Abstracts of Papers

CONFERENCE ON THE
EXPERIMENTAL ANALYSIS
OF BEHAVIOR
Sponsored by the Society for the Experimental Analysis of Behavior
at
The Meetings of the American Psychological Association,
Cincinnati, Ohio, September 9, 1959

Program Committee
J. J. BOREN
W. H. MORSE
T. VERHAVE, Chairman
 DISCRIMINATION REVERSAL AS A BASE LINE FOR
STUDYING THE ACQUISITION OF BEHAVIOR'
GEORGE A. HEISE
HOPFMANN-LA ROCHE, INC.

Discrimination reversals are being investigated in monkeys and pigeons in order to ob-
tain a reproducible, although changing, behavioral base line suitable for the study of drugs
and neurophysiological correlates as well as for the study of acquisition itself.
Two monkeys receive twice-weekly sessions in a three-key box with a chain discrimina-
tion (reversal) schedule of Fl 2 minutes. When the white light is on over the middle key, the
monkey presses this key on Fl 2 minutes to set up a red-green discrimination. A single
response on the outside key, under the correct color (shifted randomly from side to side)
produces a "banana pellet"; pressing the outside key under either color reinvokes FI 2 min-
utes. A session consists of three discriminations, e.g., red-green-red, each learned to a cri-
terion of 10 consecutive correct trials, following which the opposite color becomes correct.
In each session, the color of the initial discrimination is opposite to that of the final dis-
crimination of the preceding session.
The middle key, fixed-interval component insures that each monkey is positioned midway
between the red and green stimuli and is ready to respond to them. It also monitors the be-
havior continuously and reveals stimulant or depressant action of drugs, "frustration"
effects, etc.
The discrimination procedure for the pigeons is essentially the same as that for the mon-
keys, except that the red-green discrimination trials are given every 30 seconds in a two-key
box with no FI contingency.
Each monkey has learned 60 discriminations, and each pigeon, 280. All four subjects
usually perform steadily and consistently on their respective schedules. Neither monkeys nor
pigeons have so far shown any consistent session-by-session downward trend in errors per
discrimination after the first few sessions, nor have they shown any one-trial discrimination
reversals. Only one monkey and one pigeon show an upward trend within the session; both
of these make significantly fewer errors on the first discrimination of the session than on
the second.
Although they have received fewer than one-fourth as many discrimination problems, the
monkeys' discrimination performances are clearly superior to those of the pigeons. During
their final 30 discriminations, the two monkeys averaged, respectively, 12.6 (am = 1.5) and
14.2 (am = 1.4) errors per discrimination; but the two pigeons averaged, respectively,
18.9 (am = 2.4) and 25.1 (am = 1.6) errors per discrimination on their final 90 discrim-
inations.

'Edward Boff collected the monkey data, and Carl Scheckel, the pigeon data.

251
 BRAIN WA VES AND OPERANT BEHA VIOR
JAMES ANLIKER
HARVARD MEDICAL SCHOOL

In this study, brain waves and operant responses were measured simultaneously. A new
method for analyzing brain waves has been developed which yields data in the form of
cumulative records, thereby facilitating comparison with the usual operant-response data.
In brief, surface potentials from transoccipital electrodes are amplified, filtered, sum-
mated, and cumulated. The slope of the resulting cumulative curve may be interpreted as
an index of brain-wave activity in the frequencies passed by the filter. In the present paper,
only alpha-frequency data are reported.
The behavioral situation is extremely simple. A human lies quietly on a bed in an air-
conditioned, sound-attenuated chamber. His eyes are closed and there is complete darkness.
His preferred hand is in contact with a modified telegraph key. He is instructed to pace
his responses at approximately 3-second intervals by counting silently as follows: "one, one-
thousand, two, one-thousand, three, one-thousand," and repeat. The key is to be depressed
momentarily on the count "three." If the subject fails to respond for 2 minutes, he is
"alerted" by means of an auditory signal. Bursts (0.10 second) of white noise are delivered
at 1-second intervals, commencing at threshold intensity and increasing by 1-decibel steps
with each successive burst. A single key-response terminates the signal immediately. The re-
cording attenuator which controls the intensity of the signals then automatically returns
itself to the threshold setting. The auditory signal is represented on the cumulative records
as a hash-mark, the pen remaining in the deflected position as long as the auditory signal
is in operation. Separate cumulative recorders collect the operant responses and the brain-
wave activity. Operant responses are superimposed on the brain-wave record as hash-marks;
comparison of the two records permits discrimination of key responses and auditory sig-
nals, though both appear on the brain-wave record as hash-marks.
In the cumulative record for the operant response, it will be seen that there is usually
an orderly, progressive increase in inter-response times preceding the onset of the auditory
signal, while in the brain-wave record one may observe a corresponding fall in alpha activ-
ity. A return to lever pressing is associated with an increase in alpha activity.
Records of this type have been obtained from a large number of normal subjects. Ex-
amples of lever-pressing responses "without awareness" were presented. Subjects' reports
of their subjective responses were briefly discussed.

