You are on page 1of 30

Chapter 1

Single-Case Research Methods:


An Overview
Iver H. Iversen
Copyright American Psychological Association. Not for further distribution.

My experiments had indeed gone well. Research Methods Using Single Subjects
I was getting data from a single rat that The single-case research method involves repeated
were more orderly and reproducible than measures of one individual’s behavior before, dur-
the averages of large groups in mazes ing, and often after an experimental, educational,
and discrimination boxes, and a few or therapeutic intervention. Data are collected
principles seemed to be covering a lot of repeatedly over several observation periods or ses-
ground. (Skinner, 1979, p. 114) sions in what is customarily called a time series. The
Replication is the essence of believability. objective is to change a behavior to determine the
(Baer, Wolf, & Risley, 1968, p. 95) variables that control that behavior. When environ-
mental variables have been found that reliably
Single-case research methods refer to a vast collection
change behavior, the method can be used to control
of procedures for conducting behavioral research with
behavior. The investigator, educator, or therapist
individual subjects. Such methods are used in basic
can make the behavior start and stop and change
research and for improving behavioral problems with
frequency or duration at a specific time and place.
educational and therapeutic interventions. Analyses
Therefore, single-case methods provide a tool for a
and interpretations of data collected with research
science of behavior at the level of the individual sub-
methods for individual subjects have developed into
ject. When an educator or therapist can effectively
procedures that are considerably different from those
control the client’s behavior, then methods exist for
used in research with groups of subjects. In this chap-
helping that client acquire new behavior or over-
ter, I provide an overview of designs, analyses, and
come a problem with existing behavior. This ability
interpretations of research results and treatment out-
to help a client by using methods to control the cli-
comes using single-case research methods.
ent’s behavior is exactly what often creates contro-
versy around the use of such methods. The methods
Background and History
for controlling behavior and for helping the client
The case in single-case research methods essentially are the same, but the verbs control and help are not
refers to a unit of analysis for an individual, a few synonymous. To control behavior means to be able
people in a group, or a large group with varying to change behavior reliably, and in this context con-
membership. Single-case research methods should trol has a technical meaning originating from the
be contrasted with single-case studies that ordinarily laboratory. The word control, however, also has
consist of anecdotal narrations of what happened to political and societal meanings related to authorita-
a given person. No treatments or manipulations of tive restrictions of behavior for the individual. The
experimental variables take place in case studies. intended helping function of an applied behavior
I thank Dominik Guess and Wendon Henton for valuable comments on earlier versions of this chapter.

DOI: 10.1037/13937-001
APA Handbook of Behavior Analysis: Vol. 1. Methods and Principles, G. J. Madden (Editor-in-Chief)
3
Copyright © 2013 by the American Psychological Association. All rights reserved.
Iver H. Iversen

science is the opposite—to establish enrichment and number of data points for each individual because of
expansion of the individual’s behavioral repertoire, the repeated observations of that individual’s behav-
not to restrict it. This complex issue has lead to ior. Therefore, it is incorrect and misleading to refer
­several considerations regarding the ethics involved to the single-case research method as an N = 1
in helping a client. For example, Skinner (1978) design, falsely implying that only one data point has
argued that a client who needs help to obtain essen- been collected and that the researcher is trying to
tial goods for survival should have a right to acquire promote a finding based on a single data point.
a behavior that can provide the goods rather than
merely be provided the goods regardless of any Glimpses Into the History of
behavior. Similarly, Van Houten et al. (1988) Single-Case Research Methods
stated that I. P. Pavlov’s (1927) Conditioned Reflexes had a
major influence on B. F. Skinner’s decision to study
individuals who are recipients . . . of
psychology and on his choice of research methodol-
Copyright American Psychological Association. Not for further distribution.

treatment designed to change their


ogy (Catania & Laties, 1999; Iversen, 1992; Skinner,
behavior have the right to a therapeutic
1966). Skinner was impressed with Pavlov’s precise,
environment, services whose overriding
quantitative measures of behavior in one organism
goal is personal welfare, treatment by a
at a time. Thus, Skinner (1956) once wrote, “I had
competent behavior analyst, programs
the clue from Pavlov: Control your conditions and
that teach functional skills, behavioral
you will see order” (p. 223). Pavlov described in
assessment and ongoing evaluation, and
great detail control conditions and recording of indi-
the most effective treatment procedures
vidual drops of saliva at a given time of day for a
available. (p. 381)
­single animal. Figure 1.1 shows an early version of a
Because of the overall success of single-case single-case research method used by Pavlov (1928).
research methods, a plethora of articles, book chap- The dog, Krasavets, had previously been condi-
ters, and entire books devoted to the topic have tioned to a positive tone (positive stimulus, or S+)
appeared over the past 40 years. Recent publications with food powder. Over nine trials, Pavlov alter-
in this vast literature illustrate the wide use of nated the S+ with a negative tone (negative stimu-
single-case research methods: Specific topics are lus, or S−) that had not been conditioned to
basic methodology (Barlow, Nock, & Hersen, 2009; food. The S+ elicited several drops of saliva from
J. M. Johnston & Pennypacker, 2009), educational two glands, and the S− elicited no saliva. The
research (Kennedy, 2005; Sulzer-Azaroff & Mayer,
1991), health sciences (Morgan & Morgan, 2009),
clinical and applied settings (Kazdin, 2011), com-
munity settings (O’Neill, McDonnell, Billingsly, &
Jenson, 2010), and medicine (Janosky, Leininger,
Hoerger, & Libkuman, 2009).
The term single-case research methods is synony-
mous with a variety of related terms, the most com-
mon of which are single-subject designs and N = 1
designs. The last term should be avoided because it
is misleading. N = 1 obviously means the use of Figure 1.1.  Illustration of an early single-case
research method. Detailed data for a single dog
only one subject. However, N has an entirely differ- from Pavlov’s experiments on classical conditioning.
ent meaning in statistics, where it stands for the Positive tone is followed by food; negative tone is not.
number of data points collected, not for the number From Lectures on Conditioned Reflexes: Twenty-Five
of subjects. In customary group research, each Years of Objective Study of the Higher Nervous Activity
(Behaviour) of Animals (Vol. 1, p. 173), by I. P. Pavlov,
­subject contributes exactly one data point. In con- 1928, New York, NY: International Publishers. In the
trast, single-case research methods generate a high public domain.

4
Single-Case Research Methods

a­ lternation was irregular by design so that Pavlov (1865/1957). Nearly 150 years ago, and long before
could examine what happened to the conditioned Loeb, Pavlov, Thorndike, Watson, and Skinner, he
reflex to S+ after several presentations of S−; articulated in clear terms the need for what he called
indeed, the elicitation was reduced, as can be seen comparative experiments with individual, intact ani-
for S+ at 2:32. Thus, at the level of the individual mals. When using several animals for comparison,
dog, Pavlov first demonstrated a baseline of reliable he realized that the errors in his data stemmed from
elicitation of saliva to S+ and then demonstrated variations from animal to animal, and he wrote that
that repeated, successive presentations of S− inhib- “to remove this source of error, I was forced to
ited the flow of saliva on the next presentation make the whole experiment on the same animal . . .
of S+. because in this respect two frogs are not always
Pavlov’s work in Russia and contemporary work comparable” (p. 183). Bernard’s work appeared in
in Europe by Wundt on perception and by Weber Russian, and his work had a major influence on
and Fechner on psychophysics grew from physiol- ­Russian physiologists while Pavlov was a young
Copyright American Psychological Association. Not for further distribution.

ogy, in which the customary method was to investi- researcher. Of particular historical significance, Pav-
gate the effects of independent variables on individual lov’s main professor, Botkin, had been a student of
organisms (Boring, 1929). In Europe, Ebbinghaus Bernard’s (see Paré, 1990), and Pavlov expressed the
(1885/1913) became famous for his memory studies greatest admiration for Bernard’s experimental
using a single subject (himself). In the United States, approaches (Todes, 2002). Indeed, according to
Thorndike’s (1911) research on the law of effect Wood (2004), Pavlov was an apostle of Bernard’s.
with individual cats quickly became well-known As for single-case methods with humans, Watson
(Boakes, 1984). Apparently, Thorndike was ahead of and Rayner (1920) and Jones (1924) demonstrated
Pavlov by a few years. Thus, Pavlov (1928) wrote, conditioning and extinction of fear reactions in
infants (i.e., “little Albert” and “little Peter”). The
Some years after the beginning of the
methods were crude, and the data were qualitative
work with our new method I learned
descriptions with scant operational definitions.
that somewhat similar experiments on
However, multiple training conditions, repetition of
animals had been performed in America,
test conditions, and tests for transfer made these
and indeed not by physiologists but by
studies influential in psychology (Harris, 1979).
psychologists. Thereupon I studied in
Thorndike (1927) reported an impressive laboratory
more detail the American publications,
study with adults that serves as a very early model of
and now I must acknowledge that the
the A-B-A design (see below). Subjects received
honour of having made the first steps
instructions such as “Draw an x-inch line,” where x
along this path belongs to E. L. Thorn-
was 3, 4, 5, or 6. Such instructions were presented
dike. By two or three years his experi-
in random order, and, being blindfolded, the sub-
ments preceded ours. (pp. 39–40)
jects never saw what they drew. The consequence,
Pavlov’s work apparently did not appear in English or effect, of drawing was the experimenter saying
until a paper in Science (Pavlov, 1906) and the either “right” or “wrong.” Each subject had one
translation of Conditioned Reflexes (Pavlov, 1927), early session with no effect delivered, then seven
long after Thorndike had finished his early animal training sessions with an effect, and then one more
research. In addition, Jacque Loeb, educated in the late session without an effect. Figure 1.2 shows
German tradition of physiology of individual ani- average data for all subjects and individual data for
mals (e.g., Loeb, 1900), had a major influence on two subjects. All subjects improved accuracy of line
Pavlov, Thorndike, Watson, and Skinner (Greens- drawing during training but dropped in accuracy
pan & Baars, 2005). when the effect was removed in the late test; 16 of
At the risk of digressing too far into the his- 24 subjects had a gain compared with their scores
tory of psychology, I cannot help mentioning the in the early test (e. g., Subject 29), whereas eight
work of the French physiologist Claude Bernard subjects had no gain and instead a drop (e.g.,

5
Iver H. Iversen

S­ ubject 42) compared with the early test. Thorndike


concluded that the effect was responsible for the
gain in line-drawing accuracy. In the area of motor
learning and control, this experiment is considered
a classic method for demonstration of the effects
of feedback on performance and learning (e.g.,
Schmidt & Lee, 2005).
With the advent of Skinner’s (1938) rigorous
experimental methods featuring operationally
defined measures of behavior and highly controlled
conditions for individual animal subjects, single-
case research methods developed rapidly and laid
the foundation for successful application to humans,
Copyright American Psychological Association. Not for further distribution.

which began around the 1950s. For examples of


early behavior research and therapy using humans
(i.e., 1950–1965), see the collections of articles in
Eysenck (1960) and Ullman and Krasner (1965). In
addition, Wolpe (1958) developed single-case meth-
ods for treatment of phobias in humans.
The methods of behavior analysis for individual
subjects were laid out clearly in influential texts by
Keller and Schoenfeld (1950), Skinner (1953), and
Bijou and Baer (1961). The Journal of the Experimen-
tal Analysis of Behavior was founded in 1958 and
published experiments based on single-case meth-
odology using both human and nonhuman subjects.
When the Journal of Applied Behavior Analysis
appeared in 1968, single-case research methods
were further established as important research tools
that were not for animals only. For a more thorough
history of behavior analysis and single-case research
methods, see Barlow et al. (2009), Kazdin (2011),
and Blampied (1999).

Scientific Method
One important method, namely that of comparing
collected data from at least two different conditions,
Figure 1.2.  Example of a historically early A-B-A is common across different research disciplines.
design for all subjects (top) and for two subjects Reaching this point has not come easy. The history
(middle and bottom). Twenty-four blindfolded human
subjects drew lines. Percentage correct expresses how of science is full of anecdotes that illustrate the
many lines were drawn within criterion length (3, 4, 5 struggle a particular scientist had in convincing con-
or 6 inches). In early and late tests (Thorndike’s terms), temporary scholars of new findings. Boorstin (1983)
no consequence was presented to the subjects, whereas
related how Galileo had to go through painstaking
during training subjects were presented with the experi-
menter’s saying “right” or “wrong” depending on their steps of comparing different conditions to demon-
performance. Data from Thorndike (1927). strate that through his telescope one could in fact
see how things at a great distance looked. In about
1650, Galileo first aimed the telescope at buildings

6
Single-Case Research Methods

and other earthly objects and then made a drawing demonstration by Pascal of a scientific principle as
of what he saw. Then he walked to the location seen well as of a method of testing.
in the telescope and made a drawing of what he saw Although these anecdotes are amusing several
there. He compared such drawings over and over to hundred years later for their extreme and cumber-
demonstrate that the telescope worked. Using this some methods of testing, the practicing scientist of
method, Galileo could prepare his audience for an today still on occasion faces opposition to conclu-
important inductive step. When the telescope was sions from experiments. Thus, behavior analysts
aimed at objects in the sky, the drawing made of sometimes find themselves in a position in which it
what one saw through the telescope would reflect is difficult to convince psychologists with different
the structure of what existed at the faraway location. training that changes in contingencies of reinforce-
As is well-known from the history of science, many ment can bring about large and robust behavior
of Galileo’s contemporaries said that what they saw changes. For example, I once presented to col-
in the telescope was inside the telescope and not far leagues in different areas of psychology some data
Copyright American Psychological Association. Not for further distribution.

away. Yet, his method of repeated comparisons of that demonstrated very reliable and precise stimulus
different viewing conditions was an important step control of behavior in rats. A colleague objected that
in convincing at least some scientists of his time that such mechanistic, on–off control of behavior is not
a telescope was an instrument that could be used to psychology. Others in the audience were suspicious
view distant, unreachable objects. In a review of the of the extremely low variability in the data, which
history of experimental control, Boring (1954) gave was believed to have come from an equipment error
a related example of how Pascal in 1648 tested and not from the method of controlling behavior.
whether a new instrument could measure baromet- The method of comparing two or more series
ric pressure. When a glass tube closed at one end of readings with the same measurement instrument
was filled with mercury and then inverted with the under two or more different conditions is the
open end immersed in a cup of mercury, a vacuum ­hallmark of the scientific method (Boring, 1954).
would form at the closed end of the tube. Pascal had Whether the research design is a between-groups
the idea that the weight of air pressing on the mer- or a within-subject comparison of the effects of a
cury in the open cup influenced the height of the manipulation of a variable on behavior, the shared
column of mercury. Hence, the length of the vac- element is always the comparisons of different con-
uum at the top of the tube should change at a higher ditions using the same measurement (see also Baer,
altitude, where the weight of air was supposedly less 1993).
than on the ground. Pascal sent family members
with two such instruments made of glass tubes,
Designs for Single-Case
cups, and mercury to take the measurements. Read-
Research Methods
ings were first made for both instruments at the foot
of a mountain. Then one person stayed with one A variety of designs have been developed for
instrument and took readings throughout the day. research with animals and humans. Because this
Another person carried the second instrument up chapter is an overview, I can only describe the most
the mountain and took measurements at different common designs. Chapter 5 of this volume covers
locations; when back at the foot of the mountain, designs in considerably more detail. Experimental
measurements were taken again with both instruments. designs are specific to the problem the researcher
Clearly, the essence of the method is the comparison seeks to investigate. However, an experimental
between one condition at the foot of the mountain, design is also like an interface between the
the control condition, and the elevated conditions. researcher and the subject because information
Because the control readings stayed the same passes in both directions. An essential aspect of
throughout the day, whereas the elevated condition single-case research methods is the interaction
readings changed, proof existed that the instrument between subject and researcher, educator, or thera-
worked as intended. This experiment was a real-life pist. Pavlov, Thorndike, and Skinner, as originators

7
Iver H. Iversen

of the single-case method, modified their apparatus A-B Design


and experimental procedures for individual animals The essence of single-case research methods is the
depending on how the animal reacted to the experi- comparison of repeated measures of the same behav-
mental procedures (Skinner, 1956). The scientist ior for the same individual for at least two different
modifies the subject’s behavior by experimental experimental, educational, or treatment conditions.
manipulations that, when reliable and replicable, Figure 1.3 shows a schematic of an A-B design,
constitute the foodstuff of science; equally as using hypothetical data. Phase A is repeated baseline
important for the process of discovery is the sub- recording of behavior across several observation
ject’s influence on the researcher’s behavior. Sidman periods (sessions) before any intervention takes
(1960) described how the researcher needs to “lis- place. The intervention is a change in the individu-
ten” to how the data change in accordance with al’s environment, also across several sessions, usu-
experimental manipulations. To discover laws of ally in the form of a change in how reinforcement is
behavior, the experimenter needs to adjust methods provided contingent on behavior. The data in Phase
Copyright American Psychological Association. Not for further distribution.

depending on how they affect the subject’s behav- A serve as a comparison with behavior changes in
ior. In basic research, the experimental situation is Phase B. However, the comparison is logically from
often a dynamic exchange between researcher and an extension of data from Phase A to Phase B. Data
subject that may lead to novel discoveries and in Phase A are used to forecast or predict what
designs. In applied behavior analysis, as successful would have happened during the next sessions had
experiments and treatments are replicated over the intervention in Phase B not been introduced. If
time, designs tend to become relatively fixed and the data in Phase A are without trends and of low
standardized. The strength of standardization is a variability and are taken across several sessions,
high degree of internal validity and ease in commu- then the experienced researcher or therapist predicts
nicating the design. However, the weakness of stan- and assumes that this behavior will continue at that
dardization is that researchers or therapists may level if no interruptions in the environment occur.
primarily investigate problems that fit the standard Thus, the dotted line in Phase B in Figure 1.3 indi-
designs (Sidman, 1981). Nonetheless, the standard cates the projected trend of the behavior from Phase
designs covered here serve as basic building blocks A to Phase B. The effect on behavior of introducing
in both basic research and application. My main the change in Phase B is thus evaluated against
focus is the underlying logic of each of the covered the backdrop of the projected data from Phase A.
designs. Because forecast behavior cannot be measured, the

Figure 1.3.  A-B design using hypothetical data.

8
Single-Case Research Methods

difference between Phase A and Phase B data consti- the A-B design by itself is uncommon in clinical and
tutes the experimental effect, as indicated by the educational situations. However, the A-B design is
bracket. The validity of the statement that the inter- very common in laboratory demonstrations, espe-
vention in Phase B caused the change in behavior cially with animal subjects, where confounding vari-
from A to B thus depends not only on the change of ables can be controlled experimentally.
behavior in Phase B but also on how good the pre-
diction is that the behavior would have remained A-B-A Design
the same had the intervention not been introduced To demonstrate clear experimental control while
in Phase B. reducing the potential influence of confounding
Ideally, the behavior should not change in Phase variables, the experimenter can supplement the A-B
B unless the intervention took place. The change design with a withdrawal of the change in Phase B
should not occur by chance. Because other changes and a return to the conditions of Phase A, as a back-
in the environment might take place at the same to-baseline control. This design is therefore often
Copyright American Psychological Association. Not for further distribution.

time as the intervention is introduced, the researcher called a withdrawal design. When the baseline mea-
or therapist can never be sure whether the interven- sured in the first Phase A can be recovered in the
tion or some other factor produces the behavior second Phase A, after the intervention has changed
change. Such unintended factors are customarily the behavior in Phase B, then the researcher has
called confounding variables. For example, a child shown control of behavior and demonstrated high
may have an undesirable low rate of smiling in the internal validity. The behavior is changed by the
baseline phase (A). Then when social reinforcement intervention and changed back when the interven-
is explicitly made contingent on smiling in the inter- tion is removed. The researcher therefore has full
vention phase (B), the rate of smiling may increase control over the behavior of the individual subject
above that in the baseline (A). If the sessions stretch or client, which means that the behavior can be
out over several weeks, with one session each day, changed at the researcher’s or therapist’s discretion.
then a host of other changes could happen at the Figure 1.4 shows a cumulative record of lever press-
same time as the intervention. For example, a parent ing for a single rat during baseline, acquisition, and
could return from a trip away from home, the child extinction of lever pressing within one session. The
could recover from the flu, a bully in class may be rapid increase in lever pressing when it produces
away for a few weeks, and so on. Such other factors food contrasts with the very low rate of lever press-
could have made the rate of smiling increase by ing during the baseline phase when lever pressing
themselves or in addition to the effect of providing did not produce food. The rapid decline in lever
social reinforcement for smiling. For additional con- pressing when reinforcement is withdrawn demon-
siderations regarding the role of confounding vari- strates the control reinforcement had over the
ables, see Kazdin (1973). Because of the potential behavior. The experienced investigator comes to
difficulty in controlling such confounding variables, learn that when such clear control is obtained with

Figure 1.4.  A-B-A design: a cumulative record of lever pressing by a single rat. A = baseline with
response-independent reinforcement. B = continuous reinforcement; each lever press produced reinforce-
ment. Second A = extinction. Data from Iversen (2010).

9
Iver H. Iversen

one individual, then it will also be found with other, control over the behavior, they desire the behavior
similar individuals. change in Phase B to not remain intact when the
In clinical and educational applications, the intervention is removed.
A-B-A design is useful to show that an intervention
is effective and that a given behavior change is under A-B-A-B Design
the therapist’s control and not caused by other fac- To resolve some of these difficulties, an additional
tors. However, a critical problem arises with this phase can be added to the A-B-A design in which the
design regarding clinical or educational significance intervention (B) is repeated, thereby forming an A-B-
for the participant and for the caregivers. If the ther- A-B design, often called a reversal-replication design
apist follows an A-B-A design and changes behavior or a double-replication design. Figure 1.5 (top) shows
from a low level in Phase A to a high level in Phase B an example (R. G. Fox, Copeland, Harris, Rieth, &
and then back again to a low level in the second Hall, 1975) in which the number of math problems
Phase A to demonstrate control, in the end there completed by one eighth-grade underachieving stu-
Copyright American Psychological Association. Not for further distribution.

is no gain for the client or for the caregivers. The dent changed when the teacher paid attention to her
educator or therapist can hardly convince the care- when she worked on the assignments. The number
givers that it is a great step forward to know that of problems completed shows a gradual and very
the behavior can now be brought back to the same large change in the treatment condition. Then, when
level as when the participant came in for treatment. the baseline was reinstated, the number dropped to
The A-B-A design has shown gain in control of the baseline levels only to increase again when the treat-
behavior but has produced no clinical or educational ment was reintroduced. Because the number of prob-
gain. Therefore, the A-B-A design is not useful as a lems completed is high in both treatment (B) phases
stand-alone treatment in clinical or educational and low in both baseline (A) phases, the data dem-
situations. onstrate control by the intervention of reinforcing
A concern with the A-B-A design is that behavior completion of math problems in a single individual
that changed in Phase B may not always show a child. This design satisfies both the need to demon-
reversal back to the level seen in the first Phase A strate control of behavior and the need to help the
when the intervention is withdrawn. For example, client because the behavior is at the improved level
for motor skill acquisition such as cycling or walk- when the treatment ends after the last Phase B. This
ing, the acquired behavior may not drop back to the off–on–off–on logic of the A-B-A-B design mimics
baseline level when reinforcement is removed in the daily situations in which people determine whether
second Phase A. The individual has acquired a new something works or not by turning it on and off a
skill in Phase B that may lead to new contingent few times. The A-B-A-B design shows that control of
reinforcers that were not within reach before the behavior can be repeated for the same individual—
skill was acquired. that is, intrasubject replication. This ability of the
When behavior does not return to baseline level researcher to replicate the behavioral change in a
after the intervention is removed, the possibility also single individual provides a tremendous source of
exists that the behavior change was produced by or knowledge about the conditions that control behav-
aided by a confounding variable and not by the ior. Such results are of practical significance for
intervention. The educator and therapist therefore teachers and family or caregivers because as Baer
face a dilemma. They want to help the client acquire et al. (1968) succinctly stated, “Replication is the
a new behavior that is lacking or to suppress an essence of believability” (p. 95).
existing unwanted behavior. They also want, how- The A-B-A-B design is customarily shown as a
ever, to be able to communicate to other therapists time series with several sessions in each phase. The
and educators that they have developed a method to underlying off–on–off–on logic also exists in other
control the behavior of interest. To help the client, procedures in which two conditions alternate sev-
they desire the behavior change in Phase B to remain eral times within a session, as in multiple sched-
intact when the intervention is removed. To show ules (e.g., two or more schedules of reinforcement

10
Single-Case Research Methods
Copyright American Psychological Association. Not for further distribution.

Figure 1.5.  Top: A-B-A-B design with follow-up (post checks). From “A Computerized System for Selecting
Responsive Teaching Studies, Catalogued Along Twenty-Eight Important Dimensions, by R. G. Fox, R. E.
Copeland, J. W. Harris, H. J. Rieth, and R. V. Hall, in E. Ramp and G. Semb (Eds.), Behavior Analysis: Areas of
Research and Application (p. 131), 1975, Englewood Cliffs, NJ: Prentice-Hall, Inc. Copyright 1975 by Prentice-Hall,
Inc. Reprinted with permission. Bottom: Illustration of a within-session repeated AB * N design. Event record
showing onset of discriminative stimulus, first pen; response, second pen; reinforcement, third pen. Data are for
one rat after discrimination training showing perfect discrimination performance. Data from Iversen (2010).

alternate after a few minutes, each under a separate on occasional test trials in which a stimulus is
stimulus). Such A-B-A-B-A-B-A-B . . . or A-B * N altered, such as, for example, in the determination
designs demonstrate powerful, repeated behavior of stimulus generalization. With various modifica-
control. The A-B * N design can also be imple- tions, the repeated A-B changes within a session also
mented at the moment-to-moment level. Figure 1.5, form the basis of methods used in research and edu-
bottom, shows a sample of an event record in which cation, such as the matching-to-sample procedure
a rat promptly presses a lever each time a light turns (see Discrete-Trial Designs section, below).
on and almost never presses it when the light is off.
The A-B * N design can itself serve as a baseline for Multiple-Baseline Designs
other designs. For example, this discrimination pro- A popular expansion of the A-B design is the
cedure can serve as a baseline for assessment of the multiple-baseline design. This design is used when
effects of other factors such as food deprivation, reversing the behavior back to the baseline level is
stimulus factors, drugs, and so forth. The method of not desirable from an educational or therapeutic
repeated A-B * N changes within a session can also perspective. After an undesirable behavior has been
serve as a baseline for comparison with the outcome changed to a desirable level, educators, therapists,

11
Iver H. Iversen

and caregivers are reluctant to bring the behavior


back to its former undesirable level to prove that it
was the intervention that changed the behavior to
the acceptable level. Indeed, it may be considered
unethical to force the removal of a desirable behav-
ior acquired by a person with disability. With the
multiple-baseline design, the possible influence of
confounding variables is not assessed by withdraw-
ing the intervention condition as with the A-B-A and
A-B-A-B designs. Instead, data are compared with
simultaneously running baselines for other behav-
iors, situations, or individuals. The multiple-
baseline design probably had its formal origin in
Copyright American Psychological Association. Not for further distribution.

Baer et al. (1968). Multiple-baseline designs are not


ordinarily used with animal subjects because rever-
sals to baseline and replications are not undesirable
or unethical. Figure 1.6 illustrates with hypothetical
data the underlying logic of multiple-baseline
designs. Two children in the same environment
have the same behavioral deficit. In the top display,
the target behavior is recorded concurrently for
both children for several sessions as a baseline
before intervention. For Peter, the intervention
begins at Session 16 and continues for 10 sessions,
and the target behavior shows a clear increase com-
pared with baseline. For Allen, the behavior is still
recorded throughout the intervention phase for
Peter, but no intervention is scheduled for Allen.
However, Allen also shows a similar large increase
in the target behavior. Faced with such data, one
Figure 1.6.  Confounding variable in a multiple-­
would be forced to conclude either that the behav-
baseline design. Top two graphs: Allen’s baseline behav-
ior change for Peter may not have been the result of ior changes during the intervention for Peter, suggesting
the intervention but of other factors in the environ- the influence of a confounding variable during inter-
ment or that Allen imitated Peter’s new behavior vention. Bottom two graphs: Allen’s behavior does not
change during the intervention for Peter but changes
change. Such data never find their way to publica- when his intervention starts, suggesting control by the
tion because they do not demonstrate a clear effect intervention. Data are hypothetical.
of the intervention for him. The bottom display
shows, also with hypothetical data, the more r­ eplication of the treatment effect. The benefit of this
­customary published data pattern from such a design is that when faced with such data, the thera-
multiple-­baseline design across subjects. The pist or educator can fairly safely conclude that the
­baseline for Allen continues undisturbed, whereas intervention caused the behavior change. An addi-
Peter’s behavior changes during the intervention. tional benefit for the client is that no forced with-
When the same intervention is then also introduced drawal of treatment (return to baseline) occurs,
for Allen, his behavior shows an increase similar which would have ruined any educational or thera-
to that of Peter. The data show that the intended peutic gain for that client.
behavior change occurs only when the intervention Figure 1.7 shows an empirical example of teach-
is introduced. The data also show successful ing words to an 11-year-old boy with autism using a

12
Single-Case Research Methods
Copyright American Psychological Association. Not for further distribution.

Figure 1.7.  Multiple-baseline design across behaviors for one child. From “Increasing
Spontaneous Language in Three Autistic Children,” by J. L. Matson, J. A. Sevin, D.
Fridley, and S. R. Love, 1990, Journal of Applied Behavior Analysis, 23, p. 231. Copyright
1990 by the Society for the Experimental Analysis of Behavior, Inc., Lawrence, KS.
Reprinted with permission.

multiple-baseline design across behaviors (Matson, against ­possible confounding variables that might
Sevin, Fridley, & Love, 1990). Baselines of three occur simultaneously with the introduction of the
words were recorded concurrently. Saying “please” intervention for the first subject; similarly, the base-
was taught first while the baseline was continued line for the third subject guards against possible
for the other two target words. Saying “thank you” confounding variables associated with the introduc-
was then taught while the baseline was continued tion of the intervention for the first and second
for the last target word, which was the last to be subjects.
taught. The data show that teaching one word made The multiple-baseline design offers several levels
saying that word increase in frequency, whereas say- of comparison of effects and no effects of the inter-
ing the other words did not increase in frequency. In vention (Iversen, 2012). Figure 1.8 illustrates, with
addition, the data show that saying a given word hypothetical data, the many comparisons that can be
did not increase in frequency until teaching for that made in multiple-baseline designs, in this case a
particular word started. Thus, the data generated by multiple-baseline design across subjects. For exam-
this design demonstrate clear control over the target ple, with four clients, the baselines of the same
behaviors. In addition, the data show that the con- behavior are recorded concurrently for all clients,
trol was established by the teaching methods and and the intervention is introduced successively
not by other confounding variables. across clients. Each client’s baseline serves as a
The logic of the multiple-baseline designs dic- comparison for the effect of the intervention on that
tates that the baselines run in parallel (concur- client (as indicated with dotted arrows at A, B, C,
rently). For example, with a multiple-baseline and D). The intervention phase for one client is
design across ­subjects, one would not run the first compared with the baseline for the next client
subject in January, the next subject in February, and (­reading from the top of the chart) to determine
the third ­subject in March. The point of the design whether possible confounding variables might pro-
is that the baseline for the second subject guards duce concomitant changes in the behavior of other

13
Iver H. Iversen
Copyright American Psychological Association. Not for further distribution.

Figure 1.8.  Logic of the multiple-baseline across-subjects design.


A, B, C, and D arrows refer to change in behavior during the inter-
vention compared with the baseline for each individual; a, b, c, d, e,
and f arrows refer to the absence of a change for one individual when
the intervention has an effect at the same time for another individual.
Data are hypothetical.

clients when the behavior is made to change for the tion, there are six demonstrations of the absence of
previous client. Thus, the dotted lines (marked a, b, the possible effect of confounding variables because
c, d, e, and f ) indicate the possible assessments of the behavior score does not change for one client
effects of confounding variables. In this hypothetical when the intervention takes effect for another client.
example, there are four demonstrations of an inter- Ideally, to provide maximal internal validity, the
vention effect because the behavior score increases intervention should produce a change in behavior
each time the intervention is introduced. In addi- for the client for whom the intervention takes place

14
Single-Case Research Methods

and should have no effect on the behavior of the a two-way interaction between experimenter and
other clients. Notice that the data for the client in subject and therapist and client, respectively.
the bottom display provides the most information,
with a change when the intervention takes place for Gradual or Sudden Changes in A-B, A-B-A,
that client and three comparisons during the base- A-B-A-B, and Multiple-Baseline Designs
line to introduction of interventions for the other The literature review for this chapter revealed that
clients but without any change on the baseline for two distinctly different patterns of behavior change
the last client. appear to be associated with the A-B, A-B-A, A-B-A-B,
Baer (1975) noted that multiple-baseline designs and multiple-baseline designs. Figure 1.9 exemplifies
offer an opportunity for systematic replication this issue for the A-B-A-B design using hypothetical
(across subjects, responses, or situations) but does data so as not to highlight or critique specific stud-
not offer an opportunity for direct replication for the ies. In the top graph, the behavior change is a grad-
same subject, response, or situation (i.e., there is no ual increase in the B phase and a gradual decrease in
Copyright American Psychological Association. Not for further distribution.

return to baseline or subsequent return to interven- the second A phase with a return to baseline condi-
tion). In essence, the technique is repeated, but tions, and the last B phase also shows a gradual
­control over behavior is not. Thus, an essential com- increase in the behavior, as in the first B phase. In
ponent of functional behavior analysis is lost with the bottom graph, the behavior in the first B phase
the multiple-baseline design, yet the technique is shifts up abruptly as soon as the intervention takes
well suited for applied behavior analysis. Successful,
systematic replications of procedures across behav-
iors, subjects, and situations and across laboratories,
classrooms, and clinics over time offer important
evidence that the designs are indeed responsible for
the behavior changes.
In general, multiple-baseline designs are con-
ducted across behaviors or situations for the same
individual or across individuals (with the same
behavior measured for all individuals). The designs
have also been used for groups of individuals. Multiple-
baseline designs are appealing to educators and
therapists and have become so popular that they are
often presented in textbooks as the golden example
of modern behavior analysis tools. However, these
designs do not quite live up to the original formula-
tion of single-case research methods’ being an inter-
action between the investigator and the subject. The
investigator should be able to change the procedure
as needed depending on how the subject’s behavior
changes as a function of the investigator’s experi-
mental manipulations. To be successful, applied
behavior analysis should guard against becoming a
discipline in which participants are pushed through
a rigid protocol with a predetermined number of Figure 1.9.  Two different patterns of data in A-B-A-B
sessions and fixed set of conditions regardless of designs. The top display shows an example of gradual
how their behavior changes. In a recent interview, acquisition in both B phases and gradual extinction in
the second A phase. The bottom display shows abrupt
Sidman (as cited in Holth, 2010) pointed out that to changes in the B phases and in return to the A phase.
be effective, both basic and applied research requires Data are hypothetical.

15
Iver H. Iversen

place, stays at the same level, and then just as evaluate the frequency and root of the different
abruptly shifts down to the same level as in the first behavior patterns in these designs.
A phase with a similar abrupt shift up in the last B
phase. With animals, data customarily look as illus-
Alternating-Treatments or
trated in the top graph, because phase B is usually
Multielement Designs
contingent reinforcement and the change in Phase B
A variant of the A-B-A-B design is a more random
can be considered behavior acquisition. Similarly,
alternation of two or more components of interven-
when reinforcement is removed in the second A
tion or conditions for research. For example, the
phase, behavior will ordinarily extinguish gradually
effects of three doses of a drug on a behavioral
across sessions. In studies with humans, however,
­baseline of operant conditioning for an individual
both of these data patterns appear in the literature,
may be compared across sessions with the dose
often without comment or clarification. When data
selected randomly for each session. Thus, a possible
show a large, abrupt change in the very first session
Copyright American Psychological Association. Not for further distribution.

sequence of conditions might be B-A-A-C-B-C-C-A-


of intervention and an equally large and abrupt
B-A-C-B, and so on. This design allows for random
change back to baseline in the first session of with-
presentation of each condition and can assess
drawal, the participant either made contact with the
sequential effects in addition to determining the
changed reinforcement contingency immediately or
effect of several levels of an independent variable.
responded immediately to an instruction regarding
The basic design has its origin in Sidman (1960)
the change, such as “From now on you have to do x
under the label multielement manipulation and has
to earn the reinforcer” or “From now on you will no
since been labeled multielement design or alternating
longer earn the reinforcer when you do x.” Thus, the
treatments. This design is somewhat similar to the
two different behavior patterns observed in the A-B-
design used in functional assessment (see the sec-
A-B design with human clients, gradual versus
tion Functional Assessment later in this chapter).
abrupt changes, could conceivably reflect whether
However, an important difference is that in func-
the behavior was under control by contingencies of
tional assessment, each condition involves assess-
reinforcement or by discriminative stimuli (instruc-
ment of existing behavior under familiar conditions,
tion). Skinner (1969) drew a distinction between
whereas with the alternating-treatments design, each
contingency-shaped behavior and rule-governed
condition is an intervention seeking to change
behavior. Behavior that is rule governed already
behavior. Alternating-treatments designs are consid-
exists in the individual’s repertoire and is therefore
erably more complex than what can be covered here
switched on and off by the instructions, and the
(see, e.g., Barlow et al., 2009).
experiment or treatment becomes an exercise
in stimulus control, or rule following. However,
­contingency-shaped behavior may not exist before Changing-Criterion Designs
an intervention intended to control the behavior, When a behavior needs to be changed drastically in
and the experiment or treatment becomes a demon- topography, duration, or frequency, an effective
stration or study of acquisition. Thus, with human approach is to change the criterion for reinforce-
participants, two fundamentally different behavioral ment in small steps. This approach is essentially the
processes may underlie the different patterns of method of shaping by successive approximation
behavior seen with the use of the A-B-A-B design (Cooper, Heron, & Heward, 2007). Concrete, mea-
(and also with the A-B, A-B-A, and multiple-baseline surable criteria are applied in a stepwise fashion in
designs). Unfortunately, authors do not always accordance with behavioral changes. Each step
explain the procedure carefully enough that one can serves as a baseline for comparison to the next step.
determine whether the participant was instructed Withdrawals or reversals are rare because the goal is
about changes in procedure when a new phase was to establish a drastic behavior change. The method
initiated (see also Kazdin, 1973). Perhaps future can be characterized as an A-B-C-D-E-F-G . . .
­systematic examinations of existing literature can design, although it is rarely written that way. For

16
Single-Case Research Methods

example, as a laboratory demonstration, the the number of circles increased as well. The objec-
­duration of lever holding by a rat may be increased tive was to enable a topography change from touch-
in small steps of first 200 milliseconds across ses- lift to a continuous finger motion across circles, as
sions, then 500 milliseconds, then 1 second, and so in touch-drag. The lower diagram in Figure 1.10
forth. Eventually, the rat may steadily hold the lever shows the development of drawing for one subject
down for an extended period of time, say up to 10 for one of the four trial types on the monitor.
seconds or longer if the final criterion is set at 10 Touch-lift is indicated by a dot and touch-drag
seconds (e.g., Brenagan & Iversen, 2012). The changing- (touch and then move the finger while it still has
criterion design may also be used in the form of contact with the monitor) is indicated by a thick line
stimulus-shaping methods in educational research that connects successive circles. The figure shows
in which stimuli are faded in or faded out or modi- the interplay between procedure and results. The
fied in topography. For example, the spoken word circles come closer together, the number of circles
house is taught when a stimulus, say a schematic of increases, and the chimpanzee’s behavior changes
Copyright American Psychological Association. Not for further distribution.

a house, is presented to a child. Over time the sche- accordingly from touch-lift to touch-drag. Small
matic is modified in many small steps into the specks of not lifting the finger between consecutive
printed word HOUSE; at each step, the correct circles initially appear in Session 13, and the first
response to the stimulus remains the spoken word full sweep across all circles without lifting appears
house. A variety of other schematics of objects, each already in Session 14 and is fully developed by Ses-
with its own separate spoken word, are similarly sion 16. The time to complete each trial (vertical
modified into printed words over time. Eventually, a lines) shortens as the topography changed from
situation is created in which the child may produce touch-lift to touch-drag. Eventually, the chimpanzee
spoken words that correspond to the printed words swept over the circles in one movement for all four
(e.g., Cooper et al., 2007). McDougall, Hawkins, trial types. In the remaining steps, the stimuli were
Brady, and Jenkins (2006) described various imple- changed by fading techniques across and within ses-
mentations of the changing-criterion design in edu- sions from an array of circles to just two dots, one
cation and suggested that the changing-criterion where drawing should start and one where it should
design can profitably be combined with A-B-A-B or end. An additional aspect of the method was that the
multiple-baseline designs to establish optimal chimpanzees were also taught to end the trials them-
designs for individuals in need of large-scale behav- selves by pressing a trial termination key, which was
ior changes. introduced in Step 6. Thereby, the end of the drawn
Actual laboratory experiments and educational trace (lifting of the finger) came under control by
or clinical interventions using variations of changing- the stimuli on the monitor and not by delivery of
criterion designs are customarily highly complex reinforcement. The final performance was a highly
mixtures of different methods. For example, to gen- precise drawing behavior under visual guidance by
erate visual guidance of motor behavior in chimpan- the stimuli on the screen. Such stimulus control of
zees, Iversen and Matsuzawa (1996) introduced an complex motor performance was acquired with a
automated training protocol to bring line drawing completely automated method over a span of about
with a finger on a touch screen under stimulus con- 100 sessions (3–4 weeks) entirely by continuously
trol. Finger movement left “electronic ink” on the rearranging reinforcement contingencies and
screen surface. The top-left diagram in Figure 1.10 stimulus-fading procedures in small steps without
shows a sketch of the 10-step training procedure. the use of verbal instruction. In the final perfor-
Each session presented in mixed order four trial mance, the chimpanzees would look at the dots, aim
types as four orientations of an array of circles (as one finger at the start dot, rapidly move the finger
indicated). The chimpanzees had to touch the cir- across the monitor, lift the finger at the second dot,
cles to produce reinforcement. Stimuli changed and then press the trial termination key to produce
across and within sessions. Thus, within Steps 2 and reinforcement, all in less than 1 second for each
3, the circles were moved closer across trials, and trial. The top right graph in Figure 1.10 shows the

17
Iver H. Iversen
Copyright American Psychological Association. Not for further distribution.

Figure 1.10.  Top left: Schematic of the experimental procedure. Top right: Frequency plot of angles of
drawing, as illustrated in the top images. Bottom: Data are from one trial type, and all trials of that type are
shown in six successive sessions. Number of circles and distance between circles changed within and across ses-
sions. A dot indicates touch-lift, and a black line indicates touch-drag. From “Visually Guided Drawing in the
Chimpanzee (Pan Troglodytes),” by I. H. Iversen and T. Matsuzawa, 1996, Japanese Psychological Research, 38,
pp. 128, 131, 133. Copyright 1996 by John Wiley & Sons, Inc. Reprinted with permission.

resulting control of the angle of the drawn trace for time training the chimpanzees and added that they
each trial type. could “obviously” already draw because such
When behavior control techniques are success- smooth and precise motor performance could not at
ful and produce very reliable and smooth perfor- all have been acquired by simple shaping. Such
mance, spectators of the final performance may not comments may be somewhat amusing to behavior
quite believe that the subjects could at some point analysts. Yet, they are made by professionals in
in time not do this. In fact, a renowned develop- other areas of psychology and reveal a disturbing
mental psychologist happened to visit the labora- lack of understanding of and respect for effective
tory while the drawing experiment was ongoing. behavior control techniques developed with the use
On seeing one of the chimpanzees draw line after of single-case research methods. Unfortunately, the
line smoothly and without hesitation (in Step 10), commentaries also reveal a failure by behavior ana-
he exclaimed that the investigators had wasted their lysts to promote understanding about behavior

18
Single-Case Research Methods

analysis, even to professionals in other areas of Time Scales of Single-Case Research


psychology. Methods
The time scale can vary considerably for single-case
Discrete-Trial Designs research methods. Laboratory demonstrations with
A discrete trial is the presentation of some stimu- animals using the A-B-A design to show acquisition
lus material and implementation of a response–­ and extinction of a simple operant can usually be
reinforcer contingency that applies only in the pres- accomplished in a matter of 20 to 30 minutes (e.g.,
ence of that stimulus. The discrete-trial design has Figure 1.4). Educational interventions using A-B-A-
long been a standard procedure for use with animals B designs may last for weeks. Therapeutic interven-
in all sorts of experiments within and outside of tions for children with autism spectrum disorder
behavior analysis. Complex experiments and educa- may last a few years, with multiple changes in
tional projects, for example, involving conditional designs within this time period (e.g., Lovaas, 1987).
discriminations (e.g., matching to sample), are often Green, Brennan, and Fein (2002), for example,
Copyright American Psychological Association. Not for further distribution.

based on discrete-trial designs (see Volume 2, Chap- described a behavior analysis treatment project for
ter 6, this handbook). one toddler with autism, Catherine. Treatment con-
Historically, the discrete-trial method has tinued for 3 years with gradually increasing com-
become almost the hallmark of applied behavior plexity, beginning with instruction in home settings,
analysis, especially for its use in education and treat- through other settings, and eventually to regular
ment of children with intellectual disabilities (Baer, preschool settings with minimal instruction. Figure
2005). For example, the method may be as simple as 1.11 shows the chronological order of skill intro-
presenting a picture of an animal, and the response duction for the 1st year. The design is a gradual
that produces the reinforcer in this trial is the spo- introduction of skill complexity in which previ-
ken name of the animal on the picture; in another ously acquired skills serve as prerequisites for new
trial, the picture may be of another animal, and the skills; the logic of the design is similar to that of the
reinforced response is the name of that animal. ­changing-criterion design, mentioned earlier, except
Thus, the method is useful for teaching which that the criterion change is across topographically
­stimuli (verbal or pictorial) should control which different behaviors and situations. There is no base-
responses and when. Loosely speaking, the discrete- line for each skill other than the therapist’s knowl-
trial method teaches when a response is permitted edge that a given skill was absent or not sufficiently
and when it is not. The method may be used in a developed before it was explicitly targeted for acqui-
very informal way, as in training when normal activ- sition treatment. There are no withdrawals because,
ities should and should not occur. The method may for such real-life therapeutic interventions, they
also be presented very formally in an automated would force removal of an acquired, desirable skill.
arrangement, as in the above example with chim- Effective withdrawals may even be impossible for
panzees. In applied behavior analysis, the discrete- many of the acquired skills such as eye contact, imi-
trial method is useful for teaching single units of tation, and speech. Instead, continued progress
behavior such as acquisition of nouns but is less revealed that the overall program of intervention
useful for teaching sequential behaviors, such as was successful. Besides, the study replicated previ-
brushing teeth (Steege & Mace, 2007). Apparently, ous similar studies. Green et al. concluded that
the very term discrete-trial teaching has of late, out- “over the course of 3 years of intense, comprehen-
side the field of applied behavior analysis, come to sive treatment, Catherine progressed from exhibit-
be considered as a simple procedure for simple ing substantial delays in multiple skill domains to
behaviors. Attempts have been made, therefore, to functioning at or above her age level in all domains”
present applied behavior analysis to audiences out- (p. 97).
side of behavior analysis as a considerably richer set Multiple-baseline designs can on occasion last
of methods than just discrete-trial methods (Ghezzi, for years. D. K. Fox, Hopkins, and Anger (1987),
2007; Steege & Mace, 2007). for example, introduced a token reinforcement

19
Iver H. Iversen
Copyright American Psychological Association. Not for further distribution.

Figure 1.11.  Timeline for teaching various skills to a single


child in an intense behavior analysis program. Arrows indicate that
instruction continued beyond the time period indicated here. From
“Intensive Behavioral Treatment for a Toddler at High Risk for
Autism,” by G. Green, L. C. Brennan, and D. Fein, 2002, Behavior
Modification, 26, p. 82. Copyright 2002 by Sage Publications.
Reprinted with permission.

­ rogram for safety behaviors in two open-pit mines.


p single-case research methods, the case may not nec-
Concurrent baseline recordings of safety were essarily be one individual but a group of individuals—
recorded for both mines. After 2 years in one mine, and in this case, the group may even change
the contingencies were changed for that mine, and members over time.
the baseline was continued for the other mine for The length of time a project lasts is not an
another 3 years before the contingencies were ­essential aspect of single-case research designs. The
changed for that mine, too. For this 15-year project, essential aspects are that data consist of repeated
at each mine the contingencies were applied at the observations or recordings and that such data are
level of both the individual worker and teams of compared across different experimental or treatment
workers. This example serves as a reminder that for conditions.

20
Single-Case Research Methods

Using Single-Case Designs for from the goal of providing impetus for intervention,
Assessment with publications presenting results from assess-
ment alone without actual follow-up intervention.
A multitude of methods have been developed to
Thus, the reader is left wondering whether the
assess behavior. The general methodology is similar
causes of the problem behavior revealed by the
to single-case designs, which can be used in testing
assessment package were also causes that could be
which specific stimuli control a given client’s
manipulated and whether such manipulations
behavior.
would in fact improve the problem behavior.
Functional Assessment Functional assessment serves as a useful clini-
Functional assessment seeks to ascertain the imme- cal and educational tool to determine possible
diate causes of problem behavior by identifying immediate causes of problem behavior. How-
antecedent events that initiate the behavior and con- ever, reliability of assessment outcome cannot be
sequent events that reinforce and maintain the fully evaluated without a direct link between the
Copyright American Psychological Association. Not for further distribution.

behavior. The goal of assessment is to determine the inferred causes of problem behavior in assessment
most appropriate intervention for the given individ- and the outcome of subsequent intervention
ual. For example, a child exhibiting self-injurious using manipulations of these inferred causes for
behavior may be placed in several different situa- the same client.
tions to determine which situations and possible
consequences of the behavior in those situations Assessment Using Discrete-Trial
affect the behavior frequency. The situations may be Procedures
(a) alone, (b) alone with toys, (c) with one parent, Many children and adults with intellectual disorders
(d) with the other parent, (e) with a sibling, and have not acquired expressive language, and commu-
(f) with a teacher. The child is exposed to one situa- nication with them can be difficult or impossible. A
tion each session with situations alternating across method called facilitated communication claims to
sessions. For determination of reliability, each situa- enable communication with such clients by having a
tion is usually presented more than once. If the specially trained person, the facilitator, hold the cli-
problem behavior in this example occurs most fre- ent’s hand or arm while the client types with one
quently when either parent is present, and it is finger on a keyboard. Because some typed messages
observed that the parent interacts with the child have expressed advanced language use without the
when the problem behavior occurs, then the infer- clients ever having shown other evidence of such
ence is drawn that the problem behavior may be language, facilitated communication has been ques-
maintained by parental interaction (i.e., positive rein- tioned as a means of authentic communication. The
forcement from attention) and that the parent’s entry is question is whether the facilitator rather than the
an antecedent for the behavior (i.e., parental entrance client could be the author of the typed messages. To
is an immediate cause of initiation of the behavior). test for such a possibility, several investigators have
The therapist will then design an intervention based used single-case research methods to examine stim-
on the information collected during assessment. ulus control of typing (e.g., Montee, Miltenberger,
Causes of the problem behavior are inferred from func- & Wittrock, 1995; Wheeler, Jacobson, Paglieri, &
tional assessment methods, because no experimental Schwartz, 1993). The most common test is to first
analysis is performed in which the causes are manip- present the client and the facilitator with a series of
ulated systematically to determine how they influ- pictures in discrete trials and ask the client to type
ence behavior. Functional assessment of problem the name of the object shown on the picture. If the
behaviors has become very prevalent; special issues client types the correct object names, then the
and books have been published on how to conduct method is changed by adding test probes on some
the assessments (e.g., Dunlap & Kincaid, 2001; Neef trials. On those test trials, the pictures are arranged
& Iwata, 1994). Recently, functional assessment has such that the facilitator sees one picture, and the
appeared to have taken on a life of its own, separate client sees another picture; an important added

21
Iver H. Iversen

methodological feature is that the client cannot see


the picture the facilitator sees, and vice versa. These
studies have shown with overwhelming reliability
that the clients type the name of the picture that the
facilitators see; if the clients see a different picture,
then the clients do not type the name of the picture
they see but instead type the name of the picture the
facilitator sees. Thus, these studies have demon-
strated, client by client, that it is the facilitator who
types the messages and that the pictures do not con-
trol correct typing by the clients—the clients cannot
communicate the names of the pictures they see.
For commentaries and reviews of this literature,
Copyright American Psychological Association. Not for further distribution.

see Green (2005); Green and Shane (1994); and


Jacobson, Mulick, and Schwartz (1995). As Green
(2005) pointed out, the continued false beliefs by
facilitators, family, caregivers, and news media that
clients actually communicate with facilitated com-
munication may in fact deprive them of an opportu-
nity to receive effective, scientifically validated
treatment.

Real-Life Assessment Using Multistage


Reversal-Replication Designs
Applications of single-case research methods with
clients in real-life situations away from clinic or
school may not always follow a pure, prearranged
formula. Contextual factors, such as varying client Figure 1.12.  Top: Schematic of the events in a single
health and family situations, are additional consider- trial. The first 1.5 seconds is an observation period (in
this case, the presession instruction was to always select
ations. In such cases, the consistent replication of the noun of a noun–verb choice; new words appeared
behavior patterns in similar conditions across multi- on each trial, and the correct position varied randomly
ple sessions becomes the indicator of whether a from trial to trial). A 0.5-second baseline of electroen-
given treatment or assessment has the intended out- cephalogram (EEG) is then recorded, followed by an
active phase in which the patient can control the cursor
come. For example, completely paralyzed patients (ball) on the screen for 3 seconds. Bottom: Data from
with amyotrophic lateral sclerosis were trained to one patient with amyotrophic lateral sclerosis from one
communicate using only their brainwaves (via elec- training day with several successive tasks. T = task; T1 =
simple target; T5 = noun–verbs; T7 = color matching;
troencephalogram) to control the movement of a
T8 = addition or subtraction matching. From “A Brain–
cursor on a computer screen (Birbaumer et al., Computer Interface Tool to Assess Cognitive Functions
1999). Letters were placed on the screen, and the in Completely Paralyzed Patients With Amyotrophic
patient could move the cursor toward them and Lateral Sclerosis,” by I. H. Iversen, N. Ghanayim, A.
Kübler, N. Neumann, N. Birbaumer, and J. Kaiser, 2008,
thereby spell words to communicate. In additional Clinical Neurophysiology, 119, pp. 2217, 2220. Copyright
experiments, abilities to distinguish verbs from 2008 by Elsevier. Reprinted with permission.
nouns, odd from even numbers, and consonants
from vowels and to perform simple computations trial. The patient gets online visual feedback from
were assessed in a matching-to-sample–type task the electroencephalogram in the form of cursor
(Iversen et al., 2008). The top part of Figure 1.12 movement. If the cursor reaches the correct stimu-
shows a schematic of the events during one 5-second lus, then a smiley face appears on the screen. The

22
Single-Case Research Methods

patients lived in their private homes with constant electrode, equipment failure, or the lack of patient
care and were assessed intermittently over several cooperation. That the patient scored high again on
sessions spanning a few hours once or twice each the simple task in the last A phase, immediately after
week. The ideal scenario was always to test patients the session in the D phase, demonstrates that the
several times on a very simple task, such as steering patient had some difficulties discriminating the
the cursor to a filled box versus an open box, to stimuli presented in Task 8 in the D phase (i.e.,
make sure that the electrodes were attached cor- addition and subtraction, such as 3 + 5 or 7 − 2).
rectly, equipment worked, and the patient could still Among the many findings is that several warm-up
use the electroencephalogram feedback to control sessions were necessary at the beginning of the day
the cursor. Once the patient reached at least 85% before the patient reached the usual 85% correct on
correct on a simple task, then tasks with assessment the simple task. Several such training days with this
stimuli (e.g., nouns and verbs) were presented for very ill, speechless and motionless patient demon-
one or several sessions. strated that the patient had some deficits in basic
Copyright American Psychological Association. Not for further distribution.

If the neurodegenerative amyotrophic lateral skills. Thus, the data showed that the patient, a for-
sclerosis destroys the patient’s ability to discriminate mer banker, now had problems with very simple
words or numbers, then the patient should show addition and subtraction.
a deficit in such tasks compared with when the This example illustrates the use of single-case
patient can solve a simple task. To determine research methods in complex living situations with
whether a patient has a potential deficit in a given patients with extreme disability. To extract mean-
skill, such as odd–even discrimination, then it is ingful data from research with such patients in a
necessary to know that the patient can still do the varying environment, it is necessary to know repeat-
simple task of moving the cursor to the correct tar- edly that the complex recording and control equip-
get when the target is just a filled box. Thus, to ment works as intended and that patients’ basic
interpret the data, it was necessary to present the skills are continuously intact because no verbal
simple task several times at the beginning of, during, communication from the patient is possible (i.e., the
and end of a given day the patient was visited by the patient cannot tell the trainer that he is tired or that
testing team. It proved challenging at times to con- something else may be the matter). Indeed, both
vince family members that it was necessary to repeat trainers and family members had to be instructed in
the simple tasks several times because, as the family some detail, and often, as to why it was necessary to
members said, “You already know that he can do repeat the simple tasks several times on each visit
that, so why do you keep wasting time repeating it?” to the patient’s home. The multiphase replication
The bottom part of Figure 1.12 shows the results for design with repeated presentation of phases of sim-
each of 16 consecutive sessions for a single test day ple tasks in alternation with more complex tasks is a
for one patient. The task numbers on the x-axis refer necessary component of single-case research or test-
to the type of training task, with Task 1 (T1) being ing methods applied to complex living situations in
the simplest, which was repeated several times. The which communication may be compromised or
overall design forms a phase sequence of A-B-C-A- impossible.
C-D-A, in which the A phases present the simple
task and serve as a baseline control, and the B, C,
Data Analysis Methods
and D phases present more complex test material.
The A phases thus serve as a control for equipment A fundamental aspect of single-case designs is that
integrity and for the patient’s basic ability to move behavior is recorded repeatedly and under the same
the cursor to a correct target. For example, had the set of methods and definitions across different
training day ended after the session in Phase D, it experimental, educational, or treatment conditions.
would have been difficult to conclude that the Chapter 6 of this volume covers these issues; see
patient had a deficit in this task because the deteri- also J. M. Johnston and Pennypacker (2009) and
orating percentage correct could reflect a loose Cooper et al. (2007).

23
Iver H. Iversen

Data Recording Methods ioral data is Parsonson and Baer (1978), which
Automated recording techniques (customarily used ­covers analysis of data from basic single-case
with animal subjects) require frequent verification research designs.
that monitoring devices (switches, photocells, Fundamental issues in visual analysis are eval-
touch screens, etc.) record behavior as intended and uations of baseline stability and baseline trends.
that these monitoring devices are calibrated cor- Baselines should ideally vary little, and experi-
rectly. With observational methods, calibration of menters should analyze any conditions responsi-
criteria for response occurrence used by different ble for unexplained variation (Sidman, 1960).
observers is a crucial issue, and intraobserver as Data from an intervention cannot always be inter-
well as interobserver agreement calculations are preted if variability from the baseline carries over
essential for objective recording of behavior (see into the intervention phase. Trends in baselines
Chapter 6, this volume). Equally essential is intra- can be problematic if they are in the direction of
and interpersonnel consistency in the methods of the expected experimental or therapeutic effect.
Copyright American Psychological Association. Not for further distribution.

delivering consequences to clients in educational For example, if a baseline rate of behavior gradu-
and clinical settings. However, reports of agreement ally increases over several observation periods,
scores do not ordinarily include information about and behavior increases further during the inter-
how consistently personnel follow described vention, then it can be difficult or impossible to
methods. determine whether the intervention was responsi-
Common measures of behavior are frequency ble for the increase (e.g., Cooper et al., 2007). The
of occurrence (often converted to response rate), expressions appropriate baseline and inappropriate
response duration, and response topography (e.g., baseline have appeared in the literature to empha-
Barlow et al., 2009). A given behavior can also be size this issue. An appropriate baseline is either a
analyzed in terms of its placement in time as a flat baseline without a trend or a baseline with a
behavior pattern or its placement among other trend in the direction opposite to the expected
behaviors, as in analyses of the sequential properties effect. For example, if the rate of self-injury in a
of behavior (e.g., Henton & Iversen, 1978; Iversen, child is steadily increasing in baseline, and the
1991; see also Chapter 12, this volume). intervention is expected to decrease the behavior,
then an increasing baseline is appropriate for
Visual Data Analysis Methods intervention because the behavior is expected
Behavior analysis has a long tradition (beginning to decrease in the intervention even against an
with Pavlov, Thorndike, and Skinner) of focusing increasing baseline. There is certainly no rationale
on nonstatistical, visual analyses of data from sin- in waiting for self-injury to stabilize before inter-
gle subjects (Iversen, 1991). In Chapter 9 of this vention starts. However, an inappropriate baseline
volume, Bourret and Pietras describe a variety of is one that has a trend in the same direction as the
methods of visual data analysis. Such analyses are expected outcome of the intervention (see also
now fairly standard and are covered in most text- ­Cooper et al., 2007).
books on single-case research designs (e.g., Kennedy, Patterns of data are important in interpretations
2005; Morgan & Morgan, 2009) and in textbooks of behavior changes, in addition to descriptive sta-
on behavior analysis in general (e.g., Cooper et al., tistical evaluations. For example, the two trends
2007). previously noted here in data for A-B, A-B-A, A-B-A-
Visual analyses of data are not particular to B, and multiple-baseline designs (i.e., Figure 1.9)
experimental or applied behavior analysis and per- would not be apparent had data been presented only
meate all sciences and other forms of communica- as averages for each phase of recorded behavior.
tion, as exemplified in the books by Tufte (e.g., Plotting data for each session captures trends, vari-
1983, 1990; see also Iversen, 1988) on analyzing ability, and intervention outcomes. Without such
and presenting visual information. Within behavior data, the investigator easily misses important infor-
analysis, a classic text on visual analysis of behav- mation that might lead to subsequent changes in

24
Single-Case Research Methods

procedure. In fact, successful interaction between be ruled out). If a t test is applied, however, compar-
researcher and subject depends on visual data ing the baseline data with the treatment data, the
­analysis concomitant with project progression (e.g., standard deviation for the B phase may be very large
Sidman, 1960; Skinner, 1938) because data range from the baseline level to the
highest rate when treatment is most effective. The
Statistical Data Analysis Methods result may be a nonsignificant effect. Besides, the
Statistical analyses of behavioral data in both basic data in the B phase showing a gradual acquisition
research and application have been controversial may not be normally distributed and may not be
since Skinner’s (1938) The Behavior of Organisms, independent measures (i.e., as a series of increasing
which was devoid of traditional statistical analyses. values, data on session n influence data on session
Behavior analysis gathers information about the n + 1); the result is that the t test is not valid. How-
behavior of individual subjects, whereas the tradi- ever, the failure of a common statistical test to show
tional statistical approach gathers information about an effect certainly does not mean that such data are
Copyright American Psychological Association. Not for further distribution.

groups of subjects and offers no information about unimportant or that there is no effect of the experi-
data from individual subjects. However, visual anal- mental manipulation. Single-case research methods
yses of behavioral data are not always sufficient, and are not designed for hypothesis testing and inferen-
statistical methods can be used to supplement the tial statistics but for analysis of behavior of the indi-
analysis. Several authors have analyzed the ongoing vidual subject and for development of methods that
controversy regarding use of statistics in behavior can serve to help individuals acquire appropriate
analysis and psychology (e.g., Barlow et al., 2009; behavior. The assumptions of inferential statistics,
Kratochwill & Brody, 1978; see Chapter 7, this vol- meant for between-group comparisons, with inde-
ume). Chapters 11 and 12 of this volume provide pendent observations, random selection of subjects,
information on new statistical techniques, which random allocation to treatment, and random treat-
may prove useful for behavior analysts conducting ment onset and offset are obviously not fulfilled in
time-series single-case research designs. single-case research methods. Eventually, however,
A major issue with the use of statistics in behav- statistical tests appropriate for single-case methods
ior analysis is the treatment of variability in data. may evolve from further developments in analyses
Visual analyses take variability as informative data of interrupted time-series data (e.g., Barlow et al.,
that can prompt an experimental analysis of the 2009; Crosbie, 1993; see Chapters 11 and 12, this
sources of this variability (e.g., Sidman, 1960). With volume).
statistical analysis, however, variability is artificially Aggregation from individual data points through
compressed to a more manageable single number averages and standard deviations to statistical tests
(variance or standard deviation) without analysis of to p values and to the final binary statement “yes or
the source of the variability. Thus, visual analysis no” is, of course, common in all sciences, including
and statistical analysis can be seen as antithetical psychology. Quantitative reduction of complex
tools to uncover causes of variability in the behavior behavior patterns to yes or no eases communication
of the individual subject. of theories and ideas through publications and pre-
Consider, for example, the use of a simple t test sentations, in which actual data from individual sub-
for evaluation of data from two successive phases of jects or individual trials may be omitted. In contrast,
an A-B design (e.g., Figure 1.3) as an illustration of a focus on data linked more directly to experimental
the problems one may face with the use of a statisti- manipulations can lead to demonstrations of stun-
cal test created for entirely different experimental ning control and prediction of behavior at the
designs. If the B phase shows acquisition of behavior moment-to-moment level for the individual subject
at a gradually increasing rate compared with the (e.g., Henton & Iversen, 1978; Sidman, 1960),
baseline in the A phase, the behavior analyst has no which is much closer to everyday interactions
problems visually identifying a large effect of the between behavior and environment. In daily life,
manipulation (given that confounding variables can people respond promptly to single instances of

25
Iver H. Iversen

interpersonal and environmental cues. Such stimu- autism. One group of children (n = 19) with the
lus control of behavior is the essence of interhuman diagnosis of autism received intensive behavior anal-
communication and conduct. By demonstrating ysis procedures (40 hours/week) for 2 years, and
control of behavior at this level for the individual another group (n = 19) with the same diagnosis did
subject, behavior analysis can be both a science of not receive the same treatment. The differences
behavior and a tool for educational and therapeutic between the groups were vast; for example, 49% of
interventions. the children in the treatment group showed signifi-
cant gains in IQ and other measures of behavioral
Quantitative Descriptions of Behavior– functioning compared with the other group. This
Environment Relations study helped generate respect among parents and
Behavioral data from single-case designs have professionals for behavioral methods in treatment
invited quantitative descriptions of the relationships of children with autism. Since the Lovaas (1987)
between behavior and environmental variables. Such study, several other studies have similarly compared
Copyright American Psychological Association. Not for further distribution.

descriptions vary considerably from issue to issue two groups of children (sometimes randomly
and attract general interest in both basic research assigned) with similar levels of autism spectrum dis-
(Shull, 1991) and application (see Chapter 10, this order, in which one group received intensive behav-
volume). ior analysis treatment for 1 or several years, and the
other group received less treatment or no treatment
other than what is ordinarily provided by the child’s
Single-Case Designs and Group
community (so-called “eclectic” treatment). For
Studies Compared
example, Dawson et al. (2009) recently reported one
“Operant methods make their own use of Grand such study with 48 children (18–30 months old)
Numbers; instead of studying a thousand rats for diagnosed with autism spectrum disorder in a ran-
one hour each, or a hundred rats for ten hours each, domized controlled trial. Statistical procedures dem-
the investigator is likely to study one rat for a thou- onstrated significant gains in a variety of behavioral
sand hours” (Skinner, 1969, p. 112). Research measures for the group that received the treatment.
involving comparisons of groups of subjects, in For this study and the Lovaas study, each child went
which each group is exposed once to one level of a through complex and intense procedures based on
manipulation, is rare in behavior analysis, in partic- single-case research methods for about 2 years. The
ular with animal subjects. group comparisons mainly used standard measures
Yet, group research or studies with a large num- of the children’s performances. Enough group
ber of participants sometimes have relevance for ­studies have, in fact, been conducted that several
behavior analysis. For example, in a recent interview meta-analyses of the efficacy of behavior analysis
(Holth, 2010), Murray Sidman remarked that large- treatment of children with autism spectrum disorder
scale implementation of behavior analysis tech- have now been performed on the basis of such
niques may require prior demonstration of the studies.
effectiveness of those techniques in large popula- Group comparison methods have also increased
tions. Thus, studies using randomization and con- the visibility of behavior analysis techniques in areas
trol groups may be necessary for promulgation of of application. For example, Taub et al. (2006) used
effective behavior control techniques. A positive behavioral techniques in treatment of patients with
outcome of a group study may not make behavior paralysis of arms or legs resulting from stroke or
analysts know more about control of an individual brain injury. With the less affected or “normal”
client’s behavior, yet such an outcome may nonethe- limb restrained to prevent its use, 21 patients who
less make more people know about behavior analy- received massed practice of the affected limb (6 hours/
sis. For example, the widely cited group study by day for 10 consecutive weekdays with shaping of
Lovaas (1987) generated broad interest in behavior movement and establishment of stimulus control
analysis methods for treatment of children with of movement using social reinforcement) were

26
Single-Case Research Methods

c­ ompared with a matched control group of 20 informing the general public, therapists, scientists in
patients who received customary, standard physical other areas, educators, politicians, and so forth
therapy and general support. Treatment patients about behavior analysis techniques should, how-
showed huge and clinically significant gains in ever, be balanced with a concern for ethical treat-
motor control of the affected arm, whereas patients ment of the participants involved. When a behavior
in the control group showed no such gains. A simi- analyst, based on experience with previous results,
lar example is provided in research by Horne et al. knows full well that a particular treatment using
(2004). These investigators had first obtained reli- ­single-case methodology has proven to be effective
able single-case data that reinforcement techniques over and over, then it is indeed an ethical dilemma
could be used effectively to increase the consump- to knowingly split a population of children in need
tion of fruits and vegetables among schoolchildren. of that treatment into two groups for comparison of
To promote the findings, Horne et al. conducted a treatment versus no treatment. Children who are in
large group study with 749 schoolchildren. The a control group for a few years for a valid statistical
Copyright American Psychological Association. Not for further distribution.

children were split into an experimental group and a comparison with the treatment group may thus be
control group. A baseline of fruit and vegetable con- deprived of an opportunity for known, effective
sumption was taken first. Then at lunchtime, fruit treatment.
and vegetable consumption was encouraged by hav- Several so-called randomized controlled group
ing the children in the experimental group watch studies and meta-analyses of such studies have been
video adventures featuring heroic peers (the Food conducted over the past few decades to determine
Dudes) who enjoy eating fruits and vegetables, and whether applied behavior modification actually
the children received reinforcers for eating fruit and works (e.g., Spreckley & Boyd, 2009). For elaborate
vegetables. Children in the control group had free comments and critiques of some meta-analysis stud-
access to fruit and vegetables. Compared with the ies of applied behavior analysis methods, see Kim-
children in the control group, fruit and vegetable ball (2009). These time-consuming studies, with a
consumption was significantly higher among the presumed target audience of policymakers and
children in the experimental group. On the basis of insurance companies, would clearly not have been
such data, this program has now been implemented undertaken unless countless studies using single-
on a large scale in all schools in Ireland and in other case methods had already demonstrated that the
places in England (Lowe, 2010). behavior of an individual child can be modified and
Such group-comparison studies, published in sustained with appropriate management of rein-
journals that do not ordinarily publish studies using forcement contingencies and stimulus control tech-
single-case research methods, may be helpful in pro- niques. The sheer mass of already existing studies
moting general knowledge about behavior analysis based on single-case methodology with successful
to a much wider audience. In addition, when effec- outcomes, for literally thousands of individuals
tive, evidence-based behavior-analytic treatments across a variety of behavior problems and treatment
become broadly recognized from publications with a settings, poses the question, “How many more ran-
large number of participants and in renowned jour- domized controlled group studies and subsequent
nals, then granting agencies, insurance companies, meta-analyses are necessary before single-case meth-
journalists, and maybe even university administrators ods can be accepted in general as effective in
and politicians start to pay attention to the findings. treatment?”
For behavior analysts, group studies may not seem The use of inferential statistics as a method of
to add much basic knowledge beyond what is proof leads to the very odd situation that such meta-
already known from studies using single-case analyses may explicitly exclude the results from
research methods. Publication of group studies may, application of single-case methods with individual
however, be a tactic for promotion of basic, impor- clients (e.g., Spreckley & Boyd, 2009), even though
tant, and effective behavioral techniques beyond the the overall purpose of the meta-analyses is to decide
readership of behavior analysts. The need for whether such methods work for individual clients.

27
Iver H. Iversen

Profound misunderstandings of what can be evidence-based medicine, placed single-case


accomplished by single-case research methods in research designs with randomization of treatment
general can on occasion be heard among pedagogues highest in their hierarchy of strength of evidence for
and critics of applied behavior analysis interventions treatment decisions.
for children with developmental disorders. The Single-case research designs feature an important
argument is that the trained behavior would have component of replication of behavioral phenomena
developed anyway given enough time (without for a single individual (J. M. Johnston & Penny-
training). For example, Spreckley and Boyd (2009) packer, 2009; Sidman, 1960). With direct replica-
stated that “what is too often forgotten is that the tion, the intervention is repeated, and when behavior
overwhelming majority of children with [autism changes on each replication, the experimenter or
spectrum disorder] change over time as part of their therapist has identified one of the variables that con-
development as opposed to change resulting from trol the behavior. With systematic replication, the
an intervention” (p. 343). Commentaries such as same subject or different subjects or species may
Copyright American Psychological Association. Not for further distribution.

these present a negation of the immense accumula- be exposed to a variation of an original procedure
tion of experimental and applied hard evidence that (­Sidman, 1960), and a successful outcome fosters the
individual behavior can indeed be effectively and accumulation of knowledge (see also Chapter 7, this
reliably changed with the use of behavior control volume). Sidman’s (1971) original demonstration of
techniques. Yet such comments persist and slow the stimulus equivalence stands out as a golden example
application. Some behavior analysts appropriately of the scientific value of replication as a method of
react when professionals make such claims without proof because the remarkable results using a single
supporting data (e.g., Kimball, 2009; Morris, 2009). child, in a carefully designed experiment, have been
Group studies may have their place in behavior replicated numerous times and thereby spurred
analysis when intervention should not be withdrawn development of a whole new field of research and
because of the social significance of the target behavior. application.
Baer (1975) suggested combining methods of multiple- Behavior analysts are sometimes weary of designs
baseline designs and group comparisons, in which that compare groups of subjects because no behav-
one group of subjects first serves as a comparison to ioral phenomena are established at the level of the
another group who receives intervention; later, the individual subject with such designs. However, to
comparison group also receives the same intervention. promote behavioral phenomena using single-case
research designs, comparisons of experiments are
common, and one thereby inevitably compares
Conclusion
groups of subjects across experiments, often across
Behavior analysts have developed designs and tech- species as well. Indeed, such comparisons often
niques that can increase or decrease target behavior prove the generality of the basic principles discov-
on a certain occasion and time. These methods serve ered with one small set of subjects. Besides, estab-
as tools for an experimental analysis of behavior. lishing the efficiency of a behavioral procedure to be
The same tools are used in applied behavior analysis implemented on a large population scale may
to modify and maintain behavior for the purpose require certain types of controlled group studies.
of helping a client. Single-case research designs offer Group designs and multiple-baseline designs can
a wide range of methods, and this overview has also be profitably combined to produce powerful
merely scratched the surface. Because of their accu- and socially relevant effects on a large scale (e.g.,
mulated successes, single-case designs are now D. K. Fox et al., 1987).
being adopted in areas outside of behavior analysis With wide implementations in psychology, edu-
such as medicine (Janosky et al., 2009), occupational cation, medicine, and rehabilitation, single-case
therapy (M. V. Johnston & Smith, 2010), and pain methodology is now firmly established as a viable
management (Onghena & Edgington, 2005). means for discovery as well as for application of
Indeed, Guyatt et al. (2000), in their review of basic behavioral mechanisms. Of late, behavior

28
Single-Case Research Methods

­ anagement interventions based on single-case


m Bijou, S. W., & Baer, D. M. (1961). Child development
methodology followed up with efficiency determina- I: A systematic and empirical theory. New York, NY:
Appleton-Century-Crofts. doi:10.1037/11139-000
tions from population studies have successfully
Birbaumer, N., Ghanayim, N., Hinterberger, T., Iversen,
demonstrated how the science of the individual can I., Kotchoubey, B., Kübler, A., . . . Flor, H. (1999).
be a science for all. A spelling device for the paralysed. Nature, 398,
297–298. doi:10.1038/18581
Individual prediction is of tremendous
Blampied, N. M. (1999). A legacy neglected: Restating the
importance, so long as the organism is to case for single-case research in cognitive behavior
be treated scientifically. (Skinner, 1938, therapy. Behaviour Change, 16, 89–104. doi:10.1375/
p. 444) bech.16.2.89
Boakes, R. (1984). From Darwin to behaviorism:
Man has at his disposal yet another pow- Psychology and the minds of animals. New York, NY:
erful resource—natural science with its Cambridge University Press.
strictly objective methods. This science,
Copyright American Psychological Association. Not for further distribution.

Boorstin, D. J. (1983). The discoverers: A history of man’s


as we all know, is making big headway search to know his world and himself. New York, NY:
every day. The facts and considerations Vintage Books.
I have placed before you are one of the Boring, E. G. (1929). A history of experimental psychology.
numerous attempts to employ—in study- New York, NY: Appleton-Century-Crofts.
ing the mechanism of the highest vital Boring, E. G. (1954). The nature and history of experi-
manifestations in the dog, the represen- mental control. American Journal of Psychology, 67,
573–589. doi:10.2307/1418483
tative of the animal kingdom which is
Brenagan, W., & Iversen, I. H. (2012). Methods to differ-
man’s best friend—a consistent, purely entially reinforce response duration in rats. Manuscript
scientific method of thinking. (From in preparation.
Pavlov’s acceptance speech on receiving Catania, A. C., & Laties, V. G. (1999). Pavlov and
the Nobel Prize in 1904; Pavlov, 1955, Skinner: Two lives in science. Journal of the
p. 148) Experimental Analysis of Behavior, 72, 455–461.
doi:10.1901/jeab.1999.72-455

References Cooper, J. O., Heron, T. E., & Heward, W. L. (2007).


Applied behavior analysis (2nd ed.). Upper Saddle
Baer, D. M. (1975). In the beginning there was the River, NJ: Pearson Education.
response. In E. Ramp & G. Semb (Eds.), Behavior
analysis: Areas of research and application (pp. 16–30). Crosbie, J. (1993). Interrupted time-series analysis with
Englewood Cliffs, NJ: Prentice-Hall. brief single-subject data. Journal of Consulting and
Clinical Psychology, 61, 966–974. doi:10.1037/0022-
Baer, D. M. (1993). Advising as if for research productiv- 006X.61.6.966
ity. Clinical Psychologist, 46, 106–109.
Dawson, G., Rogers, S., Munson, J., Smith, M., Winter,
Baer, D. M. (2005). Letters to a lawyer. In W. L. Heward, J., Greenson, J., . . . Varley, J. (2009). Randomized,
T. E. Heron, N. A. Neff, S. M. Peterson, D. M. controlled trial of an intervention for toddlers with
Sainato, G. Cartledge, . . . J. C. Dardig (Eds.), Focus autism: The early start Denver model. Pediatrics, 125,
on behavior analysis in education: Achievements, chal- e17–e23. doi:10.1542/peds.2009-958
lenges, and opportunities (pp. 3–30). Upper Saddle
River, NJ: Pearson. Dunlap, G., & Kincaid, D. (2001). The widening world of
functional assessment: Comments on four manuals
Baer, D. M., Wolf, M. M., & Risley, T. R. (1968). Some and beyond. Journal of Applied Behavior Analysis, 34,
current dimensions of applied behavior analysis. 365–377. doi:10.1901/jaba.2001.34-365
Journal of Applied Behavior Analysis, 1, 91–97.
doi:10.1901/jaba.1968.1-91 Ebbinghaus, H. (1913). Memory (H. A. Rueger & C. E.
Bussenius, Trans.). New York, NY: Teachers College.
Barlow, D. H., Nock, M. K., & Hersen, M. (2009). Single (Original work published 1885)
case experimental designs: Strategies for studying
behavior change (3rd ed.). New York, NY: Pearson Eysenck, H. J. (1960). Behavior therapy and the neuroses.
Education. New York, NY: Pergamon Press.
Bernard, C. (1957). An introduction to the study of experi- Fox, D. K., Hopkins, B. L., & Anger, W. K. (1987). The
mental medicine. New York, NY: Dover. (Original long-term effects of a token economy on safety
work published 1865) performance in open-pit mining. Journal of Applied

29
Iver H. Iversen

Behavior Analysis, 20, 215–224. doi:10.1901/ Iversen, I. H. (1991). Methods of analyzing behavior
jaba.1987.20-215 patterns. In I. H. Iversen & K. A. Lattal (Eds.),
Techniques in the behavioral and neural sciences:
Fox, R. G., Copeland, R. E., Harris, J. W., Rieth, H. J., & Experimental analysis of behavior, Part 2 (pp. 193–242).
Hall, R. V. (1975). A computerized system for select- New York, NY: Elsevier.
ing responsive teaching studies, catalogued along
twenty-eight important dimensions. In E. Ramp & Iversen, I. H. (1992). Skinner’s early research: From
G. Semb (Eds.), Behavior analysis: Areas of research reflexology to operant conditioning. American
and application (pp. 124–158). Englewood Cliffs, NJ: Psychologist, 47, 1318–1328. doi:10.1037/0003-
Prentice-Hall. 066X.47.11.1318
Ghezzi, P. M. (2007). Discrete trials teaching. Psychology Iversen, I. H. (2010). [Laboratory demonstration of acqui-
in the Schools, 44, 667–679. doi:10.1002/pits.20256 sition of operant behavior]. Unpublished raw data.
Green, G. (2005). Division fellow, Gina Green, reacts to Iversen, I. H. (2012). Tutorial: Multiple baseline designs.
CNN program “Autism is a World” which focuses Manuscript in preparation.
on facilitated communication. Psychology in Mental Iversen, I. H., Ghanayim, N., Kübler, A., Neumann, N.,
Copyright American Psychological Association. Not for further distribution.

Retardation and Developmental Disabilities, 31, 7–10. Birbaumer, N., & Kaiser, J. (2008). A brain-
Green, G., Brennan, L. C., & Fein, D. (2002). Intensive computer interface tool to assess cognitive functions
behavioral treatment for a toddler at high risk in completely paralyzed patients with amyotrophic
for autism. Behavior Modification, 26, 69–102. lateral sclerosis. Clinical Neurophysiology, 119,
doi:10.1177/0145445502026001005 2214–2223. doi:10.1016/j.clinph.2008.07.001

Green, G., & Shane, H. C. (1994). Science, reason, and Iversen, I. H., & Matsuzawa, T. (1996). Visually guided
facilitated communication. Journal of the Association drawing in the chimpanzee (Pan troglodytes).
for Persons with Severe Handicaps, 19, 151–172. Japanese Psychological Research, 38, 126–135.
doi:10.1111/j.1468-5884.1996.tb00017.x
Greenspan, R. J., & Baars, B. J. (2005). Consciousness
Jacobson, J. W., Mulick, J. W., & Schwartz, A. A. (1995).
eclipsed: Jacques Loeb, Ivan P. Pavlov, and the rise
A history of facilitated communication: Science, pseu-
of reductionistic biology after 1900. Consciousness
doscience, and antiscience. American Psychologist, 50,
and Cognition, 14, 219–230. doi:10.1016/j.concog.
750–765. doi:10.1037/0003-066X.50.9.750
2004.09.004
Janosky, J. E., Leininger, S. L., Hoerger, M. P., &
Guyatt, G. H., Haynes, R. B., Jaeschke, R. Z., Cook, D. J., Libkuman, T. M. (2009). Single subject designs
Green, L., Naylor, C. D., . . . Richardson, W. S. in biomedicine. New York, NY: Springer-Verlag.
(2000). Users’ guides to the medical literature: XXV. doi:10.1007/978-90-481-2444-2
Evidence-based medicine: Principles for applying the
users’ guides to patient care. JAMA, 284, 1290–1296. Johnston, J. M., & Pennypacker, H. S. (2009). Strategies
doi:10.1001/jama.284.10.1290 and tactics in behavioral research (3rd ed.). New York,
NY: Routledge.
Harris, B. (1979). Whatever happened to little Albert?
American Psychologist, 34, 151–160. doi:10.1037/ Johnston, M. V., & Smith, R. O. (2010). Single subject
0003-066X.34.2.151 design: Current methodologies and future directions.
OTJR: Occupation,Participation and Health, 30, 4–10.
Henton, W. W., & Iversen, I. H. (1978). Classical condi- doi:10.3928/15394492-20091214-02
tioning and operant conditioning: A response pattern
analysis. New York, NY: Springer-Verlag. Jones, M. C. (1924). A laboratory study of fear: The case
of Peter. Pedagogical Seminary, 31, 308–315. doi:10.
Holth, P. (2010). A research pioneer’s wisdom: An inter- 1080/08856559.1924.9944851
view with Dr. Murray Sidman. European Journal of
Behavior Analysis, 11, 181–198. Kazdin, A. E. (1973). Methodological and assessment
considerations in evaluating reinforcement pro-
Horne, P. J., Tapper, K., Lowe, C. F., Hardman, C. A., grams in applied settings. Journal of Applied Behavior
Jackson, M. C., & Woolner, J. (2004). Increasing Analysis, 6, 517–531. doi:10.1901/jaba.1973.6-517
children’s fruit and vegetable consumption: A peer-
modeling and rewards-based intervention. European Kazdin, A. E. (2011). Single-case research designs: Methods
Journal of Clinical Nutrition, 58, 1649–1660. for clinical and applied settings (2nd ed.). New York,
doi:10.1038/sj.ejcn.1602024 NY: Oxford University Press.
Keller, F. S., & Schoenfeld, W. N. (1950). Principles of
Iversen, