 Symposium: Computers in Medicine. J. Amer.
Card sorter Off-line use Med. Assoc., vol. 191, no. II, 1965, pp. 132-133.
Paper-tape reader- Flexowriter; off-line 11. TASKETT, M.: Analog Computers in Medical Research
punch use and Diagnosis. Arch. Environ. Itealth, vol. 7,
Data-medlum Off-line use Sept. 1963, pp. 47-56.
converter 12. DELAND, E. C.: Simulation of a Biological System on
Magnetic-tape unit For analog-data an Analog Computer. IRE Trans. Electron. Com-
recording puters, 1962, pp. 17-25.
Analog-to-digital With associated 13. BROWNELL, G. L.; CAVICCHI, R. V.; AND PERRY, K. E.:
converter equipment An Electrical Analog for Analysis of Compart-
Analog computer mental Biological Systems. Rev. Sci. Instr. vol. 24,
no. 8, Aug. 1953, pp. 704-710.
14. MoRsE, R. L.: Significant Physiological Parameters of
Software (programming languages) available c
the Ballistocardiogram as Analyzed by a Mathe-
FORTRAN Without format
matical Model. Med. Res. Rept., vol. 1, Jan. 1964,
FORTRAN With format
p. 20.
FORTRAN II
15. Cnosnm, R. J.; IIARDY, J. D.; AND FESSENDEN, E.:
TABTRAN
Electrical Analog Simulation of Temperature Reg-
GOTRAN ulation in Man. IRE Trans. Biomed. Electron.,
S.P.S.
O.ct. 1961, pp. 245-252.
Machine code
16. WARNER, II. R.; AND COX, A.: A Mathematical
Model of IIeart Rate Control by Sympathetic and
i Not in order of priority.
b Maximum. Vagus Efferent Information. J. Appl. Physiol.,
c Users of the facility may prepare their programs in vol. 17, 1962, pp. 349-355.
any of these programming languages. 17. CLY_ZES, M.: Respiratory Sinus Arrhythmia: Laws
Derived from Computer Simulation. J. Appl. 33. GLASER, E. ixl'.: An Automatic System for Processing
Physiol., col. 15, 1960, pp. 863-874. Microelectrode Data. IRE Trans. Biomed. Electron.,
18. PACE, W. H., JR.,: An AnMog Computer Model for col. BME-9, no. 3, July 1962, pp. 190-194.
the Study of Water and Electrolyte Flows in the 34. VINOGRAD, S. P. : Space Medicine: Scope, Progress and
Extracellular and IntraceIlular Fluids, IRE Trans. Applicability to Other Fields. Postgrad. Med.,
Biomed. Electron., Jan. 1961, pp. 29-33. col. 36, no. 2, 1964.
19. WOOD, E. H.: Abstracts of Current Biomedical Elec- 35. TOWNSEND, .1. C.: The Mea.surement of Rate of
tronic Research Projects. IRE Trans. Biomed. Change and Rate of Rate of Change of Physiological
Electron., col. BME-8, 1961, pp. 265-267. Data and the Determination of Their Statistical
20. TABACK, L.; MARDEN, E.; MASON, It. L.; AND Significance. NASA Memorandum, 1965.
PIPBERGER, H. V.: Digital Recording of Electro- 36. TOWNSEND, J. C.: Evaluation of Rate of Change and
eardiographic Data for Analysis by a Digital Com- Rate of Rate of Change as Physiological Parameters.
puter. IRE Trans. Med. Electron., col. ME-6, NASA Space Medicine Working Paper, 1965.
no. 3, 1959, pp. 167-171. 37. BRAZIER, _{. A. B.: Some Uses of Computers in
21. BERSON, A. S.; STALLMANN, F. W.; AND PIPBERGER, Experimental Neurology. Exp. Neurol., vol. 2,
H. V.: Digital Computer Analysis of Electro- 1960, pp. 123-143.
cardiographic Data. ISA Trans., col. 1, no. 3, 38. ADEY, W. R.; AND WALTER, D. 0.: Application of
July 1962, pp. 287-289. Phase Detection and Averaging Techniques in
22. PIPBEROER, It. V.; FREIS, E. D.; TABACK, L.; AND Computer Analysis of EEG Records in the Cat.
]_IASON, II. L.: Preparation of Electrocardiographic Exp. Neurol., vo]. 7, March 1963, pp. 196-209.
Data for Analysis by Digital Electronic Computer. 39. LINDSEY, J. F.: Computer Utilization of Time-Line
Circulation, col. 21, no. 3, iX/arch 1960, pp. 413-418. Medical Data from Man in Space Flight. NASA
23. PIPBEROER, It. V.; BIALEK, S. M.; PERLOFF, J. K.; TN D-2695, 1965.
AND SCHNAPER. IT. W.: Correlation of Clinical
40. MURPHY, T. W.; AND CRANE, T. : Analog Computation.
Information in the Standard 12-Lead ECG and in of Respiratory Response Curve. Rev. Sci. Instr.,
a Corrected Orthogonal 3-Lead ECG. Amer. Heart vol. 33, May 1962, pp. 533-536.
J., col. 61, no. 1, Jan. 1961, pp. 34-43. 41. RANDOLPH, ix{. : Application of Analog Computers to
24. CosMA, J.; VOLK, M.; GREEN'OUGH, M. L.; AND ESR Spectroscopy, Rev. Sci. Instr., vol. 31, Sept.
PIPBERGER, H. V.: Automatic Method for Processing 1960, pp. 949-952.
Mass Data in Clinical Medicine (FOSDIC). Methods 42. CHANCE, B.; GREENSTEIN, D. S.; ItxGOINS, J.; AND
Inform. Med., col. 2, no. 4, Oct. 1963, pp. 125-129. YANO, C. C.: The Mechanism of Catalase Action,
25. WOODnVRY, M. A.: In Schmltt, O. H. ; and Caceres, C. II. Electric Analog Computer Studies. Arch. BiD-
A.: Electronic and Computer-Assisted Studies of chem., col. 37, 1962, pp. 322-339.
Biomedical Problems. Charles C. Thomas, Spring- 43. BERSON, A. S.; STALLMANN, F. W.; BRODERS, J. II.;
field, Ill., 1964. AND PIPBERC_ER, II. V.: Telephone Transmission of
26. BAYEVSKIY, R. M.: Transmission of Medical Informa-
Electrocardiogran_s and On-Line Computer Diag-
tion Over Limited-Capacity Telemetry Channels. nosis. Amer. J. Med. Electron., col. 4, Jan.-March
NASA TT-F-228, 1962, pp. 41-43. 1965, pp. 35-37.
27. SCHMITT, O. H.; AND CACERES, C. A.: Electronic and 44. WooD, E. lI.; SUTTERER, W. F.; IX_'ARSHALL, II. W.;
Computer-Assisted Studies of Biomedical Problems. AND NOLAN, A. C.: Use of the Human Centrifuge
Charles C. Thomas, Springfield, Ill., 1964. to Study Circulatory Respiratory and Neurologic
28. WALTER, D. O.: Spectral Analysis for Electroen- Physiology in Normal Ituman Beings and a Descrip-
cephalograms: Mathematical Determinations of
tion of an Electronic Data Processing System
Neurophysiological Relationships from Records of Designed to Facilitate these Studies. AMRL-TDR-
Limited Duration. Exp. Neurol., col. 8, 1963, pp. 63-105, Dec. 1963, pp. 1-24.
155-181.
45. STERLING, W. D.; PERRY, II.; AND I_ATZ, L. : Automa-
29. WALTER, D. O.; AND ADEY, W. R.: Spectral Analysis tion of Radiation Treatment Planning. Brit. J.
of Electroencephalograms Recorded during Learning Radiol., vol. 33, 1964, pp. 544-550.
in the Cat, before and after Subthalamic Lesions. 46. STERLING, T. D.; AND PERRY, II.: Planning Radiation
Exp. Neurol., col. 7, 1963, pp. 481-501. Treatment on the Computer. Ann. N.Y. Acad. Sci.,
30. SIAItAYA, 0.; KAHN, i. R.; AND WARE, R. U.: A col. 115, July 1964, pp. 976-997.
Digital Readout Technique Applicable to Laboratory 47. KILGOUR, F. G.; ESTERQUEST, R. T.; AND FLEMING,
and Aerospace Medical Monitoring of Physiologic
T. P.: Columbia-llarvard-Yale Computerization
Data. SAM-TDR-62-139, Feb. 1963, pp. 1-14.
Project. Bull. Yale Med. Library, col. 22, no. 2,
31: HOVEY, W. J.; GILMORE, J. E.; AND KlSSEN, A. T.:
Oct. 1963, pp. 7-13.
An Automatic Logging System for Biomedical Test
Data. AMRL-TDR-64-50, July 1964, pp. 1-13. 48. DALE, A. G.; AND DALE, N.: Some Clumping Experi-
32. CADV, L. D. : A Computer Program for Measurement ments for Associate Document Retrieval. Amer.
of ECG Wave Amplitudes, Areas and Durations. Doe., Jan. 1965, pp. 5-9.
1961.
49. WIMON, H. tI.; AND DEFRANCESCO, tI.: Automatic
l_ BIOMEDICAL RESEARCH AND COMPUTER APPLICATION
Information Processing. Med. Record News, April Lovelace Found. Rept., 1961, pp. 147-174.
1963, pp. 60-62. 52. LowLAcz, W. R., II; SCH'WICHTENBERG, A. H.;
50. Sr__viN, M.: "EDP" Input Speeds Medical Data L_, V. C.; AND SECREST, R. R.: Selection and
Output. Mod. ttosp., 1962, pp. 106-110. Maintenance Program for Astronauts for the NASA.
51. SCHWICHTENBERG, A. tI.; FLICKIN'GER, D. A.; AND Aerospace Med., vol. 33, no. 6, 1962.
LOVELACS, W. R., II: Medical Machine Record 53. LI_'DB_R(], D. A. B.: A Computer in Medicine.
Cards: Their Development and Use in the Astronaut Missouri Med., April 1964, pp. 282-284.
Selection Program. Bioassay Techniques for Human 54. LINDBERG, D. A. B.: Electronic Retrieval of Clinical
Centrifuge of Physiological Effects of Acceleration, Data. J. Med. Educ., vol. 40, no. 8, 1965.
CHAPTER 2
COMPUTER APPLICATIONS IN THE

BEHAVIORAL SCIENCES
S. B. Sells
Man has been both profoundly impressed and mental procedures, will change not only the
baffled by the prodigious advances in science and dimensions of research activities but most prob-
technology that have at the same time improved ably the topology of the entire science. The
and threatened his life since World Wax II. computer enables the research psychologist to
Whether the path leads to peace and utopian plan for data-analysis, stimulus-generation, real-
glory or to wanton destruction will undoubtedly time control of man-machine systems (including
depend oll how well and how soon man can under- computer-controlled experiments and teaching
stand and control himself. Although the sciences machines), simulation of self-regulating systems,
of man, and in particular the behavioral sciences, and learning and problem-solving programs with
are young and relatively undeveloped, they have vastly increased freedom from constraints on
started recently to make great forward strides. implementation, combined with extremely versa-
This new impetus closely coincided with the tile input and output devices. It also gives the
availability of high-speed, large-capacity, elec- research psychologist access to new axeas not
tronic, digital computers in the 1950's and has previously amenable to investigation.
kept pace with the phenomenal growth of com- Green (ref. 1) has characterized computers as
puter technology in the 1960's. The computer has giant clerks rather than giant brains; in my
proved to be a data-processor of vast speed and opinion they can be both. This discussion focuses
capacity that has extended research capabilities on the data-processing capabilities of computers
to a host of significant and previously unassailable programmed by human operators. Whether and
problems. In addition, scientists have gained ex- under what conditions the brain analogy is justi-
perience and insight into the nature of computers fied is considered later. The almost unbelievable
as information-processing systems, and have de- computational capability of a modern computer
veloped cybernetic models of behavior that can be appreciated if one example is considered:
promise to illuminate many as-yet-unanswered the time required to (1) read-in magnetic tapes
questions about human behavior. previously encoded with 120 variables per person
The purpose of this chapter is to evaluate the on a sample of 1000 cases; (2) compute inter-
impact of computers on the science of psychology, correlations (7140), principal-component-factor
but the treatment of computers as members of a analysis, and varimax rotation; and (3) print out
broad family of information-processing systems, results in tabular form. Before electronic com-
observable in nature as well as constructed by puters appeared, this would have been regarded
man, is emphasized. as an "impossible" task; with earlier, smaller
computers it may have required several hundred
NEW RESEARCH CAPABILITY hours, depending on the particular equipment
setup. However, with machines such as the
The new capability that the computer has CD-1604, the larger CD-3600, or the IBM-7090,
brought to psychological research, with its data- this job could be completed in substantially less
processing capacity, potential for simulation of than 30 minutes.
behavioral processes and automation of expert- The enormous, detailed, and tedious labor of
19
arithmetical computation has been reduced by Instruction 25 in the following way: The first
the computer to a rapid routine. At the same time, operations are performed with j equaling 3;
time, the numbers of persons sampled, of var- the next time, with j equaling 4 (3+1); the next
iables, and of occasions have been vastly extended time, with j equaling 5 (4+1); and so on until
in relation to processing time. The result is that j equals 50. After completing the repetitive series
there is no longer any cogent reason to work with of instructions in the DO "loop" with the final
small samples when it is known that the stability computation for j= 50, the computer resumes the
of parameters requires considerably larger ones regular sequence of computations following In-
to eliminate relevant sources of variance. There- struction 25. Each time, through the loop of
fore, it is also unnecessary to engage in unreliable instructions, computations are repeated with the
research practices on the grounds of expediency. then-current value of j; the value of K is specified
Computers have enabled research workers to ap- in a separate instruction. Increased flexibility can
proximate infinite series, employ truly multi- be gained by placing additional DO loops, and
variate designs, perform matrix algebra and other even IF statements, within DO loops.
previously unwieldy computations with ease, and The IF statement enables the programmer to
also to approach their scientific problems with compare two values (two means, for example)
previously unknown freedom and power. and to proceed in any of three directions depend-
PoweTful automated routines--In addition to ing on whether one is larger than, equal to, or
capacity and speed, it is important to realize that smaller than the other. The DO statement per-
the incorporation of simple logical functions mits automatic repetition of a series of computa-
permits the programming of automated sequences tions before automatic procession to the next
of operations that overcome some of the most phase when that task is completed. Statements
tedious drudgery of extended computations. This such as these, in conjunction with others, permit
can be illustrated by two types of program state- long and frequently complex sequences of logical
ments, referred to as the IF and the DO statements, decisions and computations. By utilizing nu-
that have become workhorses of computer merical coding for alphabetic characters, the
programmers. computer can handle diverse types of informa-
Consider first the following IF statement: tion-processing.
IF (5lean H-Mean L) 17, 27, 37

PSYCHOLOGICAL RESEARCH APPLICATIONS
This statement, written in FORTRAN, means
that the difference, Mean H minus Mean L, is Although psychologists pioneering the new
to be computed, and that three branching, computer applications in psychological research
alternative instructions are to be followed, de- have been relatively few (ref. 2), their accomplish-
pending on whether 5Iean H is larger than, equal ments are impressive both qualitatively and
to, or smMler than ._Iean L. If the difference is quantitatively. At the same time, thoughtful
negative, the sequence of operations branches to critics (refs. 3 and 4) have observed that other
Instruction 17; if it is zero, the operation branches disciplines have been more vigorous than psy-
to Instruction 27; and if it is positive, the branch- chology in emplo3dng this powerful tool for study
ing is to Instruction 37. The branching instruc- of many aspects of intelligent behavior; they have
tions 17, 27, 37 can be located anywhere in the warned psychologists of the possible consequences
program, and the comparison can be made of to development of their science. Acquisition of
any numbem represented by codes in the pro- computers by universities has been accelerating,
gram, such as Mean H and Mean L in the present however, and it is possible that Baker's (ref. 5)
example. more optimistic view will be realized.
An example of a DO statement is next: This survey of psychological research employ-
ing computers is necessarily brief and is confined
DO 25j= 3, 50, K
to the principal types of application. The discus-
Let us assume that K= 1. The computer repeat- sion is organized under four categories: data
edly executes all the following instructions up to processing and statistical analysis, pattern gen-
COMPUTER APPLICATIONS IN THE BEHAVIORAL SCIENCES 21
eration and recognition, computer-controlled ex- a matrix compi]er system that was programmed
periments, and simulation of adaptive behavior. originally for the Univac-ll05 and has been re-
More-detailed accounts are available (refs. 3 programmed for the small LGP-30 at. the Psycho-
to 10). metric Laboratory of the University of North
Carolina. According to Jones (ref. 4), "Each sub-
Data Processing and Statistical Analysis
routine within the system represents a single
Computers have increased the magnitude of matrix operation--addition, subtraction, multi-
problems that can be routinely included in multiplication, division, inversion, transposition, ex-
variate analyses. Thus, whereas a correlation traction of diagonal entries, read-in, print-out,
matrix of from 20 to 25 variables was considered etc. Each requires as parameters the order of
large around the end of World War II, the feasible matrices and the memory location of their initial
limit for routine operation in the early 1960's elements.., it is arranged [in the Psychometric
was around 200; on the largest computer avail- Laboratory] so that the operator may exercise
able, the CD-3600 (maximum capacity, 261 144 manual control at the Flexowriter keyboard.
48-bit words), it may approach 1000. Along with The resultant machine resembles a desk calcu-
faster access to storage, faster execution time, and lator for matrix operations." As a computer
increased accuracy derived from larger word size, readily available to graduate students, this
the advantages of increased capacity may be ex- appears to be a most. valuable training device.
pressed by tile following rough approximation: The result of such efforts, which really repre-
an increase in storage capacity from 6000 to sent the beginning of revitalization by computer
261 000 words (roughly 40 times) increases work of statistical methodology, has been to make
capacity by about 1000 times with greater speed high-powered analysis readily available to almost
and accuracy. anyone having funds for computer time. While
FORTRAN programs have been developed for this result, ha_s been a boon to scientific effort,
a wide range of statistical computing, including reflected for example by the quality of disserta-
correlation, multiple regression, canonical anal- tions using multivariate methods, there have
ysis, various methods of cluster analysis, fac- been conspicuous eases of waste and misuse as
tor analysis of R-, P-, and Q-type matrices, may be expected. Despite the (relatively small)
and rotation of factor matrices; and also for abuse, however, the computer has greatly ad-
multidimensional scaling, profile analysis, con- vanced knowledge of differences between indi-
figural-score analysis, multiple-scalogram anal- viduals, particularly in studies of personality
ysis, multiple-discriminant analysis, intraclass dimensions, attitudes, interests, and abilities.
correlation, various nonparametric analyses, and Important new studies in personnel psychology,
multivariate analysis of variance (refs. 5 to 8 and of job components and organization, of personnel
11 to 16). Shepard's development of programs for selection and utilization, and in simulation of
multidimensional scaling (refs. 17 and 18) has personnel systems also have capitalized on the
attracted much attention. expanded capabilities provided by computers.
TrTon (refs. 15 and 16) has developed an Useful summaries of these data-processing uses
executive program with cluster, factor, and with extensive references are available (refs. 5
pattern analytic components that can be called and 12).
in by command at any stage of analysis. His Nonquantitative analyses--In addition to statis-
integrated analytic program is capable of ex- tical applications, computers have facilitated
tremely versatile multidimensional analysis of several types of research involving counting and
correlated data; it was first programmed for the classification of voluminous information, such as
IBM-704 computer and more recently for the in studies of natural-langnmge data. The follow-
IB51-7090. Using such a flexible system the in- ing examples have been cited (ref. 4) as illustrative
vestigator has the option of several alternative of a particular facility of computers:
procedures at any stage of analysis, and of com- (1) Word counts, as illustrated by a study
bining various sequential patterns for empirical (ref. 20) to resolve an authorship dispute
comparison of results. Bock (ref. 19) developed (2) Studies of transitional frequencies between
pairs of word classes in speech from normal and an extremely tedious process. Fortunately ran-
aphasic speakers (ref. 21) dom-number generator programs have been
(3) Processing of speech from normals and developed for computers. If the universe is coded,
aphasics to determine functional regularities be- it is possible automatically to select random lists
tween frequency of word occurrence and rank of any length and to present them in any pro-
popularity of the word (ref. 22) grammed form of computer output. Such pro-
(4) Grammatical classification of words on the grams can also be modified in various ways; for
ba.qis of dictionary 'qook-up" procedures (ref. 23) instance, the program can be random in all re-
(5) Translation from one natural language to spects except that, if a certain stimulus occurs,
another (refs. 24 to 27) the probability of a certMn other stimulus follow-
(6) Analysis of syntactic structure (refs. 28 ing it may be made higher than for other stimuli.
to 30) It is possible to specify the probability of occur-
(7) Storage and retrieval of texts (refs. 31 rence of any stinmlus differentially with respect
to 37) to other possible stimuli, so that the computer
(8) Content-analysis of text material retrieved can prepare lists according to desired probabili-
(ref. 38) ties. With use of rectangular coordinates, com-
The study of storage, retrieval, and content- puters can select dots according to two-dimen-
classification of texts is currently of great concern sional locations and thus build two-dimensional
to librarians and bibliographers. In addition to configurations of dots and/or lines having system-
the technical problems per se, there are closely atic, random, or statistic_flly constrained random
associated and fascinating problems relating to variations. With the aid of printed, cathode-ray-
the basis of organization of knowledge. tube, and auditory outputs much flexibility in
Here is a good example of a computer's power stimulus-production can be attained.
to virtually plow through a massive accumulation Using cathode-ray-tube output to generate bar
of data to compute summary statistics: This ex- patterns of dots with different probabilities of
periment (ref. 39) involved a series of discrimina- occurrence, Green et al. (ref. 40) demonstrated
tion-reaction-time studies in which six subjects that dot-probability techniques can be used to
were tested daily, 5 days weekly, over a period. obscure any type of pattern (e.g., a square) for
For each subject there were 500 RT's daily; for which a mathematical equation can be written.
the entire experiment, more than 15 000 RT's It is easy to compute whether each dot position
weekly. Each set of RT's was analyzed in 2 (o 31 in a two-dimensional grid is within the square.
different stimulus categories, and day-to-day Perceptual studies of such stimuli, using photo-
variations in central tendency, variance, shape of graphs of the displays generated on the cathode-
distribution, error rate, error distribution by ray tube by the computer, have produced in-
stimulus category, and complex special analyses triguing results. "Subjectively it appears that as
of atypical RT's were computed. The magnitude the probability differencc decreases, the presence
of the computations was so great that if a com- of a shape is still apparent but the contours
puter had not been available the experiments cannot be perceived." These techniques can also
would have been impossible (ref. 39). This is but be used for studying contour-formation as a
one example of hundreds that could be cited. process.
White (ref. 41) used this technique with a
Pattern Generation and Recognition different procedure to study form-recognition.
Stimulus generation--Many kinds of psycho- First he generated a clear pattern, .5, using a
logical experiments require random selection of systematic configuration of dots. The pattern
stimuli from a known universe. This approach was gradually dissolved in successive manipula-
would be ideally applicable to test construction; tions by randomly moving each dot a little each
it has frequent application in perceptual a_d time; this was done by having the computer
learning studies. If the univer_ is large and the calculate a small random motion for every dot
lists must be long and numerous, the use of and then activate a camera to take a picture of
random-number tables for item-selection involves the entire display after each set of moves. The
result is a series of exposures of the number 5 These techniques involve template-matching (e.g.,
randomly walking to oblivion. White has used magnetic-ink grid characters on bank checks) in
films of this kind, played backward, to determine which prepositioned codes are "read" exactly as
at what point in the series the figures are recog- punched cards or magnetic tapes. Template-
nized. matching devices are limited to the positions and
Pattern recognition--Practical as well as theo- flexibility for which they are programmed, and
retical considerations have motivated intensive usually cannot cope with even trivial variations
study of pattern recognition by machines. Borko that human perceivers handle intuitively.
(ref. 6, p. 295) has listed several potential appli- Nevertheless, Uhr (ref. 3) reports tremendous
cations such as automatic handwriting-recog- and exciting progress toward analytic, sophisti-
nizers for the Post Office, check-readers for banks cated models that are conceptually related to
and clearing houses, photograph-interpreters for template-matching. He cites programs that
military intelligence and meteorological agencies, "learn" by storing accumulated "experience"
and text-readers for the blind; such interests have (ref. 42); utilize powerful operators such as
been responsible for considerable research and de- "edging" (turning areas into contours), averag-
velopment. This problem is fundamental to the ing, finding "connectivities" (delimiting figures
study of form-perception as well and has received as opposed to background), and counting (com-
sophisticated but limited attention from psychol- puting area and number of objects) (ref. 43);
ogists. Surveys of recent development.s in pattern- simulate nets of neuron-like elements (ref. 44);
recognition computer programs and their utility generate their o_m operators (ref. 45); and func-
as models for form-perception have been pub- tion flexibly by means of operators capable of
lished (refs. 3 and 41). Jones' (ref. 4) significant accepting inputs over a range of displacement.
comments agree with those of Uhr that psychol- Programs such as Rosenblatt's Perceptron (refs.
ogists may find themselves trailing other disci- 46 and 47) and Selfridge's Pandemonium (ref. 48)
plines in this area if they do not invest more time have employed parallel processing rather than
and effort on these problems. serial decisions, thus providing redundancy that
According to Uhr (ref. 3), "The problem posed facilitates correction of errors.
the computer or designer of a computer for It is noteworthy that these analytic programs
'pattern recognition' (or 'character recognition,' have been found to have neural-net or functional
as it is sometimes termed when a specific set of interpretations related to mechanisms of form-
predetermined patterns, usually the alphanumeric perception; according to Uhr they have been
symbols as printed in a special type font, is the instrumental in suggesting physiological and
only set to be processed) is the many-to-one psychological experiments. A most striking in-
mapping of different inputs into appropriate stance of this is the fact that Lettvin et al. (ref.
output sets." Such pattern recognition is per- 49) made a successful search for straight-line and
formed with disarming ease by human perceivers, angle operators, in the visual form-perception of
as by postmen who daily sort mail accurately the frog, on the basis of hypotheses generated by
that no existing machine could decipher. The Selfridge's Pandemonium model.
output sets referred to by Uhr are the sets into It should be apparent that, as pattern-recog-
which the human perceiver maps the inputs by nition programs advance from the simple tem-
grouping things "across an unknown set of geo- plate-matching models to the most advanced,
metric transformations and deformations." The complex, and strongest analytic models, the
pattern-recognition problem involves discovery of nature of the problem changes correspondingly.
the operations that effect this matching, whether The strong analytic models are similar to and in
by the experimenter in the laboratory or by the fact are models of cognitive processes. Very fie-
computer itself, or by both. quently they have neurophysiological interpreta-
The well-known engineering development for tions, and "A strong argument can even be made
processing of bank checks and identification sym- for the relevance of pattern recognition for
bols currently in commercial use are not pattern- learning and concept formation in machines"
recognition devices except in a restricted sense. (ref. 3).
Compufer-Confrolled Expedmenfs matter of time before this "unbelievable" setup

With the speed, storage capacity, and input- will be accepted as routine by students in experi-
output versatility that has been described, it has mental-psychology courses. Yet this is only one
not taken psychologists very long to automate example of a class of experiments involving
their laboratories. The work of Green and of stimulus-generation coupled with random-number
White (discussed above) on stimulus-pattern generators, digital-to-analog conversion, and spe-
generation was a step in this direction. Examples cial outputs for visual and auditory representa-
of more-fully automated experimental setups are tion, all integrated with the impressive capabilities
reported in this section for the purpose of illus- of computers for rapid computation and versatile
trating how far automation can be carried, and input and output of information.
also perhaps to project a vision of the psycho- Jones (ref. 4) has described a series of studies in
logical laboratory of tomorrow. This section also which the subject's essential task is estimation of
includes a brief Summary of programmed learn- an unknown population proportion, given only
ing developments, which are in fact computer- partial information about: its value. The subjects
controlled experiments in their present stage of are given a small sample of observations but are
progress. also able to profit by prolonged experience with
Automated experime_ts--One of the most im- the same parametric distribution of population
pressive efforts to use a computer in a highly proportions. The experiment is under computer
automated apparatus configuration was a study control. The computer draws each sample of
of auditory-discrimination training (ref. 50). stimuli, prints them out, and requests responses
The computer was equipped with a digital-to- and accepts them on the Flexowriter keyboard; it
analog converter, special output equipment, and provides feedback to each subject on performance
two electric typewriters for direct (on-line) input after each trial and cumulatively. Jones has noted
and output. With this setup a complete psycho- that there is greater stability of performance
physical experiment could be run under control levels witlfin subjects and less-marked individual
of the computer program. The subjects were pre- differences between subjects in these computer-
sented with a sound stimulus that could have any controlled experiments. This fact may reflect
of five values on five dimensions: frequency, greater control of experimental conditions than in
amplitude, repetition rate, duty cycle, and dura- conventional experiments in which a person per-
tion. The subject's task was to identify the sound forms the experimenter's functions. He has also
and respond by typing a series of numbers on the observed the same intense interest on the part of
typewriter. The computer informed the subject, subjects in such experiments as has frequently
by typing information on the typewriter, whether been reported in programmed learning; appar-
the response was correct, and, if not, in what re- ently this is no longer merely a function of
spects it deviated from the correct response. novelty, but rather a matter of greater personal
Having done this, the computer selected each involvement in the task.
next stimulus on the basis of the subject's pattern A possible disadvantage of the automated
of past responses; which had to be recomputed laboratory, according to Jones, is the isolation of
after each response. In this experiment it was the investigator, preventing him from making
possible to run two subjects at once. The com- personal observations that might le'_d to deeper
puter kept all records and computed all necessary insights concerning his study. However, this need
summary statistics; it also computed auditory not be true; being free from all the countless de-
wave-form patterns by producing digitized sig- mands of operation, the experimenter might be
nals corresponding to the desired sounds and then able to use his new freedom more advantageously.
playing these time-quantized data through the It is also important to note that computers are
digital-to-analog converter. still among the most expensive "gadgets" that
In this experiment the computer was used to have appeared on the budget sheets, and expan-
generate stimuli, compute the sequence of their sion of their use in laboratory work as elsewhere
presentation, feed information back to the sub- may be slowed by this fact. However, as with all
ject, and analyze the results. It seems only a equipment, production costs will undoubtedly
fall with improvements in manufacture (as in technology is illustrated by the experimental

miniaturization with the availability of tran- teaching machine built by System Development
sistors) and with mass production. Corporation (SDC) for a research project (ref.
Automated leaching--An exciting recent de- 53); it consists of three components:
velopment in educational psychology has been
(1) Bendix G-15 computer--This is a relatively
the wide interest in learning programs and de-
small general-purpose digital computer with
vices for their operation called teaching machines.
paper-tape input. A bell mounted in the com-
Although autoinstruction soared to the statits of
puter frame can be rung under computer control,
a fad for a few years and many critics predicted
permitting auditory signals. The computer, as
that the "boom" would lead to a premature
central control unit for the teaching machine,
"bust" (ref. 51), the boom appears to have
determines at all times during a programmed
slowed and a number of substantial research
session the materials to be presented to the
programs have focused serious attention on
student, analyzes students' responses received
problems that require investigation.
via the electric typewriter, compares these with
Many of the advantages of programmed learn-
stored data, and communicates information to
ing are accepted without challenge. Learning
the student.
programs must be specific and cannot be as loose
(2) Random-access slide projector--Developed
and rambling as textbooks too often are. Auto-
by SDC staff engineers, it displays instructional
instruction is self-paced and provides individually
materiMs to the student. It holds up to 600
focused feedback and greater opportunity than
35-ram slides in 15 magazines of 40 slides each.
group instruction for tailoring of lessons to indi-
Selection and projection of slides, each of which
vidual requirements. Questions have been raised
holds one item, is under computer control. (Other
about the adequacy of programs, programming
and similar devices store problem materials in
principles, and particularly evaluation of progress
internal computer storage.)
under autoinstruction as compared with com-
(3) Electric typewriter--It is linked to the
parable text material. Indeed some of the most
computer as on-line equipment and serves as a
vociferous critics of autoinstruction argue that
direct, two-way channel of communication be-
no case has yet been made for the superiority of
tween computer and student.
the program to a good text.
For programs that present their material in a Learning programs involving communication
predetermined sequence, it may be difficult to between experimenter and computer are con-
resolve this issue. However, the concept of learn- trolled by paper-tape input after conversion from
ing automation has already been extended by the punched cards to tape. This machine is being
computer to a level of flexibility that clearly goes used for study of the process of preparation of
beyond any simple comparison with good text- learning programs: for example, "What decision
books, and demands comparison only with good criteria should be used in determining when to
teachers. With the computer the sequence of branch a student to less-difficult remediable
material presented can be adapted to the student's items?"
performance to give both immediate and cumu- Expansion of the computer-controlled teaching
lative feedback, to keep records and compute machine to accommodate large numbers of
results, and, with special output equipment, to students working on a variety of programs ap-
present material in any morality or format de- pears feasible, and research programs toward
sired. Computer control of learning programs this end have been started. The SDC has begun
permits flexible, automatic operation of a program development of an expanded educational f:tcility
library and rapid access to a wide range of program built around the Philco-2000, a large-capacity
material. digital computer; it is called CLASS (Computer-
Silberman and Cou]son (ref. 52) have reviewed based Laboratory for Automate(t School Systems)
the history of computer-controlled teaching ma- and will have individual displays and communica-
chines, which has advanced impressively since tion input-output devices for each student, as
the first publication in 1959. The state of the well as monitor displays for teachers. The CLASS
facility will also receive, process, and print out have concerned themselves only with matching
data related to registration, attendance, tests, of inputs and outputs, leaving the synthesis of
students' educational backgrounds, and other in- the process to the logic and ingenuity of pro-
formation relevant to a complete educational grammer and computer (refs. 4, 5 and 59 to 61).
program. Thus CLASS may serve as a laboratory This distinction is similar in some respects to
for the study of not only school learning but also that between cognitive theory and S-R theory
many aspects of an entire school system. in psychology. Although synthetic programs may
There is little doubt that the hardware for be criticized on occasion because they perform
autonmted school systems is feasible. In fact the human-like functions in nonhuman-like ways,
development of such facilities is more likely to be they may also lead to significant insights and
slowed by the rate of progress in mastering pro- hypotheses of heuristic value. Indeed, should
gramming principles for autoinstruction, and such research result only in improving the
producing the programs required for their opera- psychologist's ability to interrogate behavior, it
tion, than by the technicM engineering wizardry will prove adequately fruitful.
involved in their design. However, experimental Simulation (and s3_lthesis) research is already
J
facilities, such as that at SDC and similar ones beginning to have a healthy impact on psycho-
on several university campuses, must be supported logical thinking and theory. As already frequently
if these goals arc to be achieved. noted, computers function in very small, step-by-
step processes, and the cold reality of detailed
Simulation oFAdoptive Behavior specification, implicit in programming, is lethal
Computers have provided an unexcelled means to slipshod workmanship and nebulous theory.
for building and testing of simulation models of As an example of this, Baker (rcf. 5) has cited
information-processing functions of living orga- Uhr's (ref. 3) comments on the major difficulties
nisms on at least two levels: neurophysiological encountered in recent efforts to express the
and behavioral. At the first level, programs have theories of even such eminent theorists as Hebb
been written that simulate or synthesize nerve and Hull in testable form. To the extent that the
nets and central information-processing functions overwhelming volume of theoretical development
of organisms. At the behavioral level, since the comes from sources not subjected to such rigorous
early work of Bush and Mosteller (refs. 54 and discipline, this observation may be a forecast of
55), investigators have focused on problems of major renovation of psychological theory as
molar behavior, such as perception, learning, computer-oriented research expands.
,=
concept-formation, decision processes, and various The interdisciplinary nature of many computer-
complex behaviors including checkers, chess, simulation enterprises is emphasized in White's
lang_lage-translation, musical composition, and review (ref. 62) of Rosenblatt's book on percep-
managerial decision making. Models of group and trons and simulation of brain mechanisms (ref.
organizational behavior also have been published 47), which sounds a note of caution regarding the
(ref. 56). Although scientists and scholars have contribution of such synthetic models to psycho-
shown concern with these problems for many logical understanding. White acknowledged the
years, a prodigious volume of work--too ex- computer program as a powerful language for the
tensive and diverse for brief summarization--has construction of "artificial intelligence" models of
appeared within the past 5 years: for example, a many human psychological processes:recognition,
survey of computer simulation of cognitive problem-solving, rote memory, and the like.
processes (refs. 57 and 58) included over 400 However, he warned that "The objective of the
published references, more than half of them whole endeavor is easily lost when the program
written since 1958. becomes an end in itself." In White's opinion
In review of different approaches to these prob- this appears to some extent to have happened
lems one important distinction stands out between with the perceptron; his evaluation is that it is an
those who have interpreted simulation more "unconvincing neurological model since few of
literally and attempted to match the computer its parameters are firmly rooted in neuroana-
process with the natural system, and those who tomical or neurophysiological data and it offers
COMPUTER
APPLICATIONS
IN THEBEHAVIORAL SCIENCES 27
little to psychologists who expect a model tailored By manipulating these weights after each input
to the known facts of human pattern recognition presentation, Clark and Farley attempted to
and discrimination." Although it is probably pre- determine whether the network could be so
mature to expect such models to solve all the organized that a given input could become re-
problems of psychology, it may be recognized liably associated with one of the two outputs.
that the requirement of interdisciplinary treat- They found that this could be done and that
ment exists, whether accomplished by a more one could classify the inputs in the desired
generalized "specialist" or by a team. manner significantly beyond chance expectation.
They also found that this was possible even when
Computer Programs for Neural Nefs the input patterns contained variable elements
In 1953 Estes and Burke (ref. 63) published a (noise). This finding resembles a prediction made
theory of stimulus variability in learning which in 1949 by Hebb (ref. 66) in discussing the func-
has been adapted to a digital-computer program. tion of cell assemblies in recognition.
They hypothesized that on successive learning More-complex programs involving randomly
trials some subset of elements, encoded in the organized networks, various methods of reinforc-
neural net, is consistently reinforced. At the out- ing the network, and various methods of orga-
set of practice all the stimulus elements impinging nizing and reducing the input signals have been
on the organism are associated with the first studied. Ashby (ref. 67) and Culbertson (ref. 68)
successful response. After the second successful have given highly sophisticated treatment to the
response, some of these elements remain associ- problems of simulation of brain and nervous
ated while others do not. Thus, after a series of system. However, the foregoing illustrates the
trials, a subset of stimulus elements should be logic of an approach to this type of simulation.
sufficiently reinforced to form an S-R connection.
Mathematical Models
This model is probably adequate for the special
conditions of certain controlled learning experi- Several investigators have attempted to formu-
ments in the laboratory but is very limited. It is late principles of neurophysiological functioning
too simple for a perceptual experiment in which and molar behavior in terms of mathematical
successive patterns with no stimulus elements in expressions. This approach predates computers
common may nevertheless properly be placed in by many years, but has increased in momentum
the same category. Clark and Farley (refs. 64 since computers came into use. Uhr (rcf. 3) has
and 65) extended the model to a two-stage process commented on mathematical analyses of pattern
in some attempts to simulate neural networks for recognition; and his remarks have general ap-
pattern recognition on a digital computer. plicability.
The computer was programmed for simulation
of a network of randomly connected nonlinear These have rarely been programmed and tested,
and it seems quite likely that they would not work
elements that. were assigned parameters to repre-
in the practical situation. They seem of value as
sent thresholds and refractory periods. The net- they explore mathematics for new approaches, as
work was composed of four groups of elements-- they classify problems, and as they suggest more
two of which represented fixed input; two, output elegant and less redundant methods for proce_ing
(defined as the difference in the number of ele- patterns, and especially for making use of the in-
ments fired in input and output, at any instant). formation obtained. But they do not seem to attack
the fundamental problem of choosing or discovering
One input consisted of firing of all the elements
the best operators for the appropriate many-one
in the first input group and of none in the second; mappings. Rather, they usually address themselves
the other input was the reverse of the first. The to the question: Given a specified set of operations,
system operated according to a rule that an ele- what are the best methods for accomplishing them
ment that received excitation above its threshold or making use of their results? Or they suggest
mathematical methods that have been thoroughly
would fire and transmit excitation to all other
developed, and hence might be powerful tools, if
elements with which it was connected. The effec- appropriate.
tiveness of this excitation (its "weight") was con-
sidered a property of each particular connection. In terms of this distinction, mathematical models
belong to the synthetic rather than the simulation important comment on this program in com-
approach. parison with learning by human subjects is that
Examples of mathematical models of behavior the computer learns any list of symbols, as pre-
that have been programmed for computers are sentcd, equally well regardless of their form.
many. Uhr has cited approaches, to the pattern- Whereas association value of symbols used as
recognition problem, using Fourier analyses (ref. stimuli is an important problem in human-
69), quantum mechanics (rcfs. 70 and 71), and learning research, this type of complication is not
integral geometry (ref. 72). Problems of optimum programmed in EPAM. In more-complex pro-
coding have been approached in terms of sta- grams, however, computers could be instructed
tistics and information theory (refs. 73 to 75), to store past experience, and such complications
and decision theory (ref. 76). Mattson (ref. 77), could be added.
extending previous work by Uttlcy (refs. 78 and Psychologists have made extensive use of the
79), has programmed a self-organizing system "thinking aloud" method to develop detailed
that will discover the proper Boolean function protocols of subjects' behavior on various prob-
for linear partitioning of an n-dimensionM space lem-solving tasks to be sinmlated by computers.
into appropriate sets. Sebestyen (ref. 80) de- A recent example (ref. 87) is the use of IPL-V to
veloped a method of transforming the space simulate the problem-solving behavior of a single
within which input patterns are coded/ it was human subject whose task was to determine the
highly successful when applied to speech-recog- pattern in which four switches were set. The
nition. "thinking aloud" protocols were studied along
with experimenters' observational notes on the
Behavior-$1mula_ion Models
subject's overt behavior; then an IPL-V program
Behavior-simulation models can be described was written to simulate the behavior of the sub-
generally as logical, noncomputational uses of ject. Both the human subject and the IPL_V
computers. They have been applied at an ac- program were then presented with an additional
celerated rate to problems of perception, learning, switch-setting problem. The outcome was that
concept formation and attainment, memory, both the reM subject and the simulated subject
tracking behavior, and such associated problems generated protocols and problem-solving be-
as teaching machines and information-retrieval. haviors that were judged highly similar. This
According to Baker (ref. 5, p. 568) "The expanse report and a similar one (refs. 59 and 60) describe
of only a part of the field could be demonstrated the experimental procedure and simulation proc-
by a bibliography of articles on cognitive processes ess in painstaking detail. Feldman also considered
which itself would readily exceed 500 entries." the criteria of judging of similarity at some
Most reported psychologicM simulation studies length. Minsky (ref. 88) has reviewed the progress
have used the fifth version of an information- with heuristic programs of this type as of 1960.
processing language approach, dcvcIoped by Edward Johnson,* a student of Jones, is re-
Newell, Shaw, and Simon (refs. 81 to 85), that is sponsible for an important innovation in this
widely known by the symbols IPL-V. This is an area. He analyzed problem-solving performance
interpretive programming system that manipu- data for 11 subjects and noted common principles
lates lists of symbols rather than numbers. An and strategies reflected in their performances. On
example of this type of program is the Elementary the basis of these principles he then developed a
Perceiving And Memorizing (EPAM) program model such that, when certain individual differ-
(ref. 86); it is designed to perform rote-memory enec parameters were taken into account, the
tasks, such as memorization of lists of nonsense "style" of problem-solution manifested by the
syllables, in a manner simulating the behavior of computer program would resemble that of one
human subjects. For this task, EPAM depends or another individual subject with a high degree
on a small amount of initial information by of accuracy.
which tile symbols are paired. This linkage is
*Johnson, E. S.: The Simulation of Human Problem
strengthened (luring tile simulation experiment
Solving from an Empirically Derived Model. Unpublished
until the list is "learned" to some criterion. An dissertation, Univ. of North Carolina, 1961.
HOW FAR CAN COMPUTERS SIMULATE havior. This was stated first., to our knowledge, by
BEHAVIOR? Charles Sanders Peirce in 1887 (ref. 90): "Pre-
cisely how much of the business of thinking
A major concern in the preceding discussion machine could possibly be made to perform, and
has been elucidation of points of similarity be- what part of it must bc left for the living mind,
tween artificial information-processing systems is a question not without conceivable practical
(computers) and natural systems (behaving or- importance; the study of it can at any rate not
ganisms). At the same time I have refrained from fail to throw needed light on the nature of the
use of anthropomorphic references, such as brain, reasoning process." Similar opinions have been
memory, thought, and the like, in order to avoid expressed (refs. 4 arid 91 to 93).
confusion in consideration of the question raised For the conservative behavioral scientist, this
in this fi/lal section. It is well known that composition may serve as sufficient justification for
puter programs, developed or being developed, pursuit of the simulation approach without com-
have performed such "intelligent" acts as proving mitment as to whether or under what conditions
mathematical theorems, playing games (such as a machine may be programmed to behave in a
chess) with greater skill than the programmers, thoroughly human-like way. However, as the
recognizing spoken language, translating from one effort, is extended increasingly to more-and-more-
naturaI language to another, composing music, complex behavior, with use of improved equip-
and even inventing new and better computers. ment of greater storage capacity and flexibility,
These accomplishments by computers go be- higher speed, and greater miniaturization, the
yond mere "giant clerical" computing routines in question of limits keeps reasserting itself. Con-
which the machines do precisely what they are sideration of this question, moreover, is not
commanded to do. Quite to the contrary they re- merely an amusing exercise, but rather a serious
flect more the quality of giant brains capable of examination of the theoretical end-points of the
discriminatioi, between alternatives, logical de- simulation process.
cisions, adaptive change as a result of "experi- Ulric Neisser, a leading opponent of the
ence," and a degree of flexibility that inspires use proposition that computers can behave in a
of the adjective intelligent even in the face of thoroughly human-like way, argued that "the
strong inhibition arising from knowledge that the root. of the difference seems to be more a matter
behavior is performed by programmed machines. of motivation than of intellect." In support, of
The controversy over whether or not machines this assertion he listed these three fundamental
do indeed "think" is truly a matter of semantics and interrelated characteristics of human thought
and reflects opinions concerning the distinction that are conspicuously absent from existing or
between manipulation of symbols and thoughts, contemplated computer programs (ref. 94, p. 195) :
as well as philosophical positions vis-&-vis the
(1) Human thinking always takes place in, and
nature of man. As might be expected, extremely
contributes to, a cumulative process of growth
divergent opinions have been expressed by highly
and development.
qualified behavioral scientists identified with
(2) Human thinking begins in an intimate
computer research. For example, Borko (ref. 6,
association with emotions and feelings which is
p. 20) has stated one extreme categorically: "The
never entirely lost.
computer performs mechanical and electronic
(3) Ahnost all human activity, including think-
operations. It is the human interpreter that
ing, serves not one but a nmltiplicity of motives
thinks"; while in an illuminating address, en-
at the same time.
titled "How to tell computers from people,"
Saunders (ref. 89) answered his own question by In Neisser's opinion these defects are im-
saying, "You can't." material in "technical applications," such as
Between these extremes lies a more pragmatic computation, in which it is of no significance
position that begs the question of a true difference whether the computer (or, for that matter, even
and concentrates instead on the value of simula- the operator) approves or disapproves of the
tion as a heuristic approach to the study of be- results, so long as they are accurate. However, he
30 BIOMEDICAL RESEARCH AND COMP'UTER APPLICATION
is deeply concerned about use of computers in of human experience, would be able to show
social decisions, "for there our criteria of ade- growth in their programmed functions in far less
quacy are as subtle and as multiple motivated as time.
human thinking itself." Such growth would of course be represented
What can bc said about these three defects symbolically in information-processing terms,
that limit the computer and reduce its "intelli- but, in the simulation frame of reference, this is
gence" ultimately below that of fallible mail? relevant to the argument. Whereas human growth
These defects are considered briefly in the follow- reflects bodily change and accommodation to the
ing sections. The reader is referred to Neisser's social requirements of a culture in addition to
competent discussion for defense of his position. expansion of cognitive capacities, these may be
regarded as magnifying the order of complexity
Growth and Development of the problem t)ut not changing it quantitatively.
In the present state of technology, computers This point recalls an observation (ref. 67) that in
have a limited behavior repertoire of instructions programming of a computer for simulation of
to execute and limited capabilities represented by behavior it is as important to specify the environ-
storage capacity and access, speed of execution ment as the characteristics of the sinmlated
and transfer of information, input-output versa- organism. While not feasible in the present state
tility, and the like. The computers of the early of knowledge, seemingly not beyond the realm of
1960's represent the second generation of elec- possibility is eventual development of computer
tronic computers, and it is reasonable to expect programs for simulation of growth experiences of
tremcndous extension of these capabilities-- a normal chiht in his social environment, while at
roughly comparable to inheritance by subsequcnt the same time an expanded repertoire of knowl-
generations of characteristics of a mature living edge, skills, attitudes, interests, tastes, and the
organism. This trend may be regarded as an like is acquired.
evolutionary development which, _dthough not
Emotionsand Feelings
"natural," is nevertheless relevant to considera-
tion of development. It is of course true that In line with such reasoning it seems equally
computer hardware may never go through phases possible to incorporate in a computer both a
within the operating life of a machine, but one storage register for feelings, and output devices
must recognize that a mechanism of "species for emotional expression. Present concerns with
development" exists. the difficulties of cognitive problems should not
Next let us consider a form of development obscure the fact that the affeetive problems are
that can be related to the operating life of a not essentially different but only complicating.
programmed machine. Such a machine would Conceptually feelings could be coded in any
necessarily be a special-purpose computer (e.g., categories desired, and optimally would be based
problem-solver) for the sake of simplicity, but on current psychological insights. Value systems
ultimately, if it were endowed with sufficient for filtering input signals could be read into
capability, its functions would bc unrestricted. storage so that the computer might decide on the
In our present state of ignorance no computer feeling code and classify it in storage accordingly,
has been given an opportunity to accumulate along with other information. On the output side,
experience over a long period; the exigencies of emotions eouht be registered by printing-out of
laboratory costs and work demands have required messages, discharge of tubes labeled for various
erasure of storage after completion of one prob- endocrine secretions, rattling of drums, or any
lem to permit reading-in of the next. However, it means desired.
is instructive to compare the lifelong exposure to The important question, of the influence that
information input and feedback in a human being affeetive factors would exercise on cognitive
with the very short-term exposure that com- functioning, cannot now be programmed for the
puters have thus far experienced. It is contended simple reason that knowledge of this interface of
that self-regaflating computer programs, if ex- the information-processing functions of the orga-
posed to "experience" of the order of magnitude nism is extremely meager. However, this may be
an important area for simulation study, for much likes, dislikes, and the like, with variable intensity
might be learned thereby. The advantage of weights and combinative rules. This problem
study of such problems in information-processing undoubtedly represents the ultimate complexity
terms is eloquently expressed in the following two discussed thus far.
statements (ref. 61, p. 362) in relation to a
Symbols versus Ideas
different problem:
One of the favorite points, made by Borko,
In the discovery of the functiona[ relations Neisser, and others of their persuasion on the
necessary for transforming a language input to a man-machine simulation issue, is that the ma-
language output for any given purpose, it is believed chine can manipulate symbols but has no ideas
that in symbolic form the functions necessary for
that are represented by the symbols. In a sense
the transformation of any other behavioral input
to output are being studied also. For example, if this is true. Nevertheless the very concept of
in hearing and answering a question, a subject must coding of information and of mechanistic process-
encode a sequence of sounds, correlate certain ing by manipulation of coded symbols makes this
aspects of this string with stored information, objection irrelevant, for there is little likelihood
evaluate the result, and encode it into an ap-
of any greater degree of ideation in the input,
propriate motor output, is it not reasonable to
conclude that these procedures are typical of the storage, control, processing, or output com-
functions used in reacting to a visually perceived ponents of the bioelectric computers than there is
situation? in the man-made nmchines. The nature of
It soon becomes apparent in the sturdy of language conscious awareness and the mechanisms mediat-
behavior that if a psychologist could understand
ing conscious experience are virtually unknown
all aspects of man's use of language, he probably
would in the process have developed a complete and remain as baffling today as to the first
functional blueprint of the relations holding the philosophers who attempted to penetrate their
vast area between stimulus and response. This mysteries. It is apparent that consciousness
functional understanding would be independent occurs, but whether it is an epiphenomenon, like
of the particular symbolism or coding used by any
the monitor set in the television studio, or integral
given sense modality.
to the on-going process is not known. What, is
known is that associative meaning can be coded
Mot;vat/on
and that symbols can represent ideas in informa-
Certainly any adequate simulation of human tion-processing terms without limitation as to
behavior must include consideration of motiva- intentional or extensional reference if such
tion. That a motivational system could be repre- references are included in the coding instructions.
sented in information-processing terms seems less
Total Simulation
important than Hunt's (refs. 95 and 96) extremely
significant and exciting prospect that motiva- This speculative discussion has necessarily
tional aspects of behavior are implicit in the taken into account the limitations of present
information-processing activities of the organism. knowledge and equipment. The question has been
Hunt has analyzed the effects of experience, ex- considered in terms of eventual possibility rather
pressed in terms of a continual interaction process, than immediate feasibility. In this frame of
in determining expectancies, sets, and adaptation reference it is assumed (quite reasonably, in view
levels (to use a few converging terms) and then of the rapid advances of science and technology)
exploited the activating, directing, and reinforc- that both knowledge and equipment will continue
ing effects of incongruity and dissonance as a the inexorable march toward realization of the
basis of a motivational mechanism. possibilities predicted.
Much more must be learned about human In speculation about future developments it
motivation before simulation programs will be seems appropriate to observe that present discus-
profitable. Such programs will undoubtedly in- sions of simulation of behavior are segmental,
volve a hierarchically organized set of facilitation relating to particular behaviors in isolation from
and inhibition commands representing inter- the total functioning of the organism. Wooldridge
related wants, needs, attitudes, interests, values, (ref. 97) in his fascinating survey of information-
32 :BIOMEDICAL RESEARCH AND COMPUTER APPLICATION
processing functions of living organisms demon- 10. WAND, J. II., Jr.: Markov Models and Monte Carlo
Techniques. Educ. Psychol. Meas., vol. 21, 1961, pp.
strates that even a simple mammal probably uses
219-225.
hundreds of computers, in a hierarchical organi- 11. DxxoN, W. J.: BIMED: A Series of Statistical Pro-
zation, for its physiological and behavioral func- grams for the IBM 709, Part 1. Behav. Sci., vol. 7,
tions. Some of these are analog computers, some 1962, pp. 264-267.
are digital, and some systems involve components 12. KELLY, E. L.; AND LINOOES, J. C.: Data Processing
in Psychological Research. Computer Applications
of both types. These develop in constant interac-
in the Behavioral Sciences, tI. Borko, ed., Prentice-
tion with an environment and are modified both
Ilall, Inc., 1962, pp. 172-199.
by growth and by experience. The organism is 13. LEESON, D. N.; AND D1MITRY, D. L.: Basic Program-
constantly programmed by its interactions in the ming Concepts and the IBM 1620 Computer.
continual environmental encounter. IIolt, Rinehart and Winston, Inc., 1962.
14. L_o_m, F. C.; ET At.: Abstracts of Statistical Com-
The hardware of the computer is gross and
puter Routines. Report no. 1056, Statistical Labora-
inefficient in comparison with the delicate micro-
tory, Case Institute of Technology, 1961.
miniaturization and efficient architecture of 15. Ta]'ON, R. C.: Domain Sampling Formulation of
nature; the physical mechanisms of information- Clusters and Factor Analysis. Psychometrika, vol.
processing are unquestionably different in sig- 24, 1959, pp. 113-135.
16. TRYON, R. C.: BC TRY System of Multidimensional
nificant aspects. Yet there seems hardly any
Analysis. Dept. of Psychology, Univ. of California,
limit to the extent to which processes identified
Berkeley, Oct. 1962.
in organisms could be simulated in computer 17. SnEPARD, R. N.: The Analysis of Proximities: Multi-
programs. As simulation studies expand in dimensional Scaling with an Unknown Distance
complexity and invade broader and more- Function: I. Psychometrika, vol. 27, 1962, pp.
125-140.
integrated segments of behavior, they will not
18. SnEPAnD, R. N.: The Analysis of Proximities: Multi-
only contribute to knowledge but also advance dimensional Scaling with an Unknown Distance
the science of man toward the ultimate goal that Function: II. Psychometrika, vol. 27, 1962, pp.
Saunders and I believe to be within reach. Of 219-246.
course a simulated man will never be human but 19. BocK, R. D.: Programming Univariate and 2klulti-
variate Analysis of Variance. Technometrics, vol. 5,
perhaps only a remarkable facsimile.
1963, pp. 95-117.
20. MOSTELL_R, F.; A,_-D WALLACE, D. L.: Inference in
REFERENCES
an Authorship Problem. J. Amer. Statist. Assoc.,
1. GREEN, B. F., Jm : Using Computers to Study Human vol. 58, 1963, pp. 275-309.
Perception. Educ. Psychol. Meas., vol. 21, 1961, 21. FILLENBAUM, S.; JONES, L. V.; AND WEPMAN, J. _.:
pp. 227-233. Some Linguistic Features of Speech from Aphasic
2. _rANDENBERO, S. G.; GREEN, B. F.; AND WRIGLEY, Patients. Language Speech, vol. 4, 1961, pp. 91-
C. F.: A Survey of Computer Usage in Departments 108.
of Psychology and Sociology. Behav. Sci., vol. 7, 22. OSGOOD, C. E.; AND MIRON, _I. S.; EDS.: Approaches
1962, pp. 108-110. to the Study of Apha._ia. Univ. of Illinois Press,
3. UHR, L.:"Pattern Recognition" Computers as Models 1963, pp. 127-130.
for Form Perception. Psychol. Bull., vol. 60, 1963, 23. JONES, L. V.; GOODMAN, M. F.; AND _'EPMAN, J. M.:
pp. 40-73. The Classification of Parts of Speech for the Char-
4. Jo,x_s, L. V.: Beyond Babbage. Psychometrika, vol. acterization of Aphasia. Language Speech, vol. 6,
.k
28, 1963, pp. 315-332. 1963, pp. 94-107.
5. BAKER, F. B.: Use of Computers in Educational 24. DELAVENAY, E.: An Introduction to Machine Transla-
Research. Rev. Edue. Res., vol. 33, 1963, pp. tion. Praeger, New York, 1960.
566-578. 25. EDMUNDSON, II. P., _.D.: Proceedings of the National
6. BoRxo, H., _D.: Computer Applications in the Be- Symposium on Machine Translation. Prentice-IIall,
havioral Sciences. Prentice-ttall, Inc., 1962. Inc., 1961.
7. COOLEY, W. W.; AND LONN_S, P. R.: Multivariate 26. OETTINCER, A. G.: Automatic Language Translation.
Procedures for the Behavioral Sciences. John Wiley IIarvard Univ. Press, 1960.
and Sons, Inc., 1962. 27. YNGV_, V. It.: Computer Programs for Translation.
8. GREEN, B. F., JR.: Digital Computers in Research. Set. Amer., vol. 206, 1962, pp. 68_7.
McGraw-Hill Book Company, Inc., 1963. 28. GREEN, B. F., JR.; WOLF, A. g.; CHOMSKY, C.; AND
9. SUPPES, P.; AND ATKINSON, R. C.: Markov Learning LAVOnEnV, K.: Baseball: An Automatic Question-
Models for Multiperson Interactions. Stanford Univ. ArLswerer. Proc. West. Joint Comput. Conf., vol.
:Press, 1960. 19, 1961, pp. 219-224.
COMPUTER
APPLICATIONS
IN THEBEHAVIORAL
SCIENCES 33
29. IIARRIS, Z. S.: Strong Analysis of Sentence Structure. Learning. in Mechanization of Thought Processes.
Mouton, The IIagne, 1962. IIer Majesty's Stationery Office, London, 1959,
30. SIMMONS, R. F.; KLEIn, S.; AND McCoNI,OGUE, K.: pp. 511-535.
Toward the Synthesis of Ihtman Language Behavior. 49. LETTWN, J. Y.; MATURA_A, H. R.; :%fcOvLI_)C_,
Behav. Sci., vol. 7, 1962, pp. 402-407. W. S.; AND PITTS, W. II.: What the Frog's Eye
31. BAKmt, F. B.: Information Retrieval Based upon Tells the Frog's Brain. Proc. IRE, vol. 47, 1959_
Latent Class Analysis. J. Assoc. Comput. Mach., pp. 1940-1953.
vol. 9, 1962, pp. 512-521. 50. SWETS, J. A.; GREEN, D. :hi.; AND WINTER, E. F.:
32. CnEXTHAM, T. E., Jl_.; AND Wn_aSIIALL, S. : Translation Learning to Identify Non-Verbal Sounds, J. Acoust.
of Retrieval Requests Counched in a "Semiformal" Soc. Amer., vol. 33, 1961, p. 855.
English-like Language. Commun. Assoc. Comput. 51. SELLS, S. B.: Personnel i%fanagement. Ann. Rev.
Mach., vol. 5, 1962, pp. 34-39. Psychol., vol. 15, P. R. Farnsworth, O. McNemar,
33. KENT, A.: Information Retrieval and Machine Trans- and Q. McNemar, eds., 1964, pp. 399-420.
lation. Vol. III, Pt. 1. Interscience, I960. 52. SILBEnMAN, H. F.; AND COULSON, J. E.: Automated
34. KENT, A.: Information Retrieval and Machine Trans- Teaching. Computer Applications in the Behavioral
lation. Vol. III, Pt. 2. Interscience, 1961. Sciences, II. Borko, ed., Prentice-IIall, Inc., 1962,
35. NEWMAN, S. N.: Information Retrieval, Toward an pp. 308-335.
Ultimate Universal System. Rev. Intern. Doc., 53. COVLSON, J. E., En.: Programmed Learning and
vol. 29, 1962, pp. 97-99. Computer-Based Instruction. John Wiley and Sons,
36. SWANSON, D. R. : Interrogating a Computer in Natural Inc., 1962.
Language. Proc. Intern. Fed. Inform. Process. 54. BUSH, R. R.; AND ]_IOSTELLER, F.: Stochastic Models
Congr., vol. 62, 1962, pp. 126-127. for Learning. John Wile)" and Sons, Inc., 1955.
37. VrCKERY, B. C.: On Retrieval System Theory. Butter- 55. BUSH, R. R; AND _IOSTE_LLER, F.: A Comparison of
worths, London, 1961. Eight Models. Studies in Mathematical Learning
38. STONE, P. J.; BALES, R. F.; NAMENV:IRTH, J. Z.; AND Theory, R. R. Bush and W. K. Estes, eds., Stanford
OG1LVm, D. M. : The General Inquirer. A Computer Univ. Press, 1959, pp. 293-307.
System for Content Analysis and Retrieval Ba._ed 56. COLEMAN, J. S.: Analysis of Social Structures and
on the Sentence as a Unit of Information. Behav. Simulation of Social Processes with Electronic
Sci., vol. 7, 1962, pp. 484-498. Computers. Educ. Psychol. Meas., vo]. 21, 1961,
39. SEIBEL, R. : Computer Solutions to Some Noncomputa- pp. 203-218.
tional Psychological Problems. Educ. Psychol. Meas., 57. SIMMONS, P. L.; AND SIMMONS, R. F.: The Simulation
voI. 21, 1961, pp. 185-201. of Cognitive Processes: An Annotated Bibliography.
40. GREEN, B. F., JR., WOLF, A. K.; Am) WIIITE, B. W.: Inst. Radio Engrs. Trans. Electron. Comput., voi.
The Detection of Statistically Defined Patterns in EC-10, I961, pp. 462-483.
a Matrix of Dots. Amer. J. Psychol., voI. 72, 1959, 58. SIMMONS, P. L.; AND SIMMONS, R. F. : The Simulation
pp. 503-520. of Cognitive Processes, II: An Annotated Bibliog-
41. WHITE, B. W.: Studies in Perception. In Computer raphy. Inst. Radio Engrs. Trans. Electron. Comput.,
Applications in the Behavioral Sciences, II. Borko, vol. EC-I1, 1962, pp. 535-552.
ed., Prentice-Hall, Inc., 1962, pp. 281-305. 59. FELDMAN, J.: Simulation of Behavior in the Binary
42. BARAN, P.; AND ESTRIN, G.: An Adaptive Character Choice Experiment. Proc. West. Joint Comput.
Reader. IRE Wescon Conv. Record, vol. 4, pt. 4, Conf., vol. 19, 1961, pp. 133-144.
1960, pp. 29-36. 60. FELDMAN, J.: Computer Simulation of Cognitive
43. KIRSCH, R. A.; Cx_s, L.; RAY, C.; AND URnXN, G. H.: Processes. Computer Applications in the Behavioral
Experiments in Processing Pictorial Information Sciences, II. Borko, ed., Prentice-Itall, Inc., 1962,
with a Digital Computer. Proc. East. Joint Comput. pp. 337-359.
Conf., vol. 12, 1957, pp. 221-229. 61. SIMMONS, R. F.: Synthex: Toward Computer Synthesis
44. LoE_R, E. E.: Image Processing and Functional of Computer Language Behavior. Computer Applica-
Retina Synthesis. Paper presented at Bionics tions in the Behavioral Sciences, II. Borko, ed.,
Symposium, Dayton, Ohio, 1960. Prentice-Hall, Inc., 1962, pp. 361-393.
45. UHR, L.; AND VoSS_R, C.: Suggestions for a General 62. W_ITE, B. W. : Review of Rosenblatt, Frank: Principles
Purpose Adaptive Computer Model of Brain of Neurodynamics: Perceptrons and Theory of
Functions. Behav. Sci., vol. 5, 1961, pp. 91-97. Brain Mechanisms. Amer. J. Psychol., vol. 76,
46. RosENBLAT'r, F.: The Perceptron: A Theory of 1963, pp. 705-707.
Statistical Separability in Cognitive Systems. Re- 63. ESTES, W. K; ANn BURKE, C. J. : A Theory of Stimulus
port No. VG-1196-G-1, Cornell Aeronautical Lab- Variability in Learning. Psychol. Rev., vol. 60,
oratory. 1958. 1953, pp. 276-286.
47. ROS_NnLXTT, F. : Principles of Neurodynamics: Per- 64. CLARK, W. A.; AND FARLEY, B. G.: Generalization
ceptrons and the Theory of Brain Mechanisms. of Pattern Recognition in a Self-Organizing System.
Spartan Books, 1962. Proc. _rest. Joint Comput. Conf., March 1955,
48. SEL._'RIDGE, O. G.: Pandemonium: A Paradigm for pp. 86-91.
34 BIOMEDICAL :RESEARCH AND COMPUTER APPLICATION
65. FARLEY, B. G.; AND CLARK, W. A.: Simulation of 81. NEWEI,L, A., ED.: Information Processing Language-V
Self-Organizing Systems by Digital Computer. Manual. Prentice-Hall, Inc. 1961.
IRE Trans. Inform. Theory, vol. PGIT-4, 1954, 82. NEWELL, A.; SttAW, J. C.; AND SIMON, II. A.: A
pp. 76-84. Variety of Intelligent Learning in a General Problem
66. IIEBB, D. 0.: The Organization of Behavior. John Solver. Self-Organizing Systems, M. T. Yovits and
Wiley and Sons, Inc., 1949. S. Cameron, eds., Pergamon Press, 1960, pp. 153-189.
67. AshBy, W. R.: Simulation of a Brain. Computer 83. NEWELL, A.; SIIAW, J. C..; AND SIMON, II. A.: Empirical
Applications in the Behavioral Science% II. Borko, Explorations of the Logic Theory Machine. Proc.
ed., Prentice-Hall, Inc., 1962, pp. 452-467. West. Joint Comput. Conf., vol. ll, 1957, pp.
68. CULBERTSON, J. T.: Nerve Net Theory. Computer 218-230.
Applications in the Behavioral Sciences, II. Borko, 84. NEWELL, A.; AND SIMON, TT. A.: The Logic Theory
ed., Prentice-Hall, Inc., 1962, pp. 468-489. Machine. IRE Trans. Inform. Theory, vol. IT-2,
69. GILMORE, II. F.: The Use of the Fourier Transform 1956, pp. 61-70.
in the Analysis of Visual Phenomena. Paper pre- 85. NEWELL, A.; ANn SIMON, II. A.: Computers in Psy-
sented at Symposium on Pattern Recognition, Ann chology. IIandbook of Mathematical Psychology,
Arbor, Mich., 1958. Vol. l, R. D. Luce, R. R. Bush, and E. Galanter,
70. GOODALL, ]Xl. C.: Performance of a Stochastic Net. eds., John Wiley and Sons, Inc., 1963, pp. 361-428.
Nature, vol. 185, 1960, p. 557. 86. FEIGENBAUM, E. A.; ANn SIMON, IT. A.: Performance
71. GREENE, P. It.: A Suggested Model for Information of a Reading Task by an Elementary Perceiving
Representation in a Computer That Perceives, and Memorizing Program. Behav. Sci., vol. 8,
Learns, and Reasons. Proc. West. Joint Comput. 1963, pp. 72-76.
Conf., vol. 17, 1960, pp. 151-164. 87. LAUGHERY, K. R.; AND GREGG, L. W.: Simulation
72. NOVIKOFF, A.: Integral Geometry, an Approach to of Human Problem-Solving Behavior. Psycho-
the Problem of Abstraction. Paper presented at metrika, vol. 27, 1962, pp. 265-282.
Bionics Symposium, Dayton, Ohio, 1960. 88. MINSKY, M.: Steps Toward Artificial Intelligence.
Proc. Inst. Radio Engrs., voI. 49, pp. 8-30.
73. FRANKEL, S.: Information-Theoretic Aspects of Char-
89. SAVNDERS, D. R. : IIow to Tell Computers from People.
acter Reading. Information Processing, UNESCO,
Educ. Psychol. Meas., vol. 21, 1961, pp. 159-183.
1960, pp. 248-251.
90. PEIRCE, C. S.: Logical Machines. Amer. J. Psychol.,
74. PUGAcn_v, V. S.: Optimum System Theory Using a
vol. 1, 1887, pp. 165-170.
General Bayes Criterion. IRE Trans. Inform.
91. YON NEUMANN, ,l.: The Computer and the Brain.
Theory, vol. 6, 1960, pp. 4-7.
Yale Univ. Press, 1958.
75. TurN, R. E.: On Dimensional Analysis. IBiXl J.
92. TtrltI_,-G, A. M.: Computing Machinery and In-
Res. Develop., vot. 4, 1960, pp. 349-363.
telligence. Mind., vol. 59, 1950, pp. 433-460. (Can a
76. CHow, C. K.: Optimum Character Recognition machine think? The World of Mathematics, J. R.
System Using Decision Function. IRE WESCON Newman, ed., Simon and Schuster, 1956, pp.
Cony. Record, vol. 1, pt. 4, 1957, pp. 121-129.
2099-2123).
77. _IATTSON, R. : A Self-Organizing Binary System. Proc. 93. BORING, E. G.: Mind and Mechanism. Amer. J.
East. Joint Comput. Conf., vol. 16, 1959, pp. Psychol., vol. 59, 1946, pp. 173 192.
212-217. 94. NEISSEn, U.: The Imitation of Man by Machine.
78. UTTLEY, A..-'_I. : The Design of Conditional Probability Science, vol. t39, 1963, pp. 193 200.
Computers. Inform. Control, vol. 2, 1959, pp. 1-2. 95. HUNT, J. _I.: Motivation Inherent in Information
79. UTTLEY, A. M. : Temporal and Spatial Patterns in a Processing and Action. Motivation and Social
Conditional Probability Machine. Automata Studies, Interaction, O. J. Itarvey, ed., The Ronald Press
C. E. Shannon and J. McCarthy, eds., Princeton Company, 1963, pp. 35-94.
Univ. Press, 1956, pp. 277-285. 96. ItvNT, J. M.: Intelligence and Experience. The
80. SERESTYEN, G. S.: Recognition of Membership in Ronald Press Company, 1961.
Classes. IRE Trans. Inform. Theory, vol. 7, 1961, 97. WOOLDRIDGE, D. E.: The Machinery of the Brain.
pp. 44-50. McGraw-Hill Book Company, Inc., 1963.
CHAPTER 3
MEDICAL DATA FROM FLIGHT IN SPACE:

OBJECTIVES AND METHODS OF ANALYSIS
Jefferson F. Lindsey, Jr.
The Medical Data Program of the National (5) Improvement of the criteria for selection
Aeronautics and Space Administration is designed and training of astronauts.
to meet three main objectives: (1) the safety of
The third objective--standardization of data
the astronauts while in flight.; (2) production of
for handling by computer--involves the first, two
new scientific information from the whole space
objectives since they cannot, be accomplished
program; and (3) the standardization of all
satisfactorily, efficiently, and expeditiously unless
medical data, derived either during flight or on
the third is met. Actor(tingly all me(Ileal data,
the ground, so that they are in a nmtually inter-
both past and future and whether derived during
changeable form for computer input and analysis.
flight or on the ground, must be recorded -rod pre-
Success in this program will faeilit.ate international
pal:e(l on magnetic tape in a standard manner so
exchange of such medical data and so contribute
that they are in a mutually interchangeable form.
to tile welfare of mankind.
By use of a proper standardized form or language,
The first objective entails the acquisition and
these data can be retrieved from comtmters and
proper utilization of all available medical data
brought to bear on specific past problems as well
bearing directly on the safety of crewmen in
as on current and future problems that may arise
flight. Such data must be in a readily interpretable
during or after future space missions. The com-
form so that physicians responsible for monitoring
puter programs must be prepared in advance for
of the medical aspects of space missions can use
the various graphic, mathematical, and statistical
them in assessment of the well-being of the crew
analyses so that interpretation can be immediate.
and take appropriate action at any moment while
Consideration must be given to the specific
a mission is in progress. Thus appropriate data,
in-flight and ground-based medical data prepared
telemetered during the flight, nmst be presented
for computer inputs. The in-flight data tele-
to the physician in such a form that he can com-
metered to Earth for medical monitoring during
pare them immediately with medical data previ-
the Mercury, Gemini, and Apollo missions are of
ously acquired during both ground-based studies
three types: physiological, spacecraft environ-
and space missions.
mental, and operational I)erformance. The physio-
Data pertinent to the second objective also
logical data for each astronaut include electro-
must be in a readily interpretable and standard
eardiographie records and data on respiration,
form for purposes of comparison, interpretation,
pulse, body temperature, blood pressure, etc.
and prediction. Yet they need not be available for
The environmental data include measurements of
simultaneous readout and immediate application,
acceleration rate, space-suit inlet and outlet tem-
because they are used for longer-range scient.ifie
peratures, carbon dioxide partial pressure, cabin
products applicable to
pressure, etc. The operational-performance data
(1) Advances in medical science and technology have been restricted for the most part to what
(2) Increase in safety of future crews each astronaut did or said, but more-elaborate
(3) More extensive flights monitoring is now being prepared. Aeromcdieal
(4) Development and design of spacecraft preparation (ref. 1) for and observations (ref. 2)
equipment involving man-machine relations front the Mercury missions have been reported.
35
Ground-based medical data are of many types; to obtain; once a space mission has started, there
some of the more important ones relate to mission can be no turning back, or second attempt, as
simulation, the clinical medical history of the would be possible in most ground-based physical
astronaut, and the base line. The mission-simula- examinations, tests, simulations, or medical ex-
tiim data for the Mercury, Gemini, and Apollo periments. Second, these in-flight data are highly
programs include information obtained during important and valid for consideration in analyses
studies in which a space mission was wholly or because they provide information about the pre-
partially duplicated on Earth; such devices as a cise reactions of crewmen under conditions of real
centrifuge, a space chamber, a mission simulator, space flight. Third, the in-flight data have a
and a procedures trainer were used in addition to direct bearing on astronauts' safety; since atten-
immobilization studies. The data acquired during tion must be focused on some selected aspect of
these simulations include those obtained during such a comprehensive program, the initial
real space missions. In most cases, however, the emphasis is better placed on data having a direct
former are far more extensive (refs. 3 and 4). bearing on safety of flight. Finally the analyses of
Data in the clinical medical history of each in-flight data largely serve to indicate future re-
astronaut are from iris cumulative record of quirements of data from both ground-based
periodic physical examinations over a number of studies and missions.
years (refs. 5 and 6); they may be supplemented The next questions to be answered were: How
by physiological and psychological test data can in-flight medical data be prepared for im-
acquired during his selection. Four phases of the mediate (instantaneous) use by the medical
astronaut-selection program, as well as a machine- monitors of each mission and for expedient post-
record system to facilitate recording and analysis flight analyses? And how can all medical data be
of medical data, have been described (ref. 5). The prepared in a form for computer inputs, incor-
third phase of the selection program, including the porating both in-flight and ground-based data?
tests administered and the results obtained, also During the search for answers to thcse specific
has been reported (rcf. 7). The clinical medical questions, the time-line presentation was de-
history includes data obtained immediately before veloped; the rest of this chapter will be devoted
and after each space mission; these are necessary to explanation and application of this concept,
for studies of any significant physiological change with attention focused on in-flight medical data.
resulting from the flight (ref. 8). The inductive leap necessary for conception of
The base-line data include those available from the way in which this approach can be extended
the literature, indicating norms and tolerances for to ground-based medical data will be left mainly
humans with respect to earlier medical measure- to the reader, since such applications are beyond
ments established under specified conditions. the scope of this chapter. Furthermore I should
Base-line data require little elaboration apart from point out that the time-line-analysis procedures
emphasis on the fact that they are based on a described herein are applicable to many other
highly selected sample, the astronauts, as well as problems of a situational-analytic type, ranging
on general base-line data av,dlable in the from problems involving simple human operations
literature. in a controlled laboratory environment to those
A prime objective in establishment of NASA's encompassing time-line analyses of airerew oper-
medical-data program was preparation of all data ations while the crew is in the process of fying a
in a standard, mutually interchangeable form supersonic aircraft on a mission lasting many
suitable for computer inputs. A question naturally
hours (ref. 9). Researchers and scientific investi-
followed: What type of data should be used
gators are therefore asked to consider the next
initially for establishing the pattern to which all
portion of this chapter with a view toward not
other types of data can be related': As a result of
only understanding the described methods of
a comprehensive study, the in-flight type of
medical data was selected for this purpose. analysis of the in-flight medical data but also
This selection was based on several considera- modification of the methods for application to
tions. First, in-flight data are the most difficult their particular disciplines.
MEDICAL DATA FRO_I FLIGHT IN SPACE .'_7
PREPARATION OF TIME-LINE MEDICAL Description of Blacks of Data Selected

DATA The first block or group of successive data
sheets, for each mission, included a series of 15
General consecutive 1-min periods covering the time from
1 hour before (T-60) to 45 rain before takeoff
In the time-line-analysis approach, data sheets (T-45). The first data sheet or tabulation in this
are constructed representing successive time group covered the l-rain period starting at T-60
periods. The approach can best be presented by and lasting until T-59. The next consecutive
describing how the in-flight medical data from l-rain period covcrcd the time between T-59 and
the manned Mercury and Gemini flights have T-58; the next, T-58 and T-57; and so on
been prepared for computer inputs in a standard, until 15 consecutive sheets were printed. This
magnetically taped format. All relevant informa- block of 15 sheets covered the time between
tion available for a given brief period of time was T-60 and T-45 for each manned space flight
printed on one data sheet by use of a computer (table 1). This procedure can be extended to in-
and its associated equipment; this included all clude additional missions and simulatcd missions.
available flight information of value to the The next group or block of data sheets included
physician concerning the well-being of the astro- the series of consecutive 10-scc periods covering
naut for the period represented. Since the physi- the time between T-120 sec (T-2) and T-zero
cian is interested in a composite presentation of (tttblc 2). This size of sample will be extensible
all relevant information during any given period, by use of data from future missions, and from
each data sheet included the astronaut's physio- past and future simulated missions.
logical data, spacecraft's environmental data, and The next blocks of consecutive data shects
astronaut's performance data. Thus the physician covered the time between T-zero aml the onset
can appraise the relations within and interactions of zero gravity (g) in 10-scc periods, since this
among these various factors. Additional data time was always stressful. The next block .of
sheets were constructed for consecutive time consecutive data sheets covered the time between
periods, each of which showed measurements of T+30 and T+45 in 1-min periods, since these
the same type as those presented on the preceding data accompanied weightlessness--less-stressful
sheet but of course differing in value because they portions of the missions. To carry this process to
pertained to different periods. its conclusion, a number of sclectcd blocks of
The requirement for duration of the periods for consecutive periods were chosen for dctailcd
data sheets was different for various portions of analyses. These blocks of data covered times
the mission because the physician is interested when the astronauts were engaged in such func-
not only in change per se but also in the rate of tions as exercising (performing identical exer-
ehangc and in the rate of rate of changes of both cises), resting or sleeping, performing or monitor-
physiological reactions and environmental condi- ing retrofire operations, exiting, or binding. The
tions. These kinds of changes are generally more specific periods chosen are shown in detail in
rapid during the stressful conditions of the last table 3.
2 rain before takeoff and during exit and reentry. Study of the information in figure 1 will pro-
Generally less stressful portions of a mission occur vide an insight, into the manner of comparison
during weightlessness, the last. 1 hour before and statistical treatment of time-line data. It
flight, and after flight. Thus data sheets, each becomes obvious that data from one mission can
covering the short period of 10 see, were selected be compared with data covering other missions,
for the stressful portions of the missions; whereas within certain lin_itations. Also, measurements
data sheets, each covering a l-rain period, were taken early during any given mission can be
selected for less-stressful portions. The reason for compared with those taken (luring the latter
this selection can be shown more clearly by de- parts of selected portions of the same mission
scription of the blocks of data sheets selected for for assessment of possible changes. Furt.hermore,
the various types of analyses to be performed when data arc prepared in the manner described,
(tables 1 and 2). the analyses need not be restricted to the period
X X X X X X X X X i;o_ _ X X X X X X X X X
X X X X X X X X X
._o_ X X X X X X X X X
bq
X X X X X X X X X
X X X X X X X X X
¢b
I .,-_ 1 X X X X X X X X X
X
X X X X X .X X X X X
X X X X X X X X X
I
X X X X X X X X X X X X X X X X X X
I
s'? X X X X X X X X X
'4D Ol_
I -.,.-.'I X X X X X X X X X
I
,_ L_ LQ
X X X X X X X X X 'r--,
_ L_w
X X X X X X X X X
ca X X X X X X X X X
X X X X X X X X X
] S t X X X X X X X X X
d_
x x x x x x x x x X X X X X X X X X
'l 2'_ x x x x x x x x x A
X X X X X X X X X
x x x x x x x x x
'T 'T X X X X X X X X X
X X X X X X X X X
r_
"e X X X X X X X X X
x x x x x x x x x
I I
@1
<
o o o _ O O © © O
MEDICAL DATA FROM FLIGHT IN SPACE 39
TABLE 3.--Types of Activities
Criticalness (C) Difficulty (D) Duly (T) Procedure (P)
1C : Highly critical 1D: Very difficult 1T: New (performed in space- 1P: Active
craft only)
2C: Medium-critical 2D : Medium-difficult 2T: Revised (combination of 2P: Passive

old and new)
3C: Noncritical 3D : Easy 3T: Old (previously performed 3P: Concurrent
in aircraft)
4P: Shared (with ground
personnel)
_R 3 L5MIN OATA 2MIN OA]A

ONTINUBUS
15 TABS 12 TABS
DATA
;S TABS
OT DATA
5 1AIBS / / / / / LANBING]5 WIN
I M_SAMPI[
5 TABS
°,,,-////
MR 4
,_ Of OAIA
5 TABS
MA-8 5MIN OATA 15MINOAT/
30 TABS 15 TABS
15MINOATI 00750OT{
15 TABS 5 TABS
// 3:09:OBT8
15 TABS
// ' ' 4 29 03 TO
//!ii!!: °'
°!
MA ,'1 3 5?OOTO
4 07 O0
15 TABS
_
o,
MA 8 OD 30 O0 _ 3:24 08
7:37:OO I, 90140 91311
15 TABS 15 TABS 70 TABS
/
2 _900 TC 1359 OOT( 32 29 30TO! ,33 53138 TC 3_ 08 3B_(

MA 9
_, _, ,_ _, 234 GO 1503 00 I, 1_ 32 44 30 1, ,_ ,34 08 38 34 19 49
;5 TABS 27 T_$ 15 TABS 15 lABS _ TABS
MA 9 TS:00:OGTO
25;15O0 I,
15 TABS
TOTALs, 1133 TABS
FmORE 1.--Example of Mercury biomedical-data requirements; tabs, tabulations (data sheets).
represented by each data sheet--10-sec or l-rain tion included on each data sheet, is now in order;
periods. If, for example, one has 15 consecutive this will be accomplished in conjunction with
l-rain data sheets, the computer can be pro- illustrations of two types of data sheets for
grammed to treat these data as either a con- selected 10-see periods (figs. 2 and 3). Data
solidated or a segmented block of data of any sheets for the l-rain periods were identical in
desired length in minutes or fractions of minutes, format except that no acceleration data are in-
such as 15, 10, 5, 3, 1, ½, }, or _; the sa,me is true eluded, since the 1-min periods are applicable to
of the 10-sec-period data sheets. times of the mission preceding flight, during
weightlessness, or after flight, when no accelera-
Descrlption of Dala-$heet Content tion forces were present. Therefore, for all prac-
Detailed examination of the specific informa- tical purposes, the following discussion of the
4O BIOMEDICAL RESEARCH AND COMPUTER APPLICATION
TIME- LiNE DATA

PROJECT P/-_RCURY MISSION ELAPSED TIME : 00_00, Z0 TO 00,00,30
,,ERc_ ,T_,S_o DATE
RESPIRATION STANDARD ACCEL. STANDARD SUIT-]_q SUIT- OUT CO2 PARTIAL CABIN
RATE SCORE Z-AXIS SCORE IEMP TEMP _ESSURE PRESSURE
IIEATS/MfN _EATHS, IMIN G-S BEG F DEG F PSIA PSIA
115.60 38.20 16.90 36,40 $ .77 50.30 65.22 E4.52 .oo_ 15.245
fl2,14 36.79 ,5.78 35,39 ! .90 50.9'6 65.18 84.46 .C000 15.219
_I2,14 36,79 22.72 4! ,63 1.77 50.30 65.22 B4.52 .oOOo 15.196
1t3.8.5 97.47 1.71 50.00 65.26 84.58 ,0000 15.229
t15.58 38.19 1,71 50.00 ._0_0 15.229
DS._.O 38,20 qp 1.64 49.65 O .0(_0 15.245
113._, 97.49 1.37 48.28 .DOC_ 15.276
1T5.60 3fi.20 1.90 50.96 41_ .0C¢-_ 15.178
I12J4 36.79 2.43 53.65 .0000 15.198
113.85 97.49 4p 2.04 51.67 000o lS.lse
i 13._L5 37'.49 1.71 50.C_ . O0(XI 15.174
t13.58 38.19 1.9C. 50.96 ,Ixl00 15.158
iii1 Illl
19.28 39.69
121.10 40.4o MEAN 41P tl" MEAN ql _
12! .t8 40.46
123.20 41.28 _" 65.22 B4.52 .O0g0 15.216
121.18 40.46
,/9.26 39.68
qp.
$23.17 45,27
123.20 4l .20
,ell.
ACTIVII Y
MI_,N = 117.07 MEAN _ 18.46 MEAN = $ .B2
SIGMA" 9.96 S1GMA = 9.72 SIGMA = .25
VARIANCE = 15.68 VARIANCE = 13.88 VARIANCE = ,06 I. PLANNED START BACKUP CLOCK.
2. INDICATED START BACKUP CLOCK.
C OMMUNICA T_C_ S
00,00,20 CC MARK.
00,00,23 P ROGER. ANO THE BACKUP CLOCK IS RUNNING
00,00,25 CC RC_ER. YOU tOOK GOOD HERE, GORDO.
00,_0,27 P ROGER. FEELS GOOD, _IDDYI
00,00,29 CC GOOD SPORT,
FIGURE 2.--Exemplary data sheet for 10-see period.
10-see type of data sheet will suffice; the discussion rate data points within a predefined period for
is keyed to figures 2 and 3. any given mission (e.g., all data points during the
Heading--The heading of cach data sheet con- combined periods of exit and reentry), and then
tained information identifying the d,tta as finding the mean and standard deviation of the
time-line data and indicating the mission from distribution of these data. Thence the z-score
which they were taken, the mission's elapsed was calculated for each heart-rate data point in
time, and its date. The time-line data (fig. 2) column I. Since z-scores contain negative num-
were taken from a Mercury-Atlas mission, bers, a new distribution (standardized score) was
MA-9; the mission's elapsed time was the time formed with a mean of 50 and a standard devia-
between 20 and 30 sec after takeoff on May 15, tion of 10 by multiplying each z-score by 10 and
1963. The reason for including the information adding 50. The standardized scores for any given
presented is obvious: each data sheet must be an mission in progress cannot be computed in-
identifiable entity in itself to be used by the stantaneously as the mission progresses since all
computer for analytic purposes. data points in the distribution (exit and reentry,
Heart rate--The rate of the astronaut's heart in this example) must be known before the calcu-
beat in beats per minute (R to R) is shown for lations can be made. However, the standardized
each beat that occurred during the 10-see period scores for each heart-rate data point can be
represented. In the example there were approxi- calculated instantaneously if the distribution has
mately two beats per second since the mean for already been established during previous missions.
the 10-sec period is very nearly 120; more pre- Details of methods for these calculations are
cisely it w,_ 117.07. Since the heart does not available (refs. 10 to 12).
beat at a constant rate, it was possible to solve Again the remsons for obtaining and printing
for the standard deviation (sigma) and the digital heart-rate data on each data sheet are obvi-
variance for the period represented. The _cond ous, but the calculation of means, standard devia-
column shows the standard score for each beat tions, and standard scores requires at least brief
represented in the first column. These standard justification. The means and the standard devia-
(z) scores were calculated by using all the heart- tions can be graphically printed by the computer
VOICE ................... -. ..........
Time, seconds
Fmv_s 3.--Exemplary data sheet for 10-sec period; ECG, electrocardiogram.
for presentation to the physician as the mission during exit and reentry are considered. If the
progresses, to indicate trends or patterns that may standard score is 40, he knows that this particular
be developing in heart rate. These statistics can beat is 1 standard deviation below the mean.
also be used in connection with various graphic, Additionally by converting to standardized scores
mathematical, and statistical analyses as will be instead of using z-scores, the need to work with
explained later. The standard scores, based on a negative numbers is eliminated; this is generally
mean of 50 and a standard deviation of 10, pro- convenient and sometimes necessary for certain
vide the physician with definite information as to types of analyses. Finally conversion of not only
how near normal is the heart's beat in comparison heart-rate but also respiration-rate and accelera-
to what should be expected at any given time tion-rate data to standard scores makes possible
and in any given circumstance for the astro- comparison of these diverse types of measure-
naut(s) participating in the mission. For ex- ments by use of a standard base (standard score);
ample, if the standard score is 50 (the mean), the accordingly various other analytic techniques
physician knows that this is the average rate for become available for use.
the astronaut when all heart beats to be expected Respiration--The respiration rate, in breaths
per minute, for the 10-sec period is shown (fig. 2) operational performance, as well as of the well-
along with the standard score for each entry. The being of the astronaut, was the voice data; they
standard scores are based on all respiration-rate- were supplemented by communications with
data points covering the same period of the mis- ground crewmen. Figure 2 shows the communi-
sion that was used to calculate the heart-rate cations between ground crewmen and the astro-
standard scores--in this instance, exit and re- naut during the 10-see period.
entry. Also, as in the case of heart-rate data, the Many types of analysis can handle voice data,
mean and standard deviation are shown for the ranging from assessment of the state of the
respiration data for the 10-see period represented. astronaut's alertness (and probable relations
The rationale for inclusion of the foregoing associated therewith) to analyses pertaining to
respiratory information in each data sheet was speech processes, audiology, and information-
similar to that previously described; when using processing (ref. 13). The physician can gain con-
several consecutive data sheets, one can begin siderable information about the condition of the
to identify existing trends, patterns, and relations. astronaut merely by listening to his voice and
Other measurements--Other physiological meas- conversing with him. Here again though, for use
urements mostly of a periodic nature were taken of voice data in the computer, voice content
but are not shown since they are not applicable to must be digitized. Therefore, operational defini-
the 10-see-period data sheet used here for illus- tions are required encompassing areas such as
trative purposes. probable attention, joy, fatigue, confusion, and
Environmental measurements--These measure- relief. (These terms originated with L. V. Surgent;
ments include acceleration, space-suit inlet and see Chapter 10 for details.) Factors considered in
outlet temperatures, carbon dioxide partial pres- classification according to these areas include
sure, and cabin pressure (fig. 2). The data sheets voice pitch, timbre, and speed, and in some cases
included rate information for each of these the quickness of the astronaut's response to ques-
parameters, plus the mean, standard deviation, tions or instructions received from the ground.
and standard scores for acceleration data. Only These questions and instructions, however, must
the mean was calculated for the other environ- be weighed with extreme care because a response
mental measurements. It can be seen now that may not be immediate because of overriding
trend information for environmental measure- operational functions in which the astronaut may
ment becomes available, as do data pertaining to happen to be engaged at the time. Consideration
relations both within and among physiological must also be given to the fact that astronauts are
and environmental aspects of space flight. a highly selected group with exceptional ability;
Activity--One type of operational-performance they are conditioned to excitement and stress.
data of a continuous nature, readily available for Therefore, very little degradation in their voice
incorporation in the selected data sheets, was a content may be more significant than greater
description of both planned and actual activity change for the normal population.
of the astronaut during the period represented. Wave-train data--These consisted of analog
The astronaut was supposed to and did start the readouts of the electrocardiogram (ECG), res-
backup clock during the 10-see period (fig. 2). piration, acceleration, and voice (fig. 3). These
By recording the planned as well as the actual are required for two main purposes: to check the
activity one can tell whether the astronaut is on, correctness of the digital data on the data sheet
ahead of, or behind schedule and assess the impli- and to use in connection with certain types of
cations thereof. However, in order to utilize these pattern and wave-form analyses.
data digitally for analytic purposes in relation to
physiological and environmental data, it was TYPES OF ANALYSES
necessary to construct operational definitions for
General
each type of activity. These definitions were sub-
divided into four areas having assigned numerical When data thus prepared in time-line format
values or weights (table 3). are available on magnetic tape, many types of
Voice--Another type of indirect indicator of analysis can be performed that apply directly to
the safety of astronauts in flight and to scientific entry on each data sheet can be used as an index
products derivable from such medical data. for this purpose. However, if one is interested in
Several aspects relative to these analyses, together rate-of-change information based on periods
with limitations, will be discussed now. different from those for which calculations appear
on the data sheets, then the instantaneous-heart-
Graph/c Analyses rate raw data on each data sheet must be used
One-dimensional graphs of the means, vari- since the summarizing effect of variance averages
ances, and standard scores can be plotted for out information concerning the variability for
comparable time periods for each astronaut on these other periods.
all measurements or for all astronauts on one In attempting to quantify rate information, it
selected measurement. For example, the graph of might be expected that the curve for the first and
heart-rate for the 5 rain immediately after takeoff second differentials of heart-rate data would pro-
consists of 30 data points--one for each 10-sec vide a measure of rate of change and rate of rate
period shown on each data sheet for 5 consecutive of change, respectively. However, this is not the
minutes. Two-dimensional graphs also have been case because of several considerations. The first
constructed showing, for example, heart-rate on consideration is that there is an assumption for
the ordinate and acceleration on the abscissa. the process of differentiation that requires meas-
Additionally three-dimensional graphs can be urements to be continuous; instantaneous heart
constructed with various colors used for the third rate is not continuous in a measurement sense.
dimension. Second, since in the case of heart rate a period of
Although most of these graphs can be plotted perhaps 15 min is regarded as being a meaningful
by the computer in real time as a mission pro- period, the difficulty of fittirg a curve to the
gresses, this information must be compared with variations in heart rate over that period must be
earlier data so that it may be meaningful to the considered. Differential calculus does not work
medical monitors. Therefore overlays need to be well under conditions of long time periods and
used so that graphic information concerning the irregularly shaped curves. Third, to quantify
current mission can be superimposed on previ- rate information, it is highly desirable and even
ously constructed graphs based on time-line data necessary to arrive at a single number to repre-
acquired from either completed or simulated sent the rate of change and the rate of rate of
missions. Overlays are also useful in comparison change of each astronaut. The question of how
of data acquired early in any given mission with to get this quantification from the equation for a
data acquired many hours after takeoff. curve poses a difficult problem. Fourth, a desir-
Often overlooked is the fact that these graphs able product would be the determination of
provide the analyst with a method for quick significance of differences existing between two
visual inspection of data. These inspections can astronauts with respect ¢o their rate-of-change
lead to clues as to the nature of relations that and rate-of-rate-of-change characteristics; here
may exist within and among physiological, en- again calculus does not provide a basis for the
vironmental, and performance factors. With these determination.
clues, hypotheses can be formulated and tested The solution seems to lie in using the concept
by use of appropriate mathematical and statistical of differentiation where it applies, and in using
methods and models. other techniques to modify the concept where it
does not apply (ref. 14). Using table 4, the method
Rare-of-Change Analyses for accomplishing this is described in the following
The rate of change and the rate of rate of paragraphs.
changes in physiological measures (such as heart Columns 1 and 4 each contain 15 items of
rate) under various environmental conditions heart-rate data for subjects A and B, respectively,
provide a sensitive index of the physiological re- over a certain period. If one of these heart rates
activity of the astronaut to his environment. If were calculated every 10 sec, the period would
one is interested in rates of change that occur span 150 sec, or 2.5 rain, for each set of data.
within 10-sec and 1-min periods, the variance (Raw data can be used instead of 10-sec averages
TABLE 4.--Hypothetical Heart-Rate (HR) Data and Their Statistical Treatment
Subjeet-A Subject-B
(1) (2)* (3) (4) (5)* (6)
d(HR) d_(HR) d(HR) d2(HR)

Heart rate Heart rate
dt dt z dt dt 2
87 87
1 2
86 0 85 1
1 1
85 1 84 3
0 4
85 1 80 1
1 3
84 0 83 0
1 3
83 1 80 1
2 4
81 2 84 2
0 2
81 0 86 1
0 1
8I 1 85 2
1 3
82 1 82 1
0 2
82 1 8O 0
1 2
83 1 82 0
2 2
81 1 80 0
1 2
82 0 82 0
1 2
8I 84
Mean, 82.9 3Jean, 0.86 :Mean, 0.76 3Jean, 82.9 Mean, 2.36 Mean, 0.92
S.D., 1.44 S.D., 2.23
*Testing of significance of differences between average Mann-Whitney U-test (ref. 15): U=19, P<.001 (differ-
rates of change (columns 2 and 5) for subjects A and B. ence is significant).
if assessment is required for a smaller incremental sumed to be a first differential, it is treated here
rate of change.) as basic data.) To follow this procedure the first
The mean heart rate of subject A for the 15 heart rate in a column is subtracted from the
entries during the recorded period is 82.9 beats second, ignoring signs, and recording the differ-
per minute; his variability is 1.44 beats per ences as shown in column 2. The second heart
minute. The mean heart rate for the 15 entries rate is then subtracted from the third and re-
for subject B is exactly the same (82.9), but the corded. Continuance of this process provides a
variability is different (2.23). column of numbers that, if plotted against time,
The question arises: Is subject A different in yields the curve produced by the first differential,
his rate of change of heart rate from subject B? or a curve of rate of change. However, since it is
To answer this, the first differential or (perhaps as desirable to avoid dealing with curves, this column
it should be stated here) the first set of differences of differences is simply averaged, thereby arriving
can be found. (While heart rate itself may be as- at the mean difference in heart rate during the 150
sec, or the mean rate of change for the subject Some Computer Programs*
during this period. The signs in this computation Time-line-data programs--This program com-
are disregarded since the analyst is interested in putes the means, variances, standard deviations,
only the rate of change and not the direction of and standard scores for all incoming digital
this rate. In terms of calculus, one would be deal-
in-flight and appropriate ground-based medical
ing with a nondirectional derivative. In a like
data. During the Mercury missions the analog
manner, table 4 shows that the column of first
medical data had to be converted to digital form
differences, as well as its mean, also has been
before this program could be used; an example of
calculated for subiect B.
its type of outputs is shown in figure 2.
The question now is: How does one arrive at a
Distribution program--This program computes
statement of the significance of difference between
the means, variances, standard deviations, stand-
the mean rates of change in the heart rates of the ard errors of the means, and critical ratios for
two subjects? Usually a t-test is used in such skewness and kurtosis for a series of variables.
situations, but a t-test cannot be applied in its The program is used mainly for assessing the
usual form here because the variability of heart shape of the distribution of selected data as
rate is related to the degree of rate of change indicated by the ratios for skewness and kurtosis.
between the two subjects. Thus, when the sub- The ratio of skewness indicates the degree of
jects differ as to the rate of change, the homoge- significance to which the particular distribution
neity-of-variance assumption of t cannot be met. varies from the normal distribution and the direc-
Furthermore, there is some question as to whether
tion in which it varies (positively or negatively
the data are indeed at more than the ordinal level
skewed). The critical ratio of kurtosis indicates
of measurement. Unless the interval level may be
the significance of the peakedness characteristic
assumed, use of t cannot be justified; and in future
of the distribution; that is, the extent to which
applications of this technique, badly skewed dis- the distribution is flat (platykurtic), medium-
tributions may be expected. Thus, for the general peaked (mesokurtic), or highly peaked (lepto-
case, use of the parametric t-test is not recom- kurtic). These two characteristics are of interest
mended as a test of significance; instead its non- due to assumptions of a normal distribution in
parametric counterpart, the Mann-Whitney many of the uses of the mean, variance, standard
U-test (ref. 15), should be used. Applying the deviation, and standard error of the mean in
U-test to the above data to determine the sig- statistical work. If the distribution is not normal,
nificance of the difference between the two mean the critical ratios of skewness and kurtosis reveal
rates of change, the absolute difference of 1.50 is this discrepancy, and a transformation of the
found to yield a U of 19. The difference in rate of
scores is required for appropriate application of
change between the two subjects is highly the analytical methods to the data.
significant (P<.001). Chi-square and frequency-distribution program--
To measure the difference between the two
This program computes the means and standard
subjects in terms of their rate of rate of change, deviations, and classifies data or subjects in up to
the second differential or set of second differences
•seven categories of response, as well as the per-
is calculated--from the column of first differ-
centage in each category. The program also com-
ences as shown in columns 3 and 6. Note that
putes a seven-category as well as a three-category
the technique for securing the second differences chi-square (ref. 16). Using the seven-category
is exactly the same as was used for securing the
chi-square, one can test either for significant
column of first differences. The mean of the
differences among up to seven astronauts with
second-differences column is secured for each
respect to one selected type of measurement
subject, and then their mean difference is tested under identical conditions, or for significant
for significance in the same manner as were the
changes taking place in one astronaut with re-
differences for rate-of-change information. The
significance of difference of the rates of rate of
*I thank Benjamin Fruchter, D. V. VeIdman, and
change has not been worked out for these data
Earl Jennings, University of Texas, Austin, who wrote
as an example. several of these computer programs.
spect to one selected type of measurement during ability, one is justified in rejecting the hypothesis
seven different periods. Tables 5 and 6, re- that the observed distribution is derived from a
spectively, illustrate these possibilities. population with equal frequencies in the three
In application of this program, two separate categories.
chi-square tests are computed. First the distribu- Three-way analysis-of-variance program--This
tion of observations in seven categories is tested program computes a three-way analysis of
for fit against the expected distribution, usually variance, printing out a complete analysis-of-
one in which the probability of the observations variance summary of all combinations of all
(data) is equal in all categories. Thus, if there is means. It permits the application of a factorial
in fact no difference in the observations, each design in which subjects or data can be classified
category has an equal opportunity of being of along three separate dimensions; for example,
the same magnitude. The statistical hypothesis two levels of performance (two levels opera-
tested is the null hypothesis, and the chi-square tionally defined), three levels of oxygen content
test then determines the probability that the (02_, 0_, and 0_3), and three time periods (first
observed distribution of data was derived from 15 min after zero-g, a 15-min sample after 1 hour
the theoretical distribution. of zero-g, and a 15-min sample after 2 hours of
Secondly the observations are classified into zero-g).
three categories. The computer program sums Analysis of variance, then, permits simultaneous
the observations (or data) in categories 1, 2, and comparison of data that are arranged in a particu-
3; next it sums the data in category 4 (neutral lar manner and classified according to certain
category) ; and then it sums the data in categories dimensions. All combinations of means are com-
5, 6, and 7. The chi-square test is then applied to pared, and contributions to variance are analyzed.
determine the significance of any difference, in Differences between means and combinations of
observations (or data) at either end or in the means are analyzed for significance beyond those
middle of the distribution, from the expected attributed to chance probability. If differences are
distribution. The resulting cAt-square is printed significant, inferences can be made for tim classi-
along with its level of signifcance. If the cAt- fications of the dimensions employed.
square is significant at a present level of prob- Correlation and regression program--This is a
comprehensive computer program that computes
the means and standard deviations for a series of
TABLE 5.--Seve_l Astronauts: One Type of
variables, further computing the intercorrelation
Measurement Under Identical Conditions
matrix and a complete multiple-regression analy-
Seven-Category Chi-Square sis that provides beta weights, multiple R-squares,
I variance, multiple correlation, corrected multiple
Astronaut 1 2 3 4 ' 5 6 7
correlation, and R-ratio. Use of this program with
in-flight medical data has been restricted to cor-
relation. The correlation matrix provides the
Frequency
Percentage --________ ---_ -- analyst with information as to how one measure
relates to another for a given period and condi-
tion over a sample of subjects; for example, one
can solve for the degree of relations among such
TABLE 6.--One Astronaut: One Type of Measure-
measures as heart rate, respiration, acceleration,
merit During Se_,en Different Periods
voice, performance, and carbon dioxide partial
Seven-Category CAt-Square pressure.
Factor-a_alysis program--This is a compre-
Time 1 2 3 4 5 6 7
hensive computer program that computes means,
standard deviations, principal-axis-factor analy-
sis, varimax rotation, multiple regression, factor-
PercentageFrequency .l@l@l @ _t@l_l" --
score weight-estimation, and standardized factor-
score computation. Useful methods employed in
factor analysis are explained (ref. 17), together pal method used in testing the serial dependency
with many examples of practical applications. of observations is autocorrelation, which is cor-
Factor analysis is used for analysis of the correla- relation of the series with itself retarded by one
tion matrix to determine the common factors or more time periods.
basic to a set of different measurements; the
solution is portrayed in the form shown in
INTEGRATED-COMPUTER-SYSTEM CONCEPT
table 7.
In addition to the time-line medical-data and
TABLE 7.--Format of Solution by Faclor Analysis associated computer programs described in this
chapter, there have been significant concurrent
Factor
Measure data-analysis programs under development, aimed
I II III h_ at having each type of data form an input into an
integrated computer system (fig. 4).
Heart rate ....
Respiration ....
Acceleration ....
MedlcaI Medlcal d I
Blood pressure .... tn-flight I | I laboratory ata I
clinical history - -
tlme-line data
Voice .... data ] [(blood chem_ stry_
Performance ....
Muscle tone ....
Electroencephalogram ....
Galvanic skin response ....
Data From medical Integrated _-41_-_ caE_ieC;rrThic J
expe riments computer system ] Ipattern-analysis]
When conducting ground-based studies--for

example, if the solution to a factor-analysis prob-
lem resulted in a high rating on factor-1 under nil
conditions, for both galvanic skin response (GSR)
and muscle tone (EMG)--one would have evi- All appropriate
. /encepha l o 9 rapb i cI
medical computer
dence that GSR could be measured without applications
measurement of EMG--one of the measurements

would not be required. This type of information
FIGURE 4.--Concept of integrated computer system.
can be very useful and valuable in view of the
current limitations on space, weight, and power
within capsules during space missions.
The purpose then is to have relevant on-line
Statistical-Model Limitations data as well as appropriate stored medical data
available and "on call" to the physician, through
Since many of the data represent time series,
the integrated computer system, to promote
there are several inherent difficulties in their
maximum safety for our astronauts and to ensure
analysis. The main difficulty is that repeated
the basis for proper analyses of all data relevant
observations of a measure over time are often
to future missions.
sequentially dependent. This dependence, indi-
cated by serial correlation, complicates the
application of statistical methods that assume REFERENCES
independence of observations.
1. BERRY, C. A.: Aeromedical Preparations. In Mercury
Increase in the time interval between observa- Proiect Summary Including Results of the Fourth
tions usually reduces the amount of sequential Manned Orbital Flight. NASA SP-45, 1963, pp.
dependency. Nevertheless tests are required for 199-209.
determination of whether the observations (or 2. Cxz'rEasoN, A. D.; McC_rcnEoN, E. P.; _IINNERS,
H. A.; ANDPOLLARD,R. A. : Aeromedical Observa-
measurements) selected for analysis are reason-
tions. In Mercury Project Summary Including
ably independent before statistical methods that Results of the Fourth Manned Orbital Flight.
assume independence can be applied. The princi- NASA SP-45, 1963, pp. 299-326.
3. CHAMBraYS, R. M.: Long-Term Acceleration and 9. LINDSEY, J. F., Ja.; ET Ab.: Evaluation of Iluman
Centrifuge Simulation Studies. U.S. Naval Accelera- Factors Aspects of the B 58 Aircraft. Air Proving
tion Laboratory, JohnsviUe, Pa., 1963. Ground Center TN 59-73, Eglin Air Force Base,
4. FRASERj W. hi.: Philosophy of Simulation in Man- Fla., Dec. 1959.
Machine Space Mission System. Lovelace Founda- 10. DIXON, W. J.; AND _[ASSEY, F. J.: Introduction to
tion for Medical Education and Research, Al- Statistical Analysis. McGraw-Itill Book Co., Inc.,
1957.
buquerque, N. Mex., 1964.
l l. EDWARDS, A. L.: Experimental Designs in Psycho-
5. LovELACE, W. R., II; SCH_ICHTENI_ERG, A. H.;
logical Research. IIolt, Rinehart and V¢inston,
LVFT, U. C.; AND SECREST, R. R.: Selection and
Inc., 1960.
Maintenance Program for Astronauts for the NASA.
12. WEINBERG, G. H.; AND SCIIUMAKER, J. A.: Statistias,
Paper presented at Second International Symposh_m
An Intuitive Approach. Wadsworth Publishing
on Submarine and Space Medicine, Stockholm,
Company, Belmont, Calif., 1962.
Sweden, Aug. 18, 1961 (Also available in Aerospace
13. STARKWEATHER, J. A.: Variations in Vocal Behavior.
Med., vol. 33, no. 6, June 1962).
U.S. Public ttealth Service, 1963.
6. LINK, hi. M.: Space Medicine in Project Mercury. 14. TOWNSE_,'D, J. C.; AND LINDSEY, Z. F., JR.: Determina-
NASA SP-4003, 1965. tion and Evaluation of Rate Measurements in the
7. W1LSON, C. L., ED.: Projeet Mercury Candidate Analysis of Space Medical Data. Multivariate
Evaluation Program. WADC TR 59-505, Wright Behav. Res., vol. 2, Jan. 1967, pp. 64-70.
Air Development Center, Ohio, 1959. 15. SIEGEL, S.: Non-Parametric Statistics. McGraw-Hill
8. _IINNERS, H. A.; DOUGLAS, W. K.; KNOBLOCK, E. C.; Book Co.. Inc. 1956.
GRA.YBIEL, A.; AND IIAWXlNS, W. R.: Aeromedical 16. JONES, R. L.; AND LINDSEY, J. F., Ja.: Ituman Factors
Preparation and Results of Post-Flight Medical Aspects of Low-Altitude Flight. PGN Document
Examinations. In Mercury Results of the First 64-I, Egtin Air Force Base, Fla., 1964.
U.S. Manned Orbital Space Flight. NASA, 1962, 17. FaucrrTEIb B.: Introduction to Factor Analysis.
pp. 83-92. D. Van Nostrand, Princeton, N.J., 1954.
CHAPTER 4
AUTOMATED MEDICAL-MONITORING AIDS

FOR SUPPORT OF OPERATIONAL FLIGHT
Robert L. Jones and Edward C. Moseley
The Apollo program implements a significant medical Data Management group is in the Bio-
development in medical support of NASA's medical Research Office, and its flmctions include
operational missions--the use of real-time, auto- (1) improvement in acquisition, reduction, stor-
mated, biomedical-data-processing, monitoring age, retrieval, and analysis of in-flight medical
techniques in flight support. Concepts of medical data; (2) establishment of a coordination/man-
monitoring have evolved through the Mercury agement focal point for biomedical data; and (3)
and Gemini programs, and the concomitant evolution and establishment of criteria and re-
sophistication of hardware, software, and pro- quirement specifications for future biomedical
ceduraI systems, coupled with multigoaled mis- data for subsequent and advanced manned
sions of increasing complexity, has resulted in missions, as well as for in-flight biomedical
evolution of a concept of biomedical mission experiments.
support requiring reM-time, automated data Efforts involved in meeting these functional
processing. Missions have lost their flight-test requirements may be better understood if they
emphasis and are becoming test-beds for a are classified as (1) for mission support of mamlcd
scientific description, analysis, and prognosis for space flight; (2) for medical support of tests of
man in space. The research mission of yesterday manned systems; (3) for ground-based bio-
is becoming the operational mission of tomorrow. environmental-research programs; or (4) for
Biomedical data have been provided to demon- support and development studies of techniques
strate man's ability to survive and perform and methodologies.
intravehicular and extravehicular tasks involving Since space does not permit discussion of all
cognitive complexity and spontaneous flexibility four areas, our major concern will be description
under conditions of severe psychophysiological of the major aspects of automated medical-data
and motor stress. Now our biomedical data must processing currently planned for support, of
allow for detailed description and analysis of man Apollo. This was a significant event in that it
as he operates in the space environment, provid- marked the beginning of automated preproeess-
ing empirical data for medical decisions having ing, reduction, storage, computation, retrieval,
impact on real-time flight-planning. display, and analysis of bioenvironmental data
With progress in the early efforts to develop for real-time flight-support.
studies for the improvement of management of Anyone conversant with the state of the art of
biomedical data by use of computerized statistical medical data may well wonder why implementa-
techniques, and as the needs of medical support tion of such techniques should be singled out as
and research programs grew, the Biomedical a significant event, in view of the many clinical
Data Management responsibility was transferred and research medical institutions having well-
to the Medical Research and Operations Di- organized, on-line, real-time, automated medical
rectorate at Manned Spacecraft Center (MSC) monitoring as part of their daily routines. The
to meet the requirenaent for an integrated program answer lies in the basic difference between two
leading to improved operational systems for orientations and recalls the old specter of ba_sic
medical-data reduction and analysis. The Bio- versus applied. While this issue has always been
49
5O BIOMEDICAL RESEARCH AND COMPUTER APPLICATION
a good academic argument, it becomes very real to do so for us. This is especially true in a long-
and binding in the operational flight-testing lead-time, time-critical program.
environment when one must justify in great Another difficulty in resolution of biomedical-
detail: data problems is achievement of optimum dialogue
between the medical community and the elec-
(1) Astronomic funding requirements to budget trical-engineering and computer communities.
personnel Many engineers have extreme difficulty in re-
(2) Complex bioinstrumentation requirements solving the dissonance generated by first glance
to the engineers at biomedical problems. Floating base lines, re-
(3) Tremendous communications requirements verse polarity, ambiguous signal-reference points,
to the already overloaded network engineer noise-filtering problems, preprocessing problems,
(4) Weight/electrical/stowage/procedural inter- and software problems are extensive; much time
face requirements to the spacecraft-hardware man and effort are required if the engineer is to be
(5) Software requirements to the real-time effective in this particular area.
programmers On the other side of the coin, the biomedical
(6) Long lead-time delay factors to the program researcher often knows very little about, the soft-
m an ager ware/hardware/procedural/lead-time interfaces
All this and much more is compounded by the involved in his effort, and far too often he re-
familiar attitude of the pilot, who dislikes monitor- fuses to make adequate efforts to correct this
ing of his performance, sees no real need for it, lack of knowledge. This problem can be extremely
and is not really convinced that you are doing critical, and the Biomedical Data Management
anything worthwhile. The simple fact of the group tries to assist by providing communications
matter is that statements about the "search for between the two communities and by becoming
truth and knowledge" have difficulty in standing familiar with spacecraft hardware and procedures.
the test as justification for the operational situa- Among the specific considerations emphasized in
tion. The justification must be very definite and this interface are the following:
very real. (1) Long lead-time (12 to 18 mon) for changes
Of the many problems involved in such an due to budget-cycle impact and hardware, soft-
effort this may be one of the most subtle yet ware, and procedural changes necessary at MSC,
outstanding. The professional bioenvironmental Goddard Space Flight Center (GSFC), and the
researcher is trained to avoid the arbitrary, dis- remote sites.
believe the absolute, and qualify his statements, (2) Requirement for early (by 12 to 18 mon)
findings, and recommendations; he deals with specification of the following data factors: (a)
measures that have floating base lines, are highly variables being measured (number and kind);
individual, for which the standard error of meas- (b) sampling rate per variable; (c) number of
urement is often great, and for which tile standard channels required; (d) real-time and/or near-
deviation often exceeds the mean. But when he real-time needs (display, printouts, plots, proc-
participates in the operational situation he is essing, computations, etc.); (e) sampling pro-
competing with disciplines of opposite orienta- cedures (how many subjects?; how often per
tion, and these opposite viewpoints are most man?; how long per man?); (f) required real-time
often in crucial management and funding orga- identification codes, event markers, time codes,
nizations. etc.; (g) real-time transmission to ground of
Consequently we must learn to make a philo- recordings aboard spacecraft, or requirement for
sophical, conceptual, and verbal transition from high-speed-dump tape recorders; and (h) priority
the research orientation to the operational situa- of medical data relative to other data (because
tion if we are to participate and to carry out our the specified pulse-coded modulation (PCM) bit
ultimate responsibility. We must not be hesitant stream can be used only in near-Earth orbit;
in stating our criteria and requirements. His- circuit margins preclude use at lunar distances,
torically when we have delayed in doing so, s(_ that data capabilities are severely limited).
someone (how well qualified?) has been quick Obviously it is often quite difficult to specify
AUTOMATED MEDICAL-MONITORING AIDS 51
such factors in great detail as early as 18 mon To gauge some of the progress we may focus
before the event. However, it is well to remember momentarily on one physiological signal--thc
that, if the lead-time is not met, an investigator electrocardiographic. In general medicine we find
may very well find himself designing hypotheses thousands of records being collected, digitized, and
to fit the limitations of the system, and this pro- summarized by cardiologists studying lead sys-
cedure has little to offer an orderly search for tems, noise-reduction, data-compression, work
knowledge. capacity, pattern-analysis, disease profiles, etc.
In most of these applications the examinee is
usually resting, engaged in a standard activity,
GENERAL MEDICAL-MONITORING SYSTEMS
or under some other controlled conditions, and
While use of digital computers has been wide- use of computers in research is extensive. In
spread in research, library, administrative, and general medicine and in operational use we find
storage activities, only recently have they been examination by single electrocardiogram (ECG)
used successfully in continuous monitoring of in the physician's office, continuous monitoring
physiological data. Various historical, economic, in a coronary-care unit, or magnetic-tape record-
clinical, and methodological reasons can be ad- ings obtained while individuals are engaged in
vanced to explain this delay in development, normal activities. In these applications the
most of which are related to the basic nature of examinee is generally under Icss-controlled con-
the analog signal (electrocardiographic, electro- ditions; the data remain in analog form for
encephalographic, galvanic skin response, etc.) individual clinical examination; and the capital
generated by some critical physiological systems. investment in hardware is quite modest.
Historically the analog computer, with its In aviation medicine and for research purposes
direct measurements and limited memory, pre- we find intense activity in electrocardiology. For
ceded the digital computer with its indirect example, in 1957 a central USAF ECG Repository
measurement by counting of numbers expressed was set up, and by 1959 about 67000 normal
as digits and its extensive memory. Economically 12-lead tracings from healthy active officers were
the all-digital method is slow and expensive be- interpreted and related to background variables,
cause the input data are not usable in their thus providing base-line data (ref. 3). Much of
original form. the current activity is in identifying and quanti-
Clinically since Einthoven introduced the string fying evidence, of central-nervous-system (CNS)
galvanometer in 1903 (ref. 1), most medical ex- arousal, from ECG's while the subjects are sub-
perience has been in visual interpretation of the jected to operational stresses in a simulator.
rhythm, patterns, rate amplitude, and duration In short, highly sophisticated monitoring sys-
from direct-analog strip charts. Such a scheme is tems have been developed for operational ground-
basic to the current Gemini system. We might based use, and others have been developed for
have quite a different set of clinical impressions study of the general problems of continuous
if relative change had traditionally been viewed as bioenvironmental monitoring in hostile environ-
opposed to the absolute measurement (ref. 2). ments. The latter activities, however, are in the
Methodologically repeated measurements from developmental rather than the operational stage
the same subject violate the assumption of inde- needed for space activity.
pendence necessary for use of some powerful
statistical models developed for other types of
MSC'S MEDICAL-MONITORING AND
applications. The number of leads, the typically
RESEARCH SYSTEM
high frequency of sampling, the lack of criteria,
and irrelevant "noise" in the data all pose addi- NASA has a highly dynamic program respond-
tional methodological and operational problems. ing to changing national goals, radical technologi-
In view of the possibilities of combinations of cal advances, and changing requirements resulting
such problems, it is surprising that any real from exploratory projects. It is within such a
progress has been made in physiological monitor- moving environment that space medicine operates
ing--yet it has. at MSC.
During the Mercury program the primary Planners of the Apollo Applications Program
emphasis was on the safety of a singly manned envisioned the capability of provision for sustained
ballistic and short-orbital flight. Launch and orbital and other extraterrestrial living and work-
reentry problems were of real engineering and ing for a fight crew of several men. The Moon is
medical concern, while monitoring was primarily merely 240000 miles away while Mars is a
by voice. As the Gemini program got underway, minimum of 37 million miles, so it seems safe to
the medical system was updated to handle the ignore nonlunar problems in this context. Even
two-man orbital missions of up to 14 days, as without the problems of distant planets the
well as the extravehicular activity and some changing informational needs will be considerable
medical experiments. for tile medical and engineering communities.
The Apollo and early Apollo Applications Before considering an automated medical-data
programs increased dramatically the need for scheme we should describe the bioinstrumentation
medical information, bringing substantial in- used in acquisition of physiological signals. How-
creases in the number of crewmen, flight com- ever, since many of NASA's publications have
plexity, mission duration, and numbers and types described in detail the Mercury, Gemini, and
of preflight tests of manned systems. All these Apollo bioinstrumentation systems, our discussion
factors had significant impact on requirements will be brief.
of medical information. Now the system had to Limitation in availability of telemetry channels
provide for the operational flight needs of three limits monitoring to physiological variables that
men in orbital and lunar flights, some as long as are considered necessary for determination of the
30 days, and some involving free-space and well-being of the flight crew. In Mercury and
lunar-surface extravehicular activity. Addition- Gemini these real-time measures were presented
ally more ground-based, preflight profile data in analog, with no analog-digital conversion or
were available for use during in-flight comparisons preprocessing. In order to provide physiological
and evaluations, such as data gathered from the data of a more comprehensive nature for post-
world's largest space-simulation chamber at flight analyses in depth, an in-flight tape recorder
MSC; furthermore profile data gathered during was provided for recording the measurements in
Mercury and Gemini flighfs were available for analog form.
in-flight analysis. Obviously the multiplicity of Table 1 lists the types of biomedical monitoring
these critical interacting requirements made the from spacecraft of the three projects mentioned.
systems more sophisticated. Figure 1 shows basic differences between the
TABLE 1.--Types of Biomedical Monitoring from Spacecraft of the Three Projects
Factor Mercury Gemini Apollo

monitored (1-man creW,') (2-man crew) (_2-man crew)
ECG Yes A&S,_ 320 sps b A&S? 200 sps b
Respiration Yes Impedance method, 40 sps; Impedance-pneumograph,

axillary ECG electrode used 40 sps
as sensor
Blood pressure Yes Manual, squeeze-bulb, brachial- Mechanical, squeeze-bulb,

occlusive system ad libitum
Body temp Yes 1.2 sps; oral thermistor probe; Oral, mechanical, ad
intermittent libitum
PKG * None Routed with EEG to tape re- 200 sps

corder (experiments data)
* Axial and sternal. Phonocardiographic.

b Samples per second.
AUTOMATED MEDICAL-MONITORING AIDS
MERCURY GEMINI
l TELEMETRY
ON BOARD ON BOARD
TELEMETRY I
' DUAL
!_PRESSURE PURPOSE
BOTTLE INFLATOR
I. CUFF INFLATION
2. WATER SUPPLY
] --4_CONTROL PRESSURIZATION
AND EVACUATION
CONTROL I J SOLENOID
SOLENOID ! • =.... .
i
POP-OFF I
SUIT -
VALVE _ED-E
SUIT
OUTSIDE
I
l I oOUTSIDE
• _ _SUIT
IOCCLUSIVE CUFF
OCCLUSIVE CUFF AND MICROPHONE
AND MICROPHONE
KEY : "_
ELECTRICAL CONNECTION
__ PNEUMATIC CONNECTION
FIGURE 1.--Blood-pressure-measuring systems for Mercury and Gemini.
Mercury and Gemini blood-pressure systems. of environmental variables also are available, such
Figures 2 and 3 show the Gemini bioinstrumenta- as cabin pressure and temperature and suit pres-
tion system, and figure 4 shows the Apollo system. sure and temperature. The sampling rates for
Voices monitored throughout all flights have such environmental data generally range from
furnished valuable information both during and about 0.5 to 1.2 samples per second, and the data
after flights. In the Apollo program a major addi- are transmitted with PCM (pulse-code modu-
tion is the TV transmission that enables monitors lation).
to view the crew at selected times. A wide variety The physiological information from Gemini
t |
FIGURE 2.--Bioinstrumentation system for Gemini.

J
J
FmURE 3.--Fitted bioinstrumentation system for Gemini.

FIGURE 4.--Bioinstrumentation system for Apollo.
flowed from each astronaut to the tracking sta- recovered after splashdown, speeded-up to ground-
tion where-it was evaluated by the station surgeon elapsed time, digitized, and used as input for
and simultaneously transmitted to GSFC. Thence computer analysis of the experiment. No in-flight
it was transmitted instantaneously to MSC in medical experiments are planned for Apollo.
Houston for use by mission control. Medical data When received at MSC the real-time telemetry
from in-flight experiments were recorded during information is recorded on magnetic tape for
Gemini flights on spacecraft-borne tape capable postflight analysis, and the physiological param-
of recording up to 100 hours; this analog tape was eters are displayed in analog form on strip-chart
recorders. Environmental parameters are moni- upper and lower limits, with a red background
tored by both a computer and humans. The illuminated when the rate goes beyond limits. A
medical monitor can also call for a digital display 30-see average respiration rate and a 30-see
of these parameters on his TV screen. representative ECG wave form are available
Since the primary purpose of this information from the pneumotachometer and ECG-wave-form
is to assist the flight surgeon in clinical assessment preprocessor, respectively.
of the crew, little real-time analysis of this With this equipment the strip charts are needed
physiological information is completed. Post-pass only (luring critical phases of the flight and are
physiological summaries are calculated from the automatically started when operationally defined
analog strip charts, as are hand plots. Immedi- limits are exceeded. In addition the hardware
ately after the flight, selected time segments of provides real-time analog-to-digital conversion
interest are converted from analog to digital for additional real-time descriptive and analytic
information for detailed analytic and reporting manipulations. An interim system, utilizing some
purposes. of this equipment,, was used in simulations of the
In general the "clinical" strategy has worked first Apollo mission. These simulations demon-
well in attempts to deal with the new hostile strated the adequacy of the revised real-time
environment for short-duration flights and few displays for medical monitoring and determined
astronauts. As any program expands in terms of some of the functional procedural and sampling
flight duration and/or number of crewmen, the requirements for automatic summaries of physio-
clinical approach reaches its effective limits and logical in-flight information.
more-automated schemes or aids become neces- During the first simulated Apollo flight the
sary. monitor had a variety of mid-pass and post:pass
We can imagine that after looking at miles of computer summaries available to him in addition
strip charts our deepest impression would recall to the instantaneous information (mean, median,
Lincoln's comment that the thing that struck him range, and variability) on heart rate, respiration
most forcibly when he first saw Niagara Falls was rate, ECG-wave-form components, and selected
"Where in the world did all the water come from !" environmental parameters, such as suit tempera-
Hardware, software, anti procedural changes are ture and cabin temperature, for the various crew-
under way in anticipation of the deluge of in- men in real time. Time plots of these parameters
formation from Apollo and its successors. Fortu- were also available; up to 12 hours of these
nately not all the expected increase will occur at summaries couhl be reviewed at any given
once, so an effective operational system can be moment.
sequentially implemented. Of considerable interest are the automatic
daily flight summaries of physiological and en-
vironmental parameters that were made available
EARLY APOLLO MONITORING SYSTEM
beginning with the interim system. As far as is
Within the general background and frame of practicable the instantaneous data are auto-
reference thus established, we can now discuss in matically tagged with independent variables of
some detail the improved biomedical-data system interest. Most tagging can be done automatically
approved for the Apollo program. Its general by tile computer (e.g., astronaut, lead, suit on or
configuration is shown in figure 5. off, day), while others (e.g., sleep, exercise, and
Several cardiotachometer and pneumotachom- stress periods) still require operational definitions
eter preprocessors, an ECG-wave-form preproc- for human push-button action.
essor, a 30-sec FM tape-loop recorder, and a As the data are buffered and processed for
number of event push buttons are currently post-pass summaries the latter are routed to
being installed for biomedical monitoring at the storage buffers containing identical tags. Thus,
mission-control center. The cardiotachometer will at the end of the day, descriptive tables and/or
provide visible digital readouts of both the in- plots can be automatically prepared for any de-
stantaneous and selectable (6-, 10-, or 30-see) pendent variable by any combination of inde-
average heart rates; it also provides selectable pendent variables. For example, heart-rate or
Aero_ed Console
Sw_tchln9 I
ControJ to Hatrlx I Control to
EGG Waveform
AnaJyzer B
Delayed ECG _B
Strlp Chart
Recorder #B
Respiration ECG #IB

Uni,ac
Heart
__T
Instantaneous Heart Rate
ECG #IA .-L-L.
FM Ground EGG #3R
EGG #IB
prom Statlon #A EGG #h.A
Remot • [ Resp. A 30-Second Tape Loop Delay

Sites ,
EGG #18
FH Ground EGG # 3B
Station #B EGG #48
I R_..
FIGITRE 5.--Configuration of biomedical-data system for Apollo.
average-wave-form plots for normal activity, displayed in a format suitable for immediate
stressful exercise, and sleep across days to date reporting on the mission. They will also provide
will be available for each crewman. Furthermore, feedback for planning of the next mission, as
automated, annotated, Ground Elapsed Time well as forming part of the basis for a cumulative
plots can be generated daily from the post-pass description of Apollo flights. The necessary and
summaries. sufficient information for application of many
One may ask why this sort of daily summary is inferential statistical techniques will be available
needed in real time when a more careful analysis in summary form and used as applicable.
could be completed after the flight. The answer,
IMPROVED PHYSIOLOGICAL MONITORING
of course, is that daily trends have important
SYSTEM FOR APOLLO
implications regarding monitoring and counter-
measures when the flight lasts roughly 1 mon or It is difficult if not impossible to describe at
longer. In general the design philosophy is to do this time the configuration of the future monitor-
as much analog-digital conversion as possible and ing system since plans are amended daily. The
to strive for successive levels of real-time data- higher frequency of Apollo flights, as well as their
reduction. longer, overlapping, orbital, and nonorbital na-
Finally the early Apollo monitoring system will ture, require that the following general goals be
provide automatic postflight summaries; essen- established: (I) increased ability to plan for and
tially they will be the daily flight summaries detect medical crises, and (2) reduction in elapsed
AUTOMATED
M_DICAL-MONITORING
AIDS 59
timebetweenacquisition ofdataandformulation Anothergeneralobjective,relevantto boththe

of interpretativereports--withindicationsof planning andthedetection goal,is increase in our
actionsrequired. real-time andnear-real-time responsiveness to all
For achievement of thesegoals,generalob- "historical"information:datafromtrainingsimu-
jectivesarecontinuallybeingredefined in terms lators,altitudechambers, laboratoryandclinic,
of the user'sneeds.After investigationof the previousflights,andnormativebase-lineresults.
physicaland intellectualtools required,the Someof this informationis examined beforethe
practicalityof eachobjectiveis determinedin flight andlater compared with postflightresults
termsof time,usefulness, andmoney.Oncethe for writingof missionreports.
objectivesare determinedpreciselyand made Onthe otherhand,somecanbefruitfully used
consistentwith the realitiesof time,relevance, for the makingof background-comparison slides
and equipment,the scheduling,specifications, for real-time display or for some phases of "auto-
building,andtestingareimplemented. With this matte monitoring"; therefore, they must be
sort of approachwe can establishattainable prepared as preflight summaries. Implementation
objectivessignificantto the program.Table 2 at MSC of contractual efforts designed to meet
showssomeidealizedhypotheticalrequirements these historical needs is currently being in-
that maybeusedascriteriafor evaluationof a vestigated.
system'seffectiveness. Still another objective calls for increased use
Oneobjective,for example, callsfor increased and flexibility of real-, near-real-, and non-real-
real-timeandnear-real-time displaysofsummary time analyses. Most of the statistical analyses
informationhavingmedicalrelevance. In addition implied earlier are univariate descriptive statistics
to the automaticdisplaysdiscussed earlier,a (e.g., mean, median, range, and variability) or
variety of othersare now beingactivelycon- univariate inferential statistics (analysis of vari-
sidered. ance, covariance, etc.). This latter class of
TABLE 2.--A Hypothetical Systems Checklist
I. System requirements--Does the system have
1. Flexible input-output capability 14. Responsiveness to multiple users

..__2. Multiple and flexible display properties 15. Adequate storage capabilities (tapes, core, drum,
m3. Analog-to-digital conversion and/or disk)
4. Digital-to-analog conversion 16. Easily implemented changes
5. Effective "interrupt" capabilities 17. Adequate documentation
6. EËficient search and "locate" 18. "State of hardware" best suited to needs
7. Versatile, selective extraction properties 19. Adequate auxiliary equipment
8. Easy proofing, updating, and correction 20. Priorities that permit reasonable turnaround time
___9. Complete information about information 21. Varied, effÉcient, and useful analytical programming
10. Varied data-compression capability capability at the descriptive, inferential,
11. Automatic and readable report predictive, and interpretive levels
12. Efficient monitoring and programs 22. Adequate expansion properties
13. Complete error messages 23. Success criterion
II. Organizational requirements--Does the systems organization provide
1. Well-defined and limited goals 4. Personnel with time to improve system

2. Coordination among information sources 5. Interdisciplinary services
3. Personnel with time to service users or to write 6. Long-range administrative and monetary support
programs
III. Operational requirements--Is the system
1. Easy to learn for one-shot, intermittent, and 3. Easy to check

repetitive applications 4. Informative by providing timely information in
2. Easy to use in one-shot, intermittent, and repetitive usable form
applications 5. Interesting to the user
statistics enabIes us to determine, for example, be generated in real time by digital computers
whether the difference in heart rate during 3 and then applied to daily and/or postflight sum-
days of sleep is simply by chance. Analysis of maries. Still other manipulation of raw data will
covarianee could be used to answer the question require analog preprocessing.
of whether there is non-chance change in heart More general objectives and approaches to the
rate when the infuence of changes in some future physiological monitoring system are being"
environmental effect (e.g., temperature) is re- considered in an effort to stabilize and reduce the
moved. Other univariate statistical models enable workload without compromising the well-being
one to describe physiological periods, determine of the astronauts. The general goal is for a better
significant trends, eliminate noise, etc. Another descriptive, inferential, and predictive system--
class of statistics, known as multivariate, answers in about that order.
the same type of question but considers all Thus the MSC biomedical-data system is in-
variables simultaneously; for example, one can deed dynamic--by the time this paper is pub-
ask whether there is a non-chance difference be- lished, changes and improvements will have been
tween suit-on and suit-off conditions when all made. However, we have described the rationale,
variables (i.e., rates of heart and respiration, and enhancement will come through optimization
wave-form components, temperature, CO2, 02, of this system within the framework of subsequent
and humidity) are considered simultaneously. mission requirements.
Other powerful multivariate models are available.
Predictive statistics is still another class of sta- REFERENCES
tistics that can be useful when applied to some 1. BURet, G.; A_l) W_NSOn, T.: A Primer of Electro-
of our problems. cardiography. Lea and Febiger, Philadelphia, 1945.
One should note that only the simplest descrip- 2. PROCTOR,L. D.; AND AI)EY, W. R.: Symposium on the
tive statistics can be applied currently in real-time Analysis of CNS and Cardiovascular Data Using
Computer l_Iethods. NASA, 1965, p. 235.
because of the workload in the Real-Time Com-
3. LAMB, L. E.: First International Symposium on
puter Complex. On the other hand, some of the Cardiology in Aviation. School of Aviation Medicine,
basic summary information for these methods can USAF, Brooks Air Force Base, Texas, 1957.
CHAPTER 5
MEDATA: A MEDICAL-INFORMATION-
MANAGEMENT SYSTEM
Tate M. Minckler and Caroline L. Horton
Significant interaction between the practice of and concurrently to provide computer-supported

medicine and the computer sciences is inevitable. records-management capability that neither dic-
This observation is predicated on the demon- tates nor restricts the medical information handled
strated advantages that accrue from proper use by it.
of the computer as a tool for processing of volu- An effort has been under way since July 1965
minous or complex data. The interdigitation of the (NSR44-012-039) to provide such management
medical and computer sciences already has re- capability for use with astronauts' medical in-
sulted in emergence of a new discipline, here formation. It is with the progress toward that
dubbed medical information management, that goal that we are now concerned.
appears destined to touch every facet of medical
practice. Medical information management may
MEDICAL INFORMATION
be defined as a supportive activity responsible
for providing coordination of informational needs Medical information in this context is the body
and resources among the patient-care, teaching, of information that is generated during patient-
research, and administrative services of the med- care activities. Its primary purpose is documenta-
ical environment. tion of the factors applied to disposition of patient
Medical information management has three care. However, these records are called upon to
functional components--record keeping, data serve research, educational, and administrative
analysis, and communication--each of which is needs as well. The most compelling reason for
at least partially amenable to computer support. development of a structured record system is not
The key to eventual integration of these now simply conversion to automation; rather it is
largely independent components rests upon the provision of an accurate, detailed, and consistent
development of medically effective, organiza- bank of data. The file structure of MEDATA is
tionally sound, computer-based, medical record designed without regard for the methods or equip-
systems. For without ready access to complete ment used in records management.
data, even the most sophisticated analysis and At least three features are essential to any
communication facilities are severely hobbled. systematic file structure: (1) provision for unique
Broad experience has demonstrated, however, identification of each individual record in the file,
that medical record systems do not "compute" (2) definition of the content of the file, and (3)
easily. Chief among many reasons for this have uniform organization of the component records.
been two: the medical information problem of We now describe these features for MEDATA
defining specifically the content and organization applications.
of the records, and the information management
Identi(;cation
problem of providing computerized support that is
flexible enough to survive the constant changes The usual orientation of medical data provides
in content and organization. The tasks then are a straightforward approach to record identifica-
to design concepts of medical documentation tion. Medical records are generally accumulated
leading toward increased utility of the stored data, in terms of three parameters: the who, the what,
61
and the when. The who may be a patient's name patient-care environment. Each functional divi-
or any of a group of unique numbers (such as sion (survey, workup, status) has a natural group
hospital number or social security number). The of subdivisions based on the type or source of
when is a date that may or may not be supple- data; for example, the several types of survey
mented by time of day. The what defines the con- documentation include electrocardiographic, lab-
tent in terms of the function and the type or source orator)', X-ray, and review of systems. Table 1
of the data. The content of the file is the axis on shows the basic pattern of record identification,
which any records system turns. listing the specific types of survey reports cur-
rently in use; each, representing a unique step in
Content
the survey process, may be generated by a dif-
In determining a pattern for medical record file ferent person at a separate time or in a special
structure, data function has been the paramount location, and usually reports on a different sub-
consideration. When the content of patient-care ject. These same criteria are applied to the desig-
information is viewed from the standpoint of nation of types of reports throughout the ME-
functional documentation, three rather distinct DATA file system.
classes of medical reports may be discerned: the In summary, unique identification is provided
survey, the specific problem workup, and the by requirement of a standardized set of state-
status report. The first two of these are commonly ments regarding the "who," "what," and "when"
intermixed in current practice, but can and should of each report in the file. In the 5IEDATA ap-
be separated for a variety of reasons. plication the patient's identity is established by
Whether the survey or screening examination is the social security number (SS NO); content is
applied to a population or an individual patient, defined by both the function of the RECORD
its purpose is the same; it identifies from the whole and the TYPE of record, and DATE indicates the
the parts having a high probability of disease and time frame.
thereby establishes a set of problems requiring
Organization
further medical investigation. Conversely but
equally significant, the survey report should also The 5IEDATA system divides each medical
serve to document the parts that are free of recog- report into two segments: the "leaders" or identi-
nizable disease. A complete or partial survey may fication portion (already described) and the
be conducted at any point in the patient-care "body." An orderly approach is followed in es-
cycle, either as an independent periodic examina- tablishment of the body of a particular report.
tion or to complement a specific workup.
The workup is usually initiated either because
the patient seeks help for a problem or because a TABLE 1.--MEDA TA Identification Pattern
survey has elicited one or more potential prob- for Survey Records
lems; it is an attempt to establish a diagnosis and Category Heading Date (examples)
to institute appropriate therapy for each problem.
The principal functional results of a workup are Who SS No. 123-45-6789
two: a plan of action that may be either diagnostic What (A) Record Survey
or therapeutic, and initial implementation of that
plan (a set of orders). What (B) Type Identification
Review of systems
The effectiveness of the plan is evaluated and (ROS)
documented periodically. The doctor's "progress Dental
note" is a prime example of such a status report, Laboratory
but also included are other reports that serve the X-ray
ECG
same basic purpose, such as nurses' notes and
Measurement
reports of procedures. Vision
This philosophic approach provides a framework Summary
for understanding and categorizing of the basic
When Date 01 Jan 68
functions of each informational component in the
A MEDICAL-INFORMATION MANAGI_MENT SYSTEM 63
First the informational content must be defined. Judgments are items for which the examiner is
Figures 1 and 2 are reproductions of the Standard expected to evaluate one or more criteria and
Report of Medical Examination (Form-88), one summarize his opinion. With the MEDATA
of the survey documents in use at Manned Space- system this expression is formatted by entering
craft Center. Its content was adopted as the of POS (positive) or NEG (negative) followed by
starting point for development of the MEDATA an explanation or amplification in prose as ap-
system. propriate. The entry POS means that in the
Terms must then be chosen as headings to opinion of the examiner the item under considera-
represent the contents. Unidentified data ob- tion is unusual, abnormal, or otherwise note-
viously are meaningless. Although many data worthy; NEG indicates normal, within acceptable
when presented in context need no overt headings, limits, or not significant. The entry POS is almost
the heading concept is implied by the context. always followed by a description of the abnormal
For this reason, most items of data in any file finding or some comment to explain why that
are accumulated as direct answers to the questions item is notable; NEG may be amplified by prose
posed by headings. This sequential association necessary. Table 2 shows a portion of the record
of heading followed by data is called a head- of a physical examination. This policy permits the
ing/data pair. rapid scanning of reports for information judged
The headings must be organized into an outline to be medically significant by the examiner.
so that relations are clear among the various The judgment format may also be combined
blocks of data. As an example of this procedure, with quantitative reporting: the judgment (POS
consider items 57 and 58 of Form 88 (fig. 2), meas- or NEG), plus commentary, is entered after the
urements of the cardiovascular system. Figure quantity and its type of unit--weight, 185 Ib;
3 is an outline arrangement of these basic items, POS (30-1b overweight for height).
with minor modifications in terminology but with Facts, the third kind of data, are not subject
all original content intact; notice the use of stand- to quantitation or judgment in the usual sense.
ard outline techniques to relate ideas. All data They are items of information, such as name,
about pulses are indented under that term, and address, history of measles, etc., often derived
other degrees of indentation clearly establish the from the patient or another informant. Facts are
relations among the remaining headings. This simply placed in the report as communicated; a
hierarchy (outline) technique provides a pattern judgment may be made as to the reliability of
by which man (and machine) can recognize the the informant, but the items themselves are not
organization of headings and therefore the as- subject to medical evaluation or quantitation.
sociation of data. The formatting of facts is not restrictive but must
The final consideration in establishment of a be uniform, item by item; for example, the ques-
systematic file structure is the policy regarding tion regarding name should always be answered
data "formatting." The format of an item of in the same pattern--last name, first name,
data refers to the detailed characterization of the middle initial.
way in which that item is to be reported. The Portions of the 5IEDATA formats for Form-88
MEDATA system recognizes three fundamental data will be used throughout the remainder of
and distinct kinds of information that must be this presentation as specific examples of the file
reported in a medical context: quantities, judg- structure currently implemented. Forms for more
ments, and facts. elaborate capture of information are being de-
Quantities are measurements generated in the veloped.
medical environment directly or indirectly from
INFORMATION MANAGEMENT
the patient; height, weight, blood pressure, and
laboratory values are obvious examples. Quanti- Information management, as applied to
ties are formatted by making the numeric value record system, means the provision of smooth
the first element in the data area or field; this is and integrated mechanisms to collect, process,
generally followed by the type of unit in which and retrieve records or parts of records. For these
the value is reported--height, 72 in.; or Hb, 15.5 g. purposes the advantages of the modern digital
Standard Form 88
(Rev, June 1956)
B..... rth_B.dg,, REPORT OF MEDICAL EXAMINATION ss,o6

Circular A-32 (Rev.)
1. LAST NAME--FIRST NAME--MIDDLE NAME Z. GRADE AND COMPONENT OR POSIt'ION
4. HOME ADDRESS (Nura_r, Jtrcd or RFD, cffll ¢r lo_n, zo'm¢ tad SMI¢) $. PURPOSE OF EXAMINATION DATE OF EXAMINATION
t: ID(NTIFICATIO#4 NO.
7. SIEX J |. RACF 9. TOTAL yEARS GOVERNMENT SERVICE
1 MlUTARY CiVlUAN
10. AGENCY [ I1. ORGANIZATION UNIT
14. NAME. RELATIONSHIP. AND ADDRESS OF NEXT OF KIN
t2_ OATE OF RIRTN I 13_ PLACE OF |fRTN
I
|$_ E×AMINIRG FACILITY OR EXAMINER_ AND ADDRESS 16. OTHER INFORMATION
17. RATING OR SPECIALTY TiME IN THIS CAPACITY (T0_O]) LAST SiX MONTHS
1 L
CLINICAL EVALUATION HOTLY. (Deacrib* everJ abnormality in detail, gnter _rflnent item number be/ore each
comment. Continue in item 73 and use addrtional iheets if necessary )
NOR. (Check each item in appropriate co/- ABNOR-
mAl u__rnn; enter "HE" ft not evRIuated ) MAL
|8. HEAD, FACI[, NECK AND SCALP

__ .u
19. NOSE
Z|. MOUTH AND THROAT
22. ARS GE ERAL _uitf Nndir i_r_ 70 asud 71)
Z3. DRUMS (per/0raf_on)
N ( V_I _ult_ ._ rqfiaeKoa

24. EYES_GE ERAL vnd_ _r_ M_. RO ¢_d ny)
25 OPHTHALMOSCOPIC
26. PUPILS (Equ4aZilp and r'tact|on)
M TI I (Ael_ia_4 parclleI _o_-

Z7, OCULAR O L TY _. n_Z_le_)
26 LUNGS AND CHEST (/n_'_U_¢ breasts)
_, HEART (TttruK, Jize, rk_l_l_m, POUnds)
30. VASCULAR SYSTEM (t'_ri¢o#it_p. etc.)
3|, AROOMEN AND VISCERA (_R¢]Ulle &¢r_)
(Hr_or, Ao_dl, 'hl._)

_, ANUS AND RECT_JM t/_.a_, I i[ind_al¢_
33 ENtX}CR_NE SYSTEM
34. G-U SYSTEM
3_. UPPER EXTREM(TIES (Slrtnolk, _nOr of
_6. FEET
37. LOWER EXTREMITIES_,_t A" e_a/_'_n I
31. SP NE. OTHER MUS._ULO_KELETAL
3_ IDENTIFYING ]_lOOY MARKS. SCARS, TA_TCOS
40. SKIN. LYMPHATICS
41. NEUROLOGtC (_vlZi_ium I_#J_ u_cr i|¢_ 7_)
43. PELVIC (Fem_Jep o_l_) (C_ck _OW ¢O_ul)
(_ VAGINAL _] RECTAL (Continue in item T3)
M. DENTAL (p_ee _ppropr_le pgn4bol# _0_ or bdo_ humor q_ _ppt_" a_d _oteer tibia, repp¢cti_cly.)
DEFECTS AND DISEASES
o--Re#tereM_ te_tli _-- _#l_ te_t4 (8,Vg)--t*L_d _id4e, br=t_eta to
I--_amrclforo_lt feeti XXX--P_oc( L._tNreP
R . L
I 1 2 3 4 S (; 7 8 9 tO El 12 13 14 15 16 E
G 32 31 30 29 2:8 Z7 _ Z5 ?.4 23 22 Zt _0 19 18 17 F REMARKS AND ADDITIONAL DENTAL
T T
IJilOIIATOIIY FINDiNgS
4S. URINALYSIS: A, SPECIFIC GRAVITY 4_5 CHEST X-RAY (Z_ce, _le, fdm _11_mb_r a_d re$1_rt)
B. ALBUMIN O MICROSCOPIC
C. SUGAR
47. S_ROtO6Y (_p_l/_ t_at w.ud u_d r*_) 'EKG 49. IBLO00 TYPE AND RH ._ OTHER TESTS
FACTOR
Fmv_ 1.--Report of medical examination CForra-88), front..

k MEDICAL-INFORMATION
MANkGEMENT
SYSTEM 65
MEASUREMENTS AND OTHER FINDINGS

Sl. HEIGHT
l S2. WEIGHT l S3. COLON HAIR_ I 5,1. COLOR EYES ..... ,55.
](Ch BUILD:
..... )t ,5LENDIDR l.i MEDIUM _ HEAVY
Lt _! OBESE iS1; .TEMPERATURE --
57.
B._o PRESSURE (Ar_m at Iml_ [¢_eD __1. PULSE (Arm at ke.rl teeeZ)
A.
SITTING
m. u,_ANTV,_ON /"_ _ --.N_RA_ON i' NEAR
RIGHT _/ co._
TO_I _-ISY s. o_ .-'" CORR --TO _Y
LEFT20/ CORN
TO_ [.Y s. ox _ _ _ RY
R. HETEROPHORIA (Spe¢ifF d/#tanc¢)
ES j EX ° R.H. L M. PRISM DIV. PRI5M CONV. PC PD
CT
63. ACCOMMODATION M. COLOR VISION ( T¢=t t_eed =n4 reatdt)
m
( Telt uJ_t and _ore)
RIGHT LEFT
____ IyRNECTEO
_. FIELD OF VISION 67. NIGHT VISION (Tul =#e4 and _ore)
I_. DEPTH F'ERCEFTION UNCORRECTED
RED LIENS TEST j_. IHTRAOCUL_R TENSION
[
70. HEARING 7II- AUDIOMETER _ PSYCHOLOGICAL AND PSYCHOMOTOR
(TcatJ uaed a_d =_ore)
lOOO 2ooo aCoo 4o00 _ e¢_oo
WV /15 SV /I$ RIGHT
I.EF'_
73. KOTES ((_tiflUCG_ AND SIGNIFICANT OR INTERVAL HISTORY
(Dec ¢ddliion_l iJiede if _c_#arF)
74. SUMMARY OF DEFECTS AND D_AGNO_S (LMf d_oee= teU.,_ _em lll_mbtrll)
75. RECOMMENDATION5--FURTHER SPECIALIST EXAMINATIONS INDICATED (Sp_-Ci[le) 71_. A PHYSICAL PflOFIIr
77. EXAMINEE (_r_ecL)
A, [] IS QUALIFIED FOR N. PHYSICAL CATEGORY
B, _] IS NOT QUALIT'IKO FOR
711. IF NOT QUALIFIED, LIST DISOUALIFYING DEFECT5 BY ITEM NUMBER
7_. TYPED OR PRINTED NAME OF PHYSICIAN SIGNATURE
IO, TYPED OR PRINTED NAME OF PHYSICIAN SIGNATURE
|1. TYPED OR PRINTED NAME OF DENTIST OR PHYSICIAN (_t_icaf¢ _kick) SIGNATUR I_
TYRED OR PRINTED NAME OF REVIEWING OFFICER OR APPROVING AUTHORITY SIGNATURE r NUMBER OF AT-
TACHED SME_rTS
Fmvm_ 2.--Report of medical examination (Form-88), back.

CARDIOVASCULAR ondary level of programming refers to instructions

BP
SITTING
SYS:
required to relate (1) data files to primary pro-
OVAS:
RECUHBEW; FIGVR_ 3.--Example of a MEDATA gram steps, and (2) the results of primary pro-
SYS:
OlAS_ outline derived from the cardio- grams to man. In the FORTRAN programming
STANDING
SYS: vascular-measurement section of language, for example, secondary-level program-
DIASt
PULSE Form-88 (fig. 2).
SITTING:
ming is represented by FORMAT STATE-
R_CU_BENT:
STANDING:
MENTS.
EXERCISE
IMBED AFTER: The operational difference between primary
2 _IN AFTER:
and secondary levels becomes clear when one
considers that primary programming deals with
computer and associated supporting equipment
concepts (standard deviation, sort, retrieve, etc.)
are obvious and need no further justification.
or generalized operations performed on many
This section describes a unique concept in com-
different data files. Secondary programming,
puter programming and then covers the opera-
quite separately, is concerned with the intimate
tional implementation of that concept for acquisi-
details of a specific file (the location and arrange-
tion, processing, retrieval, and maintenance of
ment of the numbers from which a standard
MEDATA information.
deviation is to be calculated, or the particular
The PrimaryPeripheral Programming Concept items of data to be sorted or retrieved).
Secondary programming, therefore, recognizes
The primary/peripheral concept represents a
the formatting and the sequential relations of a
significant departure from classical programming
specific data file; i.e., it defines content and
approaches. It offers great flexibility in file
organization. Since these definitions are already
structure to the user who is not familiar with
contained within the basic structure of each
computers, and yet provides a core for smooth
medical-information document already described,
integration within the acquisition, processing,
their preservation during the acquisition of
and retrieval components of the management
records in machine-readable form provides an
system. Understanding of primary/peripheral
automatic peripheral mechanism for accomplish-
programming requires background in the broad
ing most of the functions of secondary program-
responsibilities of a classical computer program in
ming.
relation to the job it is to perform.
Primary programming, then, is the concept of
The term computer program as routinely used
limiting the responsibility of a set of computer
can be loosely defined ,as a discrete set of instruc-
programs to performance of basic conceptual
tions which, when applied to the computer,
tasks. Peripheral programming is the inclusion,
controls in explicit detail the manipulations per-
within the body of each computer-stored record,
formed by the computer system on a specified
of the necessary secondary-level definitions of
file of data. The usual computer program func-
content and organization in context with the
tions at two levels. The primary level is the logical
specific data. Peripheral programming obviates
job for which the program was initially intended
the need for special secondary-level program
(e.g., to compute standard deviations or to re-
segments attached to each primary program;
trieve an existing item from the file). The sec-
instead the primary program requires only the
addition of a standard interface segment that can
TABLE 2.--Portion of the Record of a Physical interpret the peripherally programmed definitions
Examination contained in any record. The operational details
of this concept are now described under "facsimile
Item Judgment Explanation
storage."
Abdomen NEG Scaphoid; good tone Facsimile Storage

Anus and rectum POS Asymptomatic large Facsimile storage (FACS) is our name for the
hemorrhoids
construction of the computer record; it is based
GU NEG
on preparation by use of a standard typewriter
A MEDICAL-INFORMATION MANAGEMENT SYSTEM 67
keyboard of a human-readable document in data pair and can be any unique character code.
machine-readable format. Facsimile storage pre- The present terminating code is a special un-
serves the complete content of each individual printable character automatically generated by
record in the computer, including not only the system during data acquisition.
prosaic (English) headings and associated text In addition to the heading/data codes, editorial
of data but also all organizational relations. codes are generated at the time of document
The computer handles FACS in a simple preparation and stored in the record; they control
manner depending only on three position codes layout involving carriage return, tabulation, line
for each heading/data pair. The first, or hier- feed, etc. By a simple program these codes can be
archical code set, has two functions: it identifies translated into functional equivalents in the com-
the beginning of a heading/data pair, and it puter: for example, "carriage return" equals
specifies the hierarchical relation between its own "new print line," and "tabulate" equals "begin
heading and the preceding and following head- print in [specified] space." The term facsimile
ings. The set can be any series of typewriter storage therefore means just that--an exact
symbols; this presentation uses zero through 9 facsimile of an original document resides in the
(0, 1,..., 9). Each digit indicates progressively computer.
the degree of indentation from the left margin The mechanisms by which peripheral program-
in the defined outline organization of headings: ming is implemented during data acquisition are
zero means no indentation, 1 is the first level, discussed next.
2 is the second level, and so on. Table 3 shows
application of the hierarchical code to the outline Systematized TerminaI-Acquisitlon Technique
of headings given in figure 3. The systematized terminal-acquisition tech-
The second code is required to separate the nique (STAT) is a one-step procedure for tran-
variable-length heading field from the variable-
scribing information into machine-readable lan-
length data field of the heading/data pair. For
guage; it is user-oriented, being accomplished by
this purpose the typed colon (:) serves the dual the user in his own environment and under his
function of being understood by both man and
complete control. This technique replaces the
machine. The third code terminates each heading/
classical key-punch approach to data transcrip-
tion. Using a simple typing operation, it takes
advantage of the familiarity with specific medical
TABLE 3.--Application of the Hierarchical Code phraseology, spelling, and abbreviations that is
Set to Some Headings of Form-88 (fig. 3) available only among the users' staffs.
Code number Application to heading The current implementation of STAT is de-
signed to function in environments where multi-
0 CARDIOVASCULAR terminal time-sharing is not yet practical for
1 BP routine operations. However, the basic principles
2 SITTING
are equally applicable to on-line and off-line use.
3 SYS:
3 DIAS: At present, STAT functions through a com-
2 RECUMBENT mercially available terminal (IBM-1050 Tele-
3 SYS: Communications System) equipped with a stand-
3 . DIAS: ard keyboard/printer (Selectric typewriter) as
2 STANDING well as components that punch and read paper
3 SYS:
tape and cards. Four characteristics of this
3 DIAS:
1 PULSE terminal make it particularly suitable for the
2 SITTING: purposes at hand:
2 RECUMBENT:
2 STANDING: (1) Operation of the terminal is simple. Clerical
3 EXERCISE and secretarial personnel, with no background in
4 IMMED AFTER:
computer sciences, can be trained in a matter of
4 2 MIN AFTER:
hours to perform all required manipulations.
(2) The terminal can be "programmed"; that

is, instructions such as carriage return, tabulate,
tape punch "on" or "off," reader stop, etc., may
be punched and stored on paper tape in context
with alphabetic and numeric characters. As this
tape is later read through the terminal, each in-
struction is performed so that a degree of format
control is provided that is limited only by the
sequential nature of paper tape. Under program
FIOVRE 4.--Acquisition mechanics.
control, data may be accepted from any input
device (keyboard, paper-tape rcader, or card
reader) and transferred to any or all output
"programmed" as a master tape. Recall that the
devices (printer, paper-tape punch, card punch,
facsimile-storage technique of peripheral pro-
etc.).
gramming requires the use of typewriter symbols
(3) Human-readable and machine-readable
to represent indentation of the headings in the
documents can be produced automatically and
outline for retention of relations during computer
simultaneously so that data are available for
storage. In this application, zero equals no space
immediate use, even though computer support
to the left of the heading, 1 equals one space, and
may be delayed or periodic.
so on through 10 levels. The outline is typed with
(4) This terminal can function on one or both
the appropriate symbol inserted before each
of two channels or "loops." The "home loop"
heading. (Editing codes, such as carriage return
synchronizes the various components attached to
or tabulate, may be inserted to provide an or-
a specific terminal. The "line loop" adds a tele-
ganized appearance of the document..) A typed
phone line to the circuit, over which data may
colon (:) separates the heading field from the
be sent to or received from another compatible
data field. Wherever manual data are to be
terminal or computer. This two-channel feature
entered during operation of the master tape, a
is important. Documents can be drafted with the
reader-stop code is punched. A special (in our
machine on the "home" channel; additions, cor-
case, unprintable) code is used to indicate the
rections, or deletions can be manually entered
end of each data field. The finished inaster tape
from the keyboard after editing. The edited
requires meticulous preparation but provides
document (stored on paper tape) is then trans-
rapid and accurate reproduction of the original
mitted over the "line" channel to another receiv-
outline, to which a typist can add data at great
ing terminal or computer.
speed.
The direct transcription of medical information Total operational efficiency of STAT can be
into computer-readable form (peripheral program- significantly increased by use of the punch-card
ruing) is readily accomplished by medical secre- capability of the terminal system. By prepunch-
taries using the terminal features just described. ing of cards containing the necessary instructions,
Operation of STAT--The systematized termi- the time for preparation of each master rape can
nal-acquisition technique is sunmmrized in figure be reduced from several hours to 15 to 20 min
4. The typist, sitting at the tcrminal keyboard, (fig. 5). A card is prepared for each of the sequen-
places a master tape in the paper-tape reader; tial levels of a hierarchical outline (currentIy 10
this tape controls the sequential typing of a levels). These 10 cards contain all the instructions
format (an organized set of headings) to which required by the master tape except the specific
the typist adds data from a collection form. Two words of the headings. The STAT system uses the
documents result: one is the typed page; the other, simple method of marking each card with a
a paper tape called the data tape. Each contains number corresponding to the sequential degree of
an amalgam of predefined headings from the indentation for its heading. The original outline
master tape and data from the collection form. is then coded by assessment of the degree of
Preparation of master tapes--Each heading indentation and placement of that number in the
outline (described under hledicaI Information) is margin of the outline. A series of heading cards
NEDATA FILE
(HIERARCHICAL
tPR0 A,
tAROS
I RECORD:
TYPE:
SURVEY
DENTAL
SS NO: _64-80-5680 DATE: 20J_n68

NAME: Adams, John
STATUS:
X-HISSING O'RESTORABLE /-NONRESTORABLE -PROTHE$1S
s _ (7 x g)=FIXED BRIDGE (BRACKETS INCLUDE ABUTMENTS) (CODED BELOW NO)
RIGHT: O1 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 LEFT
lOF HEADINGS
I UPPER; X ( X X ) X
MONITOR
PRINTOUT J RIDHT: 32 $1 $0 29 28 27 26 25 24 2_ 22 21 20 19 1B 17 LEFT
LOWER: X 0 X
COMMENTS: Good dental function

FIGURE 5.--Preparation mechanism for
DENTAL CLASS: 2
master tape. Type 3
EXAMINER: J B Moreton, Capt, USAF OC
TYPED: jlb/01Feb6$
is ordered sequentially according to the numbers Fm_ 7.--Appearance of final typed document, a
of the outline format. This deck of cards, when facsimile of which is submitted to the computer as the
read through the machine, stops at the correct data tape.
indentation for the manual entry of each heading

in turn. The resultant paper tape is a master
tape of the outline. data over a telephone line from the paper-tape
The data.collection form--One rule governs the reader of the terminal.
design of a data-collection form: it must have the
The Information-Management Package
same headings as the corresponding master tape,
or their equivalents, and similar sequential orga- The computer manipulations necessary to
nization. It should be noted that data collection service the MEDATA record files are provided
may be accomplished with this system by dicta- in a collection of computer programs called the
tion, either directly or through a recording device, information-management package (IMP) and de-
as easily as by use of a written collection form. signed and developed on a medium-size, second-
Again, however, the sequence of the headings is generation computer (an SDS-930 with 16-K
critical for efficient transcription. Figure 6 shows core, three tape drives, a line printer, and a
a portion of Form-88 used as a manual-collection console typewriter). All programming is written
document and ready for transcription. in a subset of FORTRAN common to most com-
The results--Figure 7 shows the typed result pilers. The IMP is ultimately intended for a
from STAT. Corrections may be made by the time-shared, multiterminal environment but is
secretary either when errors occur or during a equally suitable for limited operational imple-
tape-to-tape duplication of the originM data mentation that requires less sophistication.
tape. Editing is accomplished by proofing of the Pilot projects successfully used the described
typed document, since it and the data tape have STAT but without telephonic transmission to the
exactly the same content. When correct, informa- computer. Data tapes were read by a paper-tape
tion from the data tape is fed to the computer; reader on the computer, and retrieval was ac-
this step may be accomplished by reading of the complished by use of the console typewriter,
data tape by a paper-tape reader attached di- paper tape, or punched cards to define retrieval
rectly to the computer, or by transmission of the parameters. The present version of IMP operates
, , m.
REMARKS AND ADOITIlONALDIDCfAL
44. DENTAL (P_ece dppropriate elmbole abo_e or below cumber ,,[ upJ_r a_d _er teak, rearpectitelfl.)
DERECTS AND
O-- R_torab_e teal* X-- ML,_t_ tedk (6 .'_"_f._}--_red bride, brack#J t_
[--Nonr_lorable tecta XXX--Reldaced b_ denture* ff_l_ 6btdm_
G
H 32. 31 30 29 28 27 P_ 25 I Z4 23 22 21 20 l, I, 17 F
o I _T
FIG_ 6.--Portion of Form-88 used as collection form and ready for transcription.
BIOMEDICAL RESEARCH AND COMPUTER APPLICATION
7O
concerned with recall of information from the

only in the area of record-management, but
communications and analytic capabilities are files. Both segments of the monitor act through
selective use of task-specific subroutines; each
planned as modules for the future.
Modularity is in fact the key to the IMP subroutine performs a single, unique function
concept. The IMP is organized into two types of and, during the time it is in use, is completely
computer-program components: a basic control independent of the rest of the system. This ap-
routine (monitor), and clusters of subroutines proach permits the alteration, exchange, or
addition of subroutines with minor or no changes
(fig. 8).
The monitor has three planned segments. The in the basic monitor program.
executive portion controls remote-terminal access For greatest efficiency, subroutines are orga-
nized into two classes: general and specific. General
to the system and assigns responsibility for
subroutines or "utilities" perform functions that
particular actions to one of the other two seg-
ments. The executive bridge of the monitor may serve any segment of the monitor; three
cannot be completed until either a computer is clusters of utilities are shown in figure 8. Terminal-
utility subroutines provide for translations be-
totally dedicated to the MEDATA project, or a
multiprogramming, time-shared computer sysfem tween computer-code structure and terminal-code
is available. This executive function is handled structure during transmissions, establish com-
munications-control sequences for polling and
manually in the present version. Both limbs of
the monitor are currently functional however. addressing, etc. The search-utility cluster includes
all the subroutines that search the files for specific
The "update" segment handles all modifications
of the record files, including additions, deletions, records or parts of records. This is a necessary
and changes (corrections), as well as the sorting function in order to retrieve from as well as up-
and storing of data. The "retrieve" segment is date the file. The third cluster in the utility class
is a miscellaneous group that performs a variety
of generally useful data-manipulations.
Each of the specific subroutines, also illustrated
as clusters, is tied to only one segment of the
monitor. The purposes of the sort, add, change,
and delete subroutines and the logic and action
r
CURRENTLY
MANUALLY
HANDLED I
I
options are best understood as a function of the
data-retrieval and update techniques described in
the following sections.
TEe Data-Retrieval TecEnique
The Data-Retrieval Technique (DART) is an

organized man-machine communication or "lan-
guage" by which the user defines the specific
information of interest and indicates what is to
be done with those data. The communication may
be conducted as an on-line "conversation," em-
ploying a terminal device connected to the
computer, or as an indirect "batch" operation
using punched cards or other preprepared media.
Several versions of DART have been imple-
mented, and two have been reported (refs. 1 and
2). The current version under development offers
an expanded set of options, increasing control by
HEDATA FILES ]
the user.
The basic mechanism of DART is the question-
FIOUR_ 8.--Organization of the information-management
package. answer technique similar to that used in STAT.
The computer, acting through the programmed states the definition data for the report retrieved.
management system (IMP), poses questions to The user may not want to retrieve a complete
the user. Answers entered in response to these report but only a portion or even a single item.
questions provide the definitions and the action In order to limit the amount of data retrieved,
commands which control the programmed manip- the user may specify a portion or a single item of
ulation of the records on file. Therefore, two sets a report by adding the desired heading terms
of responses are required from the user: (1) a after the TYPE definition. For example, in
definition of the data of interest, and (2) an answer to the question TYPE, the user may
instruction indicating the desired action to be enter not only the name of a report (e.g., XRAY)
performed on the defined data. but also an item heading (e.g., DIAGNOSIS):
Definition is accomplished by the same set of
TYPE XRAY--DIAGNOSIS
questions that define each report in the file: SS
NO, RECORD, TYPE, and DATE. For example, Now the search program will look within the body
in answer to the question SS NO, the user may of the report and retrieve only the specified head-
enter a particular social security number. ing DIAGNOSIS and any associated data.
RECORD is answered by the name of the The user may further define the retrieval cri-
record of interest, such as SURVEY. The re- teria by limiting the data of interest:
sponse to TYPE is the name of one of the sub-
TYPE XRAY--DIAGNOSIS: NEG
divisions of the survey system (table 1), and
DATE is answered by entry of a specific day, In this instance retrieval occurs only if the report
month, and year. The user may enter the word searched has the term NEG or NEGATIVE
ALL to indicate that a definition category is to stored in the data field associated with the
be searched completely. heading DIAGNOSIS.
Figure 9 illustrates the retrieval of a complete This capability is designated "strong search"
VISION SURVEY. In this and all subsequent logic; it allows the user to specify any series (or
illustrations, underlining indicates the informa- string) of alphabetic or numeric characters to be
tion entered by the user. Each response from the used as a model during the computer search of
computer is initiated by a line of printing that re- the stored files. The user indicates whether a
specific string is a heading or data by utilizing
the colon (:) just as it is used in the basic file
SS NO 123-k5-678_,
RECORD SURVEY. structure. A string preceded by a colon is under-
TYPE
DATE stood to be data; the absence of a preceding colon
ACTION LIST.
indicates heading.
12_-_5-6789
VISUAL ACUITY
SURVEY VISION O! dAN 68 Boolean logic adds another dimension to the
DISTANT
UNCORRECTED search definition by offering AND/OR/NOT
OD 20/15
OS 20/15 control. Several sets of strings, headings with or
NEAR
UNCORRECTED without data, may now be linked by these terms
DO 20/15
OS 20/17
ACCO/4NODATION into complex retrieval parameters. AND linkage
OD 7.2
OS 7._ requires that both conditions be met for consti-
INTRAOCULAR TENSION 15.3 HH HG 0 U
VISUAL FIELDS tution of a valid retrieval. OR linkage is satisfied
CONFRONTATION NEG
DEPTH PERCEPTION if either condition is met. NOT allows invalida-
TEST USED VTA-ND
UNCORRECTED PASSES
HETEROPHDRIA DISTANCE ES EX RH LH
tion based on a specified characteristic. For
20 v 0 0 0 O
I_" O 0 o Q demonstration of the use of Boolean logic in
NPC
PD MEDATA, assume the files to contain informa-
PRfSH DIV
PR|SH CON_
COVER TEST ORTHO
tion on only two patients whose height is exactly
RED LENS TEST
NIGHT VISION NIBH
72 in. each. One of these and a third patient each
COLOR VISION ON RECORD AS PASSES
DIAGNOSES weigh exactly 165 lb. Therefore only one of these
CONHENTS
EXM41NER GEORGE L DAILYo HD three men is both 72 in. and 165 Ib in height and
TYPED MH/O1FEB$8
weight. Figures 10 to 12 illustrate application of
FIGURE 9.--Vision survey retrieved. the AND/OR/NOT logic to such a file.
SS NO
RECORD in which data are returned to the requestor is
TYPE H[ASUREMENT-HEIGHT: 72 IN AND WEIGHT: 165 LB.
DATE ALL identical with the input format. Output is essen-
ACTION
tially a "dump" of the stored material as exempli-
|25-kS-67Rg
GENERAL
SURVEY MEASUREMENT 01 dAN 68 fied in figure 9. Whenever a definition specifies
HEIGHT 72 IN
WEIGHT 165 LB less than a complete report, as is usually the case,
a modification called diagonal-retrieval formatting
FmtraE 10.--Example of AND logic.
is employed. Diagonal-retrieval formatting refers
to recovery of all superior and all inferior head-
SS NO
RECORD _RV Y ings related to each defined heading. Figures 10
TYPE _MENT-HEIGHT: 72 IN OR WEIGHT: 165 LD.
DATE ALL.
to 14 show that this format represents facsimile
ACTION LIST.
retrieval with all extraneous information removed.
125-_5-6789 SURVEY MEASUREMENT 01 JAN 68 The purpose of diagonal formatting is to preserve
GENERAL
HEIGHT 72 IN
WEIGHT 165 LB
the most complete meaning of retrieved informa-
lk6-03-1661 SURVEY MEASUREMENT 16 NOV 67 tion. Experience has verified the value of this
GENERAL
HEIGHT 72 IN policy even though sometimes unnecessary sup-
256-75-0123 SURVEY NEASUREHENT 06 dAN 6R plementary data are returned. An example of
GENERAL
_EIGHT 165 LB
extra (but in this case probably useful) retrieval
is the differential count data printed with the
FIGURE ll.--Example of OR logic.
WBC in figure 13--because the original organiza-
tion of headings for the laboratory-sulrvey form
ss NO ElL.
RECORD SURVEY. placed the differential count one hierarchical level
TYPE MEASUREMENT-HEIGHT; 72 IN AND WEIGHT: NOT 165 LB,
DATE ALL. under the white blood count.
ACTION
lk6-O3-1661 SURVEY MEASUREMENT 16 NOV 67 ss NO ALL.

GENERAL RECORD
TYPE DO RY-WBC: kSOO TO k700.
HEIGHT 72 IN
DATE ALL.
ACTION
FIGURE 12.--Example of NOT logic,

123-_5-6789 SURVEY LABORATORY Ol JAN 68
HEMATOLOGY
CBC
WBC 4600
DIEF
NEUT 56
A third logic option--ranging--provides limited EO Ol
BASO 0O
arithmetic latitude such that a range of values LYMPH k5
MONO OR
may be specified. A particular item satisfies the

retrieval parameters if its numeric data value FIGURE 13.--Example of
falls within the stated range. Figure 13 shows ranging logic.
a simple request for a search of all individual

ss NO
LABORATORY SURVEY reports from any RECORD SURVEY,
TYPE LABORATORY-VORL:NEG AND (HD: 15 TO 16 OR HCT: NOT O TO k5),
date to find those having white blood counts DATE ALL.
ACTION LIST.
(WBC) between 4500 and 4700/mm 3. For this
125-k5-6789 SURVEY LABORATORY Ol dAN 68
example only one case is retrieved. String, I I_MUNOLOGY
VDRL NEG
Boolean, and ranging logic may be mixed in a HEMATOLOGY
CBC
single definition set (fig. 14). HB 15 GM
After the user has defined the specific informa- lk6-03-1661 SURVEY LABORATORY 16 NOV 67
I P,HUNOLOGY
tion of interest, he indicates what the computer VDRL NEG
HEMATOLOGY
is to do x_ith it. The ACTION commands avail- CBC
HB 1E GM
able to the user begin _th LIST; figures 9 to 14 HCT _7¢
have demonstrated the results of this command. 06 JAN 68

236-7B-0125 SURVEY LABORATORY
IMMUNOLOGY
LIST simply directs the computer system to VDRL NEG
HEMATOLOGY
produce the defined data in a standardized CBC
HCT kRt
format.
DART takes maximum advantage of the FmURE 14.--Example of combination of string-search,
facsimile storage of information. The basic format Boolean, and ranging logic.
A MEDICAL-INFORMATION
MANAGEMENT
SYSTEM 73
Besides LIST, new action options now being must first define the specific information of inter-
added include COUNT, which totals the number est. Tile peripheral update, therefore, uses the
of valid records matching the retrieval parameters; same communications scheme as does DART,
SPECIAL PRINT FORMATS, which are a with change in only the action-command options.
group of output, programs providing for special- The update-action commands are: ADD, followed
purpose formatting of the output of a retrieval by the new information to be added; CHANGE,
request; and several miscellaneous options to per- followed by new information which is to replace
form other special tasks. Still another kind of the current contents of the defined data field; and
action command, not illustrated, indicates the DELETE, which needs no following character
particular output instrument to be used, such as string but automatically erases the present
terminal, line printer, magnetic tape, paper tape, contents of the defined data field.
punched cards, or console typewriter. The third level or central update technique re-
mains in the development stage. Its function is
The Update Techniques accommodation of generalized changes in file
The 3IEDATA System recognizes three levels structure to allow for uniform reorganization of
of need for capability in updating. The basic up- a report format within a given file, to perform
date technique adds or deletes complete reports. universal alteration on one or more headings
A new report may be added to the established file throughout the file, or to permit any other neces-
by placing an "A" in a specific location in the sary modification of the file that is not provided
leader portion of a data tape prepared by STAT for by the basic and peripheral techniques. Be-
methodology. Since definition of tile report is cause of its broad power, the central update
already present in the data tape, recognition of capability should be controlled by those re-
the "A" allows the program to sort and file that sponsible for the filing system; it should not be
report. Deletion of a complete report uses a available directly to the general user.
similar approach: "D" is positioned in tile leader
of a data tape carrying a complete definition of THE FUTURE OF MEDATA
that specific report; here, however, no body to
the report, is required. The MEDATA System is an operational, com-
The peripheral update technique is the second puterized, medical-record system, one version of
level of capability used to modify information which was to be installed at Manned Spacecraft
already stored in the file; examples include cor- Center by June 30, 196S. Additional develop-
rection of errors or addition of new data. This ment will provide general and specific programs
level is designed for use by the original preparers for data analysis. Programming for terminal
of the data tapes for computer input and for communications depends on the availability of
users of the data-retrieval technique. The periph- computer equipment capable of multiterminal
eral update technique allows modifications only time-sharing. The most significant future con-
in the data fields of the file. This restriction is tribution, however, will be ability to provide
necessary because tlle format and organization of efficient support for continuing improvement of
all reports in a file must remain consistent; other- basic medical documentation.
wise information could be "lost" within the file
by changing of the standard headings of their REFERENCES
relations.
I. FARLEY, W. E.; I°ULIDO, A. I1.; _{INCKLERj T. M.;
In operation the peripheral update technique AND CADY, L. D., JR.: Man-Machine Communica-
is parallel with DART. In the MEDATA system tions in the Biological-Medical Research Environ-
there are two reasons for recalling material stored ment. Proceedings of 21st National Conference of
in the file. Tile first and most frequent reason is the Association for Computing Machinery, 1966,
pp. 263-267.
retrieval of information; the mechanism has been
2. HORTO-N', C. _J:.; 5[INCKLER, T. M.; AND CADY, L. D.,
discussed in detail. The second reason is updating
JR.: MEDATA: A New Concept in Medical Records
by addition _o, change in, or deletion of data Management. Proceedings of Fall Joint Computer
within tile file. To accomplish these actions one Conference, 1967, pp. 485-489.
CHAPTER 6
DEVELOPMENT OF A COMPUTER PROGRAM FOR

AUTOMATED RECOVERY OF LABORATORY DATA
Edward C. Knoblock, Mervyn R. Stein, and Stephen T. Paine
The analysis of data, including data compila- format changes are handled under the Mainte-
tions, arithmetic and statistical computations, nance Mode; the Query Mode provides for re-
graphical presentations, and data tabulations, is trieval, utilizing arithmetic and logical operations,
an integral part of laboratory operation. A com- and the generation of reports in tabular or free-
puter system capable of efficient storage and un- form output. Provision was made for the protec-
complicated recovery, with the facility for logical tion of proprietary data by prevention of inad-
and arithmetic manipulations, would be of great vertent release by unauthorized sources.
benefit to the laboratory. In a space-medicine MEDOL is a third-generation information-
program, where laboratory data are to be ana- processing system that originated front SISTRAN
lyzed, the number of critical variables increases (System Information Storage Retrieval and
so significantly that handling and analysis of Analysis). SISTRAN was specifically designed as
data become an even more complicated task. a library system for storing, retrieving, and
Medically Oriented Language (MEDOL), an analyzing aerospace documents; it provides the
information system designed around laboratory basic system programs for MEDOL. Enhance-
data, was developed to accomplish these ob- ments have been provided for the specific require-
jectives. The data from the Project Mercury ments for processing of biomedical data.
program served as a framework for development The MEDOL System is written primarily in
of a laboratory-oriented file structure for the FORTRAN IV, with a few macroassembly pro-
MEDOL system, with emphasis on current space gram (MAP) subroutines. It operates on a 7094
programs. This system includes flexibility for tape system or a 7094/7040 direct-couple system
addition of new classes of data, and provides for under IBSYS; a 16-tape drive system is recom-
storage and retrieval of data without requiring mended for maximum efficiency. MEDOL was
the user to be familiar with sophisticated com- designed to be independent of secondary storage.
puter languages. All critical parameters required In addition, it is completely modular and open-
for evaluation of quality control and to describe ended to facilitate conversion to other computer
essential detail of each procedure are described configurations including on-line computer sys-
within the file structure. tems; it was implemented in an English-like
The MEDOL system employs a higher-level syntactical and semantic source language to
interpretive source language that requires the allow the user to communicate with the system
use of a few English-language-oriented statements with maximum ease.
for file design and entry and retrieval of data. A The basic system provides functions or pro-
synonym capability is available through a system cedures for (1) generating files, updating files,
glossary, and a dimensional library provides for and retrieving data from these files; (2) generat-
interconversion of units for the convenience of ing libraries and glossaries; (3) performing ele-
the user. mentary arithmetic operations on data; (4) ex-
There are two modes of operation in the system: tracting information for array grouping; and (5)
file-generation additions and deletions of data and output of data.
75
SOURCE LANGUAGE statements, breaks them into components, and

finally codes them in the internal langnage as
An information-processing system must be
procedural directives; this is accomplished before
accessible to many users, each having different
and separately from execution of the program. If
requirements, a significant number of whom have an error is detected in a statement, an appropriate
little computer experience. Therefore, the system
error message accompanies the source-language
must provide a language (source) that can be
listing, and a procedural directive is not generated.
successfully handled by the inexperienced as well
The procedural directives supply the processor
as the experienced. The advantage of employing
with information as to which programs to call.
an English-like source langnage is that the tasks
As part of the program-execution the processor
to be performed by the computer can be declared
must relate operations, locations, and references
by individuals who are closest to the data, even
from one part of the program to those in another.
though their computer experience is linfited.
This latter activity contrasts with the work of
The MEDOL source language was designed to
increme_tal compilers used for multiprocessing
meet this objective; it has many similarities to
and on-line systems, where a statement is com-
FORTRAN but is certainly easier to learn and
piled and machine-language instructions are
use. Furthermore an individual familiar with
generated independently of the next statement.
FORTRAN will learn MEDOL faster, but this
The generation involves two n_spects: format
is not an absolute requirement. The language
design and data input. The file is based on a tree-
consists of key words and symbols combined into
structure form. The user defines all his variables
statements that describe the MEDOL procedures.
and assigns appropriate hierarchical relations be-
The key words and statements are readily compre- tween them for creation of the data file tree.
hended by the user since they are terms in com-
The tree can contain up to 15 levels with a
mon usage. Some representative procedures em-
variable number of attributes within each. The
ployed in the MEDOI, source language are
data-bearing elements are at the lowest level of
ERASE, ADD, DELETE, PRINT, and END
each branch.
OF DATA. The procedures are combined to form
In the "update" function, files, data in flies,
the source-language program (the mechanism by and formats of files can be altered. Format can
which the user indicates to the computer what
be modified to correspond with a change in tim
jobs are to be performed). The program layout is
data status. Attribute names may be added as
intended to approximate the logical thought
new data are available and may be deleted when
processes normally employed by the scientific
no longer needed. Data rearrangements within
investigator. Finally the user is relieved of many
data strings may also be accommodated. Pro-
bookkeeping chores imposed by many other
cedural directives generated in internal langnage
langalages (e.g., FORTRAN).
by the preprocessor are executed by the processor.
Data entry to the MEDOL system is via the
source-language program; they are entered as a
continuous string. The items of data are separated MODES OF OPERATION
by delimiters and arranged according to the file's
The system operates in two modes, the data-
tree structure; absent data are indicated by em-
maintenance mode and the query mode, only one
bedded commas. A hazard of this scheme is that
of which can be processed at any one time. Some
omission of a comma can result in incorrect entry
of the procedures used are unique to each mode,
of data for all items following the error.
while others may be used in common. The mainte-
nance mode provides an input function for entry
PREPROCESSOR AND PROCESSOR
of source-language information and data whereby
One of the most significant features of the files are set up and updated, and dictionaries,
MEDOL system is the modularity, which is glossaries, and libraries are provided in the data
most apparent in the operation of the preprocessor bank.
and processor. Once the source program is read The query mode is used to obtain information
into the computer, the preprocessor analyzes the from the data batik for the purpose of computa-
COMPUTER PROGRAM FOR RECOVERY OF LABORATORY DATA 77
tion or output of data as a printed report, punched subsequent data from the Gemini and Apollo
cards, or magnetic tape--retrieval may be by programs can be easily added.
item, group, or file. It is initiated by referencing Although the system handles vital-signs, fluid-
of data name(s) to the levels of structure desired. input, and food-input data, the file strueture is
Three basic types of queries exist: nonarithmetic, primarily geared toward the proeessing of labora-
arithmetic, and logical. The nonarithmetic opera- tory data. Factors common to the processing of
tion functions for retrieval of data from either routine laboratory data had to be considered in
the temporary or permanent files. design of the file, as well as those unique to
The arithmetic part performs computations by aerospace-data processing. Information eoncern-
use of the five basic arithmetic operators: expo- ing eolleetion, storage, and transportation of
nentiation, multiplication, division, addition, and specimens, including distribution to two or more
subtraction. In addition, FORTRAN IV functions laboratories, has to be included in an aerospace
can be utilized to augment the arithmetic capa- system. Each specimen must also be linked by
bility. Special subroutines are used for statistical data to the aerospace activities or experiments
analysis. under way when the specimen was obtained.
The logic query provides for decision-making The system eontains three files as outlined in
capability such as testing of the data bank Appendix: the Astronaut History File, the
against certain conditions. These expressions can Flight Data File, and the Test Deseription File.
be nested up to 15 levels to provide complex The History File lists the flight(s) in which tile
conditional expressions. In addition, arithmetic astronaut participated, his birth (late, and com-
and nonarithmetic statements can be used within ments; it can be expanded to inelude additional
conditional expressions. items of pertinence.
The Flight Data File contains the bulk of the
MULTIPLE FILES biomedical data and includes the vital-signs,
fluid-input, food-input, laboratory, and collection
The MEDOL system is capable of dynamic
array. Data can be stored and retrieved in multi- and storage data. In addition, the reference data
fill a significant position in this file. The file is
dimensional arrays located in a temporary storage
astronaut-oriented within flight.
area. Single elements, rows, or entire arrays can
The referenee data cover the significant aero-
be retrieved with one reference. The indexing
scheme to identify individual elements and rows space experiments associated with each flight and
is simple to use, and index arithmetic is available. include a detailed chronological history of the
flight. The data are divided into four basle
Arithmetic and logical computations can be per-
periods: base-line, preflight, flight, and postflight.
formed on the data in the "hold" queue as well
The base-line period begins when the astronaut
as in the permanent files. The hold queue is
enters the space program (for his first flight) or
capable of unlimited storage since, when the core
area is exhausted, data can be transferred to on completion of his last postflight period (for
scratch tape. subsequent flights). The base-line period for the
Mercury program began with the medical inter-
REPORTS rogation at Lovelaee Clinic in 1959; it exiended

to 30 days before launch. The preflight, flight,
Data can be delivered on punched cards, mag- and postflight periods last from -30 days to
netic tape, or printed reports. The system pro- launch, from launch to splashdown, and from
vides a mechanism for "formatting" of the data splashdown to +30 days, respectively. Within
for the three output media. Printed reports can eaeh period are the specific reference events that
be generated in free or formatted form (tabular ean be related to the laboratory specimens or
form). other data. Aeeess to the data may be via the
The file format was designed to accommodate basle periods or through specific events. The
aerospace laboratory and other biomedical data. basle structure of the reference data operated
The file design was based on the Mercury Pro- equally well for the Gemini and Apollo data after
gram data, but the system is so flexible that only a few minor adjustments.
The Test Description File was assembled to having to know the precise description of the
provide the parameters that would be required test within the file.
in evaluation of the data collected. The purpose During the process for data-retrieval the user
of this file was to provide a sufficiently compre- describes the desired data with instructions for
hensive description of the entire laboratory pro- computer operations which follow the logicaI
cedure for comparisons between laboratories, and sequence of the input of data into the program.
to determine whether observed responses were Table 1 shows the format of a representative
significant. For this purpose it was necessary to query. By application of such approaches, the
describe each type of sample, the analytical three files provided enable extensive evaluation
technique, the laboratory, the method, and of the experience under examination.
collection and storage of the sample, and to
identify performing personnel. Also included for
SUMMARY
method-description are the literature reference,
the precision to be expected of the method, and a The MEDOL system is not foolproof, nor does
"normal" range with a gate to identify unusual it think for the user; it is a tool for handling large
results immediately. With the various tests then numbers of medical data. If the user expends
assembled by groups of associated analyses, the enough thought on how he wants to query the
file allows ready access by the user to recover system and how he wants the medical data pre-
desired information. The subroutine of table of sented for examination, and if he devotes time to
synonyms provides access without the user planning and laying-out of the files, tree struc-
TABLE 1.--Format of a Representative Queryt
1. Extract enzyme (specify tests) test results

a. For pre- and postflight periods.
b. For astronaut X (ensure that the collection dates for these tests for this period for the
astronaut are in absolute form).
2. Extract normal range for these tests.
3. Test for out-of-range values.
4. Print out all extracted results in the following format:
Astronaut: James E. Doe

Period: Pre- and postflight
Test
Period Date,
1962 SGOT, SGPT, LDH, ALK P.,
I.U. I.U. I.U. B.U.
Preflight 2 Jan 19 6 190 --

Preflight 8 Feb 27 10 125 --
Postflight 20 Apr *68 *50 *560 11
or
Astronaut: James E. Doe
Period: Pre- and postflight
Preflight dates, 1962

Test Postflight, 20 Apr 1962
2 Jan 8 Feb
SGOT, I.U. 19 27 *68

SGPT, I.U. 6 10 *50
LDH, I.U. 190 125 *560
ALK P., B.U. -- -- 11
t All numbers are integers without signs.

COMPUTER PROGRAM FOR RECOVERY OF LABORATORY DATA 79
tures, and data formats, clear and useful reports studies from two or more laboratories had to be
can be obtained continuously. suitable for separate identification and analyses.
During recent development of MEDOL, many The files had to be so structured that they
difficult and unique problems had to be resolved. could be maintained and searched efficiently.
The system's information-retrieval and analytic Therefore, they were structured in trees of
structures already existed, but enhancements of hierarchies with nested "repeats" to accommodate
this basic program were needed to accommodate multiple flights, laboratories, specimens, and
the greater character strings typical of medical comments.
queries and files, to make updating operations The present system will not satisfy all future
easier, and to gain greater computational ability requirements within the time and cost constraints
through additional subroutines. provided, but a basic operational system has been
Around this framework it was necessary to constructed. The next step should be provision of
build a medically oriented language that would greater repeat-generating capability through plot
allow the investigator to file laboratory informa- subroutines and flexible data-association in tabu-
tion and query the files in a language familiar to lar form, so that the medical examiner can scan
him. For this purpose, glossaries of medical the printout (i.e., graphic plots and columnar
synonyms and units-conversion tables were con- listing of associative attributes) and reach a cor-
structed and placed in the system's reference files. rect conclusion quickly.
In addition, diagnostic messages designed to pin- As more data become available, greater analytic
point errors in the data, queries, and file-entry
capability must be built into the MEDOL system
operations were developed.
to provide statistical correlations between at-
Finally the most difficult task was the setting
tributes of interest to a medical investigator.
up of file formats that would provide flexibility,
Finally the internal system should be made more
beyond the Mercury-data base, for the Gemini
and Apollo laboratory analyses. Factors involv- machine-independent by elimination of the few
ing multiple astronauts on multiple flights had to assembly-language terms in the preprocessor and
be taken into account. Since all evaluations were processor, which would provide for easy transfer
not completed in a single laboratory, parallel of MEDOL from one computer system to another.
APPENDIX
NASA'S LABORATORY-DATA SYSTEM'S

TREE STRUCTURE
I. Astronaut's History File: a. Begin

b. End
a) Name
2. Comment
b) Serial number
2. Preflight period (from -30 days to
c) Flight number *implicit repeat
launch)
d) Birth date
a. Date/time
e) Comment *implicit repeat
1. Begin
2. End
II. Flight-Data File :
b. Simulations *implicit repeat
a) Flight number
1. Date/time
b) Astronaut
1. Astronaut name a. Begin
b. End
2. Serial number
2. Name
c) Reference data
3. Comment
1. Base-Iine period (from first data point,
or end of last postflight period, to be- c. Preflight physical exam (not to
include the exam during count-
ginning of next preflight period)
a. Date/time down) *implicit repeat
1. Begin 1. Date/time
2. End 2. Comment
b. Lovelaee Clinic d. Countdown, abort *implicit re-
1. Date/time peat
a. Begin 1. Date/time
b. End a. Begin
2. Comment b. End
e. Simulations *implicit repeat 2. Awaken
a. Date/time
1. Date/time
b. Comment
a. Begin
b. End 3. Prelaunch breakfast
2. Presimulation period a. Date/time

b. Comment
a. Begin
b. End 4. Don pressure suit
3. Simulation period a. Date/time
b. Comment
a. Begin
b. End 5. Insertion into spacecraft
4. Postsimulation period a. Date/time
b. Comment
a. Begin
b. End 6. Flight cancelled
5. Name a. Date/time
6. Comment b. Comment
d. Diurnal studies e. Countdown, flight (from awaken
1. Date/time to launch)
81
3. Comment *implicit repeat

1. Date/time
c. Debriefing (from arrival to re-
a. Begin
b. End lease)
1. Date/time
2. Awaken
a. Begin
a. Date/time
b. End
b. Comment
2. Period *implicit repeat
3. Prelauneh breakfast
a. Date/time
a. Date/time
1. Begin
b. Comment
2. End
4. Don pressure suit
b. Place (USA, aircraft
a. Date/time
b. Comment carrier, Grand Turk
5. Insertion into spacecraft Island, or Grand Ba-
hama Island)
a. Date/time
c. Comment
b. Comment
d. Additional activities *implicit re-
3. Flight period (from launch to splash-
peat
down)
1. Date/time
a. Date/time
a. Begin
1. Begin
b. End
2. End
2. Name
b. Lift-off (from ignition to insertion
3. Comment
into orbit)
1. Date/time (Under Base-line, Preflight, Flight, and Postflight
a. Begin periods are listed some of the reference points for
b. End the Mercury program. The list is by no means
2. Comment complete and will continue to be enlarged for the
c. In-flight (from insertion to firing Gemini and Apollo programs.)
of retro-rockets) d) Vital signs
1. Temperature *implicit repeat
1. Date/time
a. Date/time
a. Begin
b. End b. Value (*F)
c. Anatomic site (oral, rectal, or
2. Comment
axillary)
d. Reentry (from retro-rockets to
d. Comment
splashdown)
1. Date/time 2. Blood pressure *implicit repeat
a. :Begin a. Date/time
b. End b. B.P. data *implicit repeat
2. Comment 1. Value (mm-Hg)
2. Arm (right or left)
4. Postflight period (from splashdown to
3. Position (supine, standing,
+30 days)
sitting)
a. Date/time
4. Comment
1. Begin
2. End 3. Weight *implicit repeat
b. Recovery (from splashdown to a. Date/time
debriefing site) b. Value (Ib)
c. Status (after voiding and/or
1. Date/time
nude) *implicit repeat
a. Begin
b. End d. Comment
2. Recovery site (Atlantic or 4. Pulse *implicit repeat

Pacific Ocean) a. Date/time
LABORATORY DATA TREE STRUCTURE 83
b. Pulse data *implicit repeat ingredients" will be used in Gemini and Apollo
1. Position (supine, standing, programs, but not for Mercury.)
sitting) g) Specimen collection and storage *implicit
2. Value (beats per minute) repeat
3. Exercise status (before or 1. Specimen (e.g., blood)
after exercise) 2. CS data *implicit repeat
4. Comment a. Collection interval
5. Respiration *implicit repeat 1. Begin
a. Date/time 2. End
b. Value (breaths per minute) b. Collection (refers to the entire
c. Comment sample) *implicit repeat
6. Extremity measurement *implicit re- 1. Sample collection date/time
peat (end time for urine)
a. Date/time 2. Centrifugation date/time
b. Extremity data *implicit repeat (for blood specimens)
1. Value (in.) 3. Volume (for urine speci-
2. Extremity site (wrist, fore- mens)
arm, thigh, calf, and ankle) 4. Personnel
3. Side (left or right) 5. Comment
4. Comment c. Distribution (a description of di-
7. Vital capacity *implicit repeat vision of the samples and distri-
a. Date/time bution to their performing labo-
b. Value (liters) ratories) *implicit repeat
c. Comment 1. Sample [the types of blood
8. General comments *implicit repeat samples that were sent to
a. Date/time the lab. (2, below) from this
b. Comment collection; serum, plasma, or
e) Fluid input whole blood] *implicit repeat
1. Fluid data *implicit repeat 2. Laboratory (performing)
a. Date/time 3. Storage *implicit repeat
1. Begin a. Phase (initial and final,
2. End or blank)
b. Volume (ml) b. Storage method (deep
c. Type (water, tea, suspended food, freeze -10 ° F, deep
soup, coffee, juice, other, or com- freeze -40 ° F, dry ice,
binations) *implicit repeat or liquid N_)
d. Comment c. Date/time
f) Food input 1. Begin
1. Food data *implicit repeat 2. End
a. Date/time d. Comment (e.g., sample
i. Begin thawcd or lost)
2. End 4. Transport (includes trans-
b. Type (e.g., apples, potatoes) *im- portation of the sample from
plicit repeat the collection-storage area to
c. Elemental ingredients (e.g., vita- the distribution point, and
mins, calcium) *implicit repeat from the latter to the per-
d. Meal forming laboratory; or from
e. Comment the collection-storage area
(Food data for Mercury program should be directly to the performing
entered under "Meal" and "Type"; "Elemental laboratory) *implicit repeat
a. Phase (initial, final, or 5. Proteins

direct) a. Total protein *implicit
b. Storage method repeat
c. Date/time
1. Begin
2. End
d. Distribution point (ap-
plies only to initial b. Electrophoresis *im-
phase; WRAIR in this plicit repeat
1. Collection date
case)
e. Comment 2. Fractions
a. Albumin
h) Laboratory Test Performance
1. Blood b. Alpha-1
a. Chemistry globulin
1. Electrolytes c. Alpha-2
a. Sodium *implicit repeat globulin
1. Collection date d. Beta-1
2. Result globulin
3. Comment e. Beta-2
4. Laboratory (leave globulin
empty for Mercury f. Gamma
program) 3. Comment
5. Performance date 4. Laboratory
b. Potassium *implicit re- 5. Performance date
6. Steroids
peat
2. Enzymes 7. Blood gases
3. Catecholamines 8. Miscellaneous
4. Minerals b. Hematology
a. Cations 1. Routine *implicit repeat
a. Collection date
b. Tests
1. Hemoglobin
a. Result
b. Anions b. Comment
2. Hematocrit
a. Result
b. Comment
3. WBC
a. Result 3. Comment
b. Comment 4. Laboratory
4. RBC 5. Performance date
a. Result b. Electrolytes
b. Comment 1. Sodium *implicit repeat
5. WBC differential a. Collection date
a. Lymphocytes 1. Begin
b. Neutrophils 2. End
c. Stabs b. Result
d. Monocytes c. Comment
e. Eosinophils d. Laboratory
f. Basophils e. Performance date
g. Comment
6. RBC morphology
(descriptive)
7. Platelets
a. Result c. Catecholamines
b. Comment
c. Laboratory
d. Performance date
2. Miscellaneous
a. ESR *implicit repeat d. Minerals
1. Cations
c. Serology
1. VDRL *implicit repeat 2. Anions
[All blood tests not described with a special format c. Steroids

are to be treated with the standard format (see
Sodium, above).]
2. Urine
a. Urinalysis *implicit repeat
1. Collection date f. Diurnal studies *implicit repeat
a. Begin 1. Collection date
b. End a. Begin
2. Tests b. End
a. Volume 2. Sample *implicit repeat
b. Specific gravity a. Collection period
c. pH 1. Begin
d. Albumin 2. End
e. Glucose b. Catecholamines
f. Ketones 1. Test *implicit re-
g. Occult blood peat

h. Bile a. Name
i. Microscopic b. Result
c. Comment 4. Gastric analysis *implicit repeat

d. Performance a. Collection (gate
date b. Test meal
c. Steroids c. After ingestion (minutes)
1. Test *implicit re- d. Tests
peat 1. Volume
a. Name 2. Total acid
b. Result 3. Free acid
c. Comment 4. Appearance (text)
d. Performance e. Comment
date f. Laboratory
g. Miscellaneous g. Performance date
1. Albumin *implicit repeat 5. Stool *implicit repeat
a. Collection date
b. Results
1. Character
2. Direct
2. Xylose absorption test *im- 3. Concentrated
plicit repeat a. Faust •
a. Collection date b. De Rivas
b. Condition (a, b, or c) c. Comment
c. Result d. Laboratory
1. HI e. Performance date
2. H2 6. Semen *implicit repeat
3. tt3 a. Collection date
4. H4 b. Results
5. H5 1. Volume
d. Comment 2. Sperm count
e. Laboratory 3. Motility
f. Performance date a. Motile
[All urine tests not described with a special format b. Moribund
are to be treated with the standard format (see c. Inert
Sodium, above). The following tests have special 4. Morphology
formats :] a. Abnormal
3. Bone marrow *implicit repeat b. Normal
a. Collection date 5. WBC
b. Result c. Comment
1. Metamyelocyte d. Laboratory
2. Myelocyte-C e. Performance date
3. Myelocyte-B III. Test Description File
4. Myelocyte-A 1) Blood
5. Myeloblast a) Chemistry
6. Late erythroblast 1. Electrolytes
7. Normoblast a. Sodium
8. Eosinophilic myelocyte 1. Laboratory *implicit re-
9. Plasma cells peat
10. Megakaryocytes a. Name
e. Comment b. Period
d. Laboratory 1. Begin
e. Performance date 2. End
c. Method b. Electrophoresis
1. Name 1. Laboratory *implicit re-
2. Reference peat
3. Precision a. Name
(S.D. or %) . b. Period
4. Normal range 1. Begin
a. High 2. End
b. Low c. Method
5. Sample 1. Name
a. Type 2. Reference
(serum, 3. Precision
etc.) 4. Normal range
b. Anticoag- a. Albumin
ulant 1. High
6. Preservative 2. Low
7. Comment b. Alpha-1
d. Personnel *implicit 1. High
repeat _, Low
t.'
b. Potassium c. Alpha-2
1. High
2. Low
d. Beta-1
1. High
2. Low
2. Ertzymes
e. Beta-2
1. High
2. Low
f. Gamma
3. Catecholamines 1. High
2. Low
5. Sample
a. Type
b. Anticoag-
4. Minerals ulant
a. Cations 6. Preservative
7. Comment
d. Personnel *implicit
repeat
6. Steroids
b. Anions
7. Blood gases
5. Proteins
a. Total protein
8. Miscellaneous
(RBC morpholo_" not included)

8. Platelets
b) Hematology
1. Routine
a. Laboratory *implicit repeat 9. Sample
1. Name a. Type
2. Period b. Anticoagulant
a. Begin 10. Preservative
b. End l l. Personnel *implicit re-
3. Hemoglobin peat
a. Method (For HCT, WBC, RBC, and platelets, use the
1. Name format under Hemoglobin.)
2. Reference 2. Miscellaneous
3. Precision a. ESR
4. Normal range
a. High
b. Low
5. Comment
4. Hematocrit c) Serology
5. WBC 1. VDRL
6. RBC
7. WBC differential
a. Method
1. Name
2. Reference [All blood tests that do not have a special format
3. Normal range are to be treated with the standard format (see
a. Lympho- Sodium, above).]
cytes 2) Urine
1. High a) Urinalysis
2. Low 1. Laboratory *implicit repeat
b. Neutro- a. Name
phils b. Period
1. High 1. Begin
2. Low 2. End
c. Stabs c. Specific gravity
1. High 1. 5Iethod
2. Low a. Name
d. _{ono- b. Reference
cytes c. Precision
1. High d. Comment
2. Low d. pH
e. Eosino-
phils
1. High
2. Low e. Microscopic
f.Easophils
1. High
2. Low
4. Comment f. Preservative
g. Personnel *implicit re- b. End

peat 3. Method
b) Electrolytes a. Name
1. Sodium b. Reference
a. Laboratory *implicit re- c, Precision
peat d. Preservative
1. Name e. Comment
2. Period 4. Test condition *im-
a. Begin plicit repeat
b. End a. Name (plus
3. Method text)
a. Name b. Control data
b. Reference *implicit re-
c. Precision peat
d. Normal range 1. Control
1. High individual
2. Low 2. H1
e. Preservative 3. H2
f. Comment 4. H3
4. Personnel *implicit 5. H4
repeat 6. H5
2. Potassium 5. Personnel *implicit
repeat
3) Bone marrow
a) Laboratory *implicit repeat
e) Catecholamines 1. Name
2, Period
a. Begin
b. Emt
3, Method
d) Minerals a. Name
b. Ileferenee
c. Nornml range
1. Metamyelocyte
e) Steroids a. High
b. Low
2. Myelocytc C
a. High
f) Miscellaneous (volume trans- b. Low
ferred to collection group) 3. Myeloeyte-B
1. Albumin a. High
b. Low
4. Myeloeyte-A
a. High
2, Xylose absorption test b. Low
a. Laboratory *implicit re- 5. Myeloblast
peat a. High
1. Name b. Low
2. Period 6. Late erythroblast
a. Begin a. High
b. Low a. Method
7. Normoblast 1. Name
a. High 2. Reference
b. Low 3. Comment
8. Eosinophilic 4. Concentrated Faust
myelocyte a. Method
a. High 1. Name
b. Low 2. Reference
9. Plasma cell 3. Comment
a. High 5. Concentrated De Rivas
b. Low a. Method
10. Megakaryocyte 1. Name
a. High 2. Reference
b. Low 3. Comment
d. Comment 6. Personnel *implicit repeat
4. Personnal *implicit repeat 6) Semen
4) Gastric analysis a) Laboratory *implicit repeat
a) Laboratory *implicit repeat 1. Name
1. Name 2. Period
2. Period a. Begin
a. Begin b. End
b. End 3. Method
3. Test meal a. Name
4. After ingestion b. Reference
5. Total acidity c. Normal range
a. Method 1. Sperm count
1. Name a. High
2. Reference b. Low
3. Precision 2. Motility
4. Normal range a. MotUe
a. High 1. High
b. Low 2. Low
5. Comment b. Moribund
6. Free acidity 1. High
a. Method 2, Low
1. Name e. Inert
2. Reference 1. High
3. Precision 2. Low
4. Normal range 3. Morphology
a. High a. Abnormal
b. Low forms
5. Comment 1. High
7. Personnel *implicit repeat 2. Low
5) Stool b. Normal forms
a) Laboratory *implicit repeat 1. High
1. Name 2. Low
2. Period 4. WBC (occasional
a. Begin or absent)
b. End d. Comment
3. Direct 4. Personnel *implicit repeat
CHAPTER 7
CONTINUOUS MONITORING AND

INTERPRETATION OF ELECTROCARDIOGRAMS
FROM SPACE
Cesar A. Caceres, Anna Lea Weihrer, Sidney Abraham, David E. Wincr,

Larry K. Jackson, Jerome Sashin, Stuart W. Rosner, Juan B. Calatayud,
James W. McAllister, and Howard M. Hochberg
Monitoring of subjects in real time--that is, in ing of conventional time-limited medical signals
stressful situations such as during intensive care by computer, the Medical Systems Development
before or after surgery, or in space capsules--can Laboratory (MSDL, formerly the Instrumenta-
be of life-preserving necessity. Its chief limitation tion Field Station) has conducted several on-line-
is that the real time of the subject is the same as analysis projects since 1961. In most of these the
that of the human monitor--generally a highly telephone system has been used for transmission
trained physician or nurse whose time is too scarce of the wave forms as analog signals to the com-
and costly for routine, noncreative use. puter site; there they are received by dataphone
These aspects of monitoring raise timely ques- and put through an analog-to-digital converter,
tions for medical engineering: "To what extent and thence go to a digital-computer system for
can a computer system facilitate the work and analysis. The signals are processed, interpreted,
save the time of a human medical monitor?" and available for retransmission to the physician
"Can a computer system provide rapid and re- within seconds of receipt at the processing center.
liable analysis of continuous data to relieve the The measurements and interpretation are de-
human monitor of routine, noncreative tasks?" livered to a teletypewriter or remote printer for
From limited but varied experimentation in the the display. Figure 1 is a facsimile of a routine
automated monitoring of medical signals--in- printout; to demonstrate the routine function of
cluding such vital parameters as the electrocardio- an automated system for all medical signals,
gram (ECG), the electroencephalogram (EEG), electrocardiographic signals have been used as a
heart rates, heart sounds, and respiratory curves-- model.
we have evidence that an automated system can A computer-processed ECG is intended to be
transmit continuous signals, convert them to a diagnostic aid to the physician interpreting an
computer input, analyze them rapidly and ac- ECG for patient care or clinical investigation. As
curately, and display them as needed during care such the computer interpretation should be
of a patient, for other clinical or experimental pur- viewed as a screening tool providing consistent
poses, or during space flight. answers when defined wave patterns are present.
A computer interpretation, like that of the electro-
cardiographer, carries a differing weight in a
ON-LINE ANALYSIS OF LIMITED-TIME
patient's diagnosis depending on the clinical
SIGNALS
circumstances. Only the physician in charge of
On-line analysis of a signal, recorded within a the patient can determine the weight to be given
defined time period, preceded on-line analysis to the interpretation.
during open-ended time periods. As a means of, The beginner studying electrocardiography will
demonstrating the feasibility of on-line process- find that computer interpretations can be useful
¢
91
92 BIOMEDICAL RESEARCH AND COMPTJTER APPLICATION
MEDICAL SYSTEMS DEVELOPMENT LAB--HEART DISEASE CONTROL PROGRAM

COMPUTER PROCESSED ELECTROCARDIOGRAM
large. In the criteria used for infarcts, an S-wave
C ST CLINIC
preceded by an R of zero is equivalent to a
NAME DIAGNOSIS
NUMBER TAPE 0309 OPTION 000 DATE 12.05-6/ TIME DO--29--34 Q-wave. The ST-segment duration and the
HEIGHT 75 WEIGHT 20B AGE 61 MALE MEDS UNKNOWN
BP HYPERTENSION SYSTOLIC 160 ABOVE OR DIASTOLIC 95 OR ABOVE ST-amplitude are measured at three points: the
ZEi;-_ .... i_-ii_--i_--k_V-k_;--"_-'_ ....;_"7-"_;'"_;.... onset (STO), middle (STM), and end (STE).
PD •08 .OB ,06 ,ID ,DO .OB .04 ,DO .DB ,05 ,05 ,04
If waves are absent or below an arbitrary
P'A
P'D
.00 .00
,00 .DO
,DO
,DO
,00
.DO
.OD -,03 -.03 .00 .DO ,OD
,DO ,O2 ,07 .DO ,DO ,DO
.00
,DO
.DO
,DO
"suppression level," zero values are printed for
QA -.04 .00 .00 ,00 ,DO ,OO ,DO .00 .OO .00 ,00 .OO
_0 ,O2 ,00 .DO ,DO .00 ,DO .DO .OO .00 .DO •00 .DO
amplitude and duration. The following suppres-
P& 1.01 •47 .IO .DO .91 ,17 .05 .91 1.97 2.31 1.52 .g5
RD .06 .06 •03 .00 •ll .02 .02 .05 .06 .06 .07 .06 sion levels are used: P=0.025 mV--Q, 0.025
SA ._ -•37 -.$9 -•74 ,00 -,52 -•3P -.27 -.22 -._S -•OP •DO
SD ,OO .05 .08 .OB .00 ,OS ,DO •02 ,02 .02 ,0_ •00 mV; T=0.03 mV--R=0.025 m_S=0.06 mV;
R'A .DO •00 .00 .09 .00 ,00 ,00 .00 .DO ,00 .00 •00
R'D .OO .00 .DO ,03 ,DO ,00 .00 .00 ,00 ,OO •OO .00 T,T',T=0.03 mV--R, 0.1 mV; Rd, 0.05 see--S,
ST .12 .12 ,12 .12 .12 ,12 .12 .12 •12 .12 .I_ .12
STO -•01 -.04 -.03 ,05 ,DO -,05 ,08 -.OZ -.03 -.OS -.05 -.08 0.017; Sd, 0.06 see.
STM -.04 -.01 .03 .03 -.02 .01 -.01 .04 •01 -.04 -•02 -.05
STE -•02 .01 .03 -.Ol -,Of ,04 -.02 .04 .06 -.01 .Of -.04 Axis determination of P, QRS, T, Q, R, S, STO,
TA ._I .24 .06 -.23 .09 ,16 -.22 .29 .42 .34 •24 .16
ST-T, and QRS-T--A pair of limb leads, either
QRS •08 .ll ,ll ,08 ,ll ,07 ,OB ,07 ,08 .08 .OR .06 in both standard leads or in both augmented
QT .41 .41 .33 ,40 ,40 ,36 ,41 ,39 .41 ,43 .42 ,38
RATE 58 55 61 57 56 54 58 5B 57 57 54 58 leads, is selected on the basis of amplitudes of the
_G;_---'_'"-I ..... _..... i..... _..... _..... ;""_'"-;'"'i ..... _..... ;.... wave whose axis is being determined. Since
EAL B3 B3 B3 B3 83 83 B3 B3 B3 83 B3 B3
-k_;i__--_;;_a";;_; .............. _;i:_-G_;:_ ........ amplitude-determination is more accurate from

DEGREES 47 -20 37 -03 -4g 256 141 57
large waves, the pair of lea(Is selected is such that
_;i-_k_-GGGEG_G ................. :......................................
. BRADYCARDIA the smaller of the two waves is as large as pos-
B311QR_ AXIS RANGE -IO to -59
• _errA_s _wm_oN sible. For example, if the net QRS-amplitude in
leads I and II is 0.70 and 0.20 mV, respectively,
FmORE 1.--Facsimile of a routine printout.
and in aVR and aVL, 0.36 and 0.34 mV, respec-
tively, the latter pair is selected because it has
as a "self teaching" method as he reviews the the larger of the two smaller waves.
computer interpretation of the ECG, refers to the No axis is calculated unless both amplitudes
list of criteria to determine the specific abnormal- equal or exceed the following values: P, :i:0.05
ities on which each diagnosis is base(], identifies mV; ST-onset, :t:0.10 inV. The axis may be
the abnormal computer measurements, and in- checked graphically by plotting thc amplitudes
spects the ECG tracing to study the waves from on a hexaxial figure, drawing perpendiculars to
which the measurements were made. Similarly the axis, and determining the intersection of the
the practitioner shouhl first review the computer perpendiculars. The axis, clockwise from the
interpretation of the ECG and then inspect the positive horizontal through three quadrants (0 to
tracing if necessary. 270 degrees) and the final quadrant, clockwise is
In our system the 12-1cad ECG is sent by -90 to 0 degrees.
telephone or recorded on FM magnetic tape with Special information in the printout--In the code
a specially designed data-acquisition system. At line on the printout, the number in each lead
the computer center the signals are received or column is the numbcr of the clectrocardiographic
played back from magnetic tape, sampled 500 complex analyzed (QRS, counting from the left)
times per second, digitized, and entered into in 4 sec of recording. This infornmtion is useful if
computer memory. The duration of signals there is variation between heart cycles or if arti-
analyzed by the computer at one time is about 4 facts arc present• Code _etters may be printed
sec. The analysis has the following features: Mso when difficulties in processing are encountered.
I&nh_cation and measurements of waves--All The recorded calibration pulse is considered by
waves of clinical significance arc identified, and the computer program to be equivalent to 1 mV,
their amplitudes (A) are determined in mill]volts and all amplitudes are corrected to this level. The
and durations (D) in seconds. Wave terminologT calibration on the tracing is shown in the com-
is in general conventional. It shouht be noted that puter printout as a percentage. For example, a
the initial negative wave of a QRS-complex is value of 105 means that the calibration on the
termed a Q-wave when snmll and an S-wave when recorder was 10.5 ram. An interpretation of
INTERPRETING ELECTI_.OCARDIOGRAMS FROM SPACE 93
under- or overstandardization is given when the electrocardiographic analysis shouhl be replaced

calibration is more than 25 percent above or by statistically significant electrocardiographic
below 1 cm/mV. values interrelated with statistically significant
Electrocardiographic interpretation--The inter- values from other signals. But today's physicians
pretation is based on wave measurements, axes, best understand significant change in a monitored
and the other findings that are summarized in ECG by empiric diagnostic statements based on
the code line, and on their interrelation. The the conventional diagnostic tracing.
interpretations are both descriptive and inter- Our computer measures all the ECG compo-
pretive. In the left column are the wave abnormal- nents of the cardiac cycle that are necessary for
ities found in the ECG; in the right column is the these statements. The verbal interpretation is
interpretation based on these abnormalities. based on interrelation of the nmneric,d values
The objectiveness of the system tins marked following the empiric basis used in medicine. The
advantages. First, human bias is eliminated. result is the same as if a cardiologist were per-
Second, since all output data are in digital form, sonally interpreting the ECG's. The advantages
they are immediately or subsequently available of computer analysis are that the automate(t
for display and clinical interrelation with other system works around the clock and provides pre-
events. Third, methods for predictive statistical cise measurements and printout interpretations
interpretation are established. at any desired time.
In one demonstration in 1965, 1500 conven- Table 1 lists verbal interpretations available in
tional resting ECG's were sent from Las Vegas, our current computer monitoring program. The
Nevada, to Washington, D.C., and returned. In exact criteria given are subject to change as
ninny eases the interpretations were available experience dictates. New diagnoses can be added
before the electrodes were removed from the to meet the physician's needs.
subjects. In a continuing demonstration project, In our initial on-line, real-time trials, the first
tile outpatient, clinic at Hartford Hospital, Hart- instances in which complete electrocardiographic
ford, Conn., daily sends tracings to Washington, data from any human were monitored and inter-
D.C., for analysis and return of results. These and preted on an on-line, real-time basis by a digital
other demonstrations have laid the foundation computer system, continuously monitored ECG's
for on-line, real-time computer analysis and were sampled. Analog signals from an astronaut
monitoring of any medical signals. in Gemini 7 were obtained in real time and
The objectivity of the available methods sug- intermittently monitored at 30-to-60-min inter-
gested their use not only as the basis for screening vats over a 2-week perio(l.
techniques to detect disease, but also for monitor- After that initial trial, the MSDL made im-
ing subjects in intensive-care units, triggering mediate, automatic, computer analyses of the
alarm systems or control servomechanisms to ECG's of astronauts in orbit in Gemini flights 8
start therapeutic measures if necessary, amt to 12. In that project, intended to test for feas-
evaluating subjects engaged in any activity such ibility and to demonstrate the capability of
as exercise, training, or space flight. The methods on-line, real-time analysis, the MSDL received
being developed can be adapted to computers continuous data. The signals were first trans-
aboard spacecraft, to computers in communica- mitted from the capsule to the nearest tracking
tions satellites for international medical purposes, station which telemetered them to Goddard
or to small "hospital size" computers. We present Space Flight Center; thence they were trans-
some samples of what is now being done and mitted by telephone as multiplexed analog signals
suggest what. should be expected from combina- to the computer center in Washington, D.C.
tions of techniques and displays. Four electrocardiographic leads, two from each
astronaut, were available, but, because of the
limited ability of the then-existing computer
DISPLAY METHODS
system, only two signals--one from each astro-
Continuous tabulation and verbal analysis--After naut-were analyzed immediately; the other two
further research, the conventional method of were recorded on tape for postflight analysis.
TABLE 1.--Electrocardiographic Diagnosis and Criteria Used to Establish Diagnosis
P-Wave Abnormalities
Tall P-waves; P-pulmonale PA, >__0.30 mV in any three time blocks
Short PR PR, :>0.1I see; rate, <:99 in four time blocks. Other two time
blocks if present must be <0.13 sec
Prolonged PR-intervals; first-degree atrioventric- PR, 0.21 to 0.29 sec; PA, >_0.05 mV in four time blocks
ular block
No atrial activity detected; rule out nodal rhythm, All P-waves =0 in at least five time blocks analyzed
atrial fibrillation
Conduction Defects
QRS-prolongation, 0.13 sec; intraventricular con- QRS-duration, >_0.13 see in any four time blocks; no value <0.16
duction defect see
QRS-prolongation; intraventricular conduction defect QRS-duration, >_0.14 sec in any three leads; no value <0.12 sec
ST-Segment Abnormalities
Abbreviations STO--amplitude of onset of ST-segment

STM--amplitude of midpoint of segment
STE--amplitude of end of ST-segment
Junctional ST-depression; rising ST-segment STO, >_-0.10 mV negative; STM >STO. T, positive; ST segment,
>_0.08
Minor ST-depression; flat or falling ST STO between -0.05 and -0.09 mV; STM>__STO (negative) in any
two time blocks. ST-segment, >_0.08
Moderate ST-depression; flat or falling ST-segment STO, >_-0.10 mV (negative); STM>STO (negative) in any two
time blocks. ST, >0.08 any two time blocks
ST-elevatlon; rule out early repolarization STO, >_0.08 in any four time blocks; or STO, >0.06 in any five
time blocks
Marked ST-displacement, 1 4-; possible current of STO, >0.20 mV positive or negative; or STO, STM, STE, >__0.12
injury mV positive or negative
Extreme ST-displacement STO, >__0.35 mV, positive or negative, in any two time blocks
T-Wave Abnormalities
Negative T-waves, -0.10 to -0.49 mV TA, -0.10 to -0.49 mV; QRS peak-to-peak amplitude, >__0.51 mY;
in at least three time blocks
Negative T-waves, -0.50 to -0.99 mV TA, -0.50 to -0.99 mV; QRS peak-to-peak amplitude, >_0.51 mY;
in at least three time blocks. No values less negative than -0.25
mV
Negative T-waves, -1.0 to -1.49 mV TA, -1.0 to -1.49 mV; QRS peak-to-peak amplitude, >__0.51 mV; in
at least three time blocks. No values less negative than -0.25 mV
Negative T-waves, - 1.5 to - 1.99 mV TA, -1.5 to -1.99 mV; QRS peak-to-peak amplitude, >__0.51 mV; in
at least three time blocks. No values less negative than -0.50 mV
Nega_tive T-waves, -2.0 to 2.49 mV TA, -2.0 to -2.49 mV; QRS peak-to-peak amplitude, >__0.51 mV; in
at ]east three time blocks. No value less negative than -1.0 mV
Negative T-waves, -2.5 to -2.99 mV TA, -2.5 to -2.99 mV; QRS peak-to-peak amplitude, >__0.51 mV; in
at least three time blocks. No value less negative than -1.5 mY
INTERPRETING
ELECTROCARDIOGRAMS FROM SPACE 95
TABLE 1.--Electrocardiographic Diagnosis and Criteria Used to Establish Diagnosis---( Concluded)
T-Wave Abnormalities (Continued)
Negative T-waves, _-3.0 mV TA, >---3.0 mV (negative); QRS peak-to-peak amplitude, _0.51
mV; in at least three time blocks. No values less negative than
-2.0 mV
QT-Abnormalities
Prolonged QT QT, from <0.43 sec; TA, _0.10 mV positive or negative. Rate, >__61
in three time blocks
Short QT QT, -<0.29 sec. Rate, -<99 in at least two time blocks; must be
analyzed. If two blocks are less than 0.30 Dx not made
Voltage Abnormalities
Low voltage Peak-to-peak QRS-voltage, _<0.50 mV in all of at least four time

blocks analyzed
Defective data QRS-duration of zero in all time blocks suppresses other diagnoses
Rhythms
Ventrlcular rate, _>100 in three or more time Heart rate, _ 101 in three or more time blocks
blocks: tachycardia
Ventricular rate, <60 in three or more time blocks: Heart rate, _<59 in three or more time blocks
bradycardia
Variable RR-intervaI in four time blocks: rule out "A" code appears in "code line" in any five time blocks; the block
arrhythmia must not be low in voltage
Variable RR-interval in four time blocks; rule out "A" code appears in "code line" in any five time blocks; no block
artifact or atrial fibrillation can be low in voltage
Atypical QRS or artifacts in two or more time "E" appears in "code line" in any two time blocks
blocks: rule out premature contra_tions
Immediate eight-channel analysis should be artifact) only when sustained for defined time
achieved very soon. periods. Noise or artifacts occurring during a
Throughout Gemini 7, 215 separate samples specific electrocardiographic complex can also be
were received. The eleven 3.7-sec samples shown identified by similar means with adequate
in facsimile in figure 2 are typical of the teletyped recognition programs.
(TWX) output after on-line, real-time processing. Variable template--The first requirement of a
Our computerized monitoring program initially monitoring program, different from a time-limited
considers only pattern-recognition aspects of analysis, is establishment of criteria for change.
signal analysis and provides numerical values. To demonstrate what is meant, a facsimile of an
Where in the example in figure 2 significant values output display from one of our developmental
occur in the numerical values, a persistence of programs is shown in figure 3. At the onset of the
asterisks indicates those "abnormal" values. recording in figure 3 the subject's heart rate was
Asterisks identify values considered suspicious in 123; at the end, it was 144. The measurements of
normal clinical practice. In monitoring, these are PR, QRS, QT are followed by PA (P-amplitude),
important (i.e., not due to noise, movement, or PD (P-duration), STO (ST-segment onset), and
TRANSACTION NR 178
GET 282-5g-00
other electrocardiographic interpretations, noise,
014-2g-00
REV NR 177 or artifacts. From these items, indications of
GEMINI 7NAOI
CARNARVON trends to normality or abnormality can be given
INSTRi_MENTATION FIELD STATION--HEART DISEASE CONTROL PROGRAM according to clinical terms or criteria developed
COMPUTER MONITORED ELECTROCARDIOGRAM
........................................................................... for special needs. The output in figure 5 is not
GEMINI ASTRONAUT DA?A
........ !..... _....._....._...... }...... §.....Z....._..... L...!L..!! .... intended for graphic display in real clinical situa-
RA .00 .08 .09 .lO* -.04" .09 .07 -.OS* .07 .07 .06
PD .00 ,lO .07 .09 .04 .09 .07 .09 .07 .08 .09 tions; it is shown only to demonstrate output
P'A .00 .00 .00 ,00 ,06 .DO .00 ,ll .00 ,00 .DO
R'D .00 .00 .00 .00 .08 .00 .00 .09 .00 .00 .00
possibilities. Portions of the display, such as
RA .g5 .95 .89 .90 .91 .93 .95 .92 .84 .92 .gl
RD .06 ,06 .06 .06 .06 .06 .06 .06 .07 .06 .08 heart-rate or arrhythmia designations, are suit-
SA -.43 -.40 -.44 -,38 -.39 -.43 -.42 -.42 -.38 -.37 -.39
SD .05 .05 .05 .05 .04 .05 .04 .04 .04 .04 .04 able for physicians' monitoring needs in which
ST .07 .06 .08 .12 .07 .12 .07 .07 .05 .06 .12
STO -.05 .04 .00 .01 .01 .04 .00 .00 .00 .Ol .Ol some data-reduction technique nmst be intro-
STM -.02 .03 .02 .05 .04 .04 .01 .03 .Of .03 .07
ST[ .Of .06 .06 .24 .OS .27 .(]4 .04 .Of .05 .24 duced to relieve the responsible physician of
TA .32" .39 .37 ,37* .37 .39" .37 .40 .43 .39 .43
TD .15 .16 .IS .iS .15 .15 .09 .18 interpretation of vast amounts of continuous data.
T'A -.05 .00 .00 .00 .00 -.04 .00 .00 .00 .00 .00
T'D .08 .00 .00 .00 .00 .06 ,00 .00 .00 .00 .00 Other portions, such as the repetitive "no signifi-
;_-"_;_W--_;;:-U_-"7;---_T?5?---T?_;-5;;--U_---_i .... cant change" statement, can be used to trigger
QRS .ll .II .ll .ll .lO .ll .I0 .IO .ll .ID .12
QR_TE .41 .34 .3A .33 .33 .40 .34 ,33 ,37 .35 .37 alarms, lights, or other mechanisms. With the
85 82 93 82 82 Bl 78 85 79 85 75
........................................................................... analyzed data of measurements of P, QRS, etc.,
WITHIN NORMAL LIMITS
subservient-computer statistical routines can be

Fmm_E 2,--Facsimile of 11 samples of teletyped output.
used for evaluation of current results or with
previous data, or for analysis of portions of the
STM (ST-segment midpoint). "No significant data with greater scrutiny.
change" refers to the amplitudes and durations, Continuous analysis of signals can easily pro-
with the initial measurement sct as a base line duee a monumental pileup of data. The goal of
for comparison. data-reduction is to discard superfluous detail
Thus with this program the computer can make and to preserve meaningful information. A time-
statements about significant change from preced- varying template of mcasurcments is a simple,
ing beats including the presence of an arrhythmia, useful approach to solution of the problem. A
INSTRUMENTATION FIELD STATION--HEART DISEASE CONTROL PROGRAM

COMPUTER PROCESSED CO_INUOU8 ELECTROCARDIOGRAM
G_INI-TITAN 3
Measurements of initial time period

Rate PR QRS QT PA PD IRA RD STO STM TA TD
123 .10 .O6 030 .05 .06 0.31 .06 .Ok -0.02 0.06 .12
TIME -26.3 SECS IV.TE 12'2 NO SIGNIFICAYr CHANGE

TIME -22.6 SECS RATE 120 NO SIGNIFICANT CHANGE
TIME -18.9 SECS RATE ii0 NO SIGNIFICANT CHANGE
TIME -07.8 SEC8 RATE 132 NO SIGNIFICANT CHANGE
TIME -O4.1 SEC8 RATE ---
ARTIFACT RULE OUT ARRHYTHMIA
UNRECOGNIZABLE
TIME 03.3 SEOS RATE i_ NO SIGNIFICANT CHANGE
TIME 07.0 SECS RATE IM_ NO SIGNIFI_ CHANGE
TD4E 10.7 SECS RATE 140 NO SIGNIFICANT CHANGE
TIME lh._ SECS RATE 142 NO SIGNIFICANT CHANGE
TIME 18.1 SECS RATE 138 NO SIGNIFICANT CHANGE
TIME 21.8 sees RATE ---
MISSING DATA RULE OUT ECTOPIC BEAT
TIME 25.5 SECS RATE IM_ NO SIGNIFICANT CHANGE
FmVRE 3.--Facsimile of an output display.

INTERPRETING ELECTROCARDIOGRAMS F]tOM SPACE ,07
template of measurements can serve as a siandard TEMPLATE INVESTIGATIONS

to signal the occurrence of physiologically im- Continuously monitored ECG's taken during
portant change in wave form, but several tem- real surgery have been transmitted _wice weekly
plates should be available for use simultaneously for several months on a trial basis from the
as needed. operating suite of George Washington University
In figure 3 the measurements of the first beat Hospital to the MSDL for computer analysis and
served as the initial standard. If all subsequent statistical evaluation. The monitoring system at
measurements show no significant change, as in the hospital serves a cardiovascular operating
this case, only a statement to that effect appears room, seven general-purpose operating suites, a
on the printout after the measurements of the postanesthesia recovery room, and a special-care
first beat. If one or more criteria for significant unit. These locations contain only the remote
change are exceeded, the measurements of the monitors such as oscilloscopes and alarm systems,
"changed" beat can be displayed on the printout. while the signal-conditioning and recording equip-
These new measurements can serve as a new ment is in a central monitoring room; all signal-
template of measurement values. routing, recording, and display can be controlled
Safeguards are necessary to prevent introduc- by a nurse at this location.
tion of a faulty "standard" template. Measure- The data have been transmitted from the
ments other than the one(s) showing sustained operating room to the computer by applying the
significant change must be checked against the analog-ECG-amplifier output directly to a port-
previous template of measurements for transient able acoustically coupled dataphone (Bell Sys-
significant changes. In this way, short-lived tem-X603C) which fits over the transmitting end
alterations in measurement (physiologic or arti- of a regular telephone handset. The 603B data-
factual) do not affect the overall template but phone receiver at MSDL is then dialed directly
are still appropriately noted. For example, over the switched network system. The dataphone
ventricular-rate variation can occur, suggesting transmitter is basically a voltage-controlled,
artifact or arrhythmia, or ST-segment changes astable multivibrator operating at a center fre-
due to artifact, tachycardia, or myocardial injury. quency of 1988 Hz. The center frequency is
Whether the variation is transient or permanent, frequency-modulated by the ECG data with a
the cause is an important consideration before maximum deviation of +262 Hz. This is well
display to the monitoring physician. within the passband of the telephone line and
Unfortunately there are no established guide- can be heard as a high-pitched variable tone at
lines--nor even consensus--regarding the sig- the receiving terminal. Demodulation circuitry
nificance of wave changes during rest or activity. in the 603B receiver enables the original ECG to
In one of our efforts to develop tentative criteria, be recovered for presentation to the computer
a group of 2200 electrocardiographic computer- preprocessing system.
measured recordings (12-lead ECG's) were ana- Every 90 see the computer prints six 3.7-see
lyzed. These recordings, considered within normal time blocks of analyzed data processed during
limits by our electrocardiographic criteria, were that period. It. soon becomes quite evident that
derived from a population of ambulatory, free- a large mass of data is piling up; while this ac-
4
living, and thus presumably normal males. From cumulation may be desirable from a specific
these data we established the criteria (used in researcher's point of view, the operating-room
fig. 3) for what is significant change (table 2). team would be better served by data-reduction
TABLE 2._Critcria for Significant Change
Item P Q R S T Pig QItS QT
Duration, sec 0.04 0.02 0.02 0.02 0.06 0.06 0.04 0.08
Amplitude, mV 0.08 0.06 0.60 0.60 0.30

techniques that would focus on significant changes biologic and physiologic changes and some stress
or simply indicate no change. as well as measurement error; all these factors
The preliminary results of the anesthesia must be considered in review of any data. When
studies showed certain factors to be of interest in the variation exceeds these limits, the hypothesis
reference to template analyses. For example, is that close scrutiny of the subject is necessary.
amplitudes of wave forms have varied over These values are not out of line with the varia-
hourly periods by about 50 percent of the true tions noted between the 2000 subjects analyzed
values; duration values have varied by from 25 for the template-of-change analysis, previously
to 33 percent. It is of interest that sampling of a described. One advantage of computer measure-
subject as rarely as every 30 or 40 sec during ments is that they allow us to define inter- and
total anesthesia yields sufficient data for reason- intra-subject variability with a high degree of
able monitoring. Monitoring during exercise and accuracy.
stress is a separate problem now under investi- Specialized requirements--In certain instances a
gation. computer program (or subroutine) must be
The results of these studies confirm that any specially designed to cope with a situation in
template method depends on the criteria estab- which data-distortion may be expected because
lished for significant change. The criteria can be of imposed stresses, or because the subject is
based on analysis of any preselected subject popu- engaged in regular activities. Precise study of the
lation or time period; it must be tailor-made for' PR- and ST-segment regions, for example, is
the expected use. Very sensitive numerical necessary for optimal early detection of abnormal
criteria may indicate "change" so frequently thai, cardiovascular response to stress, or in coronary
the user is overwhelmed. Flexible criteria, easily care. Figure 3 is a facsimile of a printout of a
modifiable on request, are more likely to satisfy computer program that provides the necessary
his needs. detail in such circumstances.
Table 3 shows initial raw data from six subjects The printout of the computer-processed con-
monitored on-line in real time by computer from tinuous ECG (fig. 3) is for two 6-sec periods,
each of which is independently analyzed. Heart
a surgical suite; under anesthesia they underwent
rate, elapsed time from the beginning of the
uncomplicated surgery, and the data reflect 1
recording, and the particular complex analyzed
hour of each subject's surgery. Statistical tabula-
are identified; Q, R, and S amplitudes (A) and
tion of the on-line measurements was done off line.
duration (D) are calculated. Then follow the
The percentages are the degrees of variation detailed measurements of P-R and ST-T seg-
from the means during the 1-hour monitoring. ments. The calibration signal and a portion of
That is, if the mean duration of a wave was 0.0.i the analog record corresponding to each time
sec, and if it varied by 0.01 sec, the variation is interval also are sho_m; the complex selected for
expressed as 25 percent. The variation reflects analysis is identified.
TABLE 3.--Approximate Mean Values and Variations in Parameter Measurement for Six Subjects
Parameter Mean Variation, % Parameter Mean Variation, %
PA 0.14 66 STO -0.04 150
PD 0.09 33 STM -0.02 200 +
QA -0.06 66 TA 0.13 66
RA 0.58 33 TD 0.16 25
RD 0.05 20 QT 0.37 15
SA -0.19 50 QRS 0.09 20

SD 0.04 25 RATE 75 23
INTERPRETING ELECTROCARDIOGRAMS FROM SPACE 99
There is considerable interest today in electro- smaller interval is less than 75 percent of the
cardiographic monitoring for diagnostic testing larger. If there are not two consecutive regular
during exercise, for physical-fitness studies at R-R-intervals, the computer printout displays
work-evaluation centers, and for management of the word "arrhythmia" for that time block, and
acutely ill patients in coronary-care units. These analysis of the next time period begins.
activities place a great burden on electrocardiog- If an arrhythmia is absent, the program searches
raphers; in response to their needs, electrocardio- by means of a slope check for a region free of
graphic telemetry equipment and monitoring base-line shift or noise. This slope-check procedure
systems have developed rapidly. is applied to paired heart beats in every 6-see
Most current monitoring systems employ time block until a pair is located that fulfills the
analog computers which are generally limited to criteria for base-line constancy and artifact-
tracking of changes in rate and rhythm of the exclusion (fig. 4). The program identifies the
heart. But during activity the electrocardio- QRS-onsets of two adjacent cardiac cycles; the
graphic record may include shifting of the base line joining these two points is defined as the
line, noise artifacts, superposition of wave forms, indicator of base-line slope for these two heart
change in heart rate over a wide range, abnormal beats. If lines joining other corresponding points
conduction, and arrhythmias. Yet the finding elsewhere on the cardiac cycle are found parallel
that gives the exercise ECG its greatest clinical to the indicator line, the base-line slope must be
value, the ischemia-induced wave-form altera- constant for the total duration of these two heart
tions, must be detected and distinguished among beats. Since time intervals are constant, slopes
the other events. This cannot be accomplished may be compared simply by comparison of
with existing analog equipment. amplitude differences. The computer program
In our test program, the QRS-complexes in the uses the amplitude difference between QRS-onsets
4 sec of data are identified by use of minimum as the reference for base-line slope. Since physio-
derivatives, and the R-R-interval between con- logic beat-to-beat variation often prevents perfect
secutive beats is measured for detection of slope constancy, some discrepancy with the
cardiac irregularity. A cardiac irregularity is reference amplitude is allowed.
present if in two consecutive R-R-intervals the The slope-check procedure eliminates data
R-R INTERVAL JREFERENCE

• JAMPLITUDE
R-R INTERVAL
FmvP.z 4.--Application of slope-check procedure.

containing excessive base-line fluctuation or or reduction of noise in electrocardiographic

transient artifact, and also rejects a region from processing. The ECG is only an index to the
further analysis if there is disparity with the electrical activity of the heart,. Our understanding
reference amplitude beyond the established limits. of the physiology involved in production of the
If the slope check fails to find a satisfactory signal is limited. It is well known that our re-
region, the statement "artifact obscures" is cording and display devices are often poorly
printed for that. time block. standardized determinants in the representation
Once a pair of "matched" heart beats are of the activity observed. Therefore distinction
located, the program proceeds with a measure- between what is signal and what is noise is
ments routine. All vMues along the line joining necessarily poor.
QRS-onsets are computed from the R-R-interval The physician attempts to make the distinction
and the amplitude difference between QRS-onsets. by relating a particular segment of the signal in
Amplitudes are measured as vertical distances question to adjacent segments of the curve in
from the points on the curve to the indicator line. that eyele, and also to corresponding segments
This technique corrects amplitudes from a sloping in adjacent cycles. He also uses the relations
base line to a horizontal line; in this way ampli- compiled from previous viewing of other curves
tudes along the P-R and ST-T segments are as well as his knowledge of the physiology re-
measured relatively to the QRS-onset as the sponsible for the signal. Distinctions are essen-
reference level. Commencing with S-T-onset, tially based on a subjective consideration of
amplitudes are measured every 20 msee for a probabilities and may not be accurate; for this
time equal to half the R-R-interval. Four ampli- reason we have suggested increased emphasis on
tudes at 20-msec intervals preceding QRS-onset statistical considerations.
are measured on the P-R-segment. These meas- A point in the cycle is judged to be noise if it
urements and Q, R, and S measurements are represents an abrupt amplitude difference from
printed along with the time elapsed from the adjacent points in the cycle, if it is not repeated
start of the recording (fig. 3). (within limits) in a similar position in adjacent
cycles, and if it creates a pattern for the cycle
that is incompatible with the physician's pre-
NOISE
conception of a pattern.
The great noise during flight is perhaps the For purposes of discussion, noise in electro-
only differentiating feature between clinic and physiologic signals may be categorized into
space in requirements for ECG monitoring. base-line shift (low-frequency, cyclic, or non-
Typical clinical noise is shown in figure 5. The, cyclic), eyelie high-frequency (e.g., 60 Hz),
amplified version shows aetual content; conven-. relatively isolated spikes, and random noise
tionally in medical wards the pen recorder varying in frequency. Each of these types may
eliminates even large amounts. exist with various average peak-to-peak ampli-
Special problems exist in extraction of mean-. tude values. In addition, in a condition commonly
ingful data from ECG's monitored during stress, observed in transitional leads, there are marked
exercise, or certain other conditions. In electro.. changes in waves or segments from beat to beat
cardiography (luring exercise we have seen a with no predominant pattern. It is not clear in a
trend toward continuous recording during the given recording whether this condition is due to
exercise period and increase in the severity o[" true variation in the signal or to recording at a
exercise. Although noise can be reduced to some particular site.
extent by proper application of electrodes and Several techniques may be applicable to these
grounding, extraction of representative data from problems, either singly or in combination, that
the monitored ECG remains an ever-presen_ have been or could be used in both the exercise
problem both for the reading physician and for and NASA's ECG program. As already men-
a eomputcr-analysis program. tioned we have used two computer routines
Several points need emphasis before discussion identified as the amplitude-agreement check
of techniques for cxtraction of meaningful data (parallel check) and the time-interval-agreement
INTERPRETING ELECTROCARDIOGRAMS FROM SPACE I0i
FI(_URE5.--Typical clinical noise.
check; the others are (1) analog filtering, (2) frequency noise while avoiding, obscuring, or
digital-computer smoothing, (3) analog-signal falsely producing diagnostically significant tran-
averaging, and (4) digital-computer averaging. sients raises an important research question.
These different programs should be evaluated by There is still no clear definition of a significant
their utility in extraction of analyzable data from transient. The effectiveness of lowering the high-
noisy tracings with minimal distortion of signals. frequency limit of a filter for production of
A_mlogfiltering--Analog filtering is the simplest readable ECG's is also unclear.
technique for reduction of noise, but it may distort Our experience with analog filtering of the
good=quality data by obscuring or falsely produc- exercise ECG, with a bandpass filter _dlowing
ing significant wave forms when applied indis- frequencies between 0.02 and 45 Hz, indicates
criminately to the signal. The effect of filtering is, that noise remains that complicates our pattern-
among other things, a function of wave-form, recognition by computer. Many of these records
heart-rate, and filter characteristics. There is dis- are judged readable by manual methods. It is of
tortion of the ST-T segment with low-frequency interest to know (1) whether lowering of the
cutoffs higher than 0.05 Hz (at 6 db per octave). limit further (to about 25 Hz) would be effective
The proper high-frequency cutoff for exercise in producing readable data, and (2) what distor-
ECG's is less apparent. To show notching and tion of good-quality datla would accrue at these
slurring, high-frequency recording fidelity is more settings.
likely to be useful in the resting than in the Digital-computer smoothing--Several available
exercise ECG. At present no practical importance mathematical techniques may be valuable either
can be ascribed to such data. Lowering the high- alone or in conjunction with other noise-reduction
102 BIOMEDICAL RESEARCH A.ND COMPUTER APPLICATION
techniques for data. If these techniques are should be evaluated in terms of producing some
applied indiscriminately to the signal, there is a percentage change in the parameters measured,
risk that diagnostically significant characteristics changing of the signal on interpretation from one
of the signal may be either obscured or falsely diagnostic category to another, and the effect on
produced. a record's percentage of data deemed processible
Morphology of the wave forms, heart rate, the by the computer program.
type of smoothing technique used, and the number Analog signal-averaging--Averaging with a
of points involved in the smoothing of a single small analog or digital computer can be used.
point are factors that affect both these desirable Its limitations are inherent in its continuous
and undesirable tendencies. Three of these tech- effect, inability to deal with nonrandom noise,
niques-the moving-point average, the medium- averaging of homogeneous inhomogeneous tran-
point method, and the parabolic fit by the least- sients, and obscuring of diagnostically significant
squares method--have been used by us: transients. Some of these problems are related to
experimentally with the exercise-ECG program. the choice of a fiducial point in the cycle for use
Programs have been written in such a way that in triggering, which is necessary for superimposi-
the number of points used for each technique tion of complexes as is (tone with this technique.
may be varied. The amplitude threshold may shift in time with
Initial experience indicates that the medium- base-line shift, thereby effecting a requirement
point method works best on data that are distorted for a large number of complexes to be averaged
by large spikes, of relatively short duration, in order to "mean out" the contribution by the
separated by periods of relatively little noise. base-line shift to the average. Slope triggering has
Its drawback is the great tendency to eliminate been suggested, but we find no reference to ex-
important peaks and nadirs in the data. perience with this in analog averaging. Change in
The moving-point average is best suited to rate (luring the averaging period can also be a
damping of cyclic noise of high frequency as well problem.
as random noise of low average amplitude. It Digital-computer averaging--Averaging with a
reduces the amplitude of such important peaks digital computer has an advantage in that it can
and nadirs, but not so readily as does the medium- be applied to the signal only when the program
point technique. determines that the data are sufficiently poor in
By the least-squares method, the parabolic fit quality to require it. It is possible that the risk
can decrease the amplitude of low-amplitude, of obscuring diagnostically significant transients
high-frequency noise while preserving large- may be partly avoided by reduction in the num-
amplitude peaks and nadirs, but it has a tend- ber of complexes required for averaging; we have
ency to add to peaks if their amplitudes are tried this.
great. It is less effective than a moving-point The Q-onsets are identified as well as possible
average for elimination of high-frequency noise in noisy data, and each complex is corrected to
in low-frequency areas, such as ST-T regions, the horizontal by use of these Q-onsets even if
when the same number of points are used fc,r they are located somewhat in error. Probably it
each technique; another drawback is the time will be necessary to reduce the change and mag-
required for proce._ing. nitude of the error in location of the Q-wave onset
Some of the disadvantages of these programmed by refining the region of search for its location.
techniques can be overcome if they are applied Perhaps this may be done by examining a small
only when the data are automatically judged to region about the minimum derivative and identi-
be of poor quality, and only in specific areas in fying the maximum positive derivative in this
the cycle that are best suited to use of the par- band. A longer region of search would be necessary
ticular technique. Further information is needed for instances in which the maximum positive
concerning the effects of these techniques on derivative was located to the left of the minimum
various regions of the cycle, such as PPR, QPh_, derivative and vice versa. After base-line correc-
and STT regions, when rates and morphologies tion is accomplished, complexes can be super-
in these three regions are varied. The techniques imposed by alignment of minimum derivatives.
INTERPRETING _LECTROCARDIOGRAMS FROM SPACE 103
The sampling of a complex to be averaged mission to Early Bird which resides in a syn-
would end at a point located after the minimum chronous orbit 22 236 miles from Earth. The
derivative, at a distance of 75 percent of the signal was rctransmitted to the receiving station
interval between minimum derivatives. A point at Andover, Maine. At that point the original
located before the minimum derivative, at 50 frequency-modulated 2-kHz tone of the acoustic
percent of the distance between adjacent mini- coupler was recovered and placed on the private
mum derivatives, would limit the sampling region telephone lines of RCA Communications for
at the beginning of the complex. It will be noted routing to Washington. When the signal reached
that with this method there would be some over- Washington, it was transferred to domestic lines
lapping of data, with the same data appearing on for transmission to the Bell System 603B data-
both ends of the complex being averaged. Data phone receiver at the MSDL.
will be eliminated from the average, from both The signal was filtered, amplified, and con-
ends of the averaged complex, beyond the point verted from analog to digital form at a rate of
at which the shortest cycle corresponds in time 500 samples per second for presentation to our
to the other cycles. The effect of base-line shift digital computer (CDC 160-A) for measurement
may be minimized by this method, but the and analysis of the various amplitudes and dura-
method shares other disadvantages previously tions. The computer produced punched paper
mentioned of the analog averaging technique. tape that was fed into a teletypewriter (Bell
This technique, however, has an advantage System model-35 ASR) for transmission to New
over filtering and smoothing techniques when York where it was necessary to convert from
there is large beat-to-beat variation or when the domestic eight-level code to five-level Inter-
noise-deflection durations approximate the length national Telex code. The speeds of the machines
of recognizable segments and waves of interest. associated with these codes are 100 and 60 words
Both the smoothing and filtering techniques re- per minute, respectively. The information loop
duce noise by relating displaced points to other was closed when the ECG interpretation was re-
points in the region, and cannot be expected to be turned to Tours via the Early Bird satellite and
effective with problems such as beat-to-beat Telex lines.
variation in which all points in a region or seg- The quality of the signals received in Washing-
ment may be displaced. Averaging has an ad- ton was very good. The only noise showing on
vantage in this situation since with it a "displace" the paper tracings was that due to occasional
point is related to a corresponding point in other muscle movement or electrode slippage. The first
adjacent complexes. attempt at transmission was greatly affected by
Noise and satellite-data transmission--To dem- echoes; at one time 12 echoes were audible after
onstrate the feasibility of sending signals over a spoken word. This fault was soon corrected by
very long distances for on-line computer analysis, insertion of echo-suppression devices in the line
with subsequent return of the interpretation, a at the RCA terminal. During transmission it was
trial communications system was set up between also necessary to disconnect the telephone receiver
Tours, France, and MSDL in Washington, D.C., in France because the original transmitted signal
making use of the Early Bird satellite. Both was returning, after a round trip of approximately
12-lead resting and continuously monitored ECG's 0.6 sec, completely out of phase with the real-time
were transmitted for almost 2 hours daily for 5 signal. In addition to the cross talk created on
days (3-8 July 1967) with signals comparable the lines, the signal was being acoustically
in quality to local transmissions. coupled back into the transmission circuit.
The ECG signal was conditioned with a port- Insulation of the acoustic coupler from room
able ECG-encoder for patient identification and sounds eliminated the ambient noise. No diffi-
acoustically coupled to the French telephone net- culty was experienced with the teletypewriter
work for transmission to the Comsat transmitter interface, and the transition from 100 to 60 words
at Pleumeur Bodou, France. The acoustically per minute was smooth.
coupled FM signal was then used to frequency- Such a system is obviously expensive, but in
modulate a Comsat carrier frequency for trans- special situations the returns would be com-
mensurate. First is the use of satellite transmission would then be useful to compare quantitatively
for monitoring of astronauts. Underdeveloped the responses of new astronauts in training with
nations, lacking funds or technology, might benefit those of astronauts in flight. One result would be
from the far-reaching implications of such a a more quantitative basis for selection of trainees.
system to their national health. Even more During flight the results for the totM period
obvious is the potential of medical-signal analysis from one tracking station could be used to estab-
or storage centers within countries like the United lish a trend during the flight for contrast with
States with existing communications facilities. preflight information obtained from test chambers.
On-board ECG tapes or tracking-station record-
ings eould be evaluated for statistically mean-
PREDICTIVE VALUES
ingful data relating measured parameters to
Rapid statistical analysis is possible during space-flight activities. From the research view-
monitoring if data are in digital form. From such point we could quantitatively study changes
data, statistical displays can be obtained on line associated with weightlessness, particularly during
and in real time to provide, for example, a stand- ascent, orbiting, space walks, reentry, and re-
ard score value for each parameter (or a multi-. covery-in fact during any conditions that
dimensional score for all parameters) or a status cannot be tested on Earth's surface.
report on a subject, for comparison with his past To demonstrate such application of numerical
results (or for his comparison with a specified wLlues, data have been compiled from on-line
population group). By these and simiIar measures, computer measurements of the pilots in G-emini
a subject's profile can be quantitatively displayed. flights 9 to 12. The easiest parameter to under-
Statistical tabulation and analysis of data are stand is heart rate; it may not be the best from
intended to provide insight into the general, which to draw conclusions about, status of the
central, or predictive trends of the parameters subject, but it is a reasonable one to consider first..
being measured. Heart-rate change can show how Figure 6 shows the heart rates of subjects during
data can be used for predictive purposes. The different flights. Compare, for example, Gemini 11
heart rate at rest is generally used as the fidueial with flight 12 or 10. On flight 11, during the upper
point in stress, exercise, or continuous monitoring, quartile of the time, both pilots had heart rates of
or simply for pulse-determinations; the generally about 90 to 100; this did not occur in the other
accepted normal range is from 60 to 100 beats flights. Their average heart rates also were higher
per minute. than those of pilots during other flights.
Analysis of 27 000 computer-measured ECG's The heart rate was therefore variable, depend-
has allowed a better distinction of usual heart ing on the number of the flight (i.e., problems) as
rates and those that could be beyond usual well as on each individual. At the 75-percentile
limits. Our data make it apparent that two and the 50-percentile levels, for example, either
considerations are necessary for judgment of flight 10 was less stressful or perhaps the pilots
heart, rate. First, a subject's rate should be were better trained than for flight 11. Whether
within certain defined limits when at. rest. The the variation is due to function._l activity, indi-
second consideration is of particular interest; our vidual physiologic response, or undetermined
large accumulation of data has allowed insigt:t causes, it is important that. levels be determined
into variation between individuals and between
in training and in flight, for each individual pilot
disease groups even during rest. Thus an indi-
or mission, to form the basis of a template for
vidual's variation in rate must also not exceed
comparisons of one pilot with others. Subsequently
the variability within his specific category of
the levels eouhl serve to establish standards for
physical fitness or disease, or that of the "_iormal"
pilots in varying types of flight.
population's distribution.
Application of _umerical values---If the values As we have mentioned, heart rate is only one
of several astronauts on several different flights of the parameters of a subject. The availability
were surveyed, different functional time periods or of other computer-derived data nmkes of para-
different types of stress would be studied. It mount importance definition of the quantitative
Percentile • 75
• 50
Beats per • 25
minute
]00
9O
8O
7O
60
9 B A B A B A B
I0 ii 12
FIGURE 6.---Selected percentile of heart rate.
usefulness of those data in determination of the centage of QT's, }lad they been available for all
status of the subject. previous flights as for 9 to 12 in figure 9, would
Obviously many limitations are present in have shown by how much the subject deviated
these crude graphs since for this analysis we are in performance from all previous flights, preflight
not considering functions being performed, time experiences, early flight stages, or from all other
of day, etc. We are just taking data computed for astronauts. Contrasts can also be made with the
the total flight, but these are useful in showing general population or other groups. This illus-
that a trainee's reactions could be contrasted trates the fact that instantaneous, multidimen-
with this total experience. It is also useful to sional, on-line, statistical analysis can give
consider that each pilot in each subsequent flight useful indications of each subject's performance.
could be contrasted in real time with the total Another parameter studied was QRS-amplitude
experience of all previous flights. The difference which is complicated by the fact that NASA's
between one pilot's experience and the total ex- electrode placements tend to cause rapid changes
perience or that of a similar series of flights--the in polarity; but, even with this artifact, differ-
trend to or away from the total experience--might ences existed between flights 9 and 12 and 10 and
be significant (fig. 7). 11 (fig. 10). No conclusions can be drawn from
The data in reference to other measures may
these data; justification for further study is
be of equal importance. Consider the QT-duration
apparent. The onset of the ST-segment shows
shown in figure 8. Although there is similarity
similarity among subjects 10a (high quartile) and
between various subjects, we could say that sub-
ject B on flight 11 had a lower QT than others. 12a (low quartile) that may be due largely to
We could compare that fact with heart rate computer-measurement and telemetric problems,
(fig. 6) and note that the heart rate was higher. but this otherwise suggests that a range of
This subject could be assumed to be different ST-onset can be defined. Under known conditions
because of his tasks, physical attributes, or en- a subject should not vary too much from this
vironmentM circumstances. The cumulative per- range (fig. 11). The STM (midpoint of the ST)
rate
cumulative%
]00 .,L
80 popuh __
60 /s --
2O
0
60 70 80 90 100 110 120 ]30 140 150
heart rate (beats per minute)
FIGURE 7.--Cumulative percentage distribution of heart rate.
Percentile • 75
I 50
Second • 25
A8
.44 __ w _
.40 ....
36
.32
.28
A 9 AIoB A ii B AI2B
FIGURE 8.--Selected percentiles of QT.
was beset by noise in our signals, but with more useful in space medicine and warrants special
data its range could become significant. techniques for noise reduction on line.
The PR-interval is an area that shows wide It is apparent that functions of the individual
variation (fig. 12). This was the area of greatest and noise in the signal interfere with the statis-
noise in the NASA tracings, as received by us, so tical analysis to be used. The noise has been
that no conclusion can be made. Since this area is indicated to be of various types, and suggestions
an early predictor in clinical medicine, it could be for certain of the areas of improvement have
QT
cumulative %
I00
80__ : ! Po_latl, m ,j__ ....

i
_'_ _ ilot.
6O
/dr
/.,:/'
40__
20__
. un,:J i--
c) .24 .2e .3z ._s .40 .4.4 .4e .52
FmuRx 9.--Cumulative percentage distribution of QT.
Percentile • 75
• 50
second A 25
•24 =
.2o
.16
.12
.o8
.o4
A 9 B AIoB AIIB A 12 B
FIGURS 10.--Selected percentiles of QRS.
been noted. In our experimental transmission

by satellite rather than by systems currently in
from Tours, the signals during activity were re- use. The suggestion is that this change should be
markably free of noise. It would appear then made as rapidly as possible so that the statistical
that much of the problem of noise can be solved
techniques can be used and full advantage taken
Percentile • 75
• 50
• 25
.20 __
.15
.10
05
.00 --
-D5
-.10
-.15
!
:20'
-.25 --
-,30
A 9 B AIoB A ii B A 12B
FIQURE11.--Selected percentiles of ST-onset.
of the possibility of measurements by computer. (1) Empirically there is no way to analyze all
There are several preliminary recommendations the values of amplitude and duration of the con-
on the basis of the Gemini flight data. First, there tinuous electrocardiographic wave forms. Because
is need to study preflight training information of the nature of biologic variability, spot checks
and to compare it with general-population and or sampling methods of interpretation are also
other data to determine wherein special templates, generally unacceptable. Thus only crude measures
if any, are required for astronauts. This recom- of heart rate, heart rhythm, and an occasional
mendation extends to all students of special wave form are possible by conventional methods.
groups such as those in intensive-care suites. (2) When the data are finally available to the
Second, it would be reasonable to determine monitoring personnel, it is usually long after the
better ways of pilot-selection by means of quanti- fact. Thus a situation exists in which the informa-
tative data. Third, and perhaps most important, tion reaching the physician is either insufficient
the functional capability of pilots could be com- or too late.
plemented by knowledge of each other's reactions, (3) Little effort has gone into study of evalua-
in terms of quantitative variables, under various tive procedures.
conditions of stress and activity. One possible procedure to overcome these
Statistical potential--Continuous monitoring of limitations is establishment of statistical com-
the ECG is needed to provide useful clinical, puter programs that can be used to "format"
physiologic, and epidemiologic data, during exer- displays of immediate verbal statements classify-
cise or stress tests, in the management of patients ing the measurement patterns into clinically
in intensive-care units, in operating and recovery diagnostic or significant categories. These pro-
rooms, and in the routine follow-up evaluation of grams can take the electrocardiographic measure-
heart patients. With the present empiric system ments for selected time segments and relate them
of monitoring, the ECG cannot be fully used for to previous data.
prompt detection of early changes of significant The purpose of this section is to present as an
magnitude. Three factors are responsible for this: example a statistical method for determining
Percentile • 75
• 50
• 25
.24
.22
.20 ....
.m J
.16
.14
,d
.12
A 12 B
A A B A B
9 B 10 11
FIGURE 12.--Selected percentiles of PR-intervals.
significant change in the continuous monitoring measurements total at least 3002 compared to
of the ECG. The method selected is Hotelling's only about 100 possible diagnostic-statement
T2-multivariate technique; it requires only the categories. This implies that in fact only a small
extraction of information from any large pool of fraction of all possible information is being used
data such as that derived from our ECG com- for current diagnoses. One reason that can be
puter program. advanced for such nonutilization of all the data
The conventional ECG is a 12-1end signal with is that no convenient method with easy calcula-
approximately 25 measurements of amplitude and tion exists for extraction of the additional in-
duration per lead; the signal is processed by com- formation. It is in this area of electrocardiography
puter to yield amplitude (A) and duration (D) that its future utility may lie.
of individual waves. The time axis is the inde- Statistical-model projecl--We have selected elec-
pendent parameter; amplitude or height, the trocardiographic measures for input to a T 2
ordinate of the curves, is the dependent variable. program that uses these data to build a pool of
The 12-lead ECG can be represented as a predata estimates. From the predata one com-
column vector of 300 variables. Information is putes a premean column vector and a prevariance-
potentially present in each of the individual covariance matrix. If one assumes that the
variables and also in combinations of variables electrocardiographic signals are jointly distributed
from crossing leads, such as lead-1 and lead-3. as multivariate normal with a mean (_) and co-
Many of the diagnoses are based on information variance sigma, then the quantity
from both sources--individuaI variables and
N(_- _,)'S-'(_- u)
combination variables--but all possible crossover,
intra-inter, lead configurations of parametric is distributed as T 2 where N is the number of
samples used to accumulate the prepool from tained from 3.7-sec periods of continuous electro-
which _" and S were calculated. For any post- cardiographic signal, so that we have simulated a
sample pool of size N2, the quantity T _ evaluation of 10 separate column vectors of
electrocardiographic data on a continuous-moni-
N1N2
-- (_,-_2)'s-'(_1-_2) toring basis (table 4).
NI+N2
is distributed as T 2 with Nl+N2--2 degrees of TABLE 4.--Classification of 10 Patients by the

freedom where S is now the pooled eovariance T2-Technique
matrix for both states, pre and post. Normal limits
For any signal taken during an operation, a Cla._sification
Within Beyond
flight in space, or in a recovery room, a statistical
comparison can be made with the prepool data
Correct 6 2
by substituting for 22 in the expression with
Incorrect 1 1
N2=l. That is, for any electrocardiographic
signal, xo, one computes
N1 Of the three ECG's that were not within

-- (_1- xo)'S-'(_1- xo)
N1+1 normal limits, two had significant T _ values or
correct classification. Of the seven ECG's within
The T 2 value can be determined for each desig- normal limits, six were correctly cla_ssified.
nated block of time in the post period. For any Overall the T _ test showed 80-percent agreement
level alpha (a), the critical region for T _ is with the standard 12-lead ECG.
(NI+N2-- 2)p The T 2 test used 20 electrocardiographic meas-

T___> Fp.N,+N_-_-I (o0 urements for evaluation of the continuous ECG.
Nl--k N2--p--1
The measurements were few relative to the num-
where p is the number of parameters considered ber used in the standard 12-lead ECG for evalua-
in the electrocardiographic analysis. If the T _ tion. It is of interest to compare these findings
value exceeds the critical value for the 5-percent with the findings of five independent readers of
level of significance, the statement is made that the electrocardiographic tracings from which the
the current sample is significantly different from 20 measurements were taken (table 5).
the prepool estimate. To the physician, nurse, or Of the six ECG's within normal limits by both
ground crew monitoring the subject's ECG, a the standard 12-lead ECG and the T 2 test, all
significant T'- value indicates that the signal is readers agreed with this evaluation on four. At
statistically different from the electrocardio- least two readers judged the remaining ECG's
graphic prior data. The relation of statistical sig- beyond normal limits. Of the ECG's beyond
nificance must of course be determined by ex- normal limits by both methods of evaluation, all
periment al studies. readers agreed with both methods on one; at
Resti'_g electrocardiogram--Twenty electrocardi- least one reader similarly agreed regarding the
ographic measurements from 59 subjects were other.
used for the predata estimate. The ECG's were Regarding two ECG's there was disagreement
diagnosed as within normal limits both by the between the standard 12-lead ECG and the T 2
computer program and by two physicians who test. On one ECG, at least one reader agreed
used the standard 12-lead ECG as the source of with the standard method. On the remaining
information. Twenty electrocardiographic meas- ECG, at least one reader agreed with the T 2 test.
urements from each of 10 subjects, seven within In summary, at least one of the readers, using the
normal limits and three beyond, were then same limb leads from which the electrocardio-
compared with the prior-population fgures. The graphic measurements were taken for the T 2 test,
measurements of each of the 10 subjects, known agreed with the statistical method in each c_se.
to be either within or beyond normal limits on Exercise ECG---Electrocardiograms were ob-
the basis of standard 12-lead ECG's, were obtained from one subject at rest, during exercise,
INTERPRETING
ELECTROCARDIOGRAMS
FROM
SPACE 111
TABLE
5.--Electrocardiographic
Findings by Five Readers for Subjects on Which Standard 12-Lead
ECG'S and T _ Test Agree and Disagree in Evaluation
ECG and T _
Agreement Disagreement
Finding
Within Beyond ECG T 2 normal ;
normal normal normal; T 2 ECG
limits limits abnormal abnormal
All readers agreed

Within normal limits 4 0 0 0
Beyond normal limits 0 1 0 0
At least one reader agreed with T_

2 1 I 1
and after exercise (the recovery period). Twenty- book would relate them. This method considers
two column vectors of electrocardiographic data magnitude and direction of the wave form; the
were recorded from the subject seated at rest on interpretation of specific patterns is essentially
a bicycle; after exercising long enough for his empirical, resulting from clinical and autopsy
heart rate to reach 150 beats per minute, he association. These two approaches are commonly
rested. Electrocardiographic recordings continued used in clinical practice. A third approach, that
until the subject's heart rate returned to within of classic statistics (as described), is by statistical
10 beats per minute of his resting rate. There study of the joint probability distribution of the
were four blocks of exercise recording and 27 electrocardiographic observations. The distribu-
blocks of recovery-time recording. tion in a prior population is described by means,
For the T 2 test, the 22 time blocks of resting variances, and covarianees and related to observa-
ECG's were used as the predata estimate; each tions in succeeding time blocks of electrocardio-
block consisted of 20 electrocardiographic meas- graphic data.
ures of interest to the physician. The variance- It is apparent from the review of the limited
covariance matrix and the vector means were quantity of electrocardiographic data (20 vari-
calculated from the 22 time blocks of data. The ables) that not all data usually available to the
exercise and recovery column vectors were com- electrocardiographer, from his inspection of 12-
pared with the predata estimate. lead standard ECG's, were used in the T 2 test.
The T _ value was significant for the first, third, We were limited to 20 variables because of the
and fourth exercise time blocks, but not for the computer's limited storage. The absence of an
second. The first block of the recovery period was on-line computer program for automatic retrieval
significant; the remaining periods were not. The of electrocardiographic data also limited the
program could demonstrate at what time a sub- number of samples for building of a prior matrix
ject returned to resting level and differentiate the of measurements.
exercise period from the recovery period. We have not yet demonstrated to our satisfac-
Perspective--There are many possible ways to tion that smaller quantities of data than are
monitor the measurements of variables from a generally used are all that are needed to minimize
continuous electrocardiographic wave form. One false positives and false negatives in successive
approach is by detailed comparison of these with time blocks of observed data. The monitoring
measurements that have been considered stand- system, however, fills the need and offers many
ard. Another approach is to formulate values to advantages over the present means for analysis
relate measurements in a defined way similar to of the monitored ECG. The method reduces a
that in which an electrocardiographer or a text mass of data and provides the physician with
]]2 BIOMEDICAL RESEARCH AND COMPUTER APPLICATION
clinical information about important wave-form of medical data. A computer can be programmed
changes; results are stored and retrieved easily to accept, through standard commercial input-
for rapid statistical analysis. The method achieves output devices, the routine data that the nurse
its purpose by supplementing clinical judgment. and physician usually enter in the chart. Data
The additional information gained from specific from the clinical laboratory also can be entered.
time periods aids final evaluation of the ECG. This computer can be linked to another that is
A statistical technique combined with a high- able to analyze medical signals such as ECG's
speed computer provides a promising method for and EEGrs in real time. This system may be
evaluation of continuous monitoring of the ECG. applied to any intensive-care situation such as in
This combination offers the physician supple- recovery rooms and intensive-care units for stroke.
mentary clinical information to improve the The complex interrelations between clinical
diagnostic process. data, transducer data, and laboratory data can be
evaluated by the computer at each new data
entry. All data stored for defined past periods can
AUTOMATIC CARE AND EVALUATION UNITS
be reexamined in the light of the new data, and
There are many clinical spin-offs from this the computer can reduce the data to an English
project. The principles and techniques used in statement such as "stable," "hypokalemia by
monitoring astronauts in flight can be used ECG-eheck blood chemistry," or "impending
effectively for patients in a coronary-care unit or shock." This information can be immediately
undergoing surgery. Indeed our preliminary tests available on a screen in the CCU.
indicate that both these uses are entirely feasible The promptness of therapy after onset of
even though the techniques must be adapted to arrhythmias, cardiac arrest, and shock has been
a hospital environment. The first project to put shown to be directly associated with therapeutic
these monitoring techniques to clinical use is in success. This system allows rapid interpretation
the testing stage in an operating room and the of these states even when no physician is present,
coronary-care unit (CCU) at George Washington shortening the time lag between interpretation
University Hospital. and initiation of therapy.
The nurse and physician in the CCU must now Furthermore the system will perform the
evaluate the outputs of multiple electronic nurse's routine evaluation, and charting of
monitors, as well as frequent recordings of pul_, medical data will supply her more readily with
blood pressure, respiration, temperature, urine additional information when required; it sup-
output, and general status of the patient. With ports her observations by monitoring certain
the advent of other automatic constant-monitor- aspects itself (ECG, EEG, and blood pressure).
ing devices, such as for EEG's, respiration, skin Thus the nurse will have more time for care of
temperature, and cardiac output, the medical patients and for complex techniques. The im-
team will be faced with a bewildering display of mediate entry into computer memory of drug
data, making their monitoring and decision administration and procedures will allow the
function more difficult and retarding crucial computer to double-check all new orders for
time-dependent therapy. conflict, acting as a safety check to prevent
The medical team in a CCU thus devotes much overdosage or use of inappropriate drugs.
time to observation, recording, and analysis of Figure 13 shows the output of the initial
repetitive data. Monitored data on ECG's, blood monitoring program as used in a hospital ward,
pressures, pulses, respirations, and other factors and a sample of the ECG that was monitored.
are tabulated, correlated, and interpreted. In so In this printout each of the columns labeled 1 to 6
doing the medical team acts as a data-reduction is one time block of 4.8 sec. In each block one
system. All these tasks are well suited and easily typical wave form was picked and thoroughly
amenable to digital-computer manipulation. measured. As with our routine program, the whole
We suggest development of computer programs block was scanned for indication of arrhythmia.
and a hardware system for the tasks of a CCU in Certain problems have arisen in the use of this
collection, reduction, analysis, and interpretation program, but the direction of the solutions is
INTERPRETING
ELECTROCARDIOGRAMS
FROM
SPACE ]13
IN_T_UMENTA][0N FIEL0 $[ATION HEART nlSEASE CONTNOL PROq;RAN

ECG _0Ni10_ING
_OOtl .o_lo
0ATE 0_-13 ET--tg-07-111 ONLINE G. w,, 0PE_ATING ROOH HONITORING

TImE I 2 3 4 S 6
m. t
PA e3tl o3/4 434 435 434 035

P0 .10 Ol5 e13 .10 • IS alU
UA -,,23 ",,23 ".20 "-22 -.23 -.22
U0 *03 -02 e03 .02 • 03 .02
RA ,_17 .98 .89 .03 o9S 090
M0 *02 °02 ,.02 *03 .02 .02
_A "*_2 "*3q t.65 -.39 -.46 -.SO
SO .06 .07 .07 .06 .06 .06
ST ..It ,,03 .08 ,,02 .03 °03
ST0 433 o37 o36 O31 429 433
5TM e_6 a37 440 o33 429 O3a
STE ej7 a3_ 444 *33 e30 035
TA IU4 032 Ol4 o31 o33 036
TO ola *1_ .09 .16 • 15 .15
T'a -.23 -o2U ".21 ",,21 -024 -o22
[°U .09 .09 o09 ,08 .09 ol0
PM o26 o26 o29 o26 025 426

ORS =11 .11 ol2 oll • ll .10
UT .3_ .38 *39 ,,38 .39 .39
MATE 79 78 77 76 76 76
Sl ELEVATIOn; RULE OUT EANLY NEPOLARIZATION

ExTwEME ST DISPLACEMENT
TALL P NAvESt P - PU_ONALE
PROLONGEC PR INTERVALt FINST DEGREE AT_IOVENTRICULAR BLOCK
::r: :, w ":" _ T :rT r T r_ ¶ _ 4s._ _ Tl" .... r
r
LL[ 1I J: I; lii!tl ,: llxti;t
FInERy. 13.---OutpuL of the initlal monitoring program and a sample of the ECG.
clear and in progress. One problem is categorizing Logic flow for surveillance systems_A flow
of complex arrhythmias; another is the variation diagram has been developed that outlines the
in calibration in the monitoring signal. The de- system that we are designing (fig. 14). The system
signers of CCU equipment have followed NASA's mimics what the physician does, but supplements
precedents and have not used clinically conven- his logic with statistical techniques in order to
tional calibration routines. Calibration is set by arrive at its conclusion.
the nurse so that the wave form is large enough In the flow diagrams (fig. 14) the ovals enclose
to trip the rate meter and therefore be counted statements produced by the system on the screen
accurately. If we are to rely on measurement of or typewriter; the rectangles enclose various
the wave forms for statistical purposes, we logical operations that are to be performed and
should introduce a calibrated signal into the are not detailed in the flow diagram; and the
system. hexagon denotes values calculated from known
G .oo._..cLo
MONITOR
START
INPUT
e ITA% NIGH-LOW TE|T _'
Is*_l SECOND
TI_E iLOCK ( STATEMEt4T
)
COMPAgE
STATISTICS WITH
IqEVIOUS TIME
ILOCK
_ STATISTICS
Nl:
P <0.05
L/lilTSANDIOI_
AI_[:
LrMPIRIC LIMITS
TO DIAGNOSTIC LOGIC
I°_!_ I-'_7 5 TA_J_I _

/
IAGNOSTIC
_¢_ STATEMENT
I _o.,o
MONIT0_DATA
FIGURE 14,--Logic flOW for CCU surveillance: flow diagrams.

INTERPRE_NG ELECTROCARDIOGRAMS FROM SPACE 115
CORRELATE
BASICDATA6
MONITOR
RESULTS
lit i CALCULATE
PROGNOSTIC --
INDICES
No
DIAGNOSTIC CORF_ELATE
TEST STATEMENTS PROGNOSISFROM COMPLETEDATA
SUGGESTIONS WITHTHERAPY
THERAPY & PROCEDURE
CRITERIA
Yes
FIGURE 14 (Concluded).--Logic flow for CCU surveillance: flow diagrams.
formulas. The diamond-shaped boxes indicate of trends, that is, to see whether any of the
decisions to be made by the computer, ahvays of variables are changing at a rate higher than is
the "yes-no" variety. The circles indicate areas acceptable. Trends are calculated by comparing
that are to be continued on other parts of the flow means and standard deviations of a time period
diagram. The line connecting various boxes in _ith those of the previous period and noting
the flow diagrams is to be considered a line of whether there is significant difference either
action of event. In order to follow any particular empirically or on A probability basis. This is
data input one should start at the top of the flow equivalent to the physician's noting that the
diagram and follow one set of answers; the Acceleration of the heart rate is greater than
diagram should lead him sooner or later back normal. If no abnormality is found, the data can
to item-1. be passed to the so-called lumped-data trend-
The data are first compared to empirical high- analysis, a statistical method for mimicking the
low-limit tests to see whether they exceed these intuitive feeling that something is not quite
empirical limits. Then statistics are calculated correct in the mass of data. Such tests as
from the data for a period of time, and data are Hotelling's T 2 multivariate analysis can be used
compared to statistically derived high-low-limit to show that something has varied within the
tests for the past time period. This procedure data, but they cannot pinpoint the item that tms
mimics what the physician does when he says varied; they are more reliable than intuition,
that particular values are unusual for a given however, and are positive when a group of
patient. variables begins to change, although none of
These statistics can then be used for calculation the individual variables may have changed
significantly. Comparison of the monitored data par;son of the results with those from selected
with various prognostic indices can lead to a population samples can, for example, be the key
decision as to whether a prognostic statement to proper selection of personnel for space flight
can be made. or to selection of medication.
For the modern monitoring system to ac-
complish these goals, three separate items of
COMMENTARY
hardware are required. The first includes trans-
With enough data and experience we will be ducers and data-acquisition devices; modifica-
able to evaluate new relations and to begin, by tion, improvement, and development are much
statistical techniques such as muir;variant anal- needed in this field. The second incorporates
ysis, to sharpen our ability to predict clinical computer programs for routine control functions
phenomena such as shock or arrhythmias. and rapid data analysis. The third includes
Application of continuous monitoring of ECG's telemetric data-communication systems to en-
to postinfarction patients, or during anesthesia able easy storage, retrieval, and display.
and surgery, can illustrate its value. The method Immediate association of data is necessary for
can also be used in the laboratory, for example, insight into the current status of subjects and
to assess the effects of stress and the effects of for physiological research to obtain more knowl-
radiation on the cardiovascular systems of ani- edge for the improvement of care. We envision
mals (as in the production of arrhythmias). automated care and evaluation units providing
In clinical medicine, patients are intensively for analysis of many body functions, including
monitored in the operating room, the obstetrical ECG's, heart sounds, cerebral electrical activity,
suite, and other specialized intensive-care units; vital signs, and vascular, respiratory, and meta-
the procedures are not essentially different from bolic conditions. Chemical analysis also must be
those used in space flight. Personnel in these units incorporated, integrated, and related to other
perform tests and intuitively obtain results with- variables. Thus the automated care and evalua-
out the objectivity of statistical procedures. In tion unit should have facilities for on-Iine,
the clinical setting, in addition to natural human real-time statistical analysis of data, which should
bias, the major current disadvantage of function- be in a convenient format for use in real time for
ing with a human monitoring system is simply patient care.
the lack of trained personnel to perform all the These automated care and evaluation adjuncts
needed services. must serve and not supplant the human monitor.
The problems of unavailable personnel and re- The automated care unit can and should perform
duction in bias of existing personnel can be solved the tasks that are at best, routine and noncreative.
by automation. Furthermore, automated care-
evaluation units can serve throughout a hospital,
not simply the needs of specialized units. The SUMMARY AND CONCLUSIONS
automated care and evaluation unit can provide
for tile acute as well as routine phases of health Constant monitoring of subjects and data inter-
services. Indeed these two phases are not distinct pretation for evaluation or care are often humanly
since the patient may quickly change from one impossible because the data accumulate faster
status to another. Its abiIity to perform either than they can be analyzed. Use of modern com-
phase implies that the unit can carry on the puter systems and statistical techniques allows a
other. The same concepts can hold for space new dimension in the quality of medical care that
flight. Monitoring nmltidimensionally and com- the physician can give.
CHAPTER 8
PERIOD ANALYSIS OF AN
ELECTROENCEPHALOGRAM FROM AN
ORBITING COMMAND PILOT
Neil R. Burch, Ronald G. Dossett, Abbie L. Vorderman, and Boyd K. Lester
Recording of an electroencephalogram (EEG) been undertaken (ref. 5). Clinical interpretation

from a pilot in orbital flight offers an unprece- of portions of this recording has been reported
dented opportunity to inquire into neurophysio- (ref. 6) and may be considered the "standard"
logical-behavioral relations during a situation system for interpretation of states of conscious-
unique in recorded history. While the opportunity ness such as stages of sleep. However, interpreta-
is unprecedented, this very fact makes full tion by the human electroencephalographer must
interpretation of the data from any biological always suffer certain inherent difficulties. In this
system extremely difficult. The single-channel or special case he is faced with the formidable task
double-channel EEG, as a psychophysiological of transforming one or more wiggly inked lines,
measure, requires a great deal of information from recordings lasting from 1 rain to 50 hours,
about the stimulus field for optimum interpreta- into word symbols of a paragraph or a page. The
tion of neurophysiological-behavioral relations; present need is for a more efficient and exact type
but by the very nature of the flight situation the of transformation of the analog signal into
stimulus field is impossible to control, and the digital rather than word symbols.
stimulus field of ongoing events is difficult to The tracing is read in sections recorded at the
chronicle with a high degree of resolution in time. rate of 1 page or so per 20 sec and may be flipped
In the absence of stimulus-field information the and scanned in as little as 5 percent of recording
EEG is most helpful in quantitative determina- time. Processing of the tracing in this way loses
tion of the state of consciousness of pilots in much of the information carried by subtle
orbital flight. The results of analysis of over 50 changes; the necessarily qualitative nature of
hours of continual recording confirm that sleep subjective impressions results in several crucial
during this orbital flight was indeed disturbed in handicaps. It is not practical to compare directly
a way that could not have been predicted from the exquisite details of long records taken con-
laboratory experiments, although most indi- tinuously on the same subject, so that evaluation
viduals show a somewhat-disturbed sleep pattern of relative change in state of consciousness or
on the first sleep night in even a moderately un- stage of sleep is made more difficult. Qualitative
usual environment such as a strange room and data handicap evaluation of a record relative to
bed (ref. 1). At the other end of the state-of- other subjects or populations as well as make it
consciousness spectrum (ref. 2), the arousal of almost impossible to establish solid statistical
alertness and even hyperalertness may be ex- levels of confidence for impressions or diagnoses.
pected to result from a situation as novel as For future application in monitoring of space
space flight, so novel that it had then been flights, the fact that qualitative data compel the
experienced by only several dozen humans. use of an expensive high-level human "computer"
In interpretation of these EEG tracings several for the reading and interpretation of the record is
other laboratories have employed a number of important because the limited availability of time
different analytical techniques (refs. 3 and 4); from such trained personnel must always restrict
even a truncated version of period analysis has the duration and number of samples that can be
117
interpreted. Continuous, long-term, on-line inter- (ref. 6). The exact positions of the four perforated-
pretation is prohibitive. electrode sites correspond to the following
The inherent inadequacies of subjective data measurements (ref. 7):
manipulation, and certain shortcomings of several
(1) Channel 1--Midline-central site--7.8 in.
other analytical systems, are overcome by period
from the external auditory meatus in the coronal
analysis as employed in analysis of EEG's from
plane and 7.9 in. anterior to the inion in the
space in the following important aspects.
sagittal plane; midline-occipital site--l.6 in.
High resolution in time--Period analysis can
superior to the inion in the midsagittal plane
resolve changes in the signal faster than the EEG
(2) Channel 2--Left-central site--3.1 in. to the
pattern can reflect a given state of consciousness.
left of the midline-central site in the coronal
The 10-see epoch utilized in this study has proved
plane; left-occipital site--l.4 in. to the left of the
satisfactory for analysis of most experimental
midline-occipital site
sleep patterns; however, abrupt transient changes
seen in these records strongly suggest that the The master analog magnetic tapes were re-
analysis should have been in epochs of 1 sec corded on special equipment* at 0.0293 in./sec
or less. and rerecorded through four steps in a format
Long-term conti_uous analysis--Period analysis compatible with a Precision Instrument, 14-
can process the signal continually over a time channel, record-playback system. The tapes were
commensurate with the duration of the flight. replayed at 1] in./sec with output simultaneously
While it is possible that a particular system of recorded on an eight-clmnnel Grass model-III
analysis may yield information from relatively electroencephalograph and fed to the analog-to-
few and short samples, continuous analysis is pulse-width converter for period-analytic proc-
highly desirable. essing.
Multiple channels---Period analysis can inter-
relate in a meaningful way to two or more simul- Period-Analytk Processing
taneous signals. While single-channel analysis Period analysis of the EEG as a data-reduction
yields important and reliable information, analysis process has been reported (refs. 8 and 9), but one
of multiple simultaneous signals significantly must understand that period analysis yields these
increases neurophysiological information and the three basic parameters: major period, inter-
level of confidence in interpretation. mediate period, and minor period. These three
Relevance to current interpretation of tl_e EEG-- periods respectively code the base-line cross of
While analytical considerations do not require the primary EEG or dominant activity, the
that a system of analysis employ parameters peak-and-valley activity, and the very low-
having meaning in the context of EEG interpre- amplitude inflection points of the EEG signal.
tation, and while a system may process in such a They are subsequently distributed over three
way that the analyzed data cannot be directly spectra of "equivalent" frequencies in l0 bands
related to current reading, it is preferable for per spectrum (table 1).
certain of the analyzed parameters to transform The most elegant statistics of period analysis
directly into signs and interpretations now em- are the counts per second of the major, inter-
ployed by electroencephalographers. By the same mediate, and minor periods. This extremely
token, the analyzed data and the anaIytical economical process, ideal for on-line data reduc-
display should not be more complex nor require tion in real time with minimum instrumentation,
greater effort for interpretation than the primary offers a reliable index of state of consciousness
EEG signal itself. and stage of sleep. A further smoothing step,
generating a single statistic that is generally
METHOD monotonically related to arousal, is the simple
process of summing all the counts for what we
Recordir_ of the Analog Dora
refer to as the total count. As in all smoothing
The recording sites for the EEG were selected
and prepared according to reported procedures *Cook Electric Company, Morton Grove, Illinois.
ANALYSIS OF AN ELECTROENCEPHALOGRAM 119
TABLE 1.--Equivalent Frequencies in Hertz digit value, with the 30 band parameters of fre-
quency distributions expressed as percentage of
Period
Band time occupied by activity in that band, and the
Major Intermediate Minor three counts expressed as absolute counts per
second for each of the three periods. The 33 two-
1 1-4 I-4 1-10
decimal digital statistics characterizing each
2 4-6 4-6 10-20 10-sec epoch and time identification are logged on
3 6-8 6-8 20-30 incremental magnetic tape at 200 bits per inch
in a format compatible with general-purpose
4 8-10 8-10 30-40
digital computation on the IBM-7094. The flow
5 10-12 10-12 40-50 diagram for the data-analysis and data-logging
6 12-18 12-18 50-60 system used is schematized in figure 1.
While state of consciousness should be inter-
7 18-24 18--24 60-70
preted from the period-analytic parameters at
8 24-35 24-35 70-80 the time of on-line analysis, more complicated
9 35-50 35--50 80-90 classification procedures were investigated in an
attempt to improve the interpretation. Unfortu-
10 50-100 50-100 90-100
nately these procedures require programming of
a general-purpose digital computer for identifica-
procedures some information is lost in the total- tion of particular classes of electroencephalo-
count statistic, and in some cases of ambiguous graphic records in order to automate interpre-
interpretation one is forced to the independent tation; they attempt to define mathematically,
counts for clarification. In studies involving sleep still by way of period-analytic characterization,
it is convenient to weight the major-period count certain neurophysiologica] states as expressed by
by a factor of four, the intermediate-period count a set of EEG patterns. This general problem of
by a factor of two, and the minor-period count pattern recognition is becoming increasingly im-
by a factor of one in order to accentuate the portant in psychophysiological and behavioral
monotonic relation. studies because of the power of these methods.
In the 10-sec-epoch smoothing mode, period However, such manipulations suffer the disad-
analysis yields 33 parameters which statistically vantage that the complex relational patterns,
characterize each 10-see sample of the EEG that often serve as the criteria for recognition,
record. Each parameter is read as a two-decimal may be so highly abstract that interpretation in
Translator
+,me-CodolI
Generator _
S,stemL------I
Programmerl #
AT,'°S:?'I
'g'
Time-Code I Display t
I I _----][ C°nvirter_"
Tape Impedance- I IAnalog-to- I _ajor-Perio_ L--I Data l Incremental

Digital I
Recorder Matching _-- ]Pulse widthL_____] Spectral I.__.___1^ c . . / Tape I
Converter ] [Converter ] --] Computer J_ cumulator_ Recorders__ l
Jlntermedia.t_ I
Monitor l Spectral I
Analog / ---_ Display
Typewriter [
[Cornputer ]
Mi Spectrol
nor-Perio |
4
Computer I
Fm_am l.--Flow diagram for the data-anMysis and data-logging system.

behavioral or neurophysiological terms is almost presents representative 10-see analog records of

impossible. three of these different states of consciousness.
The initial set of numbers at the lower right of
Discrlminant Anah/$1s each sample gives the total counts of the major,
The technique of pattern recognition here intermediate, and minor periods; for instance,
utilized for classification of states of conscious- 11:24:37 in t}le first record (fig. 2) indicates a
ness is discriminant analysis (ref. 10). In dis- count of ll/sec in the major period, 24/sec in
criminant analysis, as in most pattern-recognition the intermediate period, and 37/see in the minor
programs, criterion groups or training sets must period. The other numbers below each record
first be established by some selection procedure convey flight time, a state of consciousness, and
that is independent of the pattern-recognition the probability associated with this record being
logic itself. Thirty-five 10-sec epochs for each of a member of the particular class identified. For
16 different states were selected by clinical example, in the first recording of figure 2 the
interpretation of the analog record. Figure 2 flight time is 030851; the state identified as
o_. 1(,#/-08 -. _v
_--- I Stco.o--I
FIGURE2.--Analog EEG for three operational states of consciousness: early stage-I sleep (1A), poorly organized
alpha (0C), and well-organized alpha (0B).
"IA" is stage-I sleep; and the probability that diseriminant-function analysis and to weighted
this 10-sec epoch belongs to class stage-I sleep is total counts of the EEG will be used to demon-
0.92, as determined by the discriminant analysis. strate the interpretive process. Where these two
The classes established by clinical interpreta- characterizations of the state of the command
tion of the EEG were grouped into 16 operational pilot are not congruent, more-specific data from
categories which represent five states of conscious- other parameters of the analysis will be presented
ness. We cannot emphasize too strongly that to clarify the interpretation. In order to follow
these classifications are based on the EEG signal the time-line profile one must interpret the
only since there was no independent measure of principal classifications of the different electro-
such states as "eyes open" or REM. Table 2 encephalographic states as behavioral states.
relates the classifications of this report to the
standard Dement-Kleitman classifications. For Principal Classificationsof the Behavioral States
adequate interpretation of the sleep data, results No classification program can be better in
of a comparative study with a simulated Apollo terms of sensitivity and selectivity of discrimina-
flight are presented. tion than the basic parameters or descriptors
that characterize the data. Therefore, it is of
RESULTS great importance to appreciate the power of the
30 period-analytic bands in characterizing each
The fact that the period counts, weighted total category.
counts, and discriminant-function classifications
Table 3 lists the mean percentage of time in
of the 50 hours of EEG have definitive relations
each band for each of the 16 categories; the total
to the events occurring during the mission shows counts are included but were not used as de-
that these parameters are sensitive indicators of
scriptors in the discriminant analysis. The cate-
the state of consciousness during space flight.
gories are designated by the operational definition
Additionally these results demonstrate the quali- used for selection of the criterion samples and by
tative and quantitative changes in the sleep the Dement-Kleitman equivalents. On the basis
patterns during flight. Of all the operations per- of the detailed findings presented in table 3, the
formed on the data, the weighted total count interpretative keys for the principal classifications
gives the most economical and reliable indicator are now outlined.
of the state of consciousness.
Artifact--Heavy-muscle artifact shifts the fre-
The relations of the mission's events to the
quency histograms of all three periods to the
right, with accentuation of all high-frequency
TABLE 2.--Equivalent Classifications components. Moderate-muscle artifact tends to
be contaminated with slow-wave-movement arti-
Classical (activity)
Discriminant, fact which shifts the major-period histogram to
Predominant Secondary the left; the high-frequency muscle components
continue to shift the intermediate and minor
1A I-Submergent
periods to the right even when the muscle artifact
2A I-Submergent with alpha is "moderate."
2B I-REM Arousal--The T_-category may be interpreted
as a state of nonspecific neurophysiological
2C II
arousal in contrast with the relatively specific
3A II visual arousal of the eyes-open category. Both
3B II III arousal states show increase in slow components,
but delta tends to predominate in the T_-category
3C III
while slow theta is dominant in eyes-open cate-
4A III IV gory. The eyes-open state shows twice as much
4B I¥ ill 24-to-35-Hz activity as does T,; this accentuation
of the relatively high-frequency beta component
4C IV
may characterize the specific visual arousal of
]22 BIOMEDICAL RESEAI_CH AND COMPUTER APPLICATION
TABLE 3.--Percentages of Time for 10 Bands Per Period and Counts Per Period
Percentage time per band Cotlnts*

Period
1 2 3 4 5 6 7 8 9 10 Period Weighted
MB--Heavymusclea_ifaa
Major 13 13 I1 6 7 15 13 15 8 7 17 68
Intermediate 1 0 0 1 2 13 16 40 23 24 39 78
Minor 1 5 19 27 11 11 15 13 7 2 50 50
(106) (196)
MA--Moderatemusc_ artifaa
28 25 14 8 5 10 8 6 2 2 10 40
Major
Intermediate 1 1 4 4 6 26 21 29 13 15 30 60
Minor 1 12 24 28 8 6 10 11 7 2 47 47
(87) (147)
Tl--lOminbefore to 15minafter lift-off
32 27 16 8 5 10 7 4 2 4 11 44
Major
Intermediate 1 2 6 7 7 32 21 27 I0 13 29 58
Minor 1 16 28 29 7 6 9 10 9 4 45 45
(85) (147)
EO---Eyes open
16 19 14 9 8 14 13 8 4 3 13 52
Major
Intermediate 1 0 1 2 3 22 20 35 16 16 32 64
Minor 0 9 26 32 8 6 9 10 6 2 46 46
(91) (162)
OB--Moderate_ weIl_rgan_ed alpha
Major 2 4 9 25 29 29 7 2 I 0 11 44
Intermediate 1 0 2 17 24 41 12 11 4 4 18 36
Minor 3 34 27 23 5 3 4 5 4 1 36 36
(65) (116)
OC--Poorly organ_ed alpha
5 7 15 20 21 24 9 3 1 0 11 44
Major
Intermediate 1 0 4 13 17 42 16 15 5 5 19 38
Minor 2 31 29 26 5 3 4 5 4 2 37 37
(67) (119)
lA--Submergent stage-I
Major 9 14 15 16 12 20 12 5 2 1 11 44
Intermediate 1 0 2 8 11 35 21 25 8 7 23 46
Minor 1 22 29 29 6 3 5 6 4 1 39 39
(73) (129)
_A--Submergentstag_I w_h alpha
7 15 16 16 14 19 9 4 2 0 11 44
Major
Intermediate 0 0 3 9 12 36 19 22 7 7 22 44
Minor 1 24 31 28 5 3 4 6 4 2 38 38
(71) (126)
_B--Stage-I REM
Major 19 19 10 7 11 10 4 2 1 10 40
Intermediate 1 4 8 9 29 23 27 8 7 24 48
Minor 0 18 33 32 6 3 4 5 5 2 39 39
(73) (127)
$C--Stage-II (light)
Major 29 28 19 8 5 7 6 3 2 0 8 32
Intermediate 1 2 7 10 11 30 19 22 8 7 23 46
Minor I 21 30 31 6 3 5 6 5 2 40 40
(71) (118)
3A--Slage-II (moderate)
7 28
Major 35 30 14 8 5 8 5 2 2 0
Intermediate 1 2 9 12 12 32 17 19 7 7 22 44
Minor 1 22 29 30 6 3 5 6 5 2 41 41
(70) (113}
TABLE 3.--Percentages of Time for 10 Bands Per Period and Counts Per Period--(Concluded)
Percentage time per band Co[lnts*

Period
1 2 3 4 5 6 7 8 9 10 Period Weighted
3B--Stage-IlI (light); stag_II(deep)

Major 40 35 14 7 4 7 4 2 2 0 7 28
Intermediate 1 5 i1 15 13 30 13 16 6 6 20 40
Minor I 24 29 28 5 3 4 6 5 2 39 39
(66) (107)
3C--Stage-III (moderate)
Major 45 29 12 6 5 8 5 2 1 0 7 28
Intermediate 2 5 11 14 14 30 14 14 6 6 20 40
Minor 2 27 26 26 5 3 5 6 5 3 39 39
(66) (107)
4A--Stage-IV(l_ht); stage-III (deep)
Major 54 26 10 6 4 6 3 1 1 0 5 20
Intermediate 2 9 18 19 15 26 10 11 4 5 17 34
Minor 3 30 27 25 5 3 4 6 6 3 37 37
(59) (91)
4B--Stage-IV(moderate)
Major 60 21 10 6 4 4 3 1 1 O 5 20
Intermediate 2 11 21 22 15 24 9 10 3 4 16 32
Minor 3 33 24 25 5 3 3 6 6 3 37 37
(58) (89)
4C--Stage-IV (deep)
Major 72 15 6 3 2 3 2 1 0 0 4 16
Intermediate 6 16 22 20 13 21 8 9 3 4 16 32
Minor 4 33 25 25 4 3 4 6 6 3 38 38
(58) (88)
*Total counts for three periods appear in parentheses.
the eyes-open situation. The predominant 30-to- characterized by a rather well behaved increase
40-Hz activity in the minor period further defines in slow components at the expense of higher
this specific visual component as being quite well frequencies. The progressive slowing first appears
organized and primarily in the fast-beta range. in the dominant primary activity and then in the
Compared to the beta component of specific superimposed activity of the intermediate period;
arousal, the T,-state tends to show increased finally, as the slow-wave components become
superimposed activity in the 12-to-18-Hz range, synchronous, they are reflected in the minor
which is demonstrated to be a well-organized period. The mean value of 4 percent in band-1
wave shape by the comparatively high 10-to-20-Hz (1-10 Hz) of the minor period, seen in deep
component of the minor period. stage-IV sleep (4C), is quite significant and
Resting--The eyes-closed (presumably resting) probably pathognomonic of a type of neuro-
categories show the expected high percentages of physiological activity that can exist only in this
alpha activity with very little in the way of slow very special state of extremely slow-wave sleep.
components. The resting state shows an un- The depth of sleep can be directly related to the
expectedly high percentage of 12-to-18-Hz activity shift to the left of the three period histograms.
so well organized that it remains modal for the An exception to this rule occurs in light sleep,
superimposed wave shapes. The very-high-fre- particularly in that presumed REM state identi-
quency components of 50-to-80 Hz have di- fied as 2B. In this stage the orderly progression of
minished in this eyes-closed state by 50 percent the shift to the left has been interrupted by a
or more when compared to the arousal categories. relative shift to the right and an increase in rela-
Sleep states--Progressively deeper sleep is tively high-frequency activity as compared to the
previous sleep stages. Of particular interest is the some normal increase in tension at lift-off (ref.
extremely high activity in the 20-to-30- and 12). The EEG recording begins approximately
30-to-40-Hz ranges in the minor period. This 10 min before lift-off (T--10 min).
component may represent a specific activation Discriminant analysis--Since the training set
state characteristic of this rather unique stage of of the Tl-category was drawn primarily from
sleep associated with rapid eye movements. this time, it is not surprising that the modal
profile finds such Tl-epochs to be the most
Time-Line Profile numerous category during these 5 min.
We may now turn to the time-line profile Weighted total counts---A relatively high level
generated from the midline EEG of the com- of arousal is indicated by the total-count reading
mand pilot and recorded from some 10 min prior of 162.
to lift-off through the first 54 hours of the flight. Band parameters--The time-line profile of delta
Figure 3 shows the discriminant-analysis time- activity is illustrated by the upper portion of
line profile, plotted according to the state of figure 5. The reading in this 5-min period is
consciousness in 5-min data points, for the first relatively high at 27 percent; this is the com-
28 hours of flight. The state-of-consciousness ponent that previously has been interpreted as a
characterization of the 5-min sample was deter- special case of arousal. The lower portion of
mined by the most frequently occurring, or figure 5 shows the minor-period activity from
modal, 10-sec epoch; muscle artifact is plotted 104o-20 Hz also to be relatively high in this state
only if it is the exclusive characterization in all of consciousness.
10-sec epochs of the 5-min sample. In majority- Mission events--Anticipation of imminent lift-
vote logic of this sort, provision must be made off and decrease in the rate of preparatory
for a tie vote if two or more categories occur with activity may have produced a special state of
the same frequency; in the case of a tie, the vigilance _th a strong behavioral component of
category representing the lower state of conscious- inhibition.
ness is taken to characterize the 5-min sample. Discrirni_mnt analysis--The discriminant classi-
Figure 4 illustrates the weighted total-count fication continues to show the Tl-states although
profile plotted as 5-min sample points against the there has been considerable change in the total
first 28 flight hours. Only time points of particular counts. The stability of the modal smoothing
interest will be discussed, but the interpretation procedure, with some loss of sensitivity, is clearly
of this index as monotonically increasing with seen in the contrast of these two indices.
increased arousal and monotonically decreasing Weighted total counts--The difference in the
with depression in state of consciousness is so level of activation, between the 5 min immedi-
straightforward that the reader is invited to ately prior to lift-off and the 10 rain before
render his own interpretation for any given lift-off, is seen as the total counts drop from 162
point in time. A paradoxical area of interpretation to 144. The mean value of the two samples
is found in stage-I ("paradoxical") sleep (ref. 11). preceding lift-off, 152, is used in this report to
The time-line profile of selected individual establish the midpoint of what we shall refer to
bands will be used occasionally to illustrate as the T_-zone which ranges from 150 to 154
certain points in particular time samples; at counts. This Tl-zone is interpreted as an arousal
points of ambiguous interpretation the inde- state reflecting special vigilance; in the weighted
pendent major-, intermediate-, and minor-period total-count graphs it is the lightly shaded bar
counts must be called on for clarification. Light- from 150 to 154.
dark cycles and orbital revolutions are indicated
Lift-off to T+5 Minutes
(circled dots) on all time-profile graphs.
Mission events---The acceleration profile of this
From T-- 10 to T-- 5 Minutes
mission shows its first peak of some 5.5 g ap-
Mission events'--A behavioral state of alertness proximately 2.66 min after lift-off.
would be expected to accompany rather intensive Discrirninant analysis---The state of conscious-
preparations for lift-off. It is kno,_m that there is ness now changes to an eyes-open classification.
ANALYSIS OF AN ELECTROENCEPHALOGRAM ]25
111
! j pI
;i' ii:
FIGURE 3.--The modal state of consciousness for every 5 min for the first 28 hours of the flight.
Weighted total counts--A count of 163 does not 5.66 min after liftoff. The astronauts report that
indicate that the event of lift-off itself has mark- this period of second-stage-engine cutoff is
edly changed the level of arousal. "crisp event" when the g-level suddenly drops to
zero (ref. 13).
From T-I-5 to T+IO Minutes
Discriminant analysis--The extremely rare ex-
Mission events--The acceleration curve shows clusive heavy-muscle state appears in response
its second and highest peak, about 7.33 g, about to maximum g-forces.
FIGITRE4.--Weighted total-count graph for the first 28 hours of flight.
Weighted total counts--The striking increase of they are adjusting to their new environment
the total counts to 198 reflects the effect of (ref. 13). The rate of verbalization by the com-
maximum g-forces. mand pilot in communication with the ground
is maximal during this early part of the mission;
From T+IO to T+60 Minutes (00: lO to 01:00) there is relative increase in rate of transmission
Mission events---It would of course be of ex- during the T_-state.
treme interest to know in exact detail what During all but 16 hours of the mission the
transpired to force a return to the level of vigi- oxygen-to-water differential-pressure warning
lance represented by the Tx-state. The astronauts light of section-2 indicated a beyond-limits
consider critical the early phase of the flight when oxygen-to-water pressure across the water sepa-
'" 1 't_t!i_iitlt'i " '! i
_!!!H!!HH! i!tlttitftiltt!t_ttlttltttttttittttitttitiitttttttttttltttttti_lttttTll
_tttftttft_fti_t_t_tttfttttttttttttttttttttttttttttWft_
_tt_tttttttl7_t7_ttttitit
Hllltttittltttlltttttttttttttttfttitttt_
¥1144i41i II 1444
Nttrstt_ _iiflt_i_IIlllftt_lilltlti!titttttttttttttfttttttltttltttt_
_t_l._tt_lt_ttli_tt!fttttt_ltltlllttltlfl_
!ifttttlttt_
I, I, , Iili {.11III11111N
_!i-II 11
H
!_!!!!!!!NNi_+_ ..........;+,+,. it 1 !ill i

ilttWtttllttfltttfJt_it_J
;:;;;:
litttttttttttttttt!tlttttttltttt
[!!,_!!_
.1_tlt_ttt_fltttttttttttttttt!,_i
!!!!!I'__i,,
tltttlf_Htttftttttttfftttttttttiifiii+_
i"iiiiIt =_,,,,,,_,.,*_
.tl_tt{tt_l_LLlt_ttittNti_{lttl_tittttitttf]_ft_t
.... Nttttttttl_Nttlttlftlllt!ttttlltttlttlttltttititttttttttftllift
iiiiiiiiiiiiiii''a'_ t
1
1Iflt;t;It74tttttttttt_tI_t_t,_m
_!_ _
_
I11Itt_ttltlt_ft1IItlttttttt,_!tl
_,,t !!i!{_
lttlttltftttttttttll_l_Ii_JjI
..... ,,,t
iii_iiii "
.... i;ii;iiii
1]
....i ltt ,irlil!il i!tl ltltlt IllliI

1 t t lt
__
1
Flavn_ 5.--Comparison of the major-period delta-index band-1 (1 to 4 Hz) with the minor-period band-2
(10 to 20 Hz) to illustrate the rather unusual slow activity occurring 10 rain before lift-off,
rators (rcf. 14). Considering the important part 01:00 the reading of 147 counts is very close to
that a fuel-cell beyond-limits warning light had the Tl-zone, and it is at approximately this time
played in Gemini V (ref. 15), it must be assumed that the discriminant analysis also indicated a
that the spurious warning light in the early part Trstate.
of this later flight was a signal of some conse-
From Ol :00 to 02:00
quence to the command pilot. We may speculate
that the differential-pressure warning light was a Mission events--The Trstate of this hour again
highly alerting signal in the behavioral sense. coincides with relatively increased verbalization
Discrirninant analysis---Most of the 1st hour from the command pilot to ground. In the light-
is occupied by the eyes-open modal state, but dark cycle, daylight had begun about 10 min
just before the end of this period there is a before 01:00 and ended about 15 min before
return to the Ti-state for 10 min. 02:00. During the first two dark cycles the down-
Weighted tolal counls--The peak arousal of 208 ward swing of the EEG cycle seems to coincide
counts occurred when the astronaut was experi- with the onset of darkness.
encing his first continuous zero-g; it may represent Discriminant analysis--This hour is almost
one phase of accommodation to this novel environ- exclusively characterized by the eyes-open state,
ment. The relatively low count value of 154 be- except for the one 5-min sample between 01:15
tween 00:45 and 00:50 occurred when the ex- and 01:20 which again shows the Trstate.
tremely high rate of verbal communication from Weighted total count,s--The weighted-count level
the command pilot to the ground diminished of 139 shows a remarkable degree of relaxation at
almost to silence. A higher rate of verbalizatiort 01:25, but the level of arousal progressively in-
begins during the subsequent 5 rain, with a con- creases during the next 25 min to peak at a
current jump to 164 counts. Between 00:55 and value of 184 before cycling down to 140 at 02:35.
128 BIOMEDICAL RESEARCH AND COMPI3TER APPLICATION
In behavioral terms, the weighted total-count since the sleep program on previous flights had
level of 140 represents relative inhibition; in not worked well because of flight-plan activities
terms of performance such relative inhibition is and the fact that the crew tended to retain their
a "good" or "bad" state depending on whether Cape Kennedy work-rest cycles, with both crew-
the stimulus field of the moment calls for con- men falling asleep during the midnight-to-0600
sidered action or relaxed inaction. period of night at Cape Kennedy (ref. 16).
Discriminant analysis--The T_-state does not
From 02 :O0 to 06 : O0 appear in this 6-hour portion of the mission. The
Mission events---The second and third orbits eyes-open state occupies the first several hours,
are seen to have been completed at 03:15 and and beginning at 08:15 we see the first indication
04:45. It is reasonable to suppose that any of relaxation with the appearance of well-orga-
change in the state of arousal, associated with nized alpha suggesting that the eyes are closed.
the orbital schedule, should now begin to be From 08:20 to 08:25 the first intimation of light
evident. At some time during the mission the sleep appears, immediately followed by an eyes-
windows were screened with foil from the food open and then by a resting eyes-closed state. Ten
packs, in addition to the Polaroid screen, to keep minutes of light sleep is noted between about
the cabin comfortable during sleep periods (ref. 08:40 and 08:50. During this first sleep cycle
16). Such screening would of course reduce what- between 08:00 and 10:00 we find five episodes of
ever EEG-arousal effects the light-dark cycle this very minimal depression in the state of
might engender. consciousness with a total llght-sleep time of
Discriminant analysis--During this 4-hour 50 min.
period the eyes-open state predominates, but Weighted total counts---The onset of light sleep
the T_-state appears seven different times for a is seen at 08:20 as a weighted count value of 122,
total Tl-time of 40 min. indicating that stage-II sleep has occurred.
Weighted total counts--A phase-locked relation Another episode of stage-II sleep apparently
appears to be developing between the weighted occurs at 09:25, but caution should be exercised
total-count profile and orbital revolutions, sug- in differentiating stage-I from stage-II sleep on
gesting that an arousal cycle is initiated at the the basis of weighted total counts only; this is
beginning of each new revolution and increases one of the paradoxical exceptions to the monotonic
for about 15 min to a peak value before relative relations between total counts and level of arousal.
relaxation as the revolution is completed. Tile A mixture of hypersynchronous alpha, associated
weighted total counts fluctuate within the Tl-zone with the lightest stage-I sleep (ref. 17), may have
for 35 rain of this time period. The fact that the weighted total counts ranging in the 80's; low-
weighted total counts do not identify exactly the voltage fast activity, also found in stage-I sleep,
same samples as those already identified by the may have weighted total counts ranging around
discriminant analysis as T1 shows that the indices 127. Such a mixture yields a combined total
are employing slightly different criteria for count that can incorrectly indicate sleep levels
recognition. deeper than stage-I. The possibility of such an
indication disappears as higher and higher time-
From 06: O0 to 12 :O0 resolution is employed in analysis of the EEG;
At the onset of sleep, slightly after 08:00, the as time epochs of 1 sec and less are employed, the
command pilot has been awake for about 15 EEG tends to become a "pure culture" representa-
hours; this time also corresponds to his "biological tion of the momentary state of consciousness.
time" of about 2300 hours. Both of the foregoing Here is an example of an ambiguous interpreta-
factors recommend interpretation of this 2-hour tion in the smoothed index of weighted total
sleep cycle as being tantamount to the first 2 counts that must be resolved by rcference to the
hours of a regular night's sleep for him. unsmoothed, independent, major-, intermediate-,
The flight plan was designed to allow the crew and minor-period count profiles. Figures 6 and 7
to sleep during hours generally corresponding to present the count profiles for the three periods
night at Cape Kennedy. This plan was followed plotted as average counts per second in 5-min
F_
_ I 7:1: i! i: _ 't---_
_*- "t : L_ : , '
i
.....;T7
i i:_ [,:± i_ i ,
: T i i : : I :
_iii' _ x !
:: mili __
:ii:
X
_ii .... _ ....... t i ,_ !, ',
;,
i:-it
:li[:i:i: f :!;I: :I:'

1
_
_
_t
_
_
_
'_
_
_r
t
t _1=_1 1::_='t
FIGURE 6.--The period-total counts for the first 28 hours of flight, displayed with the average
count for every 5 min.
data points for the full 54 hours of flight. The leads to speculation that some event was again
major-period count was 9/see while the inter- provoking a state of special vigilance.
mediate and minor counts were 22 and 42, Discriminant analysis--A 10-rain episode of
respectively. These values indicate that the stage-I sleep is seen beginning at 12:40, followed
distributions of the independent counts are in- by a return to the eyes-open state until the be-
compatible with stage-II as indicated by the ginning of a new sleep cycle at 14:15. At 14:25
weighted total counts, and corroborate the the deeper stage-I sleep of category-2B, the
interpretation that this activity was compounded presumed RES[ and dreaming state, is reached
stage-I. for the first time. Stage-2B is followed after about
10 min by stage-3B, the first occurrence of
From 12:00 to 18:00
stage-III sleep. At 14:45 the first stage-IV sleep
Missio_ e_ents--No direct information is avail- is noted with the appearance of category-4A.
able concerning mission events during this time, After only 5 min in stage-IV sleep, the state of
but the recurrence of the T_-state prior to 18:00 consciousness cycles almost immediately to the
7_ -[_T_ , T-- , ] i _ ;. i- __._.. T t ..i ._ _ : i..1_: _
7!,; !I,,_i,,,A,oAJ.
L__i- ! L_L7 _;..irrlT:! _ 1._.'4._rq._IL%,I_L .t
, m .r u ---Vi7 ,,
0
,o !_. ._
......,...... _
U: ."'-- •
:
.......
t . n , ;
, ....
,, MII .
_ ; _ _ -
__ _i _
! I1 : ! : : !;i: ;J 7tk oke
i " ' ; ' , , . : : FILI6H'i" I-toblt'_- t ; i , ..£-._ I I _ 4 t,;4.mt-.-I
:_ . i i !- -_ ', i I --"_ i !.,:;;--:T--]

--_
......
:-!-,-]:,
_F ' '___._
_" _'tli "ii ', : ._ ! t .... ? ! _ I _ ' i _ t--i ' I r-- _=t:=.'U_
l _.:'l
M
"
:
"
ii
"
;
"
t''"
"''"
l
"
;
"
--
;
'
1
1"
!
+
I,
I
I
I
1
(
I
i
t
!
1
t
_
+
'
"t
'
'
:ffm-_'_---,._--_';_-=_
t : r l " ; ; I " i ' " " I ...... ; . -l-::::'--:v--_2:_'-_2_2W__..,_.
-t ] 1 i.c_i Ll£12]
-;i {t I-_1"-'-'_
..... i; i ° -i, : i' _ ='I i......., 1..... I-,: :'.lt't t
_1 ._---w--hl _. I : _ i ----- ! I "
_-i _ i._?-'-., :_""_.I _ I,-_ I?_i+: =-F:=t=': _t:;]i.i

" -_ - _: _n_' i:' _ ..... i.......
I t;::"::P'_"
:_::_
_1 i;_;i -i :;71i;I '; " i i ! : LL_i -i i:: _- h J .__L_

L ' _ _::_t_
__Li ___L"i_-::ti::__
FIGURE7. Extension of figure 6 between 28:00 and 54:00.
lighter sleep of stage-I and early stage-iI, which close to stage-I sleep; this depression had not
continues for some 35 rain. Moderately deep been detected by the discriminant analysis.
stage-IV sleep of eategory-4B lasts for only 10 This now-familiar sleep segment accords with
rain at 15:35, and in the next 5 rain the state of previous interpretations except for the two epi-
consciousness returns to the eyes-open category. sodes of waking for 5 min at 14:30 and for 10 rain
For 5 rain, just prior to 18:00, again we see the between 14:55 and 15:05, the discriminant
now-rare Tl-state. analysis had classified these samples as stage-I
Weighted total counts--The weighted total-count sleep. The value at 14:30 is an example of a
profile comes suspiciously close to the stage-I corollary to the paradoxical exception in the
sleep level during 12:45, and the discriminant- interpretation of the weighted total count that
analysis profile also found stage-I sleep to be was discussed earlier. Relatively "pure culture,"
occurring. Forty minutes later another depression low-voltage, fast activity of st age-I sleep, par-
in the state of consciousness again comes quite ticularly of stage-I REM, may yield a relatively
high weighted total count that may incorrectly indicates a period of light sleep. Weak cycling
indicate a waking state. In such circumstances, continues in relation to revolutions.
reference to the independent period counts is
From 24 : O0 to 30 :O0
necessary for clarification of the interpretation.
The clear indication of waking with a weighted Mission eve_ls--The activity profile indicates
total-count value of 152 at 14:55 carries period that from just before 25:00 through 26:00 was a
counts of 13:28:44 which corroborate the fact mealtime, and the chewing artifact unquestion-
that this is an awake level. This second dis- ably accounts for the EEG findings of 25:00.
crepancy, between the discriminant-analysis clas- According to the flight plan a fuel cell was purged
sification and the weighted total-count assignment at 28: 40, and _indow measurements were taken
of state of consciousness, may be used to illustrate at 29: 40.
a point of considerable practical importance. This Discriminant analysis--A rather unusual range
discrepancy can also be resolved by reference to of states is seen during these 6 hours (fig. 3).
the discriminant-analysis profile of the second Extreme-muscle artifact preempts the record for
channel of EEG recorded during this flight. some 45 min from approximately 25:00. Light
The discriminant-analysis classification, of tile sleep of the 2B-categorb, is noted briefly at 20:40.
left parieto-occipital lead, agrees with the Two episodes of the Tl-state occur at 27:10
weighted total count of the midline lead that and 28: 35.
this is an eyes-open waking state. This fact Weighted total counts--The extreme-muscle ac-
demonstrates the practicality and even neces- tivity from the 25th to the 26th hour is not clearly
sity of multiple electroencephalographic leads; demonstrable in the weighted total-count profile
information from the second lead is not redundant of figure 4. Again this is an example of a smooth-
if any ambiguity is present in interpretation of ing statistic partially obscuring the information
the single-channel signal. This argument holds of the basic data. Figure 6 strikingly shows the
quite aside from additional neurophysiological muscle activity seen in the intermediate and
information that one may expect from multiple- minor periods as extremely higl_ counts, and a
site recordings and interrelations. slow-wave "che_ing '' artifact component seen
as the concurrent depression of the major-period
From 18:00 to 24:00
counts. This "mirror image" of the major-period
Mission events--The summary flight plan counts compared to the intermediate- and minor-
(ref. 18) indicates that the fuel cell was purged period counts is a characteristic artifact pattern.
at 20:50. From 21:00 to 21:30 the command It is now of increasing interest that the mission
pilot was engaged in a vision test, and at 22:25 events of this period relate strongly to the state
a radar-transponder test was scheduled, followed of consciousness reflected by the weighted total
by an exercise period. count in the T_-zone.
Discrimi_mnt analysis.--The eyes-open state
From 30: O0 to 36: O0
occupies the entire 6-hour period except for two
episodes of very light sleep appearing at 19:25 Mission events---The flight plan indicates that
and from 21:55 to 22:00. the radar transponder was turned "on" at ap-
Weighted total counts--The weighted total proximately 31:55 and turned "off" at around
counts show that the special state of vigilance, 32:10; there was another purge of the fuel cell at
referred to as the T_-zone, accompanies the 32:10. The summary flight plan schedules the
above mission events except for the period of beginning of the command pilot's sleep period
exercise. It seems most reasonable to assume at 33: 20.
that an event, such as purging of the fuel cell, Discrimi_mnt a_Talysis--Muscle artifact, alter-
would tend to reactivate the state of arousal nating with T_, occurs for some 25 min before
that earlier was associated with the differential- 32:00 (see fig. 8, a continuation of fig. 3). Early
pressure warning light; again the important part relaxation of the eyes-closed state begins around
played by the fuel cell in the flight of Gemini V 33:00 to develop into very light sleep at 33: 10;
is recalled. Shortly before 20:00 a count of 134 a brief oscillation back to the eyes-closed state
FIGURE 8.--Continuation of figure 3 between 28:00 and 54:00.
is followed at 33:25 by the beginning of the cycle of reasonable duration may be said to begin
at 34:15 and continue for 80 min. A brief arousal
deepest sleep cycle yet seen, dropping to the 4C
category of deep stage-IV sleep at 33:35. How- to eyes-open, returning through 2A and 2B cate-
ever, this deep sleep is very brief and is almost gories to a depth of 3B at 35:55, shows con-
immediately interrupted by 25 min of eyes-closed tinuance of the sleep cycle.
Weighted total counts--Figure 9 (continuation
resting and very light sleep. The first deep-sleep
Ll_4'r I
Fmv_v. 9.--Continuation of figure 4 between 28:00 and 54:00; the long sleep period between 33:00 and 41:00 is shown.
of fig. 4) shows the previously noted cycling by the entire flight. All stage-IV cycles are relatively
the state of consciousness continuing until the unstable during this time. The increased stability
onset of the deep-sleep cycle at 33:00. We see of the sleep profile in the last 25 percent of the
now, quite elegantly displayed, six episodes of sleep period argues for a return to a degree of
stage-IV sleep, with the first cycle at 33:45 neurophysiological homeostasis that had been
lasting for 10 min and the third at 35:35 lasting absent until this time in the flight. Once a_ain
only 5 min. The first and third cycles, with counts it is noted that the weighted total counts ap-
of 88 and 94, indicate a relatively light stage-IV proach or enter the T_-zone at times coincident
sleep. The second cycle of sleep, lust prior to with significant mission events.
35:00, is the deepest stage-IV seen throughout Band paranzeters----Figure 10 shows slow-wave
......... L4_,tttHNm_
_111itflttltflI ttti?lt
N4i,
tiI_tL_tltttttiiii
'I LLL IIi!I,_NI
+ ! ' ,1
i! !_lltl!ill!i!tl!ill!tittltlttfittt!!ii
_ '+ !i
!Ntii!iiiti! __ itfitiiilt
.........
_ftt_tttit
iI_l_!i!tH!f
iii;
i
.,tlil_it ll"'i'! : I t I ' il i it r l t
tt_}tttlItt:'7!
,,_ tt .L'N
tt_ttI _iNi!N!_!!!!ttii!H!!!H!!ii!f:
If!
"l!
N t_!tliHttt_f_Tjtttl_Rl
!ii'I
..... :::::
!_:tt2
Z 2 : :-- i _4-_1.
tttlltHtt.
PTl_li!itt! I_d:ll ti _J-I

_'llt,lt]l_ ' #i_i!t! lft t t4 #tlIH!_
_4.t t 14441,
tlNttt',
FzGvlt_ 10.--Comparison of the major-period delta_index band-1 (1 to 4 Hz), the intermediate-period
fast-theta-index band-3 (6 to 8 Hz), and the intermediate-period fast-activity band-10 (50 to 100 Itz).
(delta) activity from 81:30 to 37: 15, but between termination of the command pilot's sleep at
31:30 and 32:50 it is accompanied by fast fre- 41:30. Medical reports (ref. 16) of the astro-
quencies shown in intermediate-period-band-10 nauts' subjective evaluation of their sleep during
(50-100 Hz) and reflects high activity. The slow- this mission correspond nicely with the quanti-
wave activity from 33:30 to 37:15 is accom- tative evaluation of the analyzed EEG. The
panied by theta activity and reflects slow-wave medical observations are so important that we
sleep. The intermediate-period band-2 (4-6 Hz) quote them in toto:
differentiates these states very well, being high
Neither crewman slept as soundly in orbit as he
during sleep and absent during artifact.
did on the earth, and this inflight observation was
From36:00 to 42:00 confirmed in the postflight debriefing. The pilot

seemed to fall asleep more easily and could sleep
Mission events--The flight plan schedules more restfully than the command pilot. The
command pilot felt that it was unnatural to sleep complished. This was a change in the mission's
in a seated position, and he continued to awaken plan that took advantage of the excellent turn-
spontaneously during his sleep period and would
around progress at the launch site in preparation
monitor the cabin displays. He did become in-
creasingly fatigued over a period of several days, for the next launching. This deviation from the
then would sleep soundly and start his cycle of flight plan and the subsequent maneuvering may
light, intermittent sleep to the point of fatigue have been related to recurrence of a state of
all over again. vigilance.
The midline electrodes for the EEG became
Discriminant analysis---The fourth deep-sleep dislodged at 54:28. The left parieto-occipital
cycle of this period lasts for 20 min at the 4C- electrodes had apparently become dislodged
level, with lightening of sleep to the 3A-level for about 25 hours earlier (ref. 21).
15 min before waking to eyes-open at 37:15. Discriminant analysis--Eyes-open is the ex-
After some 15 rain of the eyes-open state there is clusive state except for one episode of T1 from
progressive oscillation down to the light stage-IV 43:50 to 46:05; the record then begins inter-
sleep of category-4A. Reappearance of the 2B- mit.tent oscillation between categories T1 and
category before 39:00 indicates that dreaming eyes-open that continues until the recording is
occurred during this span of moderate-to-light terminated by dislodgment of the electrodes at
sleep. The sixth and final deep-sleep cycle of this 54: 00. At 53:35 the dip to "2B" is artifact related
period is seen as a staircase progression down- to the last recorded meal.
ward from 39:00 to 40:00, reaching deep stage-IV Weighted total counls--Once again we see the
sleep of category-4C for 10 to 15 min at 40:00. state of consciousness entering into the Trzone
All evidence of sleep is absent from this index in relation to the vision test and the fuel-cell
after 42: 00. purge. The oscillation of the weighted total count
Weighted total counts--The depth of stage-IV within and about the T_-zone between about
sleep for the various cycles may be compared in 45:40 and 47:20 may well have been related to
the weighted total-count profile. The excellent the anticipation associated with the firing and
stability of the stage-I plateau during the 38th tracking of the Polaris missile. The Trstate seems
and 39th hours is rather dramatic; the stability to be closely related to events requiring special
of the final stage-IV cycle around 40:00 reaches attention and high-level performance, as evi-
a plateau at 93 counts per second. Waking occurs denced by the relation between known events
at 40:20 with a slight rebound into stage-I sleep during this flight and occurrence of this special
during the following 5 min, but with moderately elect roencephalographic state.
high arousal that peaks after 42:00. Progressive
decrease in the state of arousal culminates in Comparative Study
total weighted counts of 143 at 43:30--close to Having quantitatively analyzed the EEG re-
stage-I sleep. corded during this orbital flight and having
interpreted these data in state-of-consciousness
From 42:00 to 48:00
terms of sleep and alertness, we now turn to
Mission events---A vision test was scheduled comparison with results from other studies. Such
for 43:20 in the flight, plan and another fuel test comparison helps fractionation of the portion of
purge at 47:00. At some time during the 31st the observed disturbance of sleep that may be
revolution (ref. 19), around 48:00, a launched due to the following principal factors:
Polaris missile was sighted by the astronauts. (1) The mission profile of 14-day, 24-hour con-
Anticipation of this event and the excitement of finement and a programmed work schedule in
tracking an object under these dynamic circum- confined quarters
stances must certainly have created a special (2) The fact of deprivation of deep sleep dining
state of vigilance and arousal. the command pilot's first "night" of sleep
Some time during the 3rd day of flight, a (3) The degree of stress inherent in the ex-
partial-phasing maneuver was directed (ref. 20), tremely novel environment of orbital flight
and a posigrade burn of 12.4 ft/sec was ac- It is important to determine what fraction of
the disturbance seen in this flight profile is due deep-sleep cycles tend to stabilize at a given
to the orbital flight itself, in contrast with the level for 20 to 30 rain; no-stress sleep is expected
fraction that may be due to anticipation and the to stabilize at a given level of sleep for somewhat
circumstances of 14-day confinement in the longer periods. The last, half of the sleep profile
capsule. In terms of strictly scientific "control," tends to develop a plateau at a relatively light
the ideal situation would be for the astronaut of level of stage-II sleep; no-stress sleep also is
this flight, to simulate the flight without orbiting, expected to attain a plateau in light sleep during
so that the EEG might be recorded with all the last third of the night.
circumstances identical but for actual flight. A final transform of these data may be em-
Lacking this ideal if somewhat impractical con- ployed for evaluation of the sleep period. The
trol, we substitute a next-best set of data for lengths of time spent at various levels of sleep
comparative study. may be considered as percentages of total sleep
A series of 14-day-night Apollo missions were time. Under this normalizing procedure we find
simulated in the laboratories of Baylor Uni- the "Apollo" sleep profile to consist of the per-
versity College of Medicine. The purpose was centages indicated in table 4. The mean percent-
simulation as closely as possible of a three-man ages of no-stress sleep (ref. 24) make it clear
mission profile of the Apollo shots. One subject, that, whatever stresses and subsequent dis-
a 25-year-old candidate for Naval Officers School, turbance of neurophysiological homeostasis may
acted the role of systems engineer during one be associated with the "Apollo" mission, they
series of runs. A single-channel EEG, from are insufficient to produce more than moderate
essentially the same midline sites employed in disturbance of the first-night sleep--manifested
the orbital flight, was recorded and analyzed on primarily by a marked reduction of stage-I sleep.
line by the same period-analytic techniques. Let us now compare various episodes of sleep
While the entire 14-day run was analyzed, only in the space-flight profile with the "base line"
the weighted total-count data of the first sleep established from a simulated mission only.
period will be compared with the flight profile.
From 08:20 to 10:15
Figure 11 plots the weighted total counts in
5-min samples for the 9 hours of the first-night Considering this period as the first 2 hours of
sleep which began at about 0100 hours. The a night's sleep, we find it highly unstable and
subject had then been awake for some 15 hours, extremely light, with absence of sleep of stages
as had been the astronaut at the onset of his III and IV. This must be interpreted as a rela-
first sleep period in the 8th hour of f/ight. The tively severe disturbance of the sleep pro61e, as
extremely high value of 252 counts, 30 rain before compared with both no-stress sleep and the
the onset of sleep, reflects a required exercise "Apollo" data base. The percentages of the sleep
period. The "Apollo" sleep profile shows the levels in this episode show total deprivation of
following time-course attributes which may be stage-III and -IV sleep, with extreme deprivation
compared to expected characteristics based on of stage-II.
findings in large populations of experimental
From 14:20 to 15:40
subjects (refs. 22 and 23).
The first stage-IV-sleep cycle occurs within This sleep episode shows an inversion of the
the first 30 min from the onset of sleep; "no- expected time course, with each subsequent
stress" sleep is expected to show stage-IV slightly cycle becoming progressively deeper. The cycles
sooner. The first stage-IV-sleep cycle is the do not stabilize for any considerable length of
deepest cycle of the night; no-stress sleep also is time, and only the third and last cycle reaches
expected to show deeper stage-IV in this first stage-IV sleep. This sleep episode shows in-
slow-wave eyele than in subsequent cycles. The creased percentages of all deeper stages at the
three sequential slow-wave-sleep cycles tend to expense of stage-I sleep and waking.
become progressively less deep throughout the Interpreting this sleep episode as the first part
night; no-stress sleep is expected to show such a of the night's sleep, we still find severe disturb-
trend, with three or four cycles expected. The ance of the profile. One might, contend that the
_i!]', 1 ]J-
-- , - T
: I! :
-.T:_!t_- i
,. i; iiil!!
ff, :1 II:
.... tt
t.Z3 I.O I.I 1.2 1.3 1.4 1.5 1.6 1.7
Fmv_zz l l.--Weighted totM-count graph of the first night of the "Apollo" study.
time between 08:00 and 16:00 represents the "Apollo"; this cycle is considerably lighter than
total night's sleep that tile astronaut would have the second or third cycle, and the relatively deep
experienced had he not been on this mission; final cycle shows inversion of the expected se-
under this construction the sleep profile is seen quence. The occurrence of six cycles of deep
to be even more grossly disturbed than is indi- sleep is twice the degree of cycling of "Apollo."
cated by the previous interpretations. This degree of cycling is a disrupted profile prob-
ably produced by frequent arousals, movements,
From 33:05 to 40:35
and changes in the stage-IV sleep (ref. 25). Only
This sleep episode, the longest and deepest the final deep-sleep cycle shows the stability and
recorded on this flight, began about 40 hours duration that were present even in the sleep of
after the astronaut's last ground sleep ended. "Apollo." The last quarter of the profile shows
Thus it would be perfectly reasonable to consider an initial tendency to develop a plateau at
this to be his second "night's" sleep, but a more stage-II and st age-I sleep, but this tendency is
lenient interpretation of the degree of disturbance disrupted by the unusual final deep-sleep cycle.
is allowed in continuance of the comparison with The percentages in this sleep episode show that
the first night of "Apollo" base-line data. the deep-sleep time has considerably increased
The first-stage-IV-sleep cycle occurs 45 rain from that in any previous portion of tile flight,
after the onset of sleep, 15 min later than in and that this increase is largely at the expense of
138 BIOMEDICAL
RESEARC_ AND COMPUTER APPLICATION
TABLE 4.--Percentages of Time Spent Awake and in Four Stages of Sleep Under Four Conditions
Awake or Normal "Apollo" Night after 2 Orbital flight hours

in stage no-stress night-1 nights of stage-
of sleep sleep IV deprivation 08:20-10:05 14:20-15:40 08:20-15:40 33:05-40:35
Awake 0 0 0 57 30 71 14
Stage-I 29 7 23 39 25 15 19
Stage-II 49 63 50 4 25 7 19
Stage-III 8 15 7 0 10 4 17
Stage-IV 13 14 20 0 10 2 31
stage-I sleep which has been reduced to 19 per- the unknown stresses of orbital fight that ac-
cent. It may be argued that the degree of dis- count for the sleep profiles here reported. Sleep
turbance, represented by the above percentages, recorded during the early evening, rather than
may be simply explained by the deprivation of during the regular sleeping period, is not com-
deep sleep that has been so clearly seen during pletely adequate for establishment of the "basal"
the first "night" of the fight. This explanation sleep profile of a given individual. However, such
does not adequately explain the percentage levels base-line multichannel EEG's have been recorded
of the command pilot's second "night." Experi- from the command pilot both during sleep "and
mental studies (ref. 26) show that not even 2 under an active stimulus field; period analysis of
nights of deep-sleep deprivation produce the these records should be performed to enable
apparent "hunger" for stages III and IV seen in completion of the comparative studies and to
the flight. The percentages of the deep-sleep extend the power of behavioral interpretations.
stages are grossly increased at the expense of
stages I and II; the excessive waking time is of
DISCUSSION
course further evidence of a disrupted pattern.
We must conciude that anticipation of the The command pilot's EEG covered the first 54
14-day confinement and work schedule does not hours of orbital flight. From the point of view of
constitute enough stress to account for the dis- biological information, this was an historic occa-
turbance in the first-night sleep profile, and that sion, not only in marking the first recording of a
the deep-sleep deprivation of the first night does continuous EEG from an orbiting American
not explain the degree of disturbance of the sleep astronaut but also in providing the data base for
of the second night. It follows that the unique the most extensive analysis in the history of
circumstances of an orbital mission must produce electroencephalography. Almost every element in
additional stresses which are more disruptive of the total recording-playback system is special to
the sleep profile than are these other factors. It is this particular orbital flight: the sensors and re-
certainly known that the events and stresses of cording devices, the reproduction and dubbing
a day influence the subsequent sleep profile techniques, and the replay for analysis of copies
(ref. 27), and that simple real-life stresses such of the master analog magnetic tape. The total
as the loss of a billfold or a husband's fight with system worked remarkably well, and even the
his wife decrease the sleep percentage of stages components that failed (the electrodes that
III and IV during the subsequent night. became dislodged) contributed the valuable
All-night sleep recordings from a given indi- negative information that more research and de-
vidual, over at least 3 or 4 nights (ref. 28), are velopment was needed in this problem area of
required for quantitative determination of the electrode configuration for long-term applications.
degree of stress represented by certain percentage An event as unique as this flight is reported
decreases of the various stages of sleep. We need observationally rather than as a scientific ex-
multiple-night sleep recordings from the com- periment. This observational report emphasizes
mand pilot before we can evaluate the degree of individuality but requires comparative studies
for relation of the individual case to what is to the stimulus as well as high enough resolution
known from rigorously controlled experimental and sensitivity to allow for study and prediction
studies. For this reason, results of the Apollo of human interaction. Certainly no other single
simulation have been used for comparative psychophysiological or psychological measure can
interpretation. The depth of sleep and the be expected to suffice. After the fact it is quite
temporal evolution of the sleep profile have been clear that valuable scientific information was lost
quantitatively mea._ured by period analysis of during this flight because the pilot was not
the EEG both in this flight and in "Apollo," but instrumented for an EEG.
further experimental work is especially essential
for determination of the relation between stress
ACKNOWLEDGMENT
in an individual and the consequent degree of
disturbance in his sleep profile. We thank both the disciplines and the indi-
This study has strongly demonstrated the need viduals participating in this work, not in the
for high-resolution documentation of current usual sense of thanking but to make clear the
mission events during the flight. Period analysis degree of interdisciplinary effort that was re-
of the EEG offers a high-resolution measure of quired for its accomplishment. The results speak
alertness that can be translated into interpreta- for themselves; in their scope and detail they
tion of stimulus-response appropriateness only if are perhaps as unusual a contribution to the
the stimulus field of the moment is known. The state of the art as was this orbital flight itself.
response of the alert individual--whether by
task-performance or by EEG--should be ap- REFERENCES
propriately related to the intensity or significance
1. AGr,mw, H. W., JR.; WEBB, W. B.; ANn WILLIAMS,
of the stimulus. For a high-intensity stimulus the
R. L.: The First Night Effect: An EEG Study of
EEG response should show high arousal; for a Sleep. Psychophysiology, vol. 2, 1966, pp. 263-266.
low-intensity stimulus it should show low arousal. 2. BURCH, N. R.; ANn GREINER, T. H.: A Bioelectric
For a stimulus requiring rapid high-level per- Scale of Human Alertness: Concurrent Recordings
formance, the performance response should be of the EEG and GSR. Psychiat. Res. Repts., vol.
12, 1960, pp. 183-193.
commensurate. A stimulus allowing for relatively
3. ADEY, W. R.; KADO, ]'_.T.; A_ WALTER, D. 0.:
slow and approximate performance should evoke Computer Analysis of EEG Data from Gemini
more-relaxed behavior. In brief, the response Flight GT-7. Aerospace Med., April 1967, pp.
should be appropriate to the stimulus; it may be 345-359.
said that such a relation operationally defines 4. M^ULSRY, R. L.; FRosT, J. D., JR.; ASP GRAHAM,
good performance. M. H.: A Simple Electronic Method for Graphing

EEG Sleep Patterns. Electroencephalog. Clin.
The EEG signals from the single- and double- Neurophysiol., vol. 21, 1966, pp. 501-503.
channel placements have been shown to provide 5. PROCTOR, L. D.; ET AL.: The Analysis of Gemini-7
an index of arousal and of depth of sleep that EEG Data by Three Different Computer Oriented
has direct relevance to performance, but we are Methods. NASA report under contract NSR-
23-003-005.
just beginning to exploit the full power of multi-
6. MAULSBY, :R. L. : Electroencephalogram during Orbital
channel recording as a measure of adequate re- Flight. Aerospace Med., vol. 37, 1966, pp. 1022-1026.
sponse. The muItichannel EEG, with appropriate 7. KELLAWAY, P.; AND MAULSBY, R. L.: M-8 Experi-
analysis for interrelating the activity between ment: Inflight Sleep AnMysis Gemlni-VII Flight.
multiple sites, promises to be an extremely NASA preliminary report, Feb. 7, 1966.
8. BURCH, N. 1_.; AND CHILDERS, II. E.: Information
powerful tool for study of human behavior.
Processing in the Time Domain. Information
Important for future long-term missions are the Storage and Neural Control. Charles C. Thomas,
questions to be answered about continuing long- Springfield, Ill., 1963.
term interpersonal reactions between from three 9. SALTZBERG, B.; AND BURCtt, N. R.: A New Approach
to five crewmen. The EEG promises to aid greatly to Signal Analysis in Etectroencephalography.
I.R.E. Trans. Med. Electron., 1957, p. 24.
understanding of the interactions and dynamics
10. COOLEY, W. W.; AND LHONES, P. R.: Multivariate
in a small human group; it may well be the only Procedures for the Behavioral Sciences. John Wiley
measure of response having adequate specificity and Sons, Inc., 1962.
140 BIOMEDICAL
RESEARCH
ANDCOMPUTER
APPLICATION
ll. Jo_vET, M.: Paradoxical Sleep--A Study of Its 21. I'{ELLAWAY, P.: Experiment M-8, Inflight Sleep
Nature and :Mechanisms. Sleep Mechanisms. Progr. Analysis. NASA SP-121, vol. 45, 1966, p. 424.
Brain Res., vol. 18, 1965, pp. 20-62. 22. DEMENT, W.; AND KLEITMAN, N.: Cyclic Variations
12. BERRY, C. A.; COONS, D. O.; CATTERSON, A. D.; AND to Eye Movements, Body Movements, Body
KELLY, G. F.: Man's Response to Long-Duration Motility and Dreaming. Electroencephalog. Clin.
Flight in the Gemini Spacecraft. NASA SP-121, Neurophysiol., vol. 9, 1957, p. 673.
vol. 25, 1966, p. 235. 23. WILLIAMS, II. L.; HAMMACK, J. T.; DALY, R. L.;
13. GmssoM, V. I.; McDIVITT, J. A.; COOPE_, L. G.; DEMENT, W. C.; AND LUBIN, A.: Responses to
SCHIRRA, W. ]_{.; AND BORMAN, F.: Astronauts' Auditory Stimulation, Sleep Loss and the EEG
Reactions to Flight. NASA SP-121, vol. 27, 1966, Stages of Sleep. Electroencephalog. Clin. Neuro-
p. 272. physiol., vol. 16, 1964, pp. 269-279.
14. MmLICCO, P.; COHEN, R.; AND DEMING, J.: Electrical 24. WILLIAMS, R. L.; AGNEW, It. W.; AND WEBB, W. B.:
Power and Sequential Systems. NASA SP-121, Sleep Patterns in Young Adults: An EEG Study.
vol. 6, 1966, p. 54. Electroencephatog. Clin. Neurophysiol., vol. 17,
15. ]{ODGE, J. D.; A._D ROACH, J. W.: Flight Control 1964, pp. 376-381.
Operations. NASA SP-121, vol. 20, 1966, p. 184. 25. MONROE, L. J.: Psychological and Physiological
16. BERRY, C. A.; COONS, D. 0.; CATTERSON, A. D.;
Differences between Good and Poor Sleepers. J.
A_,_ KELLY, G. F.: Man's Response to Long-
Abnormal Psychot., vol. 72, no. 3, 1967, pp. 255-264.
Duration Flight in the Gemini Spacecraft. NASA
26. AGNEW, tt. W.; WEBS, W. B.; AND WILLIAMS, R. L.:
SP-121, vol. 25, 1966, p. 241.
The Effects of Stage Four Sleep Deprivation.
17. FouLKES, D.; AND "VOGEL, G.: Mental Activity at
Electroencephalog. Clin. Neurophysiol., vol. 17,
Sleep Onset. J. Abnormal Psychol., vol. 70, no. 4,
1964, pp. 68-70.
1965, pp. 231-243.
27. LESTEa, B. K.; BUNCH, N. R.; AND DOSSE_r, R. G.:
18. ANON.: Mission Program Profile of the Gemini-VII
Nocturnal EEG-GSR Profiles: The Inflnence of
Flight. NASA-S-66-170 (fig. 7.1.1).
Pre-Sleep States. Psychophysiology, vol. 3, no. 3,
19. BRE_,'TNALL, B. : Experiment D4/DT, Celestial Radiom-
1967, pp. 238-248.
etry and Space-Object Radiometry. NASA SP-121,
vol. 38, 1966, p. 374. 28. WILLIAMS, II.; AND WILLIAMS, C.: Nocturnal EEG
20. HODGE, J. D.; AND ROACH, J. W.: Flight Control Profiles and Performance. Psychophysiology, vol. 3,
Operations. NASA SP-121, vol. 20, 1966, p. 186. no. 2, 1966, pp. 164-175.
CHAPTER 9
ANALYSIS OF ELECTROENCEPHALOGRAPHIC
DATA FROM ORBIT BY THREE DIFFERENT
COMPUTER-ORIENTED METHODS
L. D. Proctor
The problem in assessment of electroencephalo- frequency components of the EEG but. disregards
grams (EEG's) has been the subjective element the amplitude of the signal.
in visual evaluation of such records, as well as The EEG is recorded in analog form on mag-
the need for extensive training (2 to 3 years) netic tape before the tape is played back through
before a suitable candidate can perform this task. a digitizing system. Two such systems were used:
There have been a variety of computer-type an IBM system produced a digitized record of
analyses of EEG's during the last decade, and we 312.5 points per second of real-time recording; an
are still searching for a reliable, inexpensive, ASI (Advanced Scientific Instruments) system
on-line, data-processing procedure. We are aware yielded 242 points per second of real time. In
that our Soviet colleagues routinely assess EEG's both cases the rule-of-thumb requirement of 5
by computer, using relatively inexpensive hard- points per highest frequency analyzed was met
ware. since we had agreed to ignore all frequencies
Here I endeavor to show how four different above 50 Hz in the EEG. Both digitizing systems
assessment procedures (visual assessment, zero- had sufficient core memory to permit an entire
crossings technique, smoothing and peak-counting EEG record to be digitized without interruption
technique, and Weibull statistic) may be utilized. and subsequent loss of any digits; in both, the
The advantages and disadvantages of each will range of the scale value of the digits was from 0
also be indicated. to 510. The initial base-line value for the EEG is
I wish to point out that, in the space-flight chosen at the midpoint of the range (in this case
analog records of EEG's supplied to Henry Ford at a value of 255). This base line is subsequently
Hospital, channel-II (midline-central to midline- altered as follows: The values between a present
occipital electrodes) could not be assessed re- upward crossing of the base line and the previous
liably after about 26 hours of flight (26:00). upward crossing are averaged. The value of this
Channel-I (left-central to left-occipital electrodes) recent upward crossing is then subtracted from
provided the more assessable EEG until about the average. This difference is then multiplied by
54:00. These channels during the first 26 hours 0.05, and the result is added to the present base-
of flight showed no significant difference by our line value. With use of this method the arbitrary
four assessment procedures. Therefore, this re- base line tends to follow the drifts in the average
port will be limited to the findings from analyses value of the signal. The small weighting factor of
of channel-I during its approximately 54 hours 0.05 was used so that the base line would follow
of recording. only substantial and consistent shifts.
As the values of tile digitized EEG cross the
base line, the time between each pair of sequential
ZERO-CROSSINGS TECHNIQUE
crossings is measured and stored in two sets of
The zero-crossing method for reduction of registers. In one set it is stored as a half-wave of
electroencephalographic data is a simple pattern- a particular frequency (0.5 to 50 Hz) depending
recognition program that arbitrarily defines the on the time between crossings; in the other set
141
the amount of time is stored. Thus there are two this period, but also because it seemed reasonable
sets of registers: one for frequency counts and to assume that the subject was in an alert, active
one for total time per frequency for a given state during this time rather than during the
sample of EEG. Each set contains 51 individual relatively routine hours later in the flight. It
registers (one register for each frequency from must be emphasized, though, that this procedure
0.5 to 50 Hz). From these registers we extract the incorporated several drawbacks, not the least of
frequency counts and the total time present in which was a biased standard that would not be
each frequency band that has been established present if preflight EEG's were available.
in clinical use; that is, delta (0.5-3 Hz), theta Because of the correlation between individual
(4-7 Hz), alpha (8-15 Hz), beta-1 (16-25 Hz), means and their respective variances, the para-
beta-2 (26-35 Hz), and beta-3 (36-50 Hz). metric analysis transformed each raw score
[frequency count over successive 15-sec intervals
(Z/C) or 1 min intervals (SPC)] according to
PARAMETRIC ANALYSIS OF THE
the following re!ation (ref. 1):
SPACE-FLIGHT EEG'S
X'=X+0.5
Even cursory visual examination of any electro-
encephalographic record reveals, if nothing else, where X is the raw score from either Z/C or
the extreme variability of the recorded potentials SPC, and X' is the transformed score used in
_ithin a given subject. When the data are sum- subsequent analysis.
marized by automatic methods such as zero- This transformation was carried out for
crossings (Z/C) or smoothing and peak counting homogeneity of variance over the entire recorded
(SPC), which hopefully eliminate intuitive inter- period, and all remaining parametric calculations
pretations, the problem becomes particularly were performed on these transformed scores.
acute and one must define a standard or "normal" The standard epoch (first 7 hours of flight) was
population against which the reliability of a given then ordered into four separate populations ac-
signal change may be estimated. That is, are the cording to classical delta, theta, alpha, and beta
correlations of fluctuations in an electroencepha- frequency-band criteria, with each population
lographic signal truly representative of a change described by the mean and standard deviation.
in the physiological status of the organism, or The remaining flight record was then analyzed in
simply random variations independent of the be- blocks of 15-rain (Z/C) or 10-rain (SPC) periods
havior in question? Ideally one would desire: (1) with the mean amount of activity occurring in
a library of data, taken from many subjects under each of the four frequency bands of these periods
standardized conditions, that would provide being compared to the mean and standard devia-
maximum generality and a broad base for com- tion calculated for the standard epoch.
parison; or, alternatively, (2) repeated measures
taken from the same subject that, although re-
Z/C METHOD OF DATA REDUCTION
stricted to one individual, would provide some-
what greater power for comparative techniques. Figure 1 presents the mean activity computed
However, the ideal is not often realized, and the for each of the analyzed frequency bands over
present analysis was forcibly restricted to less- 41 hours of the flight. Due to noise and extreme
desirable alternatives. fluctuations of the signal, the subsequent record
Thus, while the entire raw EEG record was was not analyzable by the Z/C method.
reduced by the methods described, the parametric Representing the initial step in the parametric
analysis dichotomized the record into two sepa- analysis, figure 1 provides an overall visual
rate epochs which respectively provide (1) a presentation of the electroencephalographic ac-
standard population (first 7 hours of flight) and tivity and demonstrates in a general manner how
(2) estimates of the subject's state of conscious- the indicated overt activities of the subject pro-
ness during the remainder of the flight. The first duce changes in the recorded potentials. For
7 hours were chosen as the standard not only be- example, the stable first 7 hours of flight (our
cause the EEG _gnal was extremely stable during standard epoch) contrast with the 8th hour when
ELECTROENCEPHALOGRAPHIC DATA FROM ORBIT 143
°>- Jll. ld ^
(nm_w _
I- hJ (nLdE.
,,_ _ _d_ ,h.I
3O
12=1[---_ ...... _ H
'.. I ',= REST
,<
/" .-_.
i:':,,....,-,,./'-'.,'-..-:,.:__
i:
....
" ',.'..,,".,
h."V:_",.:
'Al:'"-:_'_..r,,,
.L //_= l_u JJ_J _ Lu ELi _J_LLL±LI • L_Lt * • L_= _ L_ LJ L_. IJ. ' • I , , , I , , , I , , , I • • ,/_...._
PRE- I 2 :3 4 5 6 7' 8 9 I0 II I2 t3 14 1,5 16 17 18 19 20 21
FLIGHT HOUR OF FLIGHT
o
>
3O
H b--t
ii, t
I-.-
f.J ..... BETA I
_ 1 REST ,,..,
2O
i',.
....,
.... ......... .:, _ ',_
_._,-.,_..,_ ". _'_
._ _ /-_ ..... ALPHA
z
_,I,,,.-_..
/.._._.. _. _. _-"._..::--
./_._ -.M(\_
Ld :':;,..,',-'.,, _ ,,*.'.. -,,,_ _-_.'_i._ "_-',._( :, .... THETA
,:..,-q " DELTA
I0
_.I..LU.JLUJILXX_ ,_ L ,._LL_I ,_Lt 1, ,_1_ , I ,_lt • [U 1_l, , J ___.£LL*L_,IL_Ludl

22 23 24" 2_ 2(/ 27 28 29 30. 31 32 33 34 35 36 37 58:59 40 41
HOUR OF FLIGHT
Fmtra_ 1.--Mean activity computed for each of the analyzed frequency bands over
the first 41 hours of flight; raw output from zero crossings.
the astronaut was relaxed with eyes closed and parison is summarized by a histogram (fig. 2)
effecting (as expected) substantial increase in sho_5ng as individual bars the mean activity
recorded alpha activity. Due to the manner in within each of our defined bands and also the
which the individual frequencies are derived from results of the applied statistic. There is a sig-
base-line crosses (see description of Z/C methods), nificant decrease in delta activity recorded during
the quality of the EEG signal may be inferred the initial 10 min of flight from that recorded
since any sharp decrease in quantity of activity before lift-off, while alpha and beta exhibit
of all bands above delta is indicative of off-scale significant increases for the same comparison;
potential shifts or large amounts of muscle change in theta activity was insignificant. It
activity (note particularly the indicated eating must be noted, however, that a large portion of
periods) carrying the signal away from the base the lift-off period is characterized by extreme-
line. Thus, while figure 1 is quite general, it does muscle artifact that could conceivably account
indicate that gross behavioral states are indeed for a goodly portion of the increase in the high-
reflected in the automatic analysis of the recorded frequency (beta) activity.
signal. For determination of the subject's state of
With the data summarized in this manner one consciousness over the recorded period of the
can specifically compare the changes in amounts flight, the parametric analysis proceeds by initial
of activity that occurred after lift-off by simple calculation of the mean activity in each of the
t-tests for differences between the mean activities frequency bands during successive periods lasting
recorded during the 10-min period prior to lift-off approximately 12 min. The variability in the
and during the first 10 min of flight. This com- lengths of the individual periods and the number
completely agrees with visual analysis of the

electroencephalographic sleep records. It should
5
be noted that the sleep period depicted in figure 4
corresponds to the results of the visual analysis
(ref. 3) and is in disagreement by about 4 hours
o
with the initial actual sleep period indicated by
o
the flight log supplied to Henry Ford Hospital
3
along with the analog tapes of the flight. However,
o the visual analysis does indicate that there is an
li l:
2
"eyes-closed" period at the initiation of the rest

I
period, but that actual sleep did not occur until
about 14:00, with the intervening period charac-
terized by relative wakefulness. Our parametric
DELTA THETA ALPHA BETA
analysis, on the other hand, indicates that the
t= iO.T9 1= i.'/'8 f - 3.04 t= B,.3Y subject did sleep very lightly (stage-I only) as
P< .01 P= N.S. P< .OI P<.OI
shown by the flight log (see fig. 1), but the
FIOURB 2.--Comparison of mean activities during 10-rain period was classified as "eyes closed but awake"
periods preceding and following lift-off, and results of by visual criteria.
the applied statistic.
The parametric analysis has been applied with
varying success to both of the described methods
of periods in each hour result from change in of data-reduction: Z/C and SPC. While the Z/C
quality of the signal and consequent rejection of technique provided the most useful and sensitive
data. The individual means are then presented in data for parametric anal) sis, one should also note
figure 3 as they occur outside the bounds of -4-1 that the standard-deviation criteria were origi-
standard deviation of the standard epoch. Thus nally set out for this method without regard to
figure 3 shows the occurrence of activity that the characteristics of the SPC analysis. At least
deviates from what we have defined as activity for the present there seems to be no a priori
recorded when the subject is in a highly alert and reason to assume that a similar parametric
active state. analysis could not be conceived for SPC once
With the EEG activity depicted as in figure 3, the program for various parameters were estab-
one can proceed to more detailed analysis and lished from a sample of data considerably larger
determination of the various stages of sleep than from the recording of one session from one
(ref. 2) by evoking the following definitions: individual. However, parametric analysis and
its benefits can only be applied usefully to Z/C
(1) Stage-0--if all activity is within 1 standard
data with realization of the shortcomings of the
deviation of standard epoch; and if only alpha
Z/C technique, particularly with regard to analog
activity exceeds 1 standard deviation
records of poor quality.
(2) Stage-I--if alpha and thcta mean activities
Despite the problems inherent in the simple
exceed 1 standard deviation of standard epoch
Z/C analysis (inability to use poor-quality
(3) Stage-II--if only theta mean activity ex-
records such as those derived during eating),
ceeds 1 standard deviation of standard epoch
the method in combination with our standard-
(4) Stage-II!--if both theta and delta activities
deviation criterion has provided extremely high
exceed 1 standard deviation of standard epoch
resolution of the subject's state of consciousness
(5) Stage-IV--if only delta mean activity ex-
as recorded during the flight before the signal
ceeds 1 standard deviation of standard epoch
began to deteriorate at about 42:00. The benefits
Figure 3 represents the results of application of of this an_!ysis are further demonstrated by its
these criteria to the data as presented in figure 2. ability to classify extremely light sleep (for ex-
Each figure indicates the subject's state of ample, that occurring between 08: 00 and 10: 00)
consciousness during 12-min periods by the which is classified as "eyes closed but awake" by
vertical extension of the individual bars, and routine visual analysis. These facts permit speci-
BETA I
ALPHA
o%_i' THETA
t
o °
DELTA
b--q _ _ }--t H _-4 I II

REJECTED LIGHT SLEEP REJECTED REJECTED REJECTED SLEEP
DATA SLEEP DATA DATA DATA
I 2 3 4 5 IO 15 20 25 30 55 40
HOUR OF FLIGHT
FiGmm 3.--Standard-deviation analysis with Z/C data.
n
5
_l
_> 2
bJ
0 _
I 2 3 4 5 I0 15 20
HOUR OF FUGHT
...J
f/')
iu 2 i
0
Ld
0
25 50 55 40
HOUR OF FLIGHT
FmVR_ 4.--Degree of consciousness from zero crossings and standard-deviatlon criterion.
fications of reliability of the EEG signal in terms Finally it should be emphasized that evaluation
of probability of occurrence of "EEG"-indicated of the EEG signal by Z/C is independent of the
behavioral states, and have provided the electro- individual reading the record since the analysis is
encephalographer with a quantitative tool hereto- divorced from any subjective interpretations of
fore unrealized in the clinical evaluation of EEG's. the signal. This system (Z/C and standard-
SIGNAL I
deviation criterion) provides mutually exclusive
categories incorporating the full range of EEG
frequencies, the resultant benefit being a sub- VALLEY , lID
stantial reduction in ambiguity as to the state of

consciousness of the subject. Approximate real-
time analysis of electroencephalographic indicants
of states of consciousness for any individual
capable of reading numerical computer printouts
is now available.
DIGITAL SMOOTHING AND PEAK-COUNTING

TECHNIQUE
IIGNAL 2_
Let us consider one channel of "digitized"
electroencephalographic signal: a series of positive
integers proportional to the value of the signal
voltage sampled at regular intervals. The process
basic to the analysis is identification of maxima
(peaks) and minima (valleys) in the digitized
signal. A maximum is identified by a series of
three points, the middle of which is greater than
the other two; for a minimum, the middle point FmvaE 5.--Illustrations of wave properties.
must be less than the other two. (When several

successive points have the same value, and these
points are followed and preceded by points both complexity--waves 4, 5, and 6 are superimposed
of which are less or greater in value, then a on wave-3; and waves 2, 3, and 7 are superimposed
maximum or minimum, respectively, is said to on wave-1. If appropriate low-pass electronic
have occurred at the middle value.) A pair of filters were applied prior to digitizing of the
successive maxima define the boundaries of a signal, subsequent analysis might yield only
valley wave, and a pair of successive minima waves 2, 3, and 7; if a low enough filter were
define a peak wave (fig. 5). We are concerned used, analysis might yield only wave-1. We may
with four properties of individual waves: fre- define simple waves as those that are identifiable
quency, amplitude, symmetry, and complexity. without filtering and complex waves (waves 1
In addition we may consider small groups or and 3 in fig. 5) as those requiring filtering for
sequences of waves according to conditional identification. Complete specification of the com-
probabilities or in terms of larger patterns. plexity of a particular wave would require
Figure 5 illustrates the definitions of the various knowledge of all waves (revealed by all possible
wave properties. Signal-1 includes valley and filters) that overlap in time with the given wave.
peak waves. Note that the right-hand minimum, In the specific analyses described below, a
defining the peak wave, is the middle point of process of digital smoothing by computer is per-
three equal values. For the peak wave, the fre- formed, instead of electronic filtering, to reveal
quency is the reciprocal of the duration (D) in this property of complexity. In the first of several
seconds; its amplitude is the average of A and B, stages (arbitrarily chosen) of smoothing, all waves
and its symmetry is the ratio A:B (or B:A, are identified and those greater than 50 Hz in
whichever is larger). Similar definitions apply to frequency are replaced by points of constant
the valley wave. value so that they will not affect the identifica-
It is well known that, in the EEG signal, waves tion of maxima and minima. (The smoothing
of different frequencies may be present simultane- process is described in detail below.) The new
ously; fast waves are superimposed on slow waves. maxima and minima are determined, and the
Signal-2 (fig. 5) illustrates this characteristic of wave properties are computed. The frequencies,
amplitudes, and symmetries determined after UNSMOOTHED SMOOTHED
50-Hz smoothing constitute one set of data. A

Similarly the smoothed signal is now resmoothed
so that any waves having frequencies greater
than 25 Hz are replaced by constant values. The
wave properties are again determined, and a set
of data for 25-Hz smoothing is obtained. Similarly
sets of data for 15- and 8-Hz smoothing are
determined.
The details of the smoothing process are im-
portant, and many other methods were tested
before the present one was chosen. The most
obvious method might be some type of averaging
process; such methods are unsatisfactory because
they provide a very unpredictable frequency
cutoff and may lead to generation of frequencies
not present in the signal because of phase rela-
tions (kno_m as "aliasing" in power-spectrum
theory). Another problem is that a high-ampli-
tude, short-duration spike may be counted, after
averaging, as a low-amplitude slow wave. Another
possibility is simply use of successive stages of
electronic filtering; for many applications this
may be adequate, but it is probably less con- UNSM_
venient, and some of the problems associated
with averaging also apply to electronic filtering.
In the smoothing process described below, a peak
or valley wave is replaced by points of constant
value if it is of greater frequency than a specified
cutoff. Before the present method was chosen,
several variations were rejected on logical grounds
SMO_
(often after testing of revealed weaknesses) such
as smoothing on the basis of criteria for half-
waves, use of a replacement process in which the
FIGURE 6.--IUustrations of the digital smoothing process.
constant value runs from minima to minima or
from maxima to maxima (see below), and smooth-
ing of only peak waves. according to one of the four procedures illustrated
The procedure used for digital smoothing is as in figure 6A. If a peak wave is smoothed, we re-
follows: If a peak or valley wave is of greater place the points shown by values equal to the
frequency value than the specified cutoff, it is greater of the two minima bounding the wave.
smoothed; but if it is of smaller frequency, it Note that the replacement begins at the greater
remains in its original form. As long as the minima and proceeds across the wave until the
previous wave has not been smoothed, we con- wave boundary is reached. Valley waves are
sider every (overlapping) peak and valley wave. smoothed by replacement according to the
If a peak wave is smoothed, the next (adjacent) smaller of the two maxima. It is important to
peak wave (the overlapping valley wave being note that, if the initial smoothing criterion is
disregarded) is considered. Correspondingly if a chosen at too low a frequency, all waves above
valley wave is smoothed, the adjacent valley the criterion frequency may not be smoothed.
wave is next considered. Smoothing must either be repeated several times
By smoothing we mean replacement of points at a low frequency or be done by use of a series
148 BIOMEDICAL RESEARCH AND COMPIJTER APPLICATION
of successively lower stages. To clarify this point Hz,... 15.00 to 16.99 Hz, and greater than
let us say that we are initially smoothing at 20 17.00 Hz. The amplitude categories were 1 to
Hz. Two 70-Hz waves may be smoothed to form 19, 20 to 79, and $0 units or greater. The units
a 35-Hz wave which is above the cutoff. If tile come from an arbitrary scale of 0 to 510, which
smoothing process is repeated, the 35-Hz wave was used for digitization of the signal. One
disappears. hundred units, the amplitude of a large alpha
Figure 6B shows the smoothing process applied wave, was about 50 _V. Symmetry was defined
to a hypothetical signal. Six maxima and seven above with reference to figure 5; the measure of
minima are shown (by arrows) for the unsmoothed symmetry used was whichever of the ratios A:B
signal, but for the smoothed signal there are two and B:A that exceeded 1. The categories for
maxima and three minima. symmetry were 1.000 to 1.200 (symmetric), 1.201
A feature of the pattern-analysis is the ease to 3.000 (asymmetric), and greater than 3.000
with whieh eertain types of noneleetroeneephalo- (very asymmetric).
graphic signals can be recognized and rejected. Five 10-sec epochs of EEG (fig. 7), each repre-
Muscle potentials appear in the form of high- senting a stage of sleep [as interpreted by Maulsby
amplitude, high-frequency waves. Appropriate (ref. 3) and verified by me] from the second flight-
amplitude and frequency criteria can be estab- sleep period, were compared. They were chosen
lished, and, if a given number of such waves are for being relatively homogeneous throughout the
found within a period of time, that period can be 10-sec period because they contained no artifacts
disregarded. Another type of signal easily recog- by the criteria described and because they fitted
nized as noneleetroeneephalographic is a number clearly into their respective categories.
of consecutive points of extremely high or low Since there are so many categories to be con-
value; such signals are characteristic of record- sidered and because it is desirable to see all the
ings taken with loose eleetrodes. Illustrations of categories together, it was decided to represent
artifact-rejection are presented below. the number of waves falling into each category
by a circle of which the diameter is proportional
to the number. In figure 8, grouping is by sets
ANALYSES BY DIGITAL SMOOTHING AND
of nine categories (3X3 arrays) in which ampli-
PEAK-COUNTING TECHNIQUE
tude increases from left to right and asymmetry
Several different versions of the analysis de- increases from top to bottom. Thus the upper-
scribed in the General Method section were right corner of a 3 N: 3 array represents an almost
applied to the flight-EEG data (channel-I). The asymmetric wave of high amplitude. The lower-
first version to be discussed is a complete cate- left corner would be a highly asymmetric wave of
gorization of the waves aeeording to frequency, low amplitude. It is apparent from figure 8 that
amplitude, symmetry, and smoothing cutoff; it most of the activity falls in the middle-amplitude,
is applied to several representative epochs of the medium-asymmetry category (all circles except
signal. This analysis is presented first so that the the middle are filled for emphasis of deviations
reader will undemtand better the general method from the middle ranges).
and the more-summary analyses that follow. In examining figure 8 it is well to keep several
These summary analyses will cover the total ideas in mind. The first point is that eertain bands
amount of valid EEG signal. must have values of zero since smoothing is on
A_mlysis-l--In this analysis, EEG waves were the basis of frequency; for example, 8-Hz smooth-
categorized according to 10 ranges of frequency, ing removes all alpha activity from the signal.
three ranges of amplitude and symmetry, and Another point is that at high smoothing cutoffs
four criteria for smoothing. Frequencies greater only simple waves (those without higher fre-
than 50, 25, 15, and 8 Hz were successively quencies superimposed on them) are identified.
smoothed out of the signal. For each smoothing It was found, for example, that all delta activity
cutoff, valley and peak waves were classified as (1-3 Hz) is complex since the 50- and 25-Hz
follows: The frequency bands were 0 to 0.99 Hz, smoothing eategories are all zero. (The 50-Hz
1.00 to 2.99 Hz, 3.00 to 4.99 Hz, 5.00 to 6.99 smoothing category is not shox_na sinee it is very
34-05 STAGE O
_-55 STAGE I
54-17 STAGE 3
:54-24 STAGE 4 A/_ ^^ _ /_
l sec 150/JV
FIGURE 7.--Five 10-see epochs of EEG, each representing a stage of sleep

used for the analysis illustrated by figure 8.
similar to the 25-Hz category'.) The third point rhythm (around 8 to 13 Hz). Figure 8 shows
is that for slow waves the lower-amplitude cate- large amounts of 9- to ll-Hz and 11- to 13-Hz
gories have little meaning in terms of the ordinary activity at all smoothing cutoffs. Stage-I of sleep
visual analysis of EEG's. When a signal is pro- is characterized by its low-voltage activity with
gressively smoothed at lower and lower frequen- complete lack of spindling. The categories for
cies, slow waves appear even if they represent stage-I show that the activity is spread through
only some kind of statistical fluctuation or possibly the theta, alpha, and beta bands, with complete
"spindling." If we consider only higher-amplitude absence of high-amplitude waves (no circles in
categories for the slower waves (at low-frequency, the right-hand columns of the 3X3 arrays); the
smoothing cutoffs), however, these statistical highest numbers of frequencies greater than 17
fluctuations disappear. (One should note that, Hz appear in this stage. 5Iost characteristic of
although we do not readily "see" these fluctua- stage-II sleep is moderately high-voltage theta
t.ions, they are still valid properties of the signal activity; we see the reappearance of high-ampli-
and may turn out to be of interest.) The last point tude activity in figure 8 under the 3- to 5-Hz
is that in examination of this figure a given num- (low theta) category. Stages-III and -IV of sleep
ber of waves in the low-frequency categories are characterized by increasing amounts of high-
represents a greater percentage (in terms of time) voltage delta waves. The chart shows increasing
of the signal than does an equal number of high- amounts of high-amplitude activity in the 1- to
frequency waves. 3-Hz range; in addition there are relatively large
Let us first consider the obvious differences values in the 13- to 17-Hz ranges that may be
between the EEG epochs characterizing the related to the sigma (14-Hz) activity usually
various stages of sleep, disregarding symmetry associated with stage-III.
for the moment. State-0 is on the borderline be- This brief examination of figure 8 may have
tween waking and sleeping (resting with eyes shown that for some bands there is a relatively
closed) and is characterized by a strong alpha constant number of waves for all stages. 5Iost
150 BIOMEDICAL RESEARCH AND COMPErTER APPLICATION
FREQUENCY (cps) larly high for the 11- to 13-Hz band at the 50-Hz
I-3 3-5 5-7 ?-g g-II 11-[3 13-t5 15JY >17
STAGE
,(! l * i
I'1 • • • • • K[Y
smoothing cutoff (not shown in fig. 8). Apparently
there is a large amount of this asymmetric alpha
• 0-1
activity superimposed on the slower delta and
• Z theta waves.
o .o.o
°.'L_."
.° • :__l / //I I . >4 The complexity of the EEG signal may be
• .0 .0 • • • • 5-7
,_ I Io I.o .o I.o I.o I.o t.o I.G_' determined by examination of changes in the
STAGE
L-I-_ I • I''--IL'_--I-- "-I_•la_ distributions of waves after successive smooth-
'.1 / : 181.o1.o/.oI.. / I t°,.,, ings. Let us first consider stage-IV because it is
apparent even from casual observation of an
EEG chart that the slow delta waves have many
,_ I ""Io"I" o1.o I".o I"1"1"1
.o .° • higher frequencies superimposed on them. At
o 50-Hz smoothing (not shown in fig. 8) there are

STAGE " • J • • II • • • • • •
oo O" 0 -0 o • almost no waves below 7 Hz, but large numbers

i i e o • -- • EXAIdPL_
appear between 7 and 13 Hz and above 17 Hz.
• o -- WAVES
At 25-Hz smoothing, almost all the activity above
•_-_..I o Iololol o I-oI.oI 17 Hz has disappeared, and waves around 8 Hz
are appearing. This shift to lower frequencies
Oo 0 0 o o
continues until, at 8 Hz, high-amplitude 1- to
O- O• • • cO_ •
3-Hz waves predominate. Within the 9- to ll-Hz
column the number of waves decreases from the
l Oo 0 0 0 • o oo
2_ to 15-Hz cutoff even though the waves are
STAGE • • O e _ 9 I_ ° • •
4 • • •
slower than the criterion frequency; an example
I oo Oo O. 0 o
will make clear how this can occur.
Consider two adjacent "peak waves," one
e_o e1_ • _ o- o* _I e _ __
10-Hz and one 20-Hz, with the middle minima
much higher than the two outer minima. With a
FIGURE8.--Analysis of the five 10-sec epochs shown in
figure 7. 15-Hz criterion, the 20-Hz wave is smoothed,
but the 10-Hz wave is not. However, when the
maxima and minima are redetermined, the 10-Hz
constant perhaps is the 5- to 7-Hz band for 15-Hz wave also disappears, and we are left with one
smoothing. This band only shows a significant 6.7-Hz wave, the combination of the 10- and
decrease for stage-0. Other bands in the middle- 20-Hz waves.
frequency ranges and at middle smoothing Compare the smoothing process for stage-IV
cutoffs tend to remain fairly constant. This with that for stage-0. With a fairly homogeneous
constancy probably reflects the "random" charac- pattern such as the alpha rhythm, the 9- to
ter of EEG signals. In most cases there will 13-Hz activity is mostly retained even down to
always be a certain number of waves in the 15-Hz smoothing. However, for stage-0 at 8-Hz
middle categories. smoothing, some low-amplitude delta waves are
The top rows of the 3X3 arrays indicate highly recorded; these are due to slow shifts in the
symmetric waves; the bottom rows, highly asym- overall alpha pattern and to spindling. These
metric waves. As might be expected, the stage-0 "alpha associated" delta waves are easily dis-
alpha rhythm shows the greatest number of tinguished from stage-IV sleep deltas by their
symmetric waves, and high-amplitude alpha is low amplitude.
almost entirely symmetric. In stages-I and -II, A nalysis-2--In analysis-2 there is no categoriza-
low-frequency theta waves (3 to 5 Hz) seem to be tion according to amplitude and symmetry. With
relatively symmetric. More striking perhaps is the limitation of no amplitude categories, the re-
the large number of asymmetric waves that sults are more difficult to relate to those obtained
appear occasionally, such as in the alpha range by visual evaluation, and certain t3q3es of changes
for stages-II and -IV. Asymmetry was particu- in the signal are not apparent. In analysis-3 it is
demonstrated that, with the application of simple percentage of 2-sec periods that were rejected
amplitude considerations, these limitations are no according to the criteria described above is also
longer in effect. shown. The values are averages for 10-rain
Details follow of the computer program used: periods.
Epochs of 10.0 sec were regularly sampled every The events marked on the chart and data-
minute of the entire recording of valid EEG. For rejection graph will be discussed first. The
each epoch, all valley and peak waves were deter- following events were considered: pre- and post-
mined for each of the four smoothing cutoffs used lift-off periods, sleep periods (considered in detail
for analysis-1. For each such cutoff, valley and below), eyes-closed periods, meals, housekeeping,
peak waves were classified according to frequency and exercise. Some of these events are listed in a
bands: delta (1 to 3.49 Hz), theta (3.50 to 7.99 log supplied by NASA; others were interpreted
Hz), Mpha (8.00 to 14.99 Hz), beta-1 (15.00 to by visual examination of the EEG chart by
24.99 Hz), beta-2 (25.00 to 34.99 Hz), and beta-3 electroencephalographers* (ref. 3). The data-
(35.00 to 49.99 Hz). rejection graph shows no artifacts before lift-off,
As already mentioned, criteria were used for but a considerable amount of muscle-spike
rejection of epochs of electroencephalographie activity for several minutes just after lift-off.
signals containing artifacts; for this purpose Figure 11 is a detailed graph of the percentage
every 10-sec epoch was divided into 2-see periods. data-rejection for several minutes prior to and
Appropriate criteria depend, of course, on the after lift-off; it shows no rejection prior to lift-off,
method of digitization, on knowledge of the peak rejection at 6 rain after lift-off, and abrupt
characteristics of EEG's, and on the particular cessation of rejection after 23 rain. Aside from
method of recording. The criterion used for lift-off, figure 10 also shows large amounts of re-
muscle-spike rejection was 12 or more spikes ject.ion correlated with occurrence of meals and
within the 2-sec period, a spike being defined as possible correlations with exercise and house-
a minima-to-maxima rise of at least 45 units keeping. There is also a general trend of increasing
within 16 msec. The extreme value criterion for rejection near 54:00 which would correspond to
rejection of a 2-see period was a continuous the increased loosening of the electrode.
period of at least 0.5 sec in which the value of the The major changes in the frequency bands
digitized signal was either greater than 337 or shown in figure 10 are associated with sleep or
less than 174 units. These criteria were developed the eyes-closed condition. The log of the flight,
by trial and error with the advice of an electro- supplied by NASA to Henry Ford Hospital,
encephalographer* and were verified by com- shows two periods of actual sleep by the com-
parison _ith a standard reference (ref. 4). Ex- mand pilot: the second corresponds to the inter-
amples of the artifact-rejection appear in figure 9. pretation by visual examination, but the first
As already mentioned, by failure to consider does not. This lack of correspondence is also
amplitude categories, some of the changes in the consistent with the computer findings.
EEG signal are not brought out. This is particu- Consider first the second sleep period. There
larly true for delta and theta waves at 8-Hz are cyclic decreases in beta-1 (at 50-IIz smoothing
smoothing which show only very small changes eutoff) and beta-2 (at 50 Hz) which correspond
throughout the entire recording. It was shown to the deepest periods of sleep (as interpreted by
above that, when high-amplitude waves only are the visual analyses). Both theta bands (50- and
considered, there are marked increases in the 25-Hz smoothing) show cyclie increases during
number of delta waves during stages of deep the deep-sleep periocLs. Two effects seem to
sleep, and this will be demonstrated again in control the alpha activity: the alpha band, at
analysis-3. 15-Hz smoothing, peaks only during the stage-0
The results for the bands having the most or eyes-elosed, resting state; but, at 50-Hz
meaningful changes are shown in figure 10. The smoothing, increases are also seen during the
*Personal communication with W. R. McCrum, 1967. *Personal communication with W. R. McCrum, 1967.
152 BIOMEDICAL
RESEARCH
ANDCOMPUTER APPLICATION
33-5G " _ I IL'_' i :
-_1.
r
I _ 13 SPIKES J I , " "l ' 4 SPIKES ,
l J
34-o3 j_,, " . :, " :.

•v" ; _?,_;.i, i_._ "_ _ %%/'%/ I ! ' ' "
• ' _,_i_ _ _ ! ;
, • i HIgH i 17 SPIKES i I
' - '- k I_ ,S PIKES , 1"_ oH ! _ow !
I sec pO pV
FmEm_ 9.--Examples of artifact-rejection.
periods of deep sleep. These latter increases Apart from the sleep periods, most of the bands
represent waves in the alpha frequency range remain relatively constant throughout the flight;
which are superimposed on the delta and theta they are particularly constant during the first 7
activity. hours. Only small differences are seen in the EEG
The changes just described are less apparent when pre- and post-lift-off periods are compared
for the first sleep period, since there was a much (except in artifact-rejection); there is a small
smaller amount of deep sleep and its total dura- increase in beta-2 (at 50 Hz), and a small de-
tion was short. Small peaks are seen for theta cre_e in alpha (at 50 Hz) activity. In the 5-hour
(at 25 Hz); alpha (at 50 Hz) shows an increase period just prior to the second sleep period some
during th_s sleep period; and there is a small small trends in some of the bands seem to ap-
decrease for beta-1. The sleep periods are dis- pear: all alpha bands tend to decrease gradually,
cussed in much greater detail in the other analyses. while theta (at 50-Hz smoothing) and beta-2
The eyes-closed period shows clear-cut in- tend to increase. One other finding is that the
creases for both alpha bands but decreases in alpha band (at 15 Hz) seems to increase shortly
theta (at 25 Hz) activity which is just opposite after each meal.
what was seen during the sleep periods. Beta-2 A_mlysis-3--The program for analysis-3 is
decreases somewhat during the eyes-closed period. exactly the same as for analysis-2 except that
BETA-I, 50
L M Ell:. M SLEEP M M _I EH M SLEEP EM ME M

,.=,
_[ T.ETA, 25 _-_:--_--_q_. - 77 - .,, ^ A,
,,,4. eeaEcr ",4V" • v I I 1 t,,V y _ v.

_e
FLIGHT TIME {hr)
Fm_aE 10.--Activity, in percentage of time, for the frequency bands shown.
Figures 12 and 13 show the percentage of time

that the signal was in each of four frequency
bands (recorded after different smoothing cut-
offs): beta-1 at 50-Hz smoothing cutoff, alpha at
40
A 50 60
A70 80
'
25 Hz, theta at 15 Hz, and delta at 8 Hz. The
LIFT-0FF TIME [rnln}
data represent means for 3-min periods (fig. 12)
FmVRE I I.--Percentages of 2-sec periods rejected by or 5-rain periods (fig. 13), and the time of the
the artifact-rejection criteria for the periods before beginning of each period is shown. At the top of
and after lift-off.
figure 12 is a graph representing the visual
analysis of the EEG performed by McCrum for
waves with an amplitude below a certain cri- the first, sleep period. In figure 13 is a row of
terion were not counted. With this criterion, numbers representing the approximate average
changes in the EEG that were not apparent in stages of sleep (ref. 3). It is apparent that the
the previous analysis are brought out quite delta (at, 8 Hz) and theta (at 15 Hz) bands are
clearly. The effect of the cutoff is rejection of the quite closely related to the depth of sleep. The
low-amplitude waves that might be considered alpha band (at 25 Hz) shows a peak in the
"noise" or "statistical fluctuations" and essen- drowsy period just before the onset of sleep.
tially masking of the changes in higher-amplitude Some small increases in the alpha-frequency
activity. The criterion used was linearly related range, occurring at the peak of delta activity,
to the wave duration and varied from 10 to 12 represent faster activity superimposed on the
units for frequencies between 50 and 16 Hz and slow sleep waves. The beta-1 band shows an
from 50 to 80 units for frequencies between 15 almost perfect inverse relation with the depth of
and 1 Hz. These criteria were found by experience sleep.
to reveal most clearly the changes in EEG ac- To summarize, three versions of the basic SPC
tivity. This analysis was applied to the first sleep method were used for analysis of the data from
period (fig. 12) and part of the second sleep flight. The first and relatively complex version
period (fig. 13). provided complete categorization of the signal
This frequency analysis leads to some interesting

findings in the higher frequency ranges but proves
to be inadequate for showing changes in the delta
region. The source of the difficulty is what might
be called the "random character" or "noise" in
the EEG signal. As successive smoothings are
performed at lower and lower frequencies, large
amounts of low-amplitude waves appearing in
the delta range tend to mask changes in the high-
amplitude categories. The third analysis showed
that this masking could be easily eliminated by
establishment of a cutoff such that only waves
greater than a certain amplitude were counted.
It was found that the high-amplitude delta
it. waves followed very exactly the deep stages of
O sleep as interpreted by visual examination. De-
velopment of methods of data-summarization
will be emphasized in future work on tile SPC
technique.
THETA Briefly the following was learned about the
EEG (channel-l) for the flight. There were only
small changes in the data associated with lift-off.
There was, however, a period of 23 min during
lift-off in which large amounts of data were re-
jected according to the "muscle-spike criteria."
Cyclic variations in depth of sleep were clearly
shown by several of the measures used. The delta
band (at 8-Hz smoothing cutoff), with a low-
amplitude cutoff, followed very exactly the stages
of sleep as interpreted by visual analysis. It was
confirmed that the second sleep period followed
TIME the log (supplied by NASA to Henry Ford Hos-
Fm_E 12.--Activities, in percentages of time, for bands pital), but that the first sleep period occurred at
(with amplitude cutoff): beta-1 at 50-Hz smoothing a time indicated by the visual analyses and
cutoff, alpha at 25 Hz, theta at 15 tIz, and delta at different from the time indicated by the log. At
8 Hz. 08:00 there was a period of approximately 2 hours
of strong alpha activity (eyes-closed, resting
according to frequency, amplitude, symmetry, stage, and possibly light sleep). The criteria for
and smoothing cutoff. It was applied to only a data-rejection showed increased percentages of
few data and was used largely to explain the periods rejected at times corresponding to chew-
capabilities of the general method and as an aid ing and possibly to exercise and housekeeping.
in understanding of the results of the analyses Increase in alpha activity appeared to follow
that followed. The results of this analysis sug- shortly each chewing period. Near the end of the
gested which particular wave patterns character- 54 hours of valid EEG recording there was a
ized each stage of sleep, and the complexity for trend to increase in data-rejection that would
these waves was indicated by changes that re- correspond to gradual loosening of the electrode;
sulted from successive digital smoothings. it began at about 40:00. A small trend, that may
The second and third versions were summary have some importance, occurred over a period
methods. In the second version the properties of of about 5 hours prior to the second sleep period.
amplitude and symmetry were not considered. All alpha bands tended to decrease gradually in
ELECTI_OENCEPHALOGRAPHIC DATA FROM ORBIT 155
4O
3O
20
STAGE= ._00 12234 I..0!A01334334444322253 .'=',IAO07.._3

C
ALPHA
w 20
2[
i,
o 0
t_J
_3c
I--
_J
20
FmVRE 13.--Activities, in percentages of time, for each of four frequency bands (with amplitude
cutoff) : beta_l at 50-Hz smoothing, alpha at 25 Hz, theta at 15 Hz, and delta at 8 Hz.
activity, while theta (at 50-Hz smoothing) and form and magnitude. The form and magnitude
beta-2 tended to increase. of the total set from which the sample set was
drawn are called the parameters of the total
population, and these are best estimated from
WEIBULL STATISTIC
the characteristics of the cumulative distribution
Considerable work has been done in industry function of the sample. Thus any cumulative
with a nonparametric statistical method using distribution function of a total set is composed
the Weibull distribution function (ref. 5). This of an infinite number of small distributions,
method provides qualitative as well as quanti- each of which has its own unique parameters.
tative evaluation of a distribution without re- For practical purposes these small distributions
quiring any assumptions about the distribution are generalized to one large distribution (the
parameters. Experience has shown that the total sample) with one set of parameter estimates.
Weibull distribution more closely approximates If the members of this large population are essen-
physical and biological systems than does the tially similar, it is said to be a unimodal distribu-
normal distribution. tion. If, on the other hand, there are two or more
In consideration of any set of experimental subpopulations that differ greatly in one or more
observations or measurements as a population of parameters, the population is said to be bimodal
individual statistics, this population can be de- or multimodal and is called a complex distribu-
fined as an ordered set and described by a tion. From experience it is known that no sample
cumulative distribution function having both population is ideally unimodal. The variation of
1.56 BIOMEDICAL RESEARCH AND COMPUTER APPLICATION
the sample from the ideal can be defined in two has 50-percent probability of being larger or
areas: (1) sample error due to too few samples smaller than the true value and can be determined
failing to represent, a large population, and (2) from the equation
known or suspected competitive influences intro- median rank = (j-- 0.3)/@+0.4)
duced into the sample.
For most conventional statistical procedures where j is the jth-order statistic and n is the total
one must assume that the probability density sample. The cumulative distribution function
function of the total population that is sampled F(x) = 1 - e-(*/a)a
is of the form
is rearranged as
f( x) = [1/ (a'V/_) ]e-1/2t(_-_)1_|
1/[1 - E(x)] = e-(*l°)a
In such a function the mean, median, and mode
coincide, and the shape of the function is such Taking the log_ log_ of both sides results in
that the point of change of curvature lies a-dis-
In ln[1/[1--F(x)]} =/_ lnx-_ ln0
tance from the mean. It has been my experience
(and that of others) that this function is rarely making the substitution
encountered in dynamic biological situations.
Y=lnln{1./[1--E(x)]}, T=Inx, C=t31nO
Thus, conventional statistical procedures have
little real meaning and furthermore can lead to then Y=_T+C which is linear plot in Y and T
erroneous conclusions. with slope _. Note that 1-F(x) is the cumulative
Given the sample function, P(X<_x)=F(x), probability of success. We;bull-probability paper
any distribution function can be determined is so constructed that the vertical scale represents
from the equation
In ln{1/[1-F(x)]}
F(x) = 1 - e-t¢_-')m-')l_
but is graduated ill terms of F(x), the fraction
provided that -- [(x- a)/(8-a)] B is nondiminish- falling within x. The horizontal scale is a loga-
ing and vanishes at some value of a. When a is rithmic scale representing lnx. A computer pro-
assumed to be zero, gram was used that orders the sample data
points, determines their median rank, and then
F(x) = 1 - e-(_mB
computes the slope by a least-squares method.
When x=0, the function evaluated at 0 is The problem that arises in use of this computa-
tional method of determining the slope (beta) is
F(x) = 1 - e-l that all distributions are then represented as
Thus we have a distribution function character- simple or unimodal. 5Iy experience has shown
ized by three parameters: (1) alpha, the minimum that often this is not the case; rather the distribu-
life; (2) beta, tile slope or shape; and (3) theta, tions tend to be complex. This fact can be deter-
which is scalar. For most cases alpha can be mined by a graphic plot that connects each data
assumed equal to zero and ignored. point rather than a best-fit line; variations of this
Since differences ill the mean and variance of line from linearity then represent either sample
two samples have no meaning unless the sample errom or competitive modes.
populations have the same shape, the first step It is my experience that in some cases the
in analysis is determination of the shape of the evaluation of beta is the only adequate informa-
distributions. With use of the We;bull method, tion available from comparison of two sample
the first step then is evaluation of the slope beta. populations; that is, the mean and variance of
This was done by plotting of the ordered data the two samples do not differ, but there is signifi-
points on We;bull Cumulative Distribution paper. cant difference in the value of t_. It has usually
In the method presented, sample data points been found that this difference is due to intro-
are plotted against median rank values, and a duction of a second mode into one sample that
straight-line best fit is drawn to the plotted does not affect its mean and variance.
points. The median rank value is the value that Quantitative differences between two sample
ELECTROENCEPHALOGRAPHIC DATA FROM ORBIT ]57
populations can be determined by plotting of

the 90-percent confidence bands about them.
The method is the same as tile one described,
except that the 5- and 95-percent ranks are
SLEEP
substituted for the median rank of the order
statistic.
WEIBULL ANALYSIS OF FLIGHT EEG'S
The output of the Z/C technique provides a DROVSY

frequency count and normalized percentage times
for a 15-sec epoch of EEG. Each 10 sec thus yields RELAXED
one sample point or value for the We;bull graph
of a particular frequency band. After an arbitrary ALERT
decision that a sample population of 60 points i , ! , -.-
I
ILl ;oo lq:]o oo 15:3o
was sufficient to represent adequately the dis-
HOURS
tribution curve, the computer was programmed
to plot the We;bull distributions for each con- FIGURE 14.--Visual analysis of the first sleep period.
secutive 15 rain of analog EEG. On occasions
during the flight recording, this 15-rain length of
record was disadvantageous; for example, some 13:00 whereas our visual examination confirms
periods of sleep or a particular activity were less MauIsby's conclusion (ref. 3) that this sleep
than 15 rain in length. In such cases, any EEG took place between 14:21 and 15:35. The re-
information specific to that short period was mainder of the record was perused visually for
diluted in the total 15-rain sample. Shortening of gross features such as chewing movements and
the period of analysis would be costly in time and presence or absence of sleep (without classifica-
accuracy although the We;bull statistic can be tion into stages), and also for long periods of
quite powerful with a small sample of data. alpha activity suggesting relaxation. An exact
During the time before and after blast-off, the log of these observations was not supplied to
We;bull plots of 5-min periods of EEG were Henry Ford Hospital.
drawn by hand, this takes time and is quite The EEG taken during 10 min before blast-off
prohibitive for large amounts of data. was compared to that for 10 min following
One can also plot the 90-percent confidence blast-off. Figure 15 shows the 90-percent con-
bands about a We;bull distribution; that is, it is fidence bands about the delta activity (0.5 to 3
possible to plot the range within which 90 percent Hz) for the states before and after blast-off. The
of all values of the distribution will lie. The fact that the two bands are widely separated
plotting of these confidence bands has not yet indicates an extremely high probability that the
been programmed for the computer, so they were marked reduction in delta activity after blast-off
used only in analysis of the pre- and post-lift-off is a true observation. The abscissa is a log scale,
data. and the medians of these distributions indicate
The command pilot's first sleep period was ana- that the percentage time of delta activity was
lyzed visually in a fashion resembling Maulsby's halved after blast-off.
(ref. 3); we are in agreement to the minute for Figure 16 shows the 90-percent confidence band
the onset and termination of sleep (fig. 14). The about beta-1 activity (15-25 Hz) after blast-off,
only disagreement, was minor in the separation along with the simple We;bull plots of beta-I
of stages-II and -III of sleep; we did agree on activity during the two consecutive 5-rain periods
stage-IV. One should note that the Gemini 7 log just preceding blast-off. The indication is that
supplied to Henry Ford HospitaI appeared to the beta-1 activity almost doubled after blast-off,
err regarding the first actual sleep period, as it but the probability level is not quite as high as
indicates that this occurred between 09:00 and with the changes in delta activity. In the analysis
9P" 1
_90.
SD.
i • 3 4F I n
log
•
(DELTA)
i • i 4 m m i
:4L
• !
log I_ETA- I)
• • • |
i
l
l
|
Fmvm_ 15.--Ninety-percent confidence bands about

Fz(_unE 16--(1) We;bull plot of beta activity from be-
delta activity during 10-min periods preceding (1) and
tween 10 and 5 min before lift-off. (2) Beta activity
following lift-off (2).
between -5 rain and lift-off. (3) Ninety-percent
confidence band about the beta activity for the 5 min
immediately following lift-off.
of alpha activity (fig. 17) the 90-percent con-

fidence band is for the activity (8 to 15 Hz)
following bIast-off. Only the simple distribution set of 856 graphs has been derived from the SPC
of the pre-blast-off theta is drawn. Only in the technique for processing of the analog EEG. In
midrange of each sample is there significant this report only the plots derived from the Z/C
difference in the amount of alpha, with the program have been reviewed in their entirety,
post-bin.st-off record showing the greater amount; and only a few chosen graphs derived from the
the tails of the samples are not different. This SPC method have been reviewed.
fact may suggest a change in the state of physi- 5[y first attempt to compare the We;bull plots
ology of the brain that is greater than would be of a given frequency band was by simple super-
reflected by simple increase in the mean amount imposition of one graph over the other. When
of alpha. this was done the range of alpha activity (that is,
Analysis of theta activity yielded no evidence the variation in quantity) was quite small over
of change between the periods before and after the entire orbit; likewise the general slope or
blast-off. The We;bull plots of the two sample shape parameter was similar. This procedure did
periods were almost identical. yield some ideas for future analysis that will be
We;bull distribution plots were made for each discussed later, but for our present purpose it
consecutive 15 min of EEG for the entire 53.5 appeared impractical.
hours. There were separate graphs of each fre- Attention was then limited to the sleep periods.
quency band: delta, theta, alpha, and beta-1. The first sleep period lasted about 1.5 hours and
This made a total of 856 individual graphs of was interrupted frequently by periods of arousal.
data derived from the Z/C technique. A similar Each of the 15-rain periods of EEG of this sleep
99._
I-
Z
kl
0
?0.'
Q.
$0."
z
O
b-
|0."
t_
>
b-
.$.
.2-
FIGURE 17.--(1) Ninety-percent confidence band about

the alpha activity for the 10 min following llft-off. FmURE 18.--Range of Welbull plots of stage-IV sleep
(2) The simple Weibull plot of the alpha activity for for the frequency bands delta, theta, alpha, and beta-l;
the 10 min preceding lift-off. data derived from the Z/C technique.
episode was a mixture of sleep and arousal, and SPC method were used, the ranges of alpha and
the collection of Weibul] plots over this 1.5-hour theta activities narrowed considerably, but, on
period showed no features that would permit the other hand, the total quantity of each in-
labeling of any particular plot as representative creased manyfold. This observation poses some
of sleep. questions about which methods of analog proc-
The second sleep period covered a period of essing should be used; each seems to have certain
about 8 hours. The Weibull plots of the consecu- advantages over the other.
tive 15-rain EEG samples over these 8 hours The various stages of sleep other than stage-IV
were examined collectively by superimposing one could not be evaluated by this means of comparing
over another. Primarily on the basis of the plots their Weibull plots. One reason for this was that
of delta activity, four 15-rain epochs of EEG in any 10-min period, represented by a single
were separated from the total sleep group. When graph, several stages of sleep were represented;
these four periods were related to the visual thus each Weibull plot represented a mixture of
analysis of the :EEG, they coincided identically the different stages of sleep.
with the periods of stage-IV sleep that were of One other finding needs some clarification: It
10-min duration or longer. must be remembered that the Weibull plot yields
The ranges of the Weibull plots of delta, theta, a shape parameter that is some measure of the
alpha, and beta are depicted in figure 18. There operating characteristic of the system under
is complete separation of each band from the study; in this case it is the physiology of the
other, even though the ranges of alpha and brain underlying electroencephalographic activity.
theta are quite wide. When the data from the This shape parameter may change considerably
with or without alteration of the parametric

mean values of EEG analysis. When this hap-
pens, it may be assumed that. the physiology has
changed in some way that is not reflected by a
simple quantitative measurement of an EEG
frequency band.
When the slope or shape parameter of the alpha
activity, obtained from the SPC method, was Z
O
studied in detail, it became apparent that the F--
various graphs fell into groups. AII these graphs

represented compound or multimodal distribu-
tions; that is, a single Weibull plot was not a
simple straight line but a series of straight lines W
>
at various angles to one another. When the first
30 hours of the orbital flight was examined in g
this fashion, it was apparent that the 120 graphs 13
representing this time fell into eight major groups.

The important observation wa_s that all the
graphs for the first 3 hours of the flight belonged
to just one group; thereafter there was complete
intermingling of the members of the other groups
in a seemingly random fashion. The median
amount, of alpha activity w,'_s similar throughout
the entire 30-hour period. Fign]re 19 shows the
characteristic plot of the first 3 hours after
blast-off, along with some examples of the other
shapes found later.
In summing-up the Weibull analysis it can be different kinds of distribution functions can be
said that if significant quantitative differences in shown to exist in different EEG samples, these
the EEG are present they can be determined to in turn can be directly related to changes in
be significant with only a small sample of the physiology and/or behavior. When this correct
EEG. Examination of the periods immediately graphing of distribution functions has been
before and after blast-off bears this out. The accomplished, computer pattern-recognition pro-
periods of stage-IV sleep could also be determined grams must be developed to define the different
quite accurately. Examination in detail of the graphs as finitely as possible. This poses further
slopes of the individual graphs, particularly alpha problems. Since, on the basis of present observa-
activity, could be the most important finding of tions, quantitative differences in the EEG appear
all; unfortunately we still have no satisfactory significant when there are marked changes from
method of examining and classifying the hundreds alertness to deep sleep, the analysis of fine changes
of slopes that are gathered from a long period of in behavior (both affective and physiological)
EEG analysis. should be directed toward the tails of the distribu-
Several problems are involved: For one thing tions where behavioral changes are most reflected.
the exact distribution functions for most of the A proper pattern-recognition program or tech-
sample periods are unknown; none of them is nique would then require some knowledge of the
normally distributed or even symmetrically dis- "statistics of extremes."
tributed. We have accepted use of the Weibull These are only some of the problems in develop-
distribution as a close approximation. In many ment of the "nonparametric" approach to EEG
eases this seems true, but in many others it is analysis. My findings in this study, using the
obviously not; thus the graphic analysis of Weibull statistic as a first approximation, seem
distribution functions must be investigated. If to justify further efforts along this line.
GENERAL DISCUSSION tative analysis by use of computer techniques is

superior to routine visual analysis. In particular
I have reported my analyses of the flight EEG there are two major advantages: first, the com-
records by routine visual EEG analysis, by para- puter provides a consistent numerical analysis,
metric analysis of data by the zero-crossings and second, tim results are not dependent on
technique, and by a pattern analysis called subjective inference. The ideal method of analysis
smoothing and peak-counting. Finally we have by computer obviously has not been attained, but
applied Weibull statistic to the data derived we believe that through a number of techniques
from both the Z/C and the SPC techniques. A we have effected a useful analysis. Of these meth-
routine visual analysis was made of the two ods the one to be used in a particular case will be
major sleep periods, the first appearing at about determined by such factors as quality of the raw
14:00 and lasting roughly 1.5 hours and the data, cost, and time for the analysis. In upgrading
second appearing about 33:00 and lasting ap- of the computer techniques a universal system
proximately S.5 hours. will, we trust, be found that incorporates the best
The parametric and SPC analyses revealed features of the various computer programs.
another period of light sleep between 08:00 and Each of the techniques reported here lms its
10:00. Visual reevaluation of this period of the own individual merits. The Z/C technique pro-
EEG did reveal the presence of stage-I sleep vides data that, when summarized with standard
throughout and a few minutes of stage-II sleep parametric and nonparametric statistical tech-
near the end of the period (fig. 20) when the niques, reliably indicate major changes in be-
criteria of Dement and Kleitman (ref. 2) were havioral states that are currently definable and
strictly applied. We would expect the periods of now receiving significant scientific attention. The
lighter sleep to be missed on routine visual Z/C technique is the most rapid computer tech-
analysis when not followed by periods of deeper nique operating with acceptable real time; conse-
sleep that serve to cue their presence for the quently it is the lowest in cost since it also lends
clinical electroencephalographer. This fact demon- itself to a variety of small, compact, inexpensive
strates that our described computer techniques computers.
can recognize the various stages of sleep inde- The SPC technique is particularly capable of
pendently of one another. handling noise in EEG signals by rejection cri-
On review of the results of our various methods teria and by not being bound to a base line; it
of analysis of the EEG it appears that quanti- provides detailed categorization of the signal
_SLEEP_Z
5LEEPE
.SLEEP]I
SLEEP I
ALPHA
ALERT
TIME LINt
FmvR_ 20.--Visual analysis of EEG during "eyes closed" (sleep?).

according to frequency, amplitude, symmetry, Sciences, Henry Ford Hospital, contributed to

and complexity (superimposition of faster or this work: W. R. McCrum, Ph.D., T. E. LeVere,
slower waves). This technique is also a relatively Ph.D., R. 5i. Lee, Ph.D., and H. van den Ende,
rapid computer program and could be used on-line. M.S.E.E.
The Weibull statistic is a useful research tool to
assist the modeling of the neurophysiology under-
REFERENCES
lying the EEG. It served the purpose in this
report of emphasizing the complex variability in 1. BAaTLETr, M. S.: Square-Root Transformation in
data obtained by Z/C and SPC techniques. For Analysis of Variance. J. Statist. Soc. Suppl., vol. 3,
1936, pp. 68-78.
on-line analysis of sleep, the Weibull statistic
2. DEMEN% W.; AND KLEITMAN, N.: Cyclic Variations
offers no advantages. in EEG during Sleep and Their Relation to Eye
In my opinion improvement of the electrode Movements, Body Motility and Dreaming. Electro-
system and substitution of a noncontinuous encephalog. Clin. Neurophysiol., vol. 9, 1957, pp.
673-690.
program for the continuous EEG recording
3. MAULSBY, R. L. : Electroencephalogram during Orbital
would prove more efficient and less inconvenient
Flight. Aerospace Med., vol. 37, 1966, pp. 1022-1026.
to the astronauts.
4. GIBBS, F. A.; ANn Gibbs, E. L.: Atlas of Electro-
encephalography. Vol. 1. Addison-Wesley Press, Inc.,
1950.
ACKNOWLEDGMENT 5. WEIBVLL, W. : A Statistical Distribution Function of
Wide Applicability. J. Appl. Mech., vol. 18, 1951,
The follo_ing members of the staff of the pp. 293-297.
Department of Behavioral and Neurological 6. Johnson, L. G. : Unpublished notes and lectures.
CHAPTER10
TRACKING OF AN ASTRONAUT'S STATE BY

PHYSICAL MEASUREMENTS OF SPEECH
Louis V. Surgent
SUMMARY (computed as above and exclusive of pausal

silences) also showed evidence of useful validity.
This study is addressed to the question of
Group Duration and Syllables per Group were
whether it is feasible to use measurable speech
weaker but clearly responsive. Pausal silences
parameters for detection and tracking of the
within communiques were distributed poorly and
changes in state of an astronaut. Techniques for
showed no evidence of validity.
assessment of changes of state are developed to
Methodological considerations are emphasized,
serve as criteria. The most detailed of these,
and suggestions for research, development, and
Probable-state analysis, uses the "stream of
application are offered.
behavior" concepts of Barker (refs. 1-3) to or-
ganize situational _nd behavioral data for raters
INTRODUCTION
including speech communications. Procedures for
rating the state of the pilot during each commu- Detection and tracking of the changes of state
nique, on a set of 10 situationally and behaviorally of an astronaut from the physical parameters of
defined "probable states," are specified. his speech require discovery of measures that
Automatic speech-processing techniques were covary with state, and development of weighting
found inapplicable to large sections of the on- functions that transform sets of these measure-
board recordings supplied for analysis. Instead, ments into identifications of states and state-
oscillographic measuring techniques were devised intensity predictions. This is a large order, with
to simulate a feasible automatic speech-processing many unresolved problems regarding criteria
system and applied to about 68 min (5121 syl- (state) and predictor (speech). Substantial prog-
lables) of astronauts' speech. Communiques were ress toward meeting these requirements, rather
segmented into periods of uninterrupted speech, than complete fulfillment of them, was therefore
called Pause Groups or Groups, by a criterion the aim of this small, 1-year contract.
silence (170 msec) selected to distinguish between Within this frame, initial attention was given
"articulatory" and pausal silences. to criterion questions such as the following: How
Of the new measures developed, Group Highest should the state of an astronaut over a period of
Pitch and a ratio formed by dividing the duration time be specified? What set of labels should be
of the Group Last Syllable by the average dura- used? What evidence will warrant assignment of
tion of the remaining syllables, named the DURLL one label rather than another--or of two or more
Ratio from its FORTRAN designation, are most labels simultaneously? How should variations in
promising. The standard deviation of algebraic, the degree (level or intensity) of a particular
first-order, serial differences in successive syllable state be designated? Or should we settle for well-
rates--computed as the reciprocal of each syl- established physiological measures, such as heart
lable duration--showed responsiveness to situa- rate, and attempt to predict them?
tional changes, especially when normalized in a Siuce the validation methods eventually used
ratio with the standard deviation computed with- were considerably less sophisticated than those
out regard to sequential dependencies. planned, due to funding limitations, we proceed
Syllable Peak Amplitude and Syllable Rate directly to the measurement of speech parameters
163
][64 BIOMEDICAL RESEARCH AND COMPUTER APPLICATION
and to the results from the simpler methods. The function examined, including the second formant,
criterion problem is discussed more basically in contained enough information to justify its inclu-
two appendices. sion in a set of functions displayed oscillograph-
ically for human discrimination and measurement.
Second, a sizable computer-programming effort,
MEASUREMENT OF SPEECH PARAMETERS
preceded by a substantial amount of engineering,_
was required to get anything much from auto-
Possibility of Wholly Automatic Analysis of Speech
marie analysis of these recordings. Only the
To be acceptable, speech monitoring must be rectified and smoothed amplitude time function
inexpensive and convenient as welI as valid, hence, was sufficiently reliable in specifiable regions--
the requirement for investigation of wholly auto- over syllable nuclei, for example--to provide a
matic processing despite the problems foreseen. start its this direction. The third conclusion is
The main hope in view of capsule-power limita- that automatic spectral analyses by Starkweather
tions Iies in development of a spacecraft vocoder. (ref. 4) in his clinical investigations, and the
The transmission of speech measurements digi- related techniques of Popov (rcf. 5) in his study
tized on board wouht provide intelligible speech of two Russian astronauts and 15 actors rendering,
with minimum bamtwidth and power and would "This is Diamond, I read you," etc., would be
facilitate computer processing of selected param- degraded seriously by these frequency-domain
eters for pilot monitoring. distortions and sounds with nonrandom
The current use of limiting circuits to increase components.
transmitter-power utilization results in "speech
Semiautomatic Approximation to a Realizable
clipping," with consequent gross spectral dis-
Automatic System
tortion as amplitude enters their region of opera-
tion. Thus acoustical analysis is complicated. In There was good reason to continue the study
addition, the weight-comfort relation in continu- despite temporary abandonment of wholly auto-
ously worn headsets dictated the use of micro- mat ic processing and substitution of oscillographic
phones that attenuated drastically frequencies techniques. Apart from the scientific value of
in the region of the voice fundamentals; their relations that may be discovered, a semiautomatic
spectral envelopes differed, and some appeared system can be developed for digitizing of samples
marginal. of speech, either as occasional checks (in the way
The on-board recordings supplied for this study in which blood-pressure readings are now taken)
reflected these hardware limitations; they also or distributed over time to allow for processing.
contained extraneous capsule noises, tone sig- This could hasten complete automation if the
nals, and ground communiques, as well as receiver search for valid measures is confined to variables
and other electrical noise, especially in intervals digitizable in a spaceship or moon-orbiting labo-
between astronauts' speech. On the other hand, ratory either with minimum additional circuitry
the amplitude variability, spectral distortions, and data load or as part of an on-board vocoder.
and intense random noise often introduced by The oscillographic analysis reported here was
atmospheric propagation were absent. guided by these objectives. They do not in them-
The types of error made by automatic equip- selves represent feasible operational procedures
ment in digitizing such speech can be specified but are aimed at discovery of applicable measure-
quite well without laboratory trial. However, ments and relations.
speech-quality criteria normally applied to the
Performance and Digital Encoding of the Basic
synthesized output of such equipment in band-
Acoustic Measurements
width-reduction studies are replaced here by the
much less stringent and lesser-known require- Specimen oscillographic record--Figure 1 is an
ments of probabilistic state-predictions. oscillographic excerpt of a pilot's communique
Efforts to assess automatic processing for this beginning: "Main chute on green. Chute is out
application and with these recordings led to the in reef condition at 10 800 feet and beautiful
following conclusions. First, ahnost every time chute. Chute looks good. On O_ .... "The acoustic
PHYSICAL MEASUREMENTS OF SPEECH 165
Fmcam 1,--Oscillogram of pilot saying "Chute looks good,"
energy preceding the speech is thought to be a within the dynamic range of recorders and other
set of beat frequencies generated by two trans- equipment.
mitters, one aboard a rescue plane and the other The voice's fundamental frequency, attenuated
apparently aboard a rescue vessel. The recorded by microphones, was reconstructed by preprocess-
suppression of this energy, as the pilot begins his ing of circuits of the full-wave-rectification type
transmission, typifies the operation of VOX and and fed to a vocoder pitch-tracker to produce the
push-to-talk circuits on noise and extraneous "pitch time function," P(t). Pitch-tracker out-
signals processed by the receiver. puts of zero and 5.0 V were equated respectively
The "speech pressure wave form" or "speech to a 60-Hz reference, P(60 Hz), and a 300-Hz tone.
time function," S(t), gives the clearest indica- Input amplifiers to the oscillograph were then
tions of the onset, termination, and nature of adjusted to place 250 Hz 1.52 in. above the
major speech events. Vertical displacements of P(60 Hz) reference marks which are recorded
from -0.6 to 0.6 in. were set equal to a 3.0-V once each second. An empirically derived calibra-
peak-to-peak calibration tone recorded on each tion equation was used to convert measured dis-
channel. placements over a wide range to cycles per second.
The "amplitude time function," A(t), is a A BCD serial code, identifying hours, minutes,
rectified and smoothed transformation of S(t) and seconds, is along the bottom. A synchronous
with 1.5 in. equal to 1.28 V rms. Measurements 10-pulse-per-second line along the top is used with
are made from zero reference marks, A(0), and an interpolating template to measure event times
corrected by the amount of calibrated attenua- to centiseconds.
tion or amplification necessary to bring speech Pause Groups as response units--From a be-
amplitudes, over a flight phase or other period, havioral standpoint, examination of the S(t) indi-
cates the importance of distinguishing between Brady's terminology would have been adopted
what we called articulatory silence and pausal except that we read his report some time after
silence. The articulators are highly constrained submission of the project report.
with respect to position, movement, and timing The term "articulation pauses" was used once
during articulatory silences--e.g., between the by Goldman-Eisler (ref. 10) in a study of pause
"t" of "chute" and the 'T' of "looks," before distributions; she selected 250 msec as the bound-
the release of the "k" of "looks," and between the ary separating these from hesitation and breath
"s" of "looks" and the release of the "g" of pauses. She also worked extensively with breath
"good" (fig. 1). During pausal silences, on the groups, and some of the measurements and in-
other hand, the articulators are free of linguistic dices that she developed from them (refs. 11 to 13)
constraints. Inhalations occur only during pausal have analogs on Pause Groups.
silences. Pause Groups provide an excellent starting
A criterion pause was then sought that was point, in analysis of communiques for information
long enough to avoid segmentation during ar- about the state of a speaker. First, the basis of
ticulatory silences and short enough to assure segmentation is objective, as Verzeano empha-
segmentation by breath and hesitation pauses. A sized, although one cannot gloss over the prob-
value of 170 msec was found to work well for the lems of discrimination of low-level speech energy
quite rapid speech of pilots in the tapes analyzed; from "silence" in real-world systems containing
this was detected as a 0.7-in. extent on the oscillo- noise. In a wholly automated system, improve-
graphic records (4.25 in./sec). The duration of ments in segmentation will be attainable by addi-
all pauses of 170 msee or longer was recorded with tion of other, more-complex, physical criteria.
a view to investigation of the function relating Second, Pause Groups exist in every language
the optimum criterion pause (inversely) to speech and can be identified without translation. Third,
rate, but time and other constraints did not permit Pause Groups appear to have some linguistic and
this ancillary analysis. Periods of speech bounded physiological relevance. Henderson, Goldman-
by criterion silences were called Pause Groups or Eisler, and Skarbek (refs. 14 and 15) provide
simply Groups, being analogous to the breath the most interesting data. In reading, 100 percent
groups or segments bounded by inhalations. On of inhalations occurred at grammatically appro-
oscillographic readouts, Q's and G's, with "on" priate junctures, compared with 69 percent in
and "off" affixes, designate onsets and termina- spontaneous speech (cartoon stories). In reading,
tions of communiques and Groups. inhalations occurred in 77 percent of the pauses,
The concept of segmenting of speech by pause compared with 34 percent in spontaneous speech.
durations for behavioral studies was proposed The spontaneous-speech samples were further
(refs. 6 to 8) to eliminate the human judgments analyzed on an x-y plotter; the pen moved
required by Chapple's interaction chronograph. vertically during silences of 100 msec or longer
Verzeano investigated distributions (Poisson) of and horizontally (luring speech or shorter silences.
segment lengths as a function of criterion pauses Alternating periods of hesitant and fluent speech
of from 100 to 900 msec. Distributions generated were indicated by cyclic changes in slope, with
by criteria of 400 msec or more showed irregulari- speech during periods of steeper slope showing a
ties which he attributed to "respiratory rhythmic- larger number of "Ah's," false starts, and inhala-
ity"; articulatory and pausal silences were not tions at ungrammatical junctures. These were
distinguished. interpreted as periods of "planning," which
Brady's (ref. 9) concern with improvement of facilitate the fluent speech in the second phase of
techniques for measuring speech level led him to each cycle. No attempt was made to establish
analyze periods of measurable speech energy and that behavioral processes ordinarily called plan-
silence in staged telephone conversations. Hc dis- ning really occurred. The possibility that the
tinguished "intersyllabic gaps" from "listener- higher incidence of hesitation phenomena is as-
detected pauses" and found that 200 msec form sociated with the greater information content of
a boundary between the two; he called the speech the thematic decisions, required at certain points
segments marked by such pauses "spurts." in the developing cartoon story, is not mentioned.
However, it has been noted that it is more com- developed at General Dynamics and elsewhere to
patible with earlier demonstrations (ref. 16) that avoid the susceptibility of these circuits to
the length of a hesitation pause between two con- "misses" in certain phonemic environments and
secutive words varies inversely with the transi- with rapid or slurred speech, and to "false detec-
tional probability. It may well be that the princi- tions" due mainly to the extreme sensitivity of
pal difference between the two processes is that the first derivative to noise.
the transitions noted here are between "kernels," The simpler circuits were chosen as the model
phrases or larger units, instead of between words. to be simulated because they can be set to approx-
Goldman-Eisler (refs. 10 and 17) also found imate perceived syllable onsets (ref. 20). The lin-
hesitation behavior somewhat less evident when guist who scored the oscillographs used three
subjects described a cartoon than when they at- different rates of onset, equivalent to three values
tempted to summarize concisely the main point. of the first derivative, as "anchoring points" in
With practice (seven repetitions with the same his judgments and placed marks to approximate
cartoon), pauses were substantially reduced and perceived onsets. Additional guidelines were
speech rate increased. specified for marking of instrumentally trouble-
In addition to indicating the complexity of some onsets in the presence of semivowels, in-
pause-related phenomena, these data place in per- cluding the intervocalic "r", voice fricatives, and
spective Fonagy and Magdics's (ref. 18) finding 12asals.
that over 95 percent of inhalations occurred at Encoded data--The seven-digit format needed
sentence or phrase boundaries, since their analysis for encoding of Irig-B times to cent;seconds was
was based mostly on reading and highly formalized used for all event times, measurements, and other
speech such as sports broadcasts. They also data. A two-digit function code was added to
illuminate our finding--in a spot check of an identify the decoding rule. A state diagram de-
astronaut's speech during a launch--that about fining the structure and permissible sequences of
85 percent (30 of 35) of pauses longer than 170 function codes was drawn up to facilitate encoding
msec occurred at phrase boundaries as identified and programming. The event times encoded were
by a linguist before application of the 170-msec the onset and termination of each communique
pause criterion. The many simulation runs per- and of each Pause Group within it, and syllable
formed by astronauts, before flight, undoubtedly onsets. "Syllable peak amplitude" (ARMS)* was
resulted in establishment of speech responses and an obvious choice, being an easy measurement,
formats under control of the many fixed sequences with some data relating it to the state of the
of stimuli presented. Another large share of astro- speaker (ref. 21).
nauts' responses are "descriptive" and may be The literature at the time provided no guidance
expected to exhibit properties more closely re- for scoring of the noisy output of a pitch-tracker.
sembling cartoon description than summary of The discriminations necessary for the Fairbanks-
the main point. Pronovost (ref. 22) measures, for example, could
The fourth point is that the Pause Group can not be made. Intensive study of this function
serve as the "experimental unit" or "unit of led to the conclusion that its maximum excursion
analysis." This applies in the sense that ideally over a Pause Group--exclusive of second har-
all speech measurements and speaker-state as- monic seizures, which are easily recognized--was
sessments would be assembled on a Pause Group the only theoretically satisfactory, reliable, and
basis (see Appendix A). potentially instrumentable measurement avail-
Syllabic segmentation of Pause Groups--The able on every Group; this was named Group
simplest method of automatic marking of syllable Highest Pitch (PHH). To check the reliability
onsets consists in low-pass filtering, rectification of this measure and to provide estimates of its
and smoothing of the speech pressure wave form, midvalue for each Group and of its variability
and setting of a threshold on the first derivative
of the output. These circuits have also been used
*Names (of variables) assigned for computer program-
to measure speech rate (ref. 19). More complex,
ming are used throughout; A (amplitude) was measured in
automatic, segmentation techniques are being volts rms.
over Groups, the second-highest pitch and the limitations) of computer data-reductions, flight-
lowest and second-lowest "readable" pitches were phase comparisons, and graphic analyses.
also encoded. Only the PHH data were analyzed.
Group Highest Pitch
These preliminary assessments were verified
not only by data from astronauts, as will be seen, Pitch variations during countdown, launch, and
but. also by a subsequent report (ref. 23) in which initial weightlessness--The joyful anticipation ex-
grammatical sentences, uttered with simulated pressed in "Boy, can you imagine, here we go,"
emotions or culled from classroom lectures and just 2 min before lift-off (fig. 2), may have been
discussions, were rated for emotional content by suppressed somewhat by the physical discomforts
listeners. The maximum pitch for cach sentence of a long countdown and the tense readiness of
correlated well with certain ratings. Correlations the last moments. Relief from these, and the joy
were somewhat higher with simulated emotions of "We're underway," probably account for the
than for the 27 sentences (each repeated at least initial jmnp in pitch in the first TLI of launch.
once, making 60 utterances) of the classroom (This reference is to the astronaut's states of
instructor. "probable relief" and "probable joy" which are
Other information included routine identifica- defined situationally in Appendix B.) Such emo-
tions, amplitude-reference and calibrated adjust- tions ran high in Project Mercury.
ments, and provision for state-relevant data. A A drop of 30 IIz is coincident with "Little
simulation of the "voice-unvoice-silence" seg- bumpy" and the rapidly approaching region of
mentation of a typical vocoder, together with an increased vibration and acceleration. "Smoothing
alphanumeric code identifying each phoneme, was out," "Feels good," and "Through max. Q" are
tried and found feasiblc but abandoned for speed descriptive of the conditions accompanying the
in the processing. rise over TLI 8 and 9. There is a momentary drop
The measurement setup--All measurements were of 40 Hz with "Sky looking very clark outside,"
made and encoded manually since an oscillograph but this is quickly reversed as evidence of a
digitizer was not then available to us. Feed and normal insertion accumulates; a high of 230 Hz
take-up reels were secured 45 in. apart with the is reached at BECO, a nmjor event in a success-
oscillograph paper supported by a metal bridge. ful launch.
A 32-in., clear plastic strip, with a machined After TLI 14, the trend is downward toward
straightedge, was aligned with the pitch-reference more typical levels with a small, short-lived incre-
marks, P(60 Hz), and clamped. A specially dement at Tower Jettison, another important sign
signed measuring tempIate rode along this strip, of proper sequencing. The loss of this escape
permitting time, amplitude, and pitch readings mode may account for the dip that follows im-
over about 30 in. (7+ sec) of speech in one setting. mediateb' (TLI 18). A similar (tip occurs in TLI
Oscillographs were "permanized" so that measure- 24 when the pilot notes, "My pitch cheeks at -7
ment was possible under ambient illumination. at your -3"--a possible indication of trouble, to
The quantity of speech analyzed--Twenty 100-ft which the Cape responds, "Roger, seven." The
rolls of oscillograph paper, recorded at 4.25 in./sec, news that "Cape is go" and the pilot's response,
were completely processed for computer analysis; "Cape is go and I am go. Capsule is in good
they contained 497 Groups and 5121 syllables, shape," mark the rise in TLI 27, which is followed
spanning more than 64 rain of flight and the last by a dip while the last critical events of launch
two rain of two countdowns. The exact distribu- are awaited.
tion over flights is not listed here, to avoid explicit "SECO, posigradcs fired okay," "Turnaround,"
pilot-identification. and "View tremendous" send PHH to about 203
Hz. The trend line settles down a little over the
RESULTS
next. 5 rain (TLI 2 to 6 of initial weightlessness)
A mm:ber of strong and interesting relations as capsule cheeks are executed, but rises sharply
were found between the speech measurements again (TLI 7) when he contacts Canary to report,
and emotionally toned flight events, despite thc "Control check complete .... Everything go ....
incompleteness and simplicity (due to funding Capsule in fine shape." Another slow decline,
GROUP ttIGHRST PITCH (TLI Means, I_ Hz. )
I_,illliil
iiii# !
:!It
1
3
i!_i]ll!!i
4
5 ,o_ !i!I: !iii i!i!

6
7 ill
8 iii _ )
9
10
11
12 ?:!_ !!ii
13
14
15
T[! :,;
16
17
ml
IB g_!' ;;
19 ,1!_il,2_!
20 ]lAa __
21
22
23 ::_ Z!
24
2S
26
27
_,=,_ _:,
28
29 _I _ _
SO
3 _. !i I[ :ii :]!
4 £_;_ _..
5
6 '._ J!i
7 '_i ii
8
_ ._
9 ;; rrh
10
II
12
_ 222.
13
14
15
.... i,i" A
FIGURE 2.--Variutlons in Group Highest Pitch over _ countdown,
launch, and period of initial weightlessness.
here involving 25 PHH measurements, is spread due to several factors: the great significance of
over 4 min (TLI 8 to 11). Pitch builds again over this achievement as a sign of mission success;
the next 3 min as the pilot notes, "Horizon initial weightlessness is "extremely comfortable"
brilliant blue" and "Beautiful view of African and "so pleasant it tends to become addictive,"
coast." according to pilots; and the concerns ("probable
"Probable relief" and "probable joy" (Appendix apprehension") and physical stresses ("probable
B) are the major elements of the pilot's state on discomfort") of retrosequence and reentry are
attaining orbit. Their greater intensity here is remote--hours of relative safety away.
170 BIOMEDICAL RESEARCH AND 9OMPUTER APPLICATION
Heart rate throughout these phases, although The fact that PHH peaked after establishment
above average and somewhat responsive to the of contact, while heart-rate standaM deviation
events of TLI's 1, 4, 8, and 14 of launch, shows and speech rate peaked two TLI's earlier, may be
general decline after the point of maximum phys- further evidence of a relation between high PHtt
ical effort (TLI 8) while pitch is still rising. and "probable relief." Onset of the 30-see retro-
The differences between phase means--156 Hz warning light early in TLI 4 probably accounts
for countdown (C), 189 for launch (L), and 202 for the slight (lip in PHH, tempered by the good
for initial weightlessness (IW), indicated by the news that "Retro attitude is green."
respective horizontal lines--are another salient The three retrorockets fire during the first
feature. The corresponding standard deviations half of TLI 5 with some evidence of relief in the
are 13 Hz for C, with N of 7; 21 for L, with N of text of communiques for the rest of this TLI and
53; and 20 for IW, with N of 121. The t ratios, the following one--perhaps the source of the suc-
computed without pooling variances, were 6.8 for cessive increments in average PHH over the two
(L-C), 8.4 for (IW-C), and 3.8 for (IW-L), periods. The PHH then declines over the next 4
compared with 2.5, 2.4, and 2.0 for the associated rain as "Yaw keeps banging in and out" ending
Cochran-Cox (ref. 24) t' vMues computed at the with a sharp dip (TLI 9) where he decides,
5-percent level. "i'll just control it manually."
To the extent that the trends already noted ex- Then comes the moment (TLI 10) when he is
ceed the "noise," the t' test is to be doubted. If instructed to "Leave retropackage on through
it is argued on the other hand that variations reentry." When he asks, "What is the reason for
within phases are random and that the situational this?" he is told, "Not at this time; this is the
differences among phases are the prime de- judgment of Cape Flight." To this he replies,
terminers of state, the principal reason for doubt- "Roger. Say again your instructions, please.
ing the t' test is removed. The former seems closer Over." This presentation of any intense anxiety-
to the correct position. producing stimulus [designated S -_ in the defini-
Pitch perturbation during a countdown--An tion of "probable apprehension" (Appendix B)]
unusually high standard deviation of 37 Hz for in combination with a textbook example of an
one pilot during the 2-rain countdown phase was anger-producing situation (see "probable anger"
held suspect at first, and the four PHH's con- in Appendix B) which is impossible to duplicate
tributing to it were checked; they were 179, 162, in a laboratory, is accompanied by a brief rise to
63, and 151 Hz. The last two were from a single 179 Hz (TLI 10) followed by a drop to the vicinity
"sentence" with a hesitation pause of 0.19 sec of 160 Hz where it remains (with low variation)
between the subject and the verb. His voice ap- for about 4 min (TLI 11 to 14), rising slightly
parently broke as he started the sentence, and a (TLI 15) with, "I think the pack just let go ....
split-second pause was sufficient for regaining of A real fireball outside."*
full voice control--another illustration of fleeting Pitch rises and then falls as he tries to com-
emotional expression in superbly integrated and municate through the "blackout" I(TLI 1 to 19);
self-disciplined humans. reentry]. When he does get through (TLI 20) and
Pitch variations during retrosequence and reis asked, "How are you doing?" his answer, "Oh,
entry--These are shown (fig. 3) for a pilot: pretty good," is a mild reflection of the terrifying
uncertainties of a few moments before. Pitch is
(1) who, with tess than 1 min to retrofire, had
low in this region.
reason to believe that his clock was off by several
A rise begins with "Through peak-g" and con-
seconds, signalling a potentially serious recovery
problem *The drop in PItH over the latter half of this phase
(2) whose intense efforts to establish ground was checked by the nonparametric Mann-Whitney
communications just prior to this phase were U-test, the hypothesis being that TLI's 8 to 15 were
lower in mean PHH than TLI's 1 to 7. (TLI 8 was as-
unsuccessful
signed by chance.) The 0.001 significance must be treated
cautiously since sample size varies over TLI's, so that
This is a clear instance of "probable apprehen- this test is more appropriate to sequences of individual
sion" (Appendix B) for exceedingly good reasons. PHH %
PHYSICAL M:EASUREMENTS OF SPEECH 171
GROUP IIIGIIEST PITCH (TLI Means, In Hz. )
FmuaP. 3.--Variations in Group Highest Pitch over a retrosequence and reentry.

tinues through TL25 when he reports "Altimeter the 30-see light, and rises to 147 Hz when it is
off the peg 80 000"--the start of more-familiar reported; it drops as far as 113 Hz as he counts
altitudes. There is a drop over the next 50 see for the 20-sec light and tone, and rises to 180 Hz
(TLI 26-30), coincident with "contrails and when he reports them. The countdown to squib
stuff," the "rocking," and "I can't damp it arm (at 5) and to sequence (at zero) is more com-
either." plex: there is a small drop as he notes that "Tone
The intense joy and relief occasioned by the is out" and replies "That's correct" to CC's
appearance of the main chute in these circum- instruction to "Arm the squib at 5." However,
stances is eloquently expressed in the repeated his silence (luring the count provides no pitch
affirmation, "chute looks good," and "beautihd measures until he announces the culminating
chute." A skillful ground communicator recog- events--"Squib arm" and "I have se-
nized the intensity of feeling with, "Roger. Under- quence .... "--which evidence a rise. The PHH
stand the chute very good .... " Me-mwhile PHH exhibits another decline prior to retrofire and
skyrocketed to slightly over 280 Hz, a reading rises progressively as the rockets are announced.
supported by additional oscillographic and spec- The reversal of the upward trend, in connection
trographic evidence. The decline that follows is with the comment between the second and third
slow. rockets, also qualifies as an instance. The PHH
Pitch variations during retrofire in another flight-- drops prior to "Retrojettison armed" and rises
Figure 4 is a pitch plot of a different sort, based when it is announced. A favorable fuel report
on the oscillograph roll spanning the firing of momentarily reverses the downward trend toward
retrorockets in another flight. Successive com- "Retrojcttison" which terminates in a sharp rise
muniques and their subordinate Pause Groups when the event is note(t. A simiIar fail and rise
are identified Mong the abscissa by evenly spaced occurs in the neighborhood of "Scope retracts."
index numbers. Each point on the plotted line is It is regrettable that time did not permit follow-
the PHH of the single Group for which the ing this pilot through the reentry phase with this
verbatim text is given. Relations between pitch detailed analysis, or establishing the mean and
changes and events are portrayed more precisely range of his pitch excursions in other
here despite some distortion in the time dimen- circumstances.
sion. The TLI's are marked by vertical lines that Figure 4 is methodologically important; it
intersect the text at the two places where they demonstrates that single values of pitch, each
split Groups--in Groups 09-3 and 14-6. The five representing a segment (Pause Group) of a com-
TLI means, which would have appeared as five munique marked off by the 170-msec criterion
successive points if plotted on a TLI basis, are silence and without adjustment fi)r speech rate,
here represented by horizontal lines. can track key events--witil some "noise" granted.
The most striking point is the behavior of This is short of surprising only because of the
PHH during retrofire. The firing of retrorocket-1 more-striking correlations of the figure 3 reentry
is reported with a PHH of 179 Hz; retrorocket-2, which, after all, had an average of only 1.5 PHH's
with 201 Hz; and retrorocket-3, with 214 Hz. This per TLI. The TLI means (horizontal lines) do
is not a "cheer-leader pattern" with progressively reflect the low pitches (luring the count(h)wns
higher pitches on "one," "two," and "three," but preparatory to retrofire, the increase (luring retro-
an apparently spontaneous expression of "relief" fire, and the decline that follows. However, if
and "joy" at having "got three"--with the 214 the boundary of TLI 4 had occurred between
I-Iz peak over the "got" on the oscillograph.
06-2 and 06-3, say, instead of between 05-1 and
Also of interest is the Mternation that is evident
05-2, the picture wouhl have been blurred by a
between a tendency toward low PHH in situa-
plot of TLI means. The standard deviation (not
tions involving a countdown or wait for a critical
shown) being highest in TLI 5 alerts one to the
event (recall the low pitch for the countdown
phase of fig. 2) and a tendency toward higher fact that something has happened, but only ob-
pitch in reporting of the outcome: PHH drops servation of the details of figure 4 leads to any
from 217 Hz to about 120 Hz in anticipation of real insight.
PHYSICAL MEASUREMENTs OF SPEECH
173
GROUP II/GIIEST PITCH. Hz.
I_i!i
03-1
05-1
06-2
06-3
06-5
06-E
18-(
20-2:
Fz6uaE 4.--Variations in Group IIighe_ Pitch during a retrofire.

Some Variables At_ectlng PHIl during High Arousal duction in speech volume without appeal to con-
ditioned suppression. During periods when fine-
What are the controlling variables causing grain or numerous corrections must be introduced,
pitch to be sometimes high and sometimes low or when an optimum state of readiness is needed
when situational and other evidence indicates for perception of multiple or near-threshold sig-
high arousal? nals and making of appropriate responses, speech
Low pitch (PHH) during high arousal--Several output is probably restricted to those minimal
hypotheses are considered. First, individuals at- utterances (e.g., "Light is on" and "Tone is out")
tempt to bring pitch dowel to levels clcarly m_so- that were established in simulators as integral
ciated with calm, confident, competent per- parts of the S-R chains. Routine acknowledg-
formance in situations where important reinforcers ments ("Roger," "That's correct") and well-
are contingent upon demonstration to a listener established verbal sequences ("5, 4, 3, 2, 1, 0")
that all variables are under control. Pitch inflec- also may occur in these circumstances. The pilot's
tions, intensity, pausal phenomena, and other control over these multiple performances is prob-
discernible speech properties may be brought ably maintained by discriminative responses to
under similar control. deviation signals as he scans his own outputs.
The plausibility of this hypothesis is suggested When the system is relatively stable, the pilot
by (1) tile existence of speech-aural and speech- can switch to generation of more-spontaneous
kinesthetic feedback loops that provide the in- messages such as "Oh boy. She's a good little
formation cssentiaI to deveIopment of self-editing capsule, I'll clue you."
and self-controlling response; (2) the existence of However, these considerations do not identify
reinforcement contingencies that foster tile de- any process by which pitch is lowered. On the
velopment of self-control more with respect to contrary, increased glottal tension would be ex-
the expression of unpleasant or negative emotions pected from the spread of muscular tensions as-
than of pleasant or positive ones; and (3) tile oc- sociated with control of capsule attitudes and
currence of instances of obvious control: for maintenance of a high degree of readiness to
example, the quick resumption of normal speech respond discriminatively, and from the task stress
by tile pilot whose voice broke momentarily induced by concurrent critical activitics.
during a countdown, and the low pitch Ievel Fourth, in a situation where the set of possible
maintained during the several minutes (TLI 11 outcomes produces anxiety, the introduction of
to 15 of retrosequence) spanning tile landing-bag required task performances may reduce emotion
discussion. by distraction (i.e., by substituting the emotional
Second, an Estes-Skinncr (ref. 25) conditioned- responses elicited by the inserted tasks) if one
suppression paradigm is an element of situations assumes that any strongly aversive possibilities
involving countdo_ms or waits for critical events. present are not highly probable.
The probability of a highly noxious or trouble- Finally, glottal relaxation may be an integral
some outcome, on termination of the count or time part of the "orienting reflex" elicited by the se-
interval, is non-zero, and many of the important quence of discriminative stimuli culminating in
variables determining the outcome are beyond the the critical outcome and mediated perhaps by
control of tim pilot. Experimentation is needed such physiological correlates of this reflex as
to determine whether it is proper to speak of the reduced heart rate and blood pressure.
lower pitch observed in these circumstances as Elevated pitch (PHH) during high arousal--
"pitch suppression," thereby linking it with tile Situations conforming to the definitions of "prob-
well-established phenomenon of "response-rate able relief" and "prob:tble joy" (Appendix B)
suppression." The fact that speech output is yielded the highest pitches observed, for example,
diminished during countdowns is compatible with the attainment of orbit, the appearance of the
this hypothesis. main chute, and the favorable terminations of
Third, the on-going activities during count- countdowns and waits. The greater freedom of
I
downs and similar situations may explain the expression accorded pleasant emotions is another
brief, intermittent utterances and the overall re- factor.
PHYSICAL MEASUREMENTS OF SPEECH ]75
High pitch may also be expected with un- "economy size" computer program, TLI means
pleasant arousal under some conditions despite were used.
feedback loops. There is one such situation when The DURLL Ratio during countdown, launch,
the individual is attempting intensively and con- and initial weightlessness--The three horizontal
structively to cope _ith a hazardous situation lines in figure 5, at values of 0.99, 1.40, and 1.39
under time stress, and when the hazards are (or for phases C, L, and IW, respectively, are the
are thought to be) recognized by others; Berry's mean ratios computed from the corresponding
public comment on the high pitch of two Gemini light-phase means for DURLL and DURL. The
astronauts while freeing their capsule from the lines are obviously not centered on the trends,
perilously oscillating rendezvous booster is an since the ratio of the means is not equal to the
illustration. Control of voice to communicate mean of the ratios.
calm self-control would have been incongruous, The dramatic increase from lift-off through
particularly since the information communicated BECO and on to a peak of 2.26 at Tower Jettison
was enhanced by the pitch elevation. Another may well reflect the concern in Project Mercury
such situation is when arousal reaches the panic with these earl)' stages of the launch phase, and
level, i.e., when the individual's behavior comes the joy and relief on their successful completion.
to resemble the energetic, problem-irrelevant ac- This increase is not a function of acceleration
tivities of some animals in a conditioned-suppres- since the line continues to rise as acceleration
sion procedure; and a third prevails when the falls after BECO, and declines during much of
immediate social environment is not so highly the subsequent buildup of gravity (TLI 18 to 26).
constraining as this one. The downward trend is reversed with "Cape is
Finally elevated pitch is likely in communica- go .... All systems are go" and the approach of
tions under low signal-to-noise ratios, as discussed SECO.
later. A peak of about 1.58 (TLI 2 of IW) follows
SECO, turnaround, and the attainment of weight-
The DURLL Ratio lessness. The value declines over the next 7 min
Computation and significance--The end of a (TLI 3 to 9) during capsule checks, and then rises
phrase is signalled by increased duration of the sharply as he begins his very favorable report to
last syllable and by fine-grain variations in its Canary.
pitch and intensity contours. The hypothesis The DURLL Ratio during retrosequence and
that the acoustical correlates of linguistic ter- reentry--The downward trend over the 11 + rain
minals vary with changes in state of the speaker from TLI 6 of retrosequence to TLI 14 of reentry
is most readily tested with syllable-duration data. is clear in figure 6. The values over the 4+ min
To the extent that Pause Groups and phrase from the end of the communications blackout to
boundaries are coincident, the final syllables of beyond the opening of the chutes (TLI 14 to 41)
Groups tend to be longer than the same phonemic are puzzling. They are not only low but show
combinations occurring elsewhere. A strong as- their peaks one or two TLI's after the corre-
sociation between the two is expected in astro- sponding ones for pitch: TLI 25 for PHH versus
nauts' speech since, as noted earlier, most of it TLI 27 for the DURLL Ratio; and similarly 33
is rehearsed, descriptive, or concerned with mat- versus 34, 38 versus 39, 44 versus 47, and 53
ters extensively discussed before flight. versus 54. The dip at the end of phase IW (fig. 5)
DURLL, as the DURation of the group Last while PHH is peaking (fig. 3) is also noted.
syllable was called for FORTRAN programming, The most plausible explanation is that pitch
is the time between the onset of the last syllable prominence is associated with the lengthening of
of the Group and the end of the Group. DURL is syllables earlier than the last in sequences like
the average duration of the remaining syllables, "Brilliant blue," "A real fireball," "Tremendous
computed over a Group, a TLI, or a flight phase. view," and "Beautiful chute," resulting in tem-
The DURLL Ratio = DURLL/DURL. This nor- porary depression of the DURLL Ratios in these
malization sbould be made on a Pause Group TLI's. If so, substitution of the Group Maximum
basis; instead, as one of the compromises of the Syllable Duration for DURLL, as discussed later,
BIOMEDICAL RESI,:AR('tl AND COMPUTER APPLICATION
176
_'DUR LL RATIO"
1o
11
lO
11
12
,d
2o
FIGI_RE 5.--Variations in the DURLL Ratio over a countdown, launch, and period of initial
weightlessness.
"DUHLL RA TLO"
• i
eL
;I
:l f
.,
I
_L
_D
63
64
65
FmURE 6.--Variat,ions in the DURLL Ratio over a retrosequenee and reentry.

would correct for this and shouId be tried. Seg- (fig. 3). ARMS starts higher (reIative to the retro-
mentations not coincident with phrase bound- sequence mean) than PHH and continues to
aries, due to hesitation pauses or the use of a fixed climb until retrofire is over; then it drops
criterion silence regardless of speech rate, also markedly. Pitch, on the other hand, begins to
could produce low values. Finally it is possible decline as soon as contact is reestablished and
that the low values reflect the state of the astro- assurances are received that retrofire time will
naut after what must have been an emotionally be signalled properly. The initial association of
exhausting flight. elevated pitch with high ARMS may be another
Concern about the false landing-bag-deploy- illustration of reduced voice control while an
ment signal, telemetered from orbit, is a likely obviously urgent situation, as seen by the pilot,
explanation of the extraordinary value of 4.10 is coped with, or of the use of these parameters
(TLI 47) when he says "This is [blank] seven, to communicate urgency. These interpretations
standing by for impact." This is followed by are reinforced by the peaking of heart-rate varia-
another peak (1.96, TLI 54) when he confirms, bility in TLI 2. Or the higher subglottal pres-
"Landing bag is on green"--an important S +r sures, generating the loud speech by which the
(Appendices) even if it did not dispel all pilot attempted to reestablish communications,
uncertainties. may have induced glottal tensions, thereby in-
The DURML Ratio as an alternative--The pos- creasing voice frequency. This process may well
sibility of substituting the DURation of the Maxi- contribute to the high ARMS and PHH in the
mum syLlable (DURML) extent for the DURLL vicinity of TLI 62 where channel conditions
of each Group, mentioned above, was discussed occasion "Say again." However, pitch does not
in the project report. Tile measure has consid- rise when the pilot shouts (TLI 14) to get through.
erable promise. It is less dependent on the as- the communications blackout of reentry, and it
sumptions of coincidence between Pause Group falls as ARMS rises prior to retrofire (as already
and phrase or sentence boundaries; in fact the noted), suggesting that this process is at most a
DURML measure would be improved by an in- small part of the observed variations.
crease in the criterion silence to perhaps 250 msec. ARMS rises more dramatically than does pitch
Tile same is true for Group Highest Pitch but in response to the order to leave the retropackage
probably not for the DURLL Ratio. It reflects "on" and does not diminish, as pitch does, when
the expressive lengthening of syllables other than the request for an explanation is denied. The
the last, as in "... beautiful chute" and other hypothesis relating low pitch during this period
examples cited above. It is equivalent to the to the pilot's control of intense emotions is
DURLL Ratio when the Group's last syllable is recalled.
in fact the longest. There was closer agreement in general con-
On the other hand, the DURML Ratio mixes tours, when the plot of TLI heart-rate means was
two different linguistic processes and two different superimposed on the ARMS plot for retrose-
definitions of the syllable in the numerator. Tile quence, than with PHH. Heart rate starts
respective points can be identified in computer moderately high, rises a tittle to retrofire in TLI
plots, and it is possible that a level correction 5, and declines about 15 beats per minute (bpm)
may be derived to remove discontinuities between by TLI 8; it starts to rise in TLI 9 with "This
the two. DURML is typicalIy more dependent [automatic yaw control] is banging in and out
on the precision of syllable marking, being here," and continues to a peak slightly above
bounded by syllable onsets. DURLL requires 100 bpm (with rates within the TLI reaching
only one syllable-onset, but is subject to error in 120) when the explanation is finally given in
detection of the end of the Group in noisy speech. TLI 12. It then declines somewhat over the next
two TLI's and rises again in TLI 15.
Syllable Peak Amplitude
The lag between heart rate and ARMS is less
Syllable Peak Amplitude in volts rms (ARMS; than that between heart rate and PHH when these
fig. 7) seems more closely related in time to indi- variables peak in response to events culminating
cations of apprehensive arousal than is pitch in appearance of the main chute. The ARMS
SYLLABLE PEAK AMPLITUDE (TLI Means, Vrms.)
i _ !!
!Ian "I :
,I
!i
_drJi' _ ho
¢. II_AI :.:h,
Ma*,,.,
4ff
_1_ VDn
49
63
!#J',) !_'!!Ht!!] I
Fmtlm_ 7.--Variations in sylh_ble amplitude pesks over a retrosequence and reentry.

shows its major peak in TLI 36, declines sharply to estabIish communications before retrofire:
in TLI 37, and evidences only a small increment "Hawaii, did you receive? Over." The strong
when PHH attains its maximum with a lag of motivations of the moment were reflected both
20 sec in TLI 38. Heart rate peaks (138 bpm) by the extended durations (0.22 and 0.32 sec, re-
in TLI 34, drops slightly by TLI 35, is about spectively) of the two syllables in "receive" and
120 bpm when ARMS and PHH show their by the rapid articulation of the other syllables
respective peaks, and continues its decline until at a mean of 0.14 sec. Inclusion of DURLL (or
TLI 45 where it rises slightly, with ARMS DURML) in rate measures would tend to obscure
and PHH trailing it all the way in that order. these subtleties.
This lag in PHH is compatible with previously It is interesting that the mean TLI Syllable
noted associations between the "relief" and "joy" Rate for the pilot most extensively studied ranged
combination and pitch peaks. from 4.0 to 8.0 syllables per second on the basis
These observations are merely suggestive since of 24 and 23 syllables in the respective TLI's.
time plots are'intended mainly for demonstration One lower mean of 2.8 occurred, based on a single
of tim responsiveness of parameters to situational syllable. Differences between these results and
changes. Relations among variables and the as- typical published figures are due to the elimination
signment of "probable state" labels are more of DURLL, the taking of reciprocals, the elimina-
properly studied when all data are assembled tion of silences greater than 170 msec, and the
on a Pause Group basis for analysis, as discussed relatively small number of syllables in each
in tile Appendixes. computation.
The tremendous drop (0.68 to 0.38 V rms)
First-Order and Second-Order Difference Measures
between TLI's 64 and 65 is without parallel, per-
haps involving capsuIe power, bracing for im- Hypothesis and specific measures--On the as-
pact, or other physical cause. The value recovers sumption that the sequential properties of the
to 0.55 V rms in TLI 67 but dives again to 0.38 V number series, representing a given measurement
rms in TLI 68 (not plotted) when he says, on successive syllables, might vary with the state
"... seven. Impact. Rescue is manual." of the speaker, algebraic first-order and second-
order serial differences were computed for Syllable
Syllal)le Rate Duration (DL), RL, and Syllable Peak Amplitude
This measure was computed as the reciprocal (for which A is an abbreviation of ARMS). Tile
of each Syllable Duration before averaging over standard deviation was selected as the most
a TLI. This method of computation assigns higher promising measure. With SD symbolizing this
weights to faster syllables since the reciprocal statistic and F1 and F2 used for finite differences
function decreases rapidly as Syllable Duration of first-order and second order, the variable names
increases from below 0.1 to about 0.3 sec. This of interest are SD(FIRL) and SD(F2RL),
differential weighting probably explains the ap- SD(F1DL) and SD(F2DL), and SD(F1A) and
parently greater validity of Syllable Rate (RL), SD(F2A). In continuation of this symbolism the
as compared with Syll,_ble Duration, both graph- respective standard deviations, computed with-
ically and in the statistical checks employed. In out regard to sequential dependencies, are SD (RL),
this respect, RL is akin to the widely used beat- SD(DL), and SD(A). Omission of the Group
by-beat index for heart rate. The question of Last Syllable from indices involving RL and DL
whether the reciprocal transformation is appro- probably diminished the validity of these meas-
priate to the respective mean-variance relations ures a little, but reduced redundancy with the
was not investigated. DURLL Ratio to which they are somewhat
The duration of the last syllable in each Pause related.
Group (DURLL) was excluded from RL, being Some Properties--Inspection of the time plots
under the control of unique linguistic constraints. (not presented here) indicated that (1) the SD(F1)
The reasonableness of this choice is illustrated and SD(F2) measures tended to covary for each
by the first Group uttered in TLI 1 of the retro- parameter; (2) the difference measures for RL and
sequence, where the pilot was desperately trying A showed some good correlations with flight
events; (3) the lower apparent validity of Several examples of "blow ups" in SD(F1RL)
SD(F1DL) and SD(F2DL) is further evidence and SD(F2RL) were also examined, one being
that the occurrence of rapid syllables within a the now-familiar first TLI of the retrosequence
Pause Group is a more sensitive index of certain which includes the two consecutive Pause Groups
psychophysiologieal stresses than average syl- whose syllables and durations are as follows:
lable rate or duration, as ordinarily computed; "HA- (0.06 sec) WA- (0.17) II (0.14), DID (0.13)
and, just as RL weights these shorter syllable YOU (0.13) RE- (0.23) CEIVE (0.22)? O- (0.15)
durations more heavily than does DL, by virtue VER." (0.19 = DUIILL, i.e., last syllable of Pause
of the reciprocal transformation, so SD(FIRL) Group #1); and "(H)E- (0.10) LLO (0.19)
weights successive differences involving shorter HA- (0.13) WA- (0.17) II (0.14), DID (0.15)
durations more heavily than does SD(FIDL); YOU (0.13) RE- (0.22) CEIVE?" (0.32=
and (4) the TLI plots for SD(F1RL) and DURLL).
SD(F2RL) showed occasional, sudden, extreme As suspected, it is the taking of reciprocals of
values or "blow ups." occasional, very short syllables (e.g., 0.06 see)
To clarify the properties of these difference and the squaring of successive differences that
measures, a simple model was constructed con- cause the trouble, affecting SD(F2) more than
sisting of strings of U's and A's (Unaccented and SD(F1). The ratio SD(F1RL)/SD(RL) virtually
Accented) related by a Contrast ratio (C) greater eliminates "blow ups" and provides a concep-
than 1 so that A's are uniformly higher in value-- tually better measure of sequential dependencies
longer, more intense, or higher in pitch. Expres- by normalizing against the level of nonsequentially
sions for SD(F1) and SD(F2)/SD(F1) were dependent variability. Time plots of this ratio
found to converge rapidly on _f3 as a lower bound showed responsiveness to situational changes, but
as the number of "syllables" increased. this is an area requiring more research.
Considering the simplicity of the model, it was
a great surprise to find the actual values for all Other Speech Parameters
variables and phases of flight (ranging from 5 to Pause Group Duration (DURG) and Syllables
15 min) in close agreement: the mean for 17 phases per Group (ENLG) showed some promise in the
(exclusive of prelaunch countdowns) was 1.73, TLI plots; ENLG appeared more responsive, but
with 16 values ranging from 1.68 to 1.77 and one DURG made a stronger showing in the statistical
value for a launch at 1.88 (for DL). The averages comparisons made. The Coefficient of Variation
for the three variables were 1.715 for RL, 1.738 suggested that ENLG lost out in flight-phase
for DL, and 1.740 for A (ARMS). For two sepa- comparisons where it was more variable--most
rate prelaunch countdm_ms the values were 1.44 responsive. The values of this coefficient were
and 1.46 for DL and 1.51 and 1.57 for IlL, a result themselves interesting; for example, DURG was
that merits further study. The corresponding lowest (48) during a countdown and highest (74)
values for ARMS were 1.80 and 1.69.
during a retrosequence. To the extent that inhala-
Furthermore, when the computed values for tions occur during criterion pauses, ENLG ap-
SD(FIRL) were multiplied by yr_ as a scale factor
proximates Goldman-Eisler's (refs. 11 to 13)
and plotted, the correlation with SD(F2RL) was Syllable Expulsion Rate (Ell) computed as syl-
nearly perfect except for several excursions of lables per expiration. Under conditions of "rea-
SD(F2RL) where it seemed to contain additional sonably constant sound pressure level" she
information. Attention was therefore focused on
interprets the reciprocal, 1/ER, as the proportion
the expression for SD(FIRL), which is simpler. of returned air current used to produce each syl-
The constraints were redefined, and the expression lable. Topical analyses and "correlation" with
was modified to cover either phrase-like strings
or indefinitely long strings that approximate a graphical analysis of data for this project was completed
little more closely those of spoken language.* by me in my own time out of personal interest. The
computer provided the contractually required phase and
*The luxury of pursuing such questions (without regard TLI means and standard deviations, together with sums
to contractual scope, economics, or other practical values) of squares and a convenient ordering to facilitate manual
derived from the fact that most of the statistical and analysis.
some psychiatric ratings suggested that "high against an average syllabic duration (DURL)
values belong to content implying free flowing or derived from the Syllable Count and the Group
outgoing effect" and "low indices.., belong to Duration with the last (or longest) syllable ex-
topics of restricted emotionality, which implies cluded from both; (3) PHH; (4) RL; (5) Syllables
tension states or intellectualized speech." per Group; or (6) a validated selection of these.
Duration of pausal silences (inter-Group si- This is a vast simplification of the procedures used
lences greater than the 170-msec criterion pause) in this rescarch, particularly with respect to
was poorly distributed over TLI's and showed no syllable marking.
promise of validity. Level-2 scoring would be applied whenever a
more detailed analysis seemed warranted by
RECOMMENDATIONS level-1 results, medical data, or other on-line
information: (I) Syllable Peak Amplitude
A Semlau_omofi¢ Monlfor_ng System (ARMS) is a promising example, but requires a
General description--Detection of changes in normalization to correct for level changes during
speech parameters by human monitors would be propagation and elsewhere; this could be accom-
enhanced by a real-time, CRT display of func- plished by digitizing of each peak as we have
tions such as S(t), A(t), and P(t) (fig. 1). clone; then, instead of computing the average, the
On detecting or suspecting a change in state, computer would select the Group's highest peak
the monitor would initiate analysis of selected and normalize it against the average of the re-
speech samples. This would be done at a console maining ones, forming an AI-II-I Ratio analogous
equipped for oscillographic readout, for computer to the DURLL and DURML ratios and (except
access via an x-y oscillograph digitizer capable for normalization) to PHH. (2) A computer
of coordinated keyboard entries in an efficient operation on the same amplitude data could yield
code, and for playback over a system such as the the difference measure SD(F1A)/SD(A). Finally,
one used in this research, which enables repetitive if the marking of syllables could be accomplished
audio and CRT display of short speech selections or facilitated automatically or if the considerable
without tile inconvenience of tape loops or the labor of manual marking could be justified, (3)
long recycle times of Tape Search Units. the DURLL or DURML Ratio and (4) the RL
Two levels of scoring are envisaged: Measures measures could be improved; and (5) the dif-
in level-1 wouht be selected by further research ference measure SD(F1RL)/SD(RL) on LR could
for both validity and speed. The raw inputs might be added.
include: (1) Group "on" and Group "off" times Requisite research--Assembly of a laboratory
encoded as distances along x from a reference version of the proposed operational console, with
point; (2) the onset of the DURLL or, if a dif- computer access, and its use for developing pro-
ferent. DURML were readily apparent, the two cedures and for conducting research is rccom-
onsets bounding it; (3) PHH and its calibrated mended. Crews of Apollo, MOL's, or high-per-
base line, P(60 Hz); and (4) a Group Syllable formance aircraft could serve as subjects in
Count derived by tapping of the perceived syl- simulators or in flight. Some development and
lables into a digital counter, perhaps with check- improvement of speech-processing equipment
ing of the oscillograph to avoid the error of would be necessary for the recommended traces
perception of elided syllabics. A less precise count, and for promising new ones.
probably adequate for level-1 scoring, can be Criterion pauses should be studied as a func-
made directly oil the oscillograph without auditing tion of speech rate, with the results applied to
of the communique. segment communiques into Pause Groups--for
With these simple inputs, time plots (preferably example, by employing a very short value (such
on a Pause Group basis as in fig. 4) can be gener- as 150 msec) and programming the computer to
ated by a small computer for the following apply and regroup entries by the longer criteria
variables: (1) Group Duration (DURG); (2) a derived as a function of speech rate.
DURLL Ratio, modified to permit either DURLL Other questions merit research: Does segmenta-
or DURML in the numerator and normalized tion by a rate-proportionate criterion pause coin-
cide any better with listener-detected pauses? serves most automatically analyzable speaker-
With breath pauses? (A respiration time function state information in addition to making better
recorded on a channel parallel with speech would use of available power for transmitting intelligible
be desirable.) With linguistically appropriate speech over great distances. The tradc-offs will
boundaries? Does the distribution of articulatory be particularly clear if speaker-state analysis
pauses vary with emotional state when normalized proves its worth in further validations.
for speech rate? How do these variations affect Hybrid (voiced-excited) vocoders digitize a base
the optimum criterion pause which is essentially band of low-frequency voice energy as an analog
an upper limit on or at least a very high value function and employ channel-vocoder techniques
within this distribution? for the rest of the transmitted spectrum. At the
To what extent would a combination of cri- receiving end, the narrow base band is distorted
teria-for example, the requirement of an in- to yield a wide, flat spectrum capable of energizing
creased DURLL Ratio on the syllable preceding all channels of the synthesizer; this improves
a criterion pause--improve segmentation? Im- speech quality and preserves more aurally dis-
provement appears likely, since syllable prolonga- criminable information on state, but complicates
tion is one of the linguistic "terminals" that automatic speaker-state analysis. Vocoders have
signal the end of a phrase (ref. 26), even one inter- been produced by General Dynamics, for example,
rupted by hesitations. With good-quality speech, with selectable channels and voice-excited modes.
the addition of terminal pitch and amplitude
criteria would merit consideration. Supplementary Laboratory Research
Exploratory validations should be continued, General--The search for speech parameters and
utilizing time plots, trend tests, and other simple weighting functions that discriminate among prob-
alternatives to Probable State Analysis. Finally, able states would also be facilitated by a simple
when procedures are stabilized and measures have laboratory set-up for testing and clarifying rela-
been selected, a full-scale validation study should tions in data from simulators and actual flight.
be executed, using a "streamlined" version of A set of standard individual and group tasks
Probable State Analysis (Appendices A and B). must be developed for generation of speech under
laboratory control. The several classes of verbal
Feasibility of a Wholly Automatic System responses, differentiated on the basis of con-
The main hope for an automatic speaker-state trolling stimuli (ref. 28), should be represented.
monitoring system, applicable to space flight Some should generate "constrained" speech--
within the next 5 years or so, lies in development for example, propositional functions or formats in
of a spacecraft channel vocodcr (ref. 27), and which the subject inserts the appropriate word
computer programs for analysis of the digital from a prescribed set; others should generate
transmissions to identify Pause Groups, Group relatively "unconstrained" or "free style" speech.
Highest Pitch, and other parameters including The verbal output should be an integral part of
some new measures not possible with current the task; thus the experimenter's interest in it
analog transmissions or on-board recordings. In is concealed. The tasks should facilitate applica-
fact, programs of this type--for example, for tion of various reinforcement contingencies for
use of spectral information for improvement of manipulating the state of the speaker.
syllable-onset detection during voiced segments, Generation o] "unconstrained" speech--A signal-
particularly in the presence of semivowels, voiced detection task is probably the most efficient way
fricatives, and nasals, and for formant tracking-- to generate quantities of comparable verbal re-
have _ been developed in my laboratory and sponses of the "constrained" type for each experi-
elsewhere. mental condition. Experimental error variance
The speech synthesized at the receiving end of is reduced by the use of two stimulus conditions--
a ehannel-vocoder system has an artificial quality, Signal plus Noise (SN) and Noise Alone (N)--
seeming to lack much of the aurally discriminable and the requirement of only two verbal responses.
state information ordinarily important to medical Presentations may be initiated by the subject or
monitors. On the other hand, this system pre- the experimenter.
The S/N ratio may be decreased either to in- This would bring key-press responding under the
crease the difficulty of the task or to deprive the control of the respective indicator lights and
subject of a dependable basis for his decisions remove or drastically reduce subvocal control
and permit administration of a random or planned which is essential to the inferences of interest.
sequence of successes and failures. However, Manipulation of speaker's state in the laboratory--
this procedure shifts response control from the The concepts and stimulus operations of rein-
stimulus to sequential or other adventitious forcement theory provide the main basis. Con-
variables (ref. 29). The suggested inclusion (ref. tingent reinforcements (those determined by the
30) of easier discriminations in near-threshold correctness of the decisions) can be accommodated
studies maintains discriminative responding by by signal-detection theory to the extent that they
a periodic reinforcement. can be expressed in a "Ioss function." Noncontin-
Tile basic verbal report may be Signal/No gent reinforcements used to manipulate state are
Signal, Affirmative/Negative, Yes/No, etc. In outside the formal structure of decision theory as
addition the subject may be required to request applied to the detection task (but not to the sub-
each stimulus presentation, to assign a "confidence ject's decision to ]cave or go along with the situa-
level" to each decision, or to keep track of and tion). The task lends itself readily to either typc
report presentation numbers and outcomes for of reinforcement.
the record. The possibilities can be expanded "Probable apprehension" (Appendix B) can be
by combining these with any one of a variety of simulated by conditioning of a small panel light
simple numerical, geometric, or verbal tasks, as a sign (S-') of impending, unavoidable, aversive
while establishing contingencies on only the de- stimulation at the end of about 3 rain in accord-
tection task. ance with the Estes-Skinner "conditioned sup-
Reduction in the number of decisions, and con- pression" procedure (ref. 25). Termination of the
sequently in the value of signal-detection statistics warning light by either an avers;re or a pleasant
as a source of information about the state of the outcome, in some probability mix, is more
subject, is the price paid for this greater variety analogous to real countdowns.
of verbalizations and the greater task complexity; A type of "task stress" can be induced by a
this is directly compensated by data on a more modification of the Sidman avoidance technique
diverse range of task performances. (refs. 33 and 34). In the standard procedure the
Provision for pressure-key responding--A pres- subject receives one shock for failure to respond
sure key was installed as part of the signal-detec- within n seconds of his last reponse (the "re-
tion setup that was nearly completed before this sponse-shock interval") and an additional shock
project arrived (refs. 31 and 32). Tile intention every m seconds (the "shock-shock interval")
was to determine whether changes in the dura- until he responds again; the parameters m and n
tional and intensive properties of speech during and shock intensity are important. As applied to
state changes are accompanied by analogous the signal-detection task, shock-avoidance must
changes in the responses of a different musculature be made contingent not on the mere occurrence
not subject to specific social conditioning and but on the correctness of decisions during the
styling. response-shock interval--for example, q correct
This requires freeing pressure-key response decisions every t seconds, with a new interval
from the control of subvocal speech which occurs, initiated when the criterion is met or at the end
for example, in tapping-out the syIIables of "af- of t seconds. With certain values of q, t, and S/N,
firmative" and "negative," or in counting to press the subject can increase the probability of post-
a specified number of times for "yes" or "no." ponement of the shock more by crowding extra
Dot-dash patterns encourage the dit-dah verbali- responses into the response-shock interval than
zation of the Morse code beginner. A possible by taking extra time with each decision because
solution is to press a variable number of times of the nature of the task. The procedure would be
(say 3 to 5) until a light indicating "affirmative" superimposed on a schedule providing payment
goes "off," and an independently varied additional for each correct response. With knowledge of the
number of times for a negative-decision indicator. distribution of task-cycle times for regular rein-
forcement, the probability of drawing a sample of Evidence of the effectiveness of these laboratory
n responses, with an average time not exceeding a "situational" changes would be sought in measur-
specified value, can be used to establish t. The able changes in task-performance data, intro-
speed incentive can be removed by a modification, spective reports and checklists, psychophysiolog-
that may be termed the "block correctness ratio ical measures, and other response information.
method," in which the subject must produce
correct responses in each block of n to avoid aver- ACKNOWLEDGMENTS
sive stimulation at the end of the block. The electronic equipment was engineered by
Boredom may be occasioned by long sessions, L. C. Stewart and W. D. Larkin, and the cali-
easy decisions, a fiat hourly rate with no payoffs brated oscillographs were painstakingly produced
or punishments, or long waits between series. It by S. B. Swackhamer. Drs. R. A. Houde and W.
can be relieved at least temporarily by more- B. Newcomb provided occasional consultation in
complex contingencies. linguistics, as did Mrs. Marian MacEachron who
These were among the paradigms examined and also helped with the computer programming. Dr.
adapted for assessment of the versatility of the Willis marked syllable onsets and performed the
signal-detection task for investigation of changes tedious oscillographic measurements and numeric
of state relevant to space flight as they affect encoding in the summer of 1966 when he was _,
"constrained" speech. The applicability of these graduate student in linguistics at the University
contingencies to some tasks suitable for generating of Rochester. The favorable responses of Drs.
"constrained" speech has also been explored. In C. A. Berry and A. D. Catterson, Manned Space-
addition, several experiments aimed at clarifica- craft Center, to a proposal in 1964 resulted in
tion of the speech-parameter variations observed sponsorship by Dr. J. F. Lindsey who saw the
during countdowns and waits have been sketched. possibility of incorporating the proposed "prob-
An informal report (ref. 35) is available from the able state" concepts and acoustical techniques
writer on request. with his Time-Line-Analysis approach.
APPENDIX A
ASSESSMENT OF ASTRONAUTS' CHANGES OF

STATE FOR SPEECH RESEARCH
PROBABLE-STATE ANALYSIS its report is usually short. These factors facilitate

General--Probable-state analysis is a method setting of state-related data into approximate
for organizing situational and response evidence time correspondence with speech measurements
sequentially, and for using it for generation of for analysis.
10 separate time functions--one for each of the The Pause Group as a unit of analysis]or valida-
probable states selected--depicting variations in tion studies--As reported in the main text, physical
the state of the pilot,. measurements have been made from the Group
The rationale was described in a preproposal as a subdivision of the communique and from the
document (re]. 36) and further developed during syllahle as a subdivision of the Group. For bring-
this study. The modifier "probable" was intro- ing all measurements into proper time corre-
duced to emphasize the incompleteness of evidence spondence for a validation study, those from the
ordinarily available and the consequent need to syllable must be converted to a Group basis;
avoid direct and unqualified assertions about separate state vector must be generated for each
the state of a pilot--except perhaps where the Group. It is the state of the pilot over the time
evidence is written plainly for all to see. span of each Group, not the Group or its text
Spacing and time span of state and speech meas- per se, that is assessed in the light of all available
urements-The speech-pressure wave form is a evidence except the physical properties of his
time function, so are its various transformations speech! This sets the stage for the statistical
into voltages representing amplitude, pitch, for- analysis of results and is quite compatible with
mants, and other parameters. These functions the Time-Line Analysis approach. For, if success-
exhibit marked discontinuities within periods of fuI, state predictions made from the acoustical
so-called continuous speech and go to zero between measurements from each Pause Group could be
communiques. assigned to _he TLI in which all or the greater
Measurement operations on these transformed number of s)_llables occurred, or distributed with
functions yield sets of numbers or measurement appropriate weights.
vectors, each representing the pilot's speech be- The main problem is the number of judgments
havior over the time span of the communique, required of raters. Summing of all acoustical
or a segment thereof (Pause Group), from which measures over communiques and generation of
the measurements are made. Speech segments state assessments over their respective time spans
are spaced unevenly over time, and measurements ease this task somewhat but add new problems.
from them are identically spaced. Probable-state Communiques vary in length from a "Roger" to
vectors must have the same-spacing and time status reports running many minutes in orbital
span as have the speech measurements if relations flights. Even if these are subdivided topically,
between the two are to be explored--a type of the range is too great from the standpoint of
asynchronous "sampling" that may be called statistical analysis or TLI application.
"event paced." However, for testing of the procedures and
Fortunately the text of space-ground commu- concepts of probable-state analysis, this expedient
nications is a very valuable source of information was adopted. Data on state were organized se-
about situational changes and pilots' responses quentially by communiques or their topical sub-
to them, and the interval between an event and divisions with the proviso that further subdivi-
187
][S_ BIOMEDICAL RESEARCH AND COMPUTER APPLICATION
sions could be effected whenever necessary to pies the central portion. The pilot's communiques
portray changes in state that were considered (Q's) are in solid capitals. Left and right por-
significant. A suborbital flight was chosen for tions of the original llX17-in, data sheets have
the trial analysis. been deleted since some of the information--
Sequential organization of probable-state informa- introspections, retrospections, task-performance
tion-An adaptation of the techniques of Barker details, situational factors, and medical data--
(refs. 1 to 3), for recording and portrayifig the cannot be released without approvals.
"stream of behavior" as a first step in analysis The first of three concepts requiring discussion
of it, was found most suitable for our purposes. is Barker's "episode," symbolized here by E.
Figure 8, one of 14 worksheets required to "Although continuous, the stream of behavior
represent a suborbital flight, illustrates the occurs in perceptible units. One of these units is
adapted procedures and format. A verbatim the behavior episode. Episodes are the 'things'
transcript of pilot-ground communications oceu- that people normally see themselves doing [Eating
]. Had been unable to see horizon and stars ] "This wa:

I REALLY
as expected. [ that .0[_
2. A little behind in scheduled activities due ] control,
to extra time looking out "windows." [ I figure,
I was la_
3. Altitude down to about 40_ miles (230,000 ft) I between
had in f:
4.
-- The .05G Q45.
started light came on at O7:48, just as he I
have tim,
up be for_
SF6 .O5G LIGHT ]
G's starl
_, P- STATUS.
MONITORS SEQUENCES, S/C-
SFq HE-ENTRY SEQUENCE ] KEEPS MC INFORMED.
_OLL PROGRAM [ cONTIN_FES ASCS, NORMAL.
-BUILDUP [ CONTINUES HF _HECK.
-BUILDUP [
Q45 P I'M HEADING YOU LOUD AND CLEAR, HF HIGH.
123456789 123456789
cc Back to UHF.
OKAY. THIS IS BLANK 7. 6 : "I

Q46
1234_56789 123456789 ought
CS whel
ii pro!
tates
is) to
I MANUAL
RIES RE-ENTRY
(MP). DAMPING ON
RATE THE
Q47 p AH, G-BUILDUP 3, 6, 9.
12345--6789 123456789 Note parl
G-decline
cc
G-MAX G-DECLINE I Cent ri fu_

out "O.K
Q48 P 0KAY. OKAY.
Center h(
123456789 12345__6789
after th_
CC Coming through loud and clear. P swit chf
Q49 P OKAY.
12345-6789 123456789 (One aria:
MP & swi_
QSO P OKAY. thereaftc
123456789 123456789 reductio,
Q-MAX Q-DECLINE
Q5] P THIS IS SEVEN. OKAY.

123456789 123456789
F,eURE 8.--Excerpt from a "stream of behavior" representation of a suborbital flight.

APPENDIX A 189
an apple and playing cricket are given as exam- those not charted are sufficiently conspicuous in
ples.].., the action of an episode is directed the communiques.
toward a single goal." The 10 "probable states"--These are probable
Barker and his co-workers prefer a format in discomfort, apprehension, joy, urgency, relief,
which the central portion contains short para- anger, conflict, sorrow, activation, and effective-
graphs by on-the-spot observers, describing in ness; this is an extension of an earlier list (ref.
everyday language the behavior of a single 36). The first eight are distinguished (Appendix B)
individual, usually in a social setting. An excep- on the situational, stimulus, or input side. Out-
tion is the study by Soskin and John (ref. 37) who put or response data are examined for com-
present the recorded conversations of a man and patibility information on situational and back-
wife encountering river traffic in a rowboat; ground variables and in estimation of intensity.
episodes are occasionally marked, but incidentally Defining situational changes are specified for
to their main objectives. each of these probable states in terms of the as-
Episodes are indexed in order of occurrence, sociated stimulus operations of reinforcement
with new ones added as innermost brackets. For theory (refs. 28 and 41 to 46).
convenience, monitored sequences, reporting the Large, gaping holes wouhl be evident in any
status of spacecraft and pilot and maintaining analysis restricted to observables. They must be
communications, are grouped in a single block filled by assertions having lower warranty--by
with the bracket extending the length of the inferences from available data drawn by the
phase. analyst with varying degrees of subjective prob-
Episodes correspond to a chain or series of ability. Some of these statements will be couched
responses under the control of a single but possibly in terms of events and processes presumed to
complex set of reinforcement contingencies, or to have occurred within the pilot, and warranted
a response sequence under the control of a more by a knowledge of the culture and specific
"situation-and-goal set" in Woodworth's (ref. backgrounds known to have shaped the observ-
38) phenomenologically oriented terminology. able and unobservable behaviors of the particular
Blocks and brackets on the left side of figure 8 individual, than by data from the immediate
denote "situational factors" (SF). Schoggen (ref. situation.
39) calls these "environmental force units" There is no escaping inferences of this type.
(EFU's) (reflecting the Gestalt-Lewin background The only question is the choice of a conceptual
of this work) and defines them to include the framework for their admission, of which there
physical and social process or events that "spur, are many. By one type of approach, sometimes
guide or restrain behavior." In Mercury flights, called "phenomenological" (ref. 47), situational
SF's were mostly physical--lift-off, build up of and response evidence would bc examined along
gravity, BECO, etc.--with social interactions with other data and used for reconstruction of
confined to communications channels. the pilot's "psychological world" in terms of how
The mere occurrence of an environmental he perceived, thought, and felt at the moment.
change is not sufficient. A false landing-bag- One "reinforcement theory" approach, also
deployment signal, for example, lacks the full widcly held, assmnes that these inner processes,
credentials of an SF (or EFU) until the pilot insofar as they are behavioral, consist of stimulus-
evidences some "awareness" of it by his responses. response sequences and interactions that function
Barker, in the Gestalt tradition, requires evidence in accordance with the same laws as do the ob-
that the change has "penetrated" the individual's servable sequences on which the principles were
"psychological world"; this parallels the be- established in the first place. These processes are
haviorist view that a change is a "stinmlus" only known as "mediational' since they provide a
if followed by a detectable response or an alterna- presumptive link between observable stimuli and
tion in response parameters (ref. 40). The small responses. The greater complexity of human be-
subset of stimuli, singled out for charting, is havior is attributed largely to the great number
determined by the framework and objectives of and variety of intervening and interacting
the analysis. In the format of figure 8 many of stimulus-response chains evolving from man's
190 BIOMEDICAL RESEARCH AND COMPYrTER APPLICATION
capacity to manipulate words and other surrogates between. Many joyful, angry, apprehensive, "con-
for things-in-themselves. flictful," and energetic responses can be identified
Apart from the comparative theoretical merits without situational evidence, particularly if the
of these contending views, the reinforcement- individuals involved are known. When the se-
theory approach has practical advantages in this quence of situations in which the responses are
application. Most important is the fact that it embedded is also known, judgments of com-
reinforces patterns of analysis which make the patibility are relatively easy. Detection of in-
most of observables before proceeding with reluc- compatible responses is the most important part
tance and caution to inferential assertions of of it.
lower warranty when they are needed to fill the Raters make intensity estimates also, mapping
gaps in making a prediction as to the probable situational and response evidence onto integer
state of the pilot. Moreover, pilots and practically scales with due allowances for the residual and
oriented personnel are more likely to accept cumulative effects of situations earlier in the se-
analyses emphasizing observable situational and quence and for pilot-selection and training
response data than one that presumes to know procedures.
how the pilot is "seeing things," "thinking," and The rating sheet has a list of communique
"feeling." numbers in one column, followed by 10 columns
This use of reinforcement theory in combination each headed by a probable-state name, with each
with Barker's "stream of behavior" concepts and row containing the digit set 123456789. The rater
techniques, with their roots in the "phenomenolog- is directed to consider previously identified sets
ical" approach, is methodologically defensible. of communiques bracketed by a "situational
The first, is concerned exclusively with the search factor" on the left or an "episode" on the right,
for relationships or "functions" exhibited by alI to examine the evidence, and to chart first the
behavior; the latter begins by observing sequences changes in the states most clearly indicated by
of behavioral events as they occur, in a given the evidence. Other states are then adjusted to
spatial, temporal, and social setting, and orders allow for the passage of time and for interactions
the situational and response data in a manner with the new situation.
that can facilitate a "functional analysis."
Assessing the evidence--Three types of judg-
ALTERNATIVES TO COMPLETE
ments are required. The first are "stimulus class-
PROBABLE-STATE ANALYSIS
membership judgments." The rater examines the
situational changes and stimuli occurring im- Probable-state analysis (PSA) is tedious and
mediately before and during a communique, as time-consuming. Gathering the evidence and plac-
portrayed in the "stream of behavior" format ing it in correspondence with communiques is
(fig. 8), to detect equivalencies between these the most demanding task. But failure in this is a
and the defining situations of the respective prob- crucial failure.
able states. The complexity of these judgments The approach recommended in the main text
at times is evident in Appendix B. reserves complete PSA for the final stage of the
"Response compatibility judgments" are the validation process, at which time it is performed
second type. The fact that idiosyncratic patterns with the Group as the unit of analysis. The ex-
of expressive and coping behaviors are common ploratory search for valid speech measures is
among human adults is not surprising in view of conducted with simpler criterion techniques ap-
differences in genetics and conditioning histories. plied to simulator and flight situations. Co-
It is for this reason that probable states are most ordinated laboratory studies are employed to
readily distinguished on the basis of situational tease out relations confounded in the complexity
evidence (and that generalizing of speech-state of operational sequences. This section identifies
relations from one pilot to another without sup- some of these simpler methods.
porting data has been avoided). On the other Restricting the number of ratings--An attempt
hand, if response evidence were of no value, it to substitute two state dimensions--Activation
would be ignored. The truth lies somewhere in Level and Hedonic Tone--disclosed that weight-
APPENDIX A 191
ing and averaging of the intense pleasant and un- classified to represent the clearest contrasts among
pleasant aspects of launch, reentry, and some the several states. Speech-parameter differences
other situations equated them to periods more are examined for patterns.
nearly average in both respects. Splitting of this Other--The intensive examination (main text)
bipolar scale into two unipolar ones--probable or relations between excursions of a time plot
pleasantness and probable unpleasantness--seems and concurrent events typifies what may be called
desirable. "coincidence methods"--a term reflecting legit-
Restricting the information considered to that in imate interests in co-occurrences and covariation,
communiques--After some familiarization with the as well as the inferential hazards of these tech-
terminology of space flight and with the circum- niques. The mood ratings and task-performance
stances of a particular flight, one can rate the criteria used to assess responses of Mercury astro-
state of a pilot over the time span of a Group or nauts to stress (ref. 49) offer interesting possibili-
communique, using only the transcriptions. How- ties. The comparison of flight phases within the
ever, Davitz (ref. 48) found that speaker-state framework of Time-Line Analysis is illustrated
judgments from spoken sentences are different in the main text. Correlations with heart rate and
(presumably better) than those from text alone. other physiological measures is another worth-
Restricting the analysis to periods of flight clearly while direction, particularly if speech measures
evidencing state changes oJ interest--Emotionally are computed on a Group basis and if heart rate
toned incidents can be identified, rated, and is averaged over the time span of each Group.
APPENDIXB
PROBABLE-STATE DEFINITIONS
Interruption of the text is avoided by group which places a premium on achievement and
citation of most references: refs. 28, 41 to 46, success. One pilot stated forthrightly, "I thought
50, and 51. The following are the easiest sources this was a chance for immortality"--the ultimate
of the concepts and nomenclature of reinforce- in social approval.
ment theory: refs. 43, 52, and 53.
Probable discomfort--In one defining situation Central in their personalities are strong needs
the individual is exposed to aversive stimulation for achievement and mastery. These are men who
such as environmental extremes, noxious sub- must do, and do well, and this quality is obvious
early in their histories .... By the same token, their
stances, aches, pains, irritations, or other physio-
potential for disappointment is high. With strong
logical conditions. In the other he is deprived needs for achievement and strong feelings of
of such things as food, water, activity, rest, or individual responsibility, failure can be disturbing,
sex. These two classes of "primary reinforcing for it cannot readily be minimized or rationalized ....
stimuli" are termed negative (S -R) and positive As committed men, disappointments are keenly felt;
as ego-strong men, hope is sustained and disappoint-
(S+R), respectively.
ment leads to renewed effort. Similarly, effects are
The degree of discomfort or "stress" resulting readily aroused and strongly felt, but there is good
from the specific exposure or deprivation is control of potentially disabling effects on behavior
estimated from the situational change itself, from (ref. 49).
the pilot's responses to it, and from general
information on the probable time course of its A flight surgeon said of one pilot that he was
effects. Definitely it is not assumed that all more concerned about efficient performance than
deprivations and all noxious stimulations produce about external dangers. For these reasons any-
the same "state," but only that the degree of thing that threatens or appears to threaten the
discomfort resulting from the set of those oc- full and competent achievement of both major
curring at any moment can be estimated, and and minor flight goals is interpreted as situa-
that these estimates will correlate with speech- tional evidence of some degree of "probable
parameter changes. apprehension" even if the only consequence is
Probable apprehension--The defining situation disapproval (S -a') (or withholding of approval)
is the occurrence of any stimulus which, through by associates. Situations and performances likely
prior conditioning, has become a sign of punish- to generate self-disapprovals are also in this
ing "consequences that the individual cannot category. Negative self-evaluations become fairly
avoid or terminate. The signal is known as a reliable indicators of probable social disap-
"negative, secondary reinforcing stimulus" (S-r). provals, and, by virtue of their occurrence over a
Two types of threatening situations are dis- wide range of situations and consequences, they
tinguished when the ratings are made. The first become "generalized" and functions as S-o*'s in
concerns physical dangem and discomforts such their own right.
as the onset of a serious thruster problem, a The common element that gives this category
landing-bag-deployment signal of unknown origin, its theoretical integrity is the S -_ signal, S -a_
a premature 0.05-gravity light, or a substantial being a subset. This reduction enhances confidence
rise in capsule _emperature after failure of all that ratings, based on situational and response
efforts to control it. The second type contains evidence of such apparent diversity, may reflect a
threats of negative social evaluation or self- unitary state.
evaluation. Astronauts have been selected from Probable joy--The defining situation is the
a culture, by a process, and for a task, each of occurrence of any stimulus that, through prior
193
conditioning, has become a sign (S +') of probable, both S-" and S +" components of a special class.
positive reinforcement--of the occurrence of The S -_ components signal the highly probable
attainment of an S÷R; or of the cessation, termina- onset, at the end of a relatively fixed period, of
tion, or avoidance of an S -R or an antecedent S -r. either a severely punishing consequence or an
The two types of situational evidence considered immutable sequence culminating in one. The
parallel exactly those of "probable apprehension." S +_ components signal concurrently the prob-
The first type includes an acceleration profile ability that the required performances can be
duplicating the expected one experienced in discovered if necessary and executed in time for
simulation; onset of a green retroattitude light, successful avoidance of the undesirable conse-
the appearance of the chutes, and a message from quences. Siegel and Wolf (ref. 54) suggest the
a recovery plane that "I have you visually." ratio of the time required for completion of
These S+r's are obviously related in the opera- essential tasks, to the time available, as the
tional system to primary positive reinforcements. appropriate index.
The acceleration example illustrates the fact that When the S -_ elements justify a high "probable
some strong S-ms can function sinmltaneously as urgency" rating, they also increase "probable
S+"s. Other examples include the temporary en- apprehension" and "probable activation" and
velopment of the capsule in flames as booster could be absorbed by them. As a separate item,
engines drop away, which would be terrifying the "probable emergency" charts one important
had not photographic evidence established it as class of tension-producing situations.
verification (S +_) of proper staging. The "fireball" Probable relief--These situations are identified
during reentry ordinarily signifies (S +*) proper by substantial reduction or termination of S -R or
functioning of the ablative heat shield. The S -r S -_ stimulation. The combination of "relief and
components of these situations probably still joy," encountered for example as launch and
evoke some "probable apprehension." Allowance reentry sequences near completion, is due to the
must be made for these complexities in rating of fact that the reductions and terminations occa-
all three states discussed so far. sioning the relief also signify (S +') successful
In situations of the second type the S +', regard- completion of the phase or mission. When the
less of origin, signals positive social evaluations pilot has contributed to the situational change,
(S+g"s). When tile pilot has the requisite feed- self-approval or social approval enters the picture,
back, the mere occurrence of a successful or as in the example of suit-temperature control.
outstanding performance may be taken as pre- Probable anger--The interruption of a stimulus-
sumptive evidence of the occurrence of favorable response sequence that has customarily resulted
self-evaluative responses, and of the joy occa- in positive reinforcement (either procurement of
sioned by these as predictors of social approvals-- positive reinforcements or escape from, termina-
if one assumes that the attainment is a signifi- tion of, or reduction of negative reinforcements)
cant one. is the defining situationaI input. The breaking of
In one flight the achievement of control over sequences leading to social approval or self-
suit temperature by following a carefully de- approval is a special case.
veloped plan is an illustration. As evidence The holds and delays of countdown are an
(S+_'s) of successful control accumulated, there example, with one pilot providing substantiating
was justifiable "pride and joy" in the accomplish- evidence on the response side. During one of
ment, followed by approvals from the ground many long holds he is reported to have gone on
(S+_r's). Ill this case, primary reinforcements the ia_ercom with "his only terse remark of the
(S+R's) also resulted (here without social media- day": "I'm cooler than you are. Why don't you
tion) from control over suit temperature for the fix your little problem and light this candle:"
rest of the flight. (The joyful exhilaration on Pressure suits and other physical constraints
attainment of orbit, and its intensification by qualify by interfering _ith normal responses for
other factors, is discussed in the main text.) relieving discomforts and procuring positive re-
Probable urgency--This is an estimate of task inforcements. Requesting an explanation and not
pressures. The high-urgency situation includes getting it in the customary manner is another.
APPENDIX B 195
Probable conflict--Conflict situations present "ego" or "self." To the extent that the situation
stimuli requiring two or more incompatible reis highly aversive, anticipation of "probable
sponses or response sequences. Conflict between relief" is an additional component.
the attractiveness of the view (or of other phe- Probable aclivation--This is the "arousal syn-
nomena) and the programmed task requirements drome." It is estimated mainly from response
is one example. Attempts at self-control also (output) information, with heart rate serving as
evidence conflict. They are responses to situations an available index for most pilots. Other evidence
containing two types of reinforcement con- considered includes the range of stimuli to which
tingencies: one evoking the angry, anxious, responses are being made at a given time; the
dozing, or other responses to be controlled; the latency, rate, and adequacy of the responses;
other, the controlling responses. Successful con- introspective or retrospective reports of elation,
trolling responses generate stimulus:response drowsiness, or not being "on top of things"; and
sequences that are incompatible with the to-bc- the text of communiques. Situational evidence is
controlled responses; for example, one pilot examined to ensure compatibility with response
calmed himself during a hold by the self-com- evidence and to support intensity estimates; this
mand, "You're building up too fast. Slow down. procedure reverses that of the eight situationally
Relax." defined states.
Probable sorrow--The situational change in- Probable effectiveness--This is an overall assess-
volves irretrievable loss of the S+r's that have in ment of the pilot's ability to process information
the past indicated a high degree of probability and perform required ta_sks; it is based on all the
that one's responses would be positively rein- evidence, whether included in the preceding
forced. When the opportunity to achieve a highly ratings or not. Job-performance information is
desirable flight objective is irretrievably lost, the given most weight : for example, switch-throwing
states that follow are likely to include "probable errors; double-authority errors in attitude-control;
anger" when the scheduled sequence is inter- time taken to detect an error; looking for
rupted, some "probable apprehension" if a hazard constellation where it would have been if the
or social disapproval is indicated, and finally flight had not been delayed; and pursuing lower-
"probable sorrow" (regret) when the permanence priority tasks to the detriment of higher-priority
of the loss is recognized. ones. Introspective reports also are relevant.
The sadness that alternates with anxiety when "Probable confusion," deleted from the pre-
death is inevitable, and death's approach is slow proposal suggestions, is absorbed by the informa-
enough to permit such alternations, also fits the tion-processing aspects of "probable effective-
definition. The S+_'s about to be lost involve ness," and by the situational aspects of "probable
stimuli generated in one's own body, including urgency." Boredom and other states not spe-
those stimulus-response sequences called the cifically listed are similarly encompassed.
REFERENCES
1. BARKER, R. G., ANn WRIGHT, H. F.: Midwest and Contexts. Quart. J. Exp. Psychol., vol. 10, 1958,
Its Children. Harper and Row Publishers, Inc., 1955. pp. 96-106.
2. BARKER, R. G.; WRmHT, It. F.; BARKmt, L. S.; AND 17. GoLDMAN-EISLER, F.: Significance of Changes in
SCHOQGEN, M'.: Specimen Records of American and the Rate of Articulation. Langu. Speech, vol. 4,
English Children. University of Kansas Publications, 1961, pp. 171-174.
1961. 18. FONARGY, I.; AND ]_{AGDICS, K.: Speed of Utterance
3. BAI_KEa, R. G., E_).: The Stream of Behavior. in Phs.s¢_ of Different Lengths. Langu. Speech, vol.
Appleton-Century-Crofts, Inc., 1963. 3, i960, pp. 179-192.
4. S_AaKWEXTHER, J. A.: Measurable Dimensions of 19. STARKWEATHER, J. A.: A Speech Rate Meter for
Vocal Behavior. Presented at meeting of American Vocal Behavior Analysis. J. Exp. Ann. Behav.,
Association for the Advancement of Science (Denver, vol. 3, no. 2, 1960, pp. lll-ll4.
Colo.), Dec. 1961. 20. NEWCOMn, W. B.: The Perceptual Basis of Syllable
5. POPOV, V. A.; SIMONOV, P. V.; TISnCn_NKO, A. G.; Bmmdary. Presented at Summer Meeting, Linguistic
FROLOV, M. V.; AND KHACHATUR'YANTS, L. S.: Society of America (Austin, Texas), 1963. (Available
Analysis of International Characteristics of Speech from Electronics Division, General Dynamics Corp.,
as a Criterion of the Emotional State of Man Rochester, N.Y.)
under Conditions of Space Flight. Zh. Vysshei 21. HARaTS, J. S.: Voice Communication: Effect of Stress
Nervnoi Deyatel'nosti im. I. P. Pavlova, vol. 16, Conditions on Speaker Intelligibility. Contract
no. 6, 1966, pp. 974-983. (Translation TT 67-30554 N6ori-104, Purdue University, I948.
available from CFSTI.) 22. FAmBXNKS, G.; ANI) PRONOVOST, W. ; An Experimental
6. VERZEANO, .-/_I.; AND FINESINGER, J. E. : An Automatic Study of the Pitch Characteristics of the Voice
Analyzer for the Study of Speech in Interaction and during the Expression of Emotion. Speech Monogr.,
in Free Association. Science, vol. 109, 1949, pp. vol. 6, no. 1, 1939, pp. 87-104.
45-46. 23. IIva_rxR, G. L.: Some Relations Between Emotions
7. VERZEANO, _I.: Time-Patterns of Speech in Normal and the Prosodic Parameters of Speech. SCRL
Subjects. J. Speech IIearing Disorders, vol. 15, Monograph No. 1, 1967.
1950, pp. 197-201. 24. Co(_rI_x_, W. G.; ANt) Cox, G. M.: Experimental
8. VERZEANO, M.: Time-Patterns of Speech in Normal Designs. John Wiley and Sons, Inc., 1950.
Subjects. J. Speech IIearing Disorders, vol. I6, 1951, 25. ESTES, W.; AND SKINNER, B. F.: Some Quantitative
pp. 346-350. Properties of Anxiety. J. Exp. Psychol., vol. 29,
9. BEADY, P. T.: A Technique for Investigating On-Off 1941, pp 390-400.
Patterns of Speech. Bell System Tech. J., vol. 44, 26. NEWCOMB, W. B.: Some Tempo Manifestations of
1965, pp. 1-22. the Terminals of English. University Microfilms,
10. GOLO_IAN-EISL_R, F.: The Distribution of Pause Inc., 1960.
Durations in Speech. Langu. Speech., vol. 4, 1961, 27. SCHROEDEa, _l _. R.: Vocoders: Analysis and Synthesis
pp. 232-239. of Speech. Proc. IEEE, vol. 54, no. 5, 1966, pp.
ll. GOLDMAN-EIsLER, F.: Speech-Breathlng Activity--A 720-734.
Measure of Tension and Effect During Interviews. 28. SKINNER, B. F.: Verbal Behavior. Appleton-Century-
Brit. J. Psychol., vol. 46, 1955, pp. 53-63. Crofts, Inc., 1957.
12. GoLn_tAN-EIsLEr_, F.: Speech-Breathing Activity and 29. GUEE_, D. M.: Consistency of Auditory Detection
Content in Psychiatric Interviews. Brit. J. Med. Judgments. Psychol. Rev., vol. 71, 1964, pp. 392-407.
Psychol., vol. 29, 1956, pp. 35-48. 30. SENDERS, V. L.; AND SOWARD, A. : Analysis of Response
13. GOLDMAN-EISLER, F.: The Determinants of Rate of Frequencies in the Setting of a Psychophysical
Speech Output and Their Mutual Relations. J. Experiment. Amer. J. Psychol., vol. 65, 1952,
Psychosomat. Res., vol. 1, 1956, pp. 137-143. pp. 358-374.
14. ItENI)ERSON, A.; GOLDMAN-EIsI_ER, F.; AND SKARBEK, 31. SI3RGENT, L. V.: The Speech-Analysis and Speaker-
A.: Temporal Patterns of Cognitive Activity and State Study Capability of the Comm_'.nications
Breath Control in Speech. Langu. Speech, vol. 8, Sciences Laboratory. Electronics Division, General
1965, pp. 236-242. Dynamics Corp., Rochester, N.Y., Feb. 1965.
15. I_IENDERSON_ A.; GOLDMAN-EISLER, F.; AND SKARBEK, 32. SUR(;ENT, L. V.: Acoustical Analysis of A_tronaut
A.: Sequential Temporal Patterns in Spontaneous Speech. Vol. I (main text). Vol. II Appendices.
Speech. Langu. Speech, vol. 9, 1966, pp. 207-216. Electronics Division, General Dynamics Corp.,
16. GoLD_As-Eisler, F.: The Predictability of Words in Rochester, N.Y., Feb. 1967 (NASw-1252).
197
33. HEFFERLIh _E, R. F.: An Experimental Study of Psychology, Appleton-Century-Crofts, Inc., 1950.
Avoidance. Genet. Psychol. _Ionogr., vol. 42, 44. MOWRER, O. II.: Learning Theory and Behavior.
John Wiley and Sonsp Inc., 1960.
1950, pp. 231-334.
34. SIDMAN, M.: Two Temporal Parameters of the Main- 45. MOWRER, O. H.: Learning Theory and the Symbolic
tenance of Avoidance Behavior by the White Rat. Processes. John Wiley and Sons, Inc., 1960.
J. Comp. Physiol. Psychol., vol. 46, 1953, pp. 46. SKINNER, B. F.: Science and IIuman Behavior. The
253-261. MacMillan Co., 1953.
35. SURGENT, L. V.: Some Directions in Speaker-State 47. WAIN, T. W., EO.: Behaviorism and Phenomenology.
Research. Electronics Division, General Dynamics Contrasting Bases for Modern Psychology, Uni-
Corp., Rochester, N.Y., Aug. 1968 (Report No. 68- versity of Chicago Press, 1964.
126). 48. DAVITZ, J. R., ED.: The Communication of Emotional
36. SURGENT, L. V.: The Applicability of Acoustical Meaning. McGraw-tiill Book Co., 1964.
Speech Analysis Techniques to the Monitoring of 49. KORCHIN, S. J.; ANn RUFF, G. E. : Personality Char-
acteristics of the Mercury Astronauts. The Threat
Spacecraft Crews: A Preprosal Submittal. Electronics
Division, General Dynamics Corp., 1960, Rochester, of Impending Disaster. Contributions to the Psy-
N.Y., June 1964. chology of Stress, G. It. Grosser; H. Wechsler; and
37. SOSKIN, W. F.; AND JO_N, "V. P.: The Study of Spon- lXl. Greenblatt, eds., M.I.T. Press, 1964.
taneous Talk. The Stream of Behavior, R. G. 50. I_IcCLELLAND, D. C.; ATKINSON, J. W.; CLARK,
Barker, ed., Appleton-Century-Crofts, Inc., 1963. R. A.; AND LOWELL, E. L. : The Achievement Motive.
38. WOOl)WORTH, R. S.: Situation- and Goal-Set. Amer. Appleton-Century-Crofts, Inc., 1953.
J. Psychol., vol. 50, 1937, pp. 130-140. 51. I'_{cCLELLAND, D. C. : Notes for a Revised Theory of
39. SCHOGGEN, P. : Environmental Forces in the Everyday Motivation. Studies in Motivation, D. C. McClelland,
Lives of Children. The Stream of Behavior, R. G. ed., Appleton-Century-Crofts, Inc., 1955.
Barker, ed., Appleton-Century-Crofts, Inc., 1963. 52. ItOLLAND, J. G.; AND SKINNER, B. F.: The Analysis
40. SKINNER, B. F. : The Behavior of Organisms. Appleton- of Behavior. McGraw-Hill Book Co., 1961.
Century-Crofts, Inc., 1938. 53. ULRICH, R.; STACHNIK, T.; AND _,IABRY, J.: Control
41. HUNT, W. A.: Recent Development._ in the Field of of Human Behavior. Scott, Foresman and Co., 1966.
Emotion. Psychol. Bull., vol. 38, 1941, pp. 249-276. 54. SIEOEL, A. T.; AND WOLF, J. J. : Computer Simulation
42. HUNT, J. lX,IcV.; COLE, ._,{. L.; AND REIS, E. E. S.: of Man-Machine Syster_. Unusual Environments
Situational Cues Distinguishing Anger, Fear and and Ituman Behavior, N. M. Burns; R. M.
Sorrow. Amer. J. Psychol., vol. 71, 1958, pp. 136-151. Chambers; and E. Hendler, eds., The Free Press
43. KELLER, f. S.; AND SCHOENFELD, W. N.: Principles of of Glencoe, 1963.
"_ U.S. GOVERNMENT PRINTING OFFICE: 1971 0-412-732

Biomedical Research in Space Flight

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Biomedical Research in Space Flight

Uploaded by

Copyright:

Available Formats

NASA SP-5078

BIOMEDICAL RESEARCH AND

For sale by the Superintendent of Documents,

The National Aeronautics and Space Administration has established a Tech-

RONALD J. PHILIPS, Director

M[ajor General, USAF, AIC

Cha pter Page

Utilization of Computers in MedlcaJ Diagnosis ................................. 1

Computer Applications in the Behavioral Sclences--S. B. Sells ........................ 19

New Research Capability ................................................. 19

Medical Data from Flight in Space: Objectives and Methods of AnalysTs--]e_Terson F.

Preparation of Time-Line Medical Data ...................................... 37

Automated Medlcai-Monitorlng Aids for Support of Operational Flight--Robert L. Jones

General MedlcaI-Monltoring Systems........................................ 51

MEDATA: A Medlcal-lnformatlon-Management System--Tare M. Minckler and Caroline L

Medical Information ...................................................... 61

Development of a Computer Program for Automated Recovery of Laboratory Data--

Preprocessor and Processor................................................ 76

Continuous Monitoring and Interpretation of Electrocardiograms from Space--Cesar A.

On-Line Analysis of Limited-Time Signals ..................................... 91

Display Methods ......................................................... 93

Predictive Values ........................................................ 104

Summary and Conclusions ................................................. 116

Period Analysis of an Electroencephalogram from an Orbiting Command Pilot--Nell R.

Burch, Ronald G. Dossett, Abbie L. Vorderman, and Bayd K. Lester ..................... 1 t7

Results ................................................................. 121

Discussion .............................................................. 138

Acknowledgment ........................................................ 139

Anafysls of E|ectroencephafographlc Data from Orbff by Three Di_'erent Computer-Oriented

Zero-Crossings Technique .................................................. 141

Parametric Analysis of the Space-Fllght EEG's ................................. 142

Digital Smoothing and Peak-Countlng Technique ............................... 146

Analyses by Digital Smoothing and Peak-Countlng Technique ..................... 148

Welbull Analysis of Flight EEG's ............................................ 157

Acknowledgment ........................................................ 162

References ............................................................. 162

10 Tracklng of an Astronaut's State by Physical Measurements of Speech--Louis V. Surgent 163

Summary ............................................................... 163

Measurement of Speech Parameters ......................................... 164

Recommendations ........................................................ 182

Acknowledgments ........................................................ 185

Probable-State Anafysls ........................................... 187

Alternatives to Complete Probable-State Analysis ...................... 190

References .................................................................. 197

MEDICAL USAGE OF COMPUTER SCIENCE

Since the early 1940's, the development of UTILIZATION OF COMPUTERS IN MEDICAL

responses to sudden change in environmental from a simulated system as hypotheses to be

A good many scientific libraries are computer- PATIENTS' MEDICAL RECORDS

TABLE 1.--Equipmenl at One Typical Laboratory REFERENCES

COMPUTER APPLICATIONS IN THE

IF (5lean H-Mean L) 17, 27, 37

Compufer-Confrolled Expedmenfs matter of time before this "unbelievable" setup

fall with improvements in manufacture (as in technology is illustrated by the experimental

MEDICAL DATA FROM FLIGHT IN SPACE:

Jefferson F. Lindsey, Jr.

PREPARATION OF TIME-LINE MEDICAL Description of Blacks of Data Selected

TABLE 3.--Types of Activities

Criticalness (C) Difficulty (D) Duly (T) Procedure (P)

2C: Medium-critical 2D : Medium-difficult 2T: Revised (combination of 2P: Passive

3C: Noncritical 3D : Easy 3T: Old (previously performed 3P: Concurrent

_R 3 L5MIN OATA 2MIN OA]A

MA-8 5MIN OATA 15MINOAT/

2 _900 TC 1359 OOT( 32 29 30TO! ,33 53138 TC 3_ 08 3B_(

TOTALs, 1133 TABS