252
 AN ADJUSTING FIXED-RATIO SCHEDULE FOR A TIME-OUT FROM
AN A VOIDANCE SCHEDULE AS REINFORCEMENT
THOM VERHAVE
THE LILLY RESEARCH LABORATORIES

Four rats were first trained to press a modified Switchcraft lever to avoid electric shock.
According to Sidman's terminology, SS was 3 seconds and RS was 30 seconds. The shock
was provided by a constant-current generator passing half-wave 60-cycle current at 1.5 mil-
liamperes and 400 volts. After avoidance responding was well-established, a second lever
was made available to the animal. Each press on this lever produced a 10-minute time out
from the avoidance schedule. The TO was indicated by a 433-cycle-per-second tone. No
separate phase of discrimination training between TO and avoidance periods was given be-
fore introduction of the TO lever. All animals simultaneously (1) came to produce the TO
on a regular basis, and (2) learned to discriminate between tone and no-tone conditions as
indicated by the absence of responding on either lever during TO. After TO-producing be-
havior was well-established, the animals were trained to produce TO on a fixed-ratio
schedule. One animal, for example, was moved from crf to FR 5, to FR 10, and so on to
FR 35. Responding on the TO lever was then put on an adjusting-ratio schedule. After
each TO the fixed ratio for the next TO was increased by one. However, 5 minutes passed
since the end of a TO without the animal obtaining another TO, the FR schedule was auto-
matically reduced by one step for each successive 5-minute period of no responding. The
main datum recorded was the length of the ratio schedule maintained by the animal. This
was done by monitoring the position of the two-way stepper which controlled the number
of responses to be emitted for each particular TO. All animals "selected" their own particu-
lar length of the ratio schedule during the first session on the adjusting schedule. For
example, with one rat the fixed ratio slowly increased from crf to FR 34, after which the
animal occasionally would not produce a TO within the 5-minute adjusting period. The
schedule would consequently oscillate near FR 34, varying between FR 33 to FR 35 during
successive TO's. These levels were maintained during many successive sessions.
The performance on this adjusting schedule demonstrates that:
1) Frequency and duration of FR pausing is also a function of the length of the ratio
schedule in the case of escape behavior.
2) Fixed-ratio "straining" can occur at low FR values.
3) Very fine control exists between FR schedule length and performance. An increase
from FR 35 to FR 36 may produce enough of an increase in straining during the next FR
run to bring the schedule down to FR 35 again.
The use of the adjusting schedule serves to:
1) Eliminate the variability in straining usually observed with ordinary FR schedules.
2) Give a quantitative index of straining.
3) Amplify the functional relationship existing between the variables on which its suc-
cessful use is based.

253
 VARIABLE-INTERVAL ESCAPE FROM A SERIES OF SHOCK PULSES
JAMES A. DINSMOOR
INDIANA UNIVERSITY

White rats were placed on a shock grid and provided with a bar to press. The primary
aversive stimulus was a series of 0.05-second pulses of 2-milliampere direct current. A given
series of shocks was terminated when the animal first pressed the bar following the lapse of
a predetermined but variable interval of time (mean duration, 30 seconds). The factors
varied were: a) the mean interval between successive pulses of shock (7.5, 15, 30, 60, or
120 seconds); b) the length of time the shock series remained off when terminated by the
animal (15, 30, 60, 120, or 240 seconds); c) provision or omission of light and tone sig-
nals. When provided, one of these signals indicated the continuation of the series of shocks,
while the other indicated that no shocks were being given.
Data have been obtained for three animals, each observed during 10-hour sessions at
most combinations of the foregoing values. At the beginning of a given day's session, little
or no retention is observed; the rate of responding rises from zero, with an acceleration
related chiefly to the frequency with which the shocks are presented, to a terminal value.
The same holds true when signals are provided. A number of shocks evidently must be ad-
ministered to provide a support of some kind for the subsequent behavior.
Terminal rates have been calculated for responding in the presence of the shocks during
the final 7 hours of each session. For the most part, these rates vary directly in magnitude
with the frequency with which the shock is delivered, but with a slope considerably less than
unity, so that more responses are given per shock at the lower frequencies. The length of
time the shocks remain off is important mainly at the briefer durations; here, rates of re-
sponding tend to be depressed, particularly where less than one or more than eight shocks
are given per minute.
The ratio of rates in the presence and absence of the shocks provides a rough measure of
discriminability. When signals are provided, this ratio holds at ten to one or better. With-
out signals, the same ratio may be maintained at the highest frequency of shock. delivery;
but as the frequency of shock declines, the ratio drops and eventually falls below unity. The
lower values are probably an artifact derived from the dependence of time sampling on the
animal's behavior.

254
 INTERACTIONS IN A MULTIPLE SCHEDULE:
REVERSAL OF AN EXTINCTION CURVE
G. S. REYNOLDS
HARVARD UNIVERSITY

Two pigeons were used daily in a 3-hour experimental session consisting of 30 repeti-
tions of the following cycle: 3 minutes of a red key-light followed by 3 minutes of a green
key-light. The procedure that was correlated with each of these exteroceptive stimuli was
either a variable-interval schedule of reinforcement with a mean interval of 3 minutes or
extinction.
Interactions were found under conditions when responding was reinforced in both com-
ponents of the schedule, in neither component, or in just one component.
In one part of the experiment, both members of the multiple schedule were maintained on
extinction for several sessions, and the rate of key pecking declined in each member. When
only one member was changed to a variable-interval schedule of reinforcement, the rate of
responding increased in both members. That is to say, when the multiple schedule
EXT EXT changed to either mult VI ext or mult ext VI, the rate of responding increased
in both members although reinforcements were received in only one member of the multiple
schedule. Three such examples of reversals in extinction curves were obtained for each of
the two birds.

THE BEHAVIORAL EFFECTS OF REPEATED EXPOSURE

TO THREE MIXED EXTINCTION SCHEDULES
ALBERT WEISSMAN
CHAS. PFIZER & CO., INC.

Rats were stabilized on each of three reinforcement schedules: mix FR 24 ext 10 min-
utes, mix FI 2 minutes, ext 10 minutes, and DRL 20 seconds ext 10 minutes. The extinc-
tion component had little effect upon expected responding during the respective reinforce-
ment components, but came to differ markedly from one trial extinction. When the
reinforcement schedule was FR 24, extinction was characterized by initial ratio "priming
runs," followed by a plateau of no responding. As extinction progressed, rapid bursts of
responding arose, differing considerably in frequency among rats. Under Fl 2 minutes, ex-
tinction began with typical fixed-interval priming runs, which ceased abruptly. Mostly short
bursts and long pauses were in evidence throughout the remainder of extinction. When re-
sponses were reinforced by DRL 20 seconds, extinction consisted (again after an appro-
priate priming run) of long pauses; but relatively prolonged response bursts at high rates
(invariably absent during DRL itself) were observed.
A ratio of responses in each 10-minute extinction period to responses per reinforcement
showed that rats under DRL emitted far more extinction responses (7.5-10.0) than did sub-
jects under either FR (2.6-5.3) or FI (2.8-3.2) in proportion to the mean number of re-
sponses per reinforcement during reinforcement components.

255
 THE EFFECTS OF SEVERAL CONDITIONS OF MOTIVATION
ON TWO TYPES OF REINFORCEMENT
WILLIAM HODOS
WALTER REED ARMY INSTITUTE OF RESEARCH

Data are reported from an investigation of the effects of several conditions of motiva-
tion on the rate of lever pressing reinforced by one or the other of two types of events:
intracranial stimulation (ICS) and food.
Female rats were ovariectomized and implanted with chronic electrodes in the septal
region of the brain. Reinforcements were programmed on a multiple schedule consisting of
15 minutes of VI 30 seconds (reinforcement, sweetened condensed milk) followed by 15 min-
utes of VI 15 seconds (reinforcement, ICS) during a 2-hour session.
Data were obtained under four conditions of motivation. (1) In the control condition,
free feeding (food and water) was provided in the home cage, and, in addition, free-
feeding milk was given 60 minutes before the start of the testing session. (2) In the estrus
condition, everything was the same as in the control condition except that estrus was in-
duced by the injection of steroid hormones. (3) In the food-deprived condition, the animals
were provided with free-feeding water in the home cage, but no food or milk was given for
48 hours prior to testing. (4) In the food-deprived plus estrus condition, everything was the
same as in the food-deprived condition except that estrus had been induced in addition.
The data supports the following conclusions. (1) Food deprivation increases the rate of
lever pressing reinforced by food and by ICS. (2) Estrus does not affect the rate of lever
pressing reinforced by food or by ICS.

PERFORMANCE DECREMENT DURING A 7-DAY AVOIDANCE SESSION

JOHN J. BOREN
MERCK INSTITUTE FOR THEAPEUTIC RESEARCH

Two rhesus monkeys, previously trained to lever press on a multiple schedule of avoid-
ance and fixed-interval reinforcement, were subjects in an experiment to determine the
decrement in performance generated by a prolonged avoidance session. The monkeys were
kept predominantly on a Sidman avoidance procedure for 7 days, during which the animals
were forced to respond at least once every 20 seconds. The avoidance procedure was inter-
rupted only for a feeding period every 2 days and for occasional daily tests of fixed-
interval behavior. Changes in performance occurred slowly and gradually, but from the
fifth day on, rather clear effects could be seen. The rate of avoidance responding decreased
to about one-half the original value, and the number of shocks increased to more than ten
times the original amount. The fixed-interval rate tended to decrease, although, in some
cases, a late increase was observed. The decrement in fixed-interval behavior was rather
small compared with the decrement in avoidance behavior.

256
 SOME STUDIES ON THE DIFFERENTIATION OF
HUMA N TIMING BEHA VIOR '
JULIAN I. TABER, LLOYD E. HOMME,
and
ATTILA P. CSANYI
UNIVERSITY OF PITTSBURGH

S s were required to hold back an ergograph trigger for a preselected interval of time.
On each trial, the S was differentially reinforced or corrected with one of three lights.
These lights were labeled "Over," "Hit," and "Short." S s responded freely except for a
3.00-second SA period following each response. During this time, one of the correction
lights, produced by the preceding response, remained on. No external SD'S were available to
S upon which a temporal discrimination could have been based. Each response duration
was recorded to the nearest 0.01 second.
In a study of the rate of learning, three naive college students were run on alternate
50 trial blocks of "training" and "estimation." During "estimation" blocks the reinforcing
lights were turned off, and S was instructed to estimate the training interval as accurately as
possible. The two experimental sessions run on each S began and ended with a 50-trial
block of estimation trials. S s were told that the interval was 1 second and they were
reinforced during training for R-durations of 1.00 i 0.10 second. R's of less than 0.90 sec-
ond (T,) caused the "Short" light to come on, while R's longer than 1.10 seconds (T2)
resulted in the "Over" light. The "Hit" range (AT) was thus 0.20 second long. Accuracy
improved rapidly with training, and showed little loss between the two experimental ses-
sions. During the second experimental session, accuracy on estimation trials became as high
as that shown on reinforced trials. During each session, there was a slight general upward
"drift" in R-durations, more marked on estimation than on training trials.
In order to evaluate the effects of T, on relative accuracy, a second experiment was be-
gun, with two well-practiced laboratory workers as S s. Each day for 6 days the S s prac-
ticed on a new interval giving 300 to 400 reinforced responses. Six intervals between
0.50 and 5.00 seconds were used and given in a random order for each S. For each Tp, AT
was a constant fraction (one-tenth) of T,. The relative variability of terminal responses as
a function of T, yielded curves with slight negative acceleration which were much closer to
"ideal" Weber functions than were the curves for the relatively unpracticed early training
trials.
Other experiments have yielded data consistent with the findings of Stelter, Barnes, and
Homme (1959)2 with respect to the effects of force requirement changes on the timing be-
havior of rats. Testing humans with a light trigger after training with a heavy one led to
increases in R-durations, while light-to-heavy changes led to decreases. It was found that
increasing or decreasing the loudness of a response-produced auditory signal led to similar
changes. Bilateral transfer of the skill has also been found.
The data gathered by this technique suggested a number of theoretical and practical con-
siderations of relevance to the study of DRL schedules, discrimination learning, response
differentiation, and psychophysics.
'This research was supported in part by grants from the National Science Foundation and from the National
Institute for Mental Health, U. S. Public Health Service.
2Stelter, C. J., Barnes, H. W., and Homme, L. E. An investigation of precise timing behavior in the rat. Amer.
Psychol., 1959, 14, 421 (Abstract).
257
 BEHAVIOR OUTPUT UNDER CHAINED FIXED-RATIO REQUIREMENTS
IN A 24-HOUR EXPERIMENTAL SPACE'
JACK D. FINDLEY
UNIVERSITY OF MARYLAND

For approximately 6 months, one male Mangebey monkey worked in a 24-hour experi-
mental space (Fig. 1) where it obtained food pellets, water, and 1-minute illuminations of
an overhead light. All food and water were given in this space except for a quarter slice of
orange supplied by the experimenter each day. Food, water, and light reinforcements were
obtained via a three-member chain permitting selection of the terminal reinforcement in the
second member. A white light above a push button defined the first member, three colored
lights above three momentary switches defined the second member, and a colored light
above another push button defined the last member. The sequence of contingencies was as
follows: In the presence of the white light, completing an FR on RI extinguished that
light and illuminated the colored lights of the second member; closure of a switch under
one of the lights extinguished these lights and illuminated the light above R3 with a color
identical with the one previously selected in the second member; completing an FR on R3
delivered a terminal reinforcement appropriate to the color, and illuminated the white
light above RI.
One result obtained with this monkey is outlined in Fig. 2, where the mean number of
reinforcements obtained in 24 hours is expressed as a function of the FR requirements on
RI and R3. As the FR was systematically increased from 5 to 80, the number of food,
water, and light reinforcements declined, approaching an asymptote near 80 responses;
beyond an FR of 80, the food and water intake again declined sharply. The total response
output increased almost linearly from FR 5 to FR 80, then declined. The amount of food
and water intake at FR 320 was approximately 20% of that obtained at FR 5. Performance
was characteristic FR, with pauses primarily following terminal reinforcement.

SMOKE Y 120*

IC 80 FOOD

APPARATUS
9
0

Figure 1.
0~~~~4

Ii.
o~~~~~
4 0

5 40 80
0

16 0
FR ON BOTH R I a R3
Figure 2.
0

3 20

'This work was supported in part by Grant No. M 1604 from the U. S. Public Health to the University of
Maryland.
258
 RESPONSE LATENCIES AS A MEASURE OF THE INTERACTION
OF COMPONENTS ON A MULTIPLE FIXED-RATIO SCHEDULE
C. SCHUSTER
UNIVERSITY OF MARYLAND

Two White Carneau pigeons, maintained at 75% of normal body weight, were trained to
peck a lucite key to gain access to grain for a 4-second period. Following a single session of
continuous reinforcement, the animals were transferred to a multiple fixed-ratio schedule,
consisting of two alternated fixed ratios each separately associated with a different discrim-
inative stimulus: green and white illumination of the lucite key. A total of 70 reinforcements
was earned in each daily session, 35 under each of the discriminative stimuli. A 30-second
blackout period followed each reinforcement.
Response latencies were measured in 0.01-second intervals from the onset of a discrimi-
native stimulus to the first response emitted, and were separately recorded for each of the
70 ratios.
The base-line ratio requirement was 20 responses under both discriminative stimuli
(FR 20-20). Following stabilization of the latency measures at the base-line requirements,
the size of the ratio under one of the discriminative stimuli was systematically varied over
a period of weeks, while the ratio size under the alternated discriminative stimulus was
held constant. Three different combinations of ratio requirements were studied: FR 10-20,
FR 40-20, and FR 80-20. Animals were returned to the base-line schedule (FR 20-20) be-
tween each of the schedule changes. The animals were run for a minimum period of 7 days,
or until the behavior had stabilized under each of the experimental conditions.
The results obtained demonstrate that the latencies shown on a multiple fixed-ratio
schedule are a function of both the ratio requirement of a particular component and the
ratio requirements of the other components of the schedule. In general, as the ratio require-
ment of one component is increased while the other is held constant, latencies shown prior
to the latter ratio decrease. For example, when the schedule was changed from an FR 20-20
to an FR 40-20, the latencies shown prior to the FR 20 component decreased significantly.
In addition, as the ratio requirement of one component is decreased while the other is held
constant, latencies shown prior to the latter ratio increase significantly. This was found
when the schedule was changed from an FR 20-20 to an FR 10-20, where the latencies prior
to the FR 20 component show a marked increase. These interactions, which may be viewed
as a "contrast effect," were observed to be stable for periods as long as 30 consecutive
sessions.
A further study followed this general procedure, but with a base-line ratio requirement
of 80 responses. The ratio combinations investigated were: FR 80-20, 8040, 80-10, and
80-160. Changes in the latency shown prior to the ratio maintained at a requirement of
80 responses corroborated the findings at the lower base-line values. However, the contrast
effect at these higher values did not remain for periods longer than 10-14 days. Rather,
the latencies on the FR 80 component showed a gradual return to the values obtained under
the base-line condition.

259
 SOME NOTES ON PUNISHMENT AND A VOIDANCE'
NATHAN H. AZRIN
ANNA STATE HOSPITAL

Pigeons were subjects in a program of research on aversive stimuli. A method of de-

livering aversive electric shock through implanted electrodes has provided precise control
of the duration and intensity of electric shock the subjects actually received. When used as a
punishment, this method has made it possible to observe orderly behavioral relationships
that were confused and variable when studied with rats on electrified grids.

MINUTES

As an example, Fig. 1 shows the effects of moderate punishment (3 milliamperes for

0.02 second) delivered following each response of a pigeon under a schedule of food rein-
forcement of VI 1 minute. It can be seen that the rate is markedly reduced during the
first day; but recovery occurs within each day. After 10 more days, recovery of the responses
is almost complete, but responding is still reduced at the initial introduction to the pun-
ishment condition. In contrast, extremely intense punishment (10 milliamperes for 0.02 sec-
ond) allows little or no recovery. The rate under punishment is about one-tenth of that
without punishment, with no signs of recovery after 20 days of continued exposure.
This method of delivering electric shock has also been used to condition experimentally
naive pigeons to respond (a 10-gram peck) under the Sidman-type avoidance procedure.
This experiment was one of the first successful attempts to condition an avoidance response
that was not already part of the respondent pattern to shock, and which has an operant
level that is essentially zero. The main difficulty in conditioning such a response under
shock-avoidance is that a sufficiently intense shock produces respondents that reduce the
likelihood of the response. By continuously varying the shock intensity (via a potentiom-
eter) in accordance with the subject's behavior at each moment, this and other difficulties
were overcome, and the new avoidance response could be shaped.
'This investigation was supported by a grant from the Psychiatric Training and Research Fund of the Illinois
Department of Public Welfare.
260
 AVOIDANCE OF TIME OUT FROM FIXED-INTERVAL REINFORCEMENT
FRANCIS MECHNER and RONALD RAY
SCHERING CORPORATION

Six male Albino rats were trained on a procedure where the first response following the
end of a 15-second fixed time interval was reinforced with 0.05 cubic centimeter of water.
This 15-second fixed interval was always initiated by the first response occurring after a
reinforcement. However, the fixed interval was only allowed to go to completion if the
animal never paused longer than 3 seconds once it had initiated the interval. Then, a
2-minute time out was imposed, accompanied by a blackout. Four of the six animals readily
stabilized on this procedure. The conditions were then changed so that reinforcement came
automatically at the end of the 15-second interval, but, again, only if the animal had suc-
cessfully staved off the time out. Following this, the avoidance condition was also with-
drawn, so that the only remaining requirement for reinforcement was the response that
initiates the fixed interval.
Alongside each procedural diagram shown in the figure are the inter-response time distri-
butions obtained under that condition for each of the four animals. The average response
rates in the first and second halves of the 15-second interval are also shown. As the figures
indicate, the inter-response times shift from short to long as the reinforcement and avoid-
ance contingencies are successively withdrawn. The response rates drop correspondingly,
the greater drop occurring in the second half of the interval. A further finding for the
second of the three conditions was that all six animals showed a rate peak approximately
3 seconds after the beginning of the 15-second interval, in the vicinity of the point where
the 2-minute time out could first occur.

P2 R4
.4 u

iR7 I .0. R-4-

sp-.

LLK
43.2- R2 RR 6 2 L41 M. a
.se s

lxi
I/

L 32 U,
hi.

.S.- . l
,.
O-.3. w

2
I RT

lb 2
I

3 z
3
*CSI

i.I

IKUATERALP
0*t t t ; " s 2 s d d INTERVAL INTERVAL
INTEQ-RPS,pO3E TIME (SEC.)

261
 OPERANT BEHA VIOR IN PSYCHIATRIC PA TIENTS'
DONALD H. BULLOCK
INSTITUTE OF THE PENNSYLVANIA HOSPITAL

Psychoses, particularly the schizophrenias, are characterized clinically by a disturbance

in reality contact. Therefore, when compared with nonpsychotics and nonpatients, psy-
chotic patients may be expected to be less amenable to the control ordinarily exerted over
operant behavior by various reinforcement contingencies. This "hypothesis" has received
preliminary experimental support (e.g., Lindsley).
The typical operant procedure (repeated sessions until performances stabilize) is difficult
with acute patients: they are hospitalized for shorter periods and exhibit greater clinical
variability. One approach to the problem is to develop briefer operant tests.
Hospitalized acute and chronic psychotic and nonpsychotic patients and nonpatients
(student nurses, university students) were tested. The subject, alone in a cubicle, sat facing
a panel containing two knobs separated by a reward tray. Knob-pulling responses were rein-
forced with a type of money: poker chips, which patients exchanged for Snack Bar credit
slips; and cash for nonpatients (Bullock & Brunt, Psychol. Rec., Sept. 1959).
Subjects were given a discrimination learning problem consisting of 20 discriminated
fixed intervals. Each DFI "trial" consisted of a 26-second fixed interval during which the
left-knob lamp was lighted and neither knob produced chips. At the end of the 26 seconds,
the left light went out and the right knob lamp was lighted. The next right-knob response
produced a chip and started the next interval. The 20 DFIs were preceded first by con-
tinuous reinforcement, then by Fl 26 seconds (right knob lighted and reinforced).
Three criteria of discriminative control were applied: (a) decreased total responding dur-
ing left-knob lighted periods (extinction); (b) predominance of left-knob responding in the
presence of the lighted left knob ("stimulus-response generalization"), and (c) inter-rein-
forcement times approaching the 26-second minimum. Criterion values were based upon the
nonpatient data. The proportions of subjects attaining at least two of the three criterion
values were: 88% of 17 nonpatients, 92% of 12 nonpsychotic patients, 64% of 14 acute psy-
chotic patients (ill for 1 year or less), 40% of 15 chronic psychotic patients (ill for more
than a year).
'This research was supported by U. S. Public Health, National Institute of Mental Health Small Grant
NI- 1464, matching grants from Lilly and Wyeth Laboratories, the Theano Foundation (Mrs. E. Paul DuPont), and
the Pennsylvania Hospital. Manly Y. Brunt, Jr., M.D., collaborated in the research.

262
 DRUG EFFECTS ON POSITIVELY AND NEGATIVELY MOTIVATED HABITS
STUDIED SIMULTANEOUSLY IN RA TS
OAKLEY S. RAY and LARRY STEIN
VA RESEARCH LABORATORIES IN NEUROPSYC'iIATRY
and
WYETH INSTITUTE- FOR MEDICAL RESEARCH

This study was designed to investigate the relative effect of drugs on behavior under
positive and under aversive control. Thus, with tones as discriminative stimuli, positively
and negatively motivated habits were simultaneously trained in rats.
The animals were trained in a double-lever cage, with an electrifiable grid floor and a
dipper mechanism that delivered a 0. l-cubic centimeter milk reinforcement. In separate ses-
sions during preliminary training, the animals learned to press one lever for milk and the
other lever to avoid shock. Tones of one frequency indicated when the milk reward could
be obtained; and tones of a different frequency, when shocks were to be avoided. In both
cases the correct response had to occur within 10 seconds after the onset of the tone in
order to deliver the food or avoid the shock.
After performance of both responses had stabilized in the separate sessions, training
proper was initiated, in which both positive and negative trials were combined in the same
session. After several weeks of such training, performance of both habits had stabilized at
about 95% correct responses, and the drug series was begun.
Two procedures were used. In one, drugs were administered prior to the session; the
daily sessions were typically composed of 100 trials, and the shocks remained in during drug
testing. In a later modification, the S was run for 3-400 trials in a session and drugs were
administered after a stable base line was established for a given day; for drug testing, the
S was run without shock reinforcement.
Preliminary data on reserpine, chlorpromazine, meprobamate, LSD-25, amphetamine,
and phenyl ethyl amine are reported.
In low doses on both procedures, reserpine (0.2 milligram per kilogram per day), chlor-
promazine (0.75-2.0 milligrams per kilogram), and LSD-25 (60-125 gamma) showed similar
effects and decreased the number of avoidance responses without notably changing the
latency or number of correct approach responses. In one S the major effect of chlorpro-
mazine and reserpine was to decrease considerably the number of approach responses while
only moderately influencing the latency and number of avoidance responses.
Meprobamate (80 milligrams per kilogram) eliminated avoidance responding in the sec-
ond procedure for over an hour, and caused an increase in the latency of the approach
response but never blocked it completely.
Racemic amphetamine (1.0 milligram per kilogram), on the second procedure, eliminated
the food-motivated response for an hour, without modifying the latency of the avoidance
response. At three dose levels under the second procedure, phenyl ethyl amine had no selec-
tive effect on the two response systems.

-263
 AN EXPERIMENTAL ANALYSIS OF CERTAIN EMOTIONS
B. F. SKINNER
HARVARD UNIVERSITY

A hungry magazine-trained pigeon (A) is placed in a limited space with a nonhungry

pigeon (B). A is reinforced with food for "attacking" B. At first, A merely gets close
enough to B so that defensive pecking occurs, but eventually A develops full-scale "aggres-
sion." In general, B fights more successfully because it need not force the fight, but A de-
velops various successful leads. As "cold" fighting gives way to attack in A's behavior, all
the "ethological" signs of aggressive emotion emerge: A struts about the cage, coos in a
characteristic fashion, and its feathers, particularly on its neck, become erect. This can
hardly be simply a conditioned emotional respondent because it does not occur in B, the
aggressee.
A stimulus in the presence of which fighting is not reinforced becomes an effective SA.
Emotional and operant patterns of aggression appear under one stimulus, peaceful behavior
under another.
Avoidance behavior can be set up in B if pecking a key is followed by (a) the operation
of the magazine (which "appeases" A) or (b) SA.
Modifications of the experiment should make it possible to study a wide range of emo-
tional behaviors in the laboratory. Pharmacological implications are obvious.

264
 THE EFFECTS OF A TIME-OUT PERIOD DURING
FIXED-RATIO REINFORCEMEINT
ROBERT CLARK
WALTER REED ARMY INSTITUTE OF RESEARCH

A 3-minute time-out (TO) period occurred after varying numbers of fixed ratios. The
number of reinforcements preceding each TO period was varied between 2 and 8.
The S s were four male albino Wistar rats, maintained at 80% body weight and trained to
press a lever. The reinforcement was 6 seconds of access to sweetened milk. Following dip-
per training and a period of crf, each animal was placed on FR 11 for 75 days. The cage
light and a faint clicking sound were present during each daily session, ending after the
70th reinforcement. During the TO, the cage light and clicker were switched off. The FR TO
condition was in effect for 33 days.
The FR schedule generated high response rates and short pauses after reinforcement.
Cumulative-response records showed the typical "burst" and "break" pattern. The TO pro-
duced a marked and almost immediate decrease in response rate. The rate decrease was
more pronounced for Rat 1. Rats 2 and 4 also showed increases in the PAR time. Ferster
and Skinner, using pigeons as subjects, have reported increases in rate and decreases in the
PAR under comparable conditions.
In all cases, the transitions to the lower rates occurred within the first few sessions fol-
lowing the introduction of the TO. The behavior was quite stable thereafter.
The cumulative records indicated that the typical FR behavior pattern was maintained
during SA. No "warm-up" effect was noted within any block of ratio runs for any animal.
Also, there were no progressive changes in rate during the daily sessions. One rat, No. 4,
after approximately 20 days of running, showed progressive decreases in rate as the TO was
approached; but this phenomenon disappeared within a few days.
On the 109th day of running, all animals were returned to FR 11 for 28 days. Complete
recoverability of data was obtained within the first two or three sessions.
The data indicate that the quantitative effects of the FR schedule studied in isolation are
quite different from those of the multiple FR TO schedule.

265
 RESPONSE DIFFERENTIATION DURING A SIMPLE DISCRIMINATION
A. H. BLOCK and J. M. NOTTERMAN
PRINCETON UNIVERSITY

In this exploratory study, apparatus was used which allowed for a measure of response
differentiation (i.e., level of force magnitude) during a simple light-dark discrimination.
This study is an attempt to overcome the limitations of the conventional measures (e.g.,
rate, latency), which have failed to provide an adequate response measure in the presence of
discriminative stimulus.
The SD_SA disparity in rate was similar to that typical of other findings; namely, SD
was considerably higher than St. The force data, however, reveal that force decreases dur-
ing SD and increases in magnitude during SI, indicating general similarity to prior force
data gathered under conditions of regular reinforcement and subsequent extinction.

THE EFFECT OF FORCE-PROPORTIONAL REINFORCEMENT

UPON RESPONSE DIFFERENTIA TION
G. A. CICALA and J. M. NOTTERMAN
PRINCETON UNIVERSITY

The present study is a preliminary attempt to obtain response differentiation on the basis
of quantitative reinforcement proportional to the amount of force that the animal applies
to the bar. The study was also designed to examine the premise that increasing food depriva-
tion subsequent to the establishment of a stable conditioned response produces significant
changes in the intensive aspects of the response.
In general, two statements can be made about the results of the present study.
1. Reinforcement proportional to the force with which the animal depresses the bar
tends to increase the mean peak force of response.
2. Major increases in deprivation, at least of the order used in this study (46.5 hours),
serve to decrease both the force and frequency of a conditioned operant response.

266
 EFFECTS OF D-AMPHETAMINE, MEPROBAMATE, PHENOBARBITAL,
MEPHENESIN, OR CHLORPROMAZINE
ON DRL AND FR SCHEDULES OF REINFORCEMENT WITH RATS
ROGER T. KELLEHER and LEONARD COOK
SMITH, KLINE & FRENCH LABORATORIES

One group of food-deprived rats was trained on a modified DRL schedule of food rein-
forcement. On this schedule, each response was reinforced that occurred between 18 and
21 seconds after the preceding response. Response rates and the relative frequencies of
times between successive responses (inter-response times) were computed. In control ses-
sions, the animals responded at stable rates of about 3 responses per minute, and most
inter-response times were between 15 and 21 seconds. Meprobamate (25-100 milligrams per
kilogram, p.o.) or d-amphetamine (0.75-3 milligrams per kilogram, p.o.) produced more
short inter-response times, and the total number of responses was markedly increased.
Phenobarbital (10-80 milligrams per kilogram, p.o.) or mephenesin (100400 milligrams per
kilogram, p.o.) also produced more short inter-response times; however, both drugs de-
creased response rates because of periods in which all responding was depressed. Chlor-
promazine (2.5-5 milligrams per kilogram, p.o.) produced more long inter-response times
even though stable response rates were maintained.
Another group of rats was trained on a 50-response, fixed-ratio schedule of food rein-
forcement (FR 50). In control sessions, these animals responded at stable rates of about
200 responses per minute. Chlorpromazine, meprobamate, mephenesin, or phenobarbital
did not affect behavior on FR 50 at doses that did affect the DRL inter-response time dis-
tribution; however, a wide range of doses of d-amphetamine depressed response rates on
FR 50.

STEREOTYPY OF RESPONSE DURING

CONTINUOUS AND INTERMITTENT REINFORCEMENT
R. J. HERRNSTEIN
HARVARD UNIVERSITY

Pigeons were trained to peck at a narrow rubber strip that extended across almost the
entire width of one wall of the experimental box. The location of pecks in the horizontal
dimension was recorded. There was no differential reinforcement for particular locations of
pecking. The effect of reinforcement (food) for every peck was compared with that of rein-
forcement on a variable-interval schedule. Under both procedures, the responding is stereo-
typed in that it is not uniformly distributed on the strip. The degree of stereotypy is much
more marked under intermittent than under continuous reinforcement.

267
 MULTIPLE SCHEDULING IN CHILDREN
E. R. LONG
UNIVERSITY OF NORTH CAROLINA

Attempts were made to establish multiple-schedule control in 32 preschool children. Two

schedules (Fl's and FR's) were paired with different discriminative stimuli and presented
in an irregular order. Most children were begun on multiples of Fl 1.5 FR 10 or FI 2 FR 10.
A few were begun on Fl 1 and then shifted to one of the multiple schedules after three
sessions on the Fl.
After approximately eight sessions, a series of changes was made for all subjects, inas-
much as none had been brought under multiple control. The first change entailed the use of
different reinforcers with the two schedules. Half of the subjects were given a penny plus a
highly valued trinket when reinforced on FR's and a single, less valued trinket when rein-
forced on FI's. Reinforcing conditions for the remaining subjects were the reverse of these.
Within two to four sessions, recognizable multiple control was produced in 10 subjects.
Responding during the ratios was usually at a high rate, with little initial pausing. A range
of rate patterns was obtained with the intervals. Those intervals following a series of fixed
ratios usually contained long initial pauses. Initial pausing occasionally occurred in intervals
following other intervals, although running through the interval was much more frequent.
In some instances, there was a partial running through followed by a pause.
Very few instances of smooth acceleration of rate were observed. In some cases, there
was an abrupt transition from a pause to a high constant rate. In others, a low constant
rate was maintained throughout the interval. Multiple inflections and runs of high rate,
"ratio-like" responding were also observed. The latter records seemed more characteristic
of mixed than of multiple schedules, and suggested that any existing discrimination was
based on behaviorially generated stimuli rather than the external visual ones.
After approximately 12 sessions, a Lindsley manipulandum was substituted for the en-
cased telegraph key then in use. This produced multiple control in five more subjects and
strengthened the control in six of the ten subjects who had previously shown evidences of
multiple control. The new manipulandum reduced rates for both ratios and intervals; the
interval rate, however, was reduced much more. Much of the ratio-like grain and many of
the multiple inflections and runs previously seen in intervals disappeared, indicating per-
haps stronger external stimulus control.
Finally, three additional subjects were brought under some degree of multiple control
when the experimenter gave each child approximately 25 trinkets at the beginning of his
session. As a result, the FT rates were lower after four sessions with this procedure.
Fourteen subjects failed to show any evidence of multiple control in spite of these manip-
ulations as well as a number of adjustments in the size of the schedules.

268
 REDUCTION IN RATE OF VOCAL PSYCHOTIC SYMPTOMS BY
DIFFERENTIAL POSITIVE REINFORCEMENT
OGDEN R. LINDSLEY
HARVARD MEDICAL SCHOOL BEHAVIOR RESEARCH LAORATORY

Over the past 6 years of intensive analysis of the free-operant behavior of chronic psy-
chotics, we have found that psychotic episodes are most directly measured functionally. By
separately and simultaneously recording the psychotic symptom on one recorder and a non-
symptomatic response on another recorder, the functional suppression of normal behavior
by psychotic symptoms can- be directly measured. Such a two-channel recording system is
demanded because symptoms can occur without disturbing nonsymptomatic behavior, and
nonsymptomatic behavior can be disturbed by other events than psychotic symptoms.
Vocal symptoms of hallucinatory patients were recorded through a filtered voice key on
one cumulative recorder. Nonsymptomatic manual responses were recorded through a
plunger on another recorder and reinforced with candy and cigarettes. The never-reinforced
vocal symptoms of one chronic schizophrenic occurred at a rate of 500 grammatical stresses
per hour while he operated the plunger at 100 pulis per hour on FR 20. Similar rates had
been maintained for over 130 daily, 1-hour experimental sessions.
When the reinforcement was switched from the nonsymptomatic to the symptomatic be-
havior, the extinguished nonsymptomatic responses decreased from 100 to 10 pulls per hour
over 170 hours. However, the "reinforced" vocal symptoms did not increase in rate, but
also declined from 500 to 10 grammatical stresses per hour. When again reinforced, the
iionsymptomatic responses increased from 10 to 50 pulls per hour over 110 hours. However,
the "extinguished" vocal symptoms increased from 10 to 600 grammatical stresses per hour.
Grammatical stresses of normals increased when positively reinforced, and decreased
when extinguished in this situation. Thus, the "negative" effect of positive reinforcement
on vocal psychotic symptoms appears to be due to their symptomatic nature. The rate of
vocal symptoms was probably increased by induction from the positively reinforced non-
symptomatic behavior and decreased by the distracting effect of the "reinforcing" stimuli.
Vocal psychotic symptoms appear to be under some strong control that resists direct
differential positive reinforcement.

INSTA TEMENT OF LABORATORY STUTTERING IN NORMALLY

FLUENT INDIVIDUALS THROUGH OPERANT PROCEDURES
I. GOLDIAMOND
SOTHERN ILLINOIS UNIVERSITY

Normally fluent S s were required to read from printed pages, with recordings made of
nonfluencies until a stable rate was established. A persistent shock was then introduced.
Its cessation for a limited period was made contingent upon nonfluency. Chronic stuttering
was instated and eliminated as a function of such stimulus change.

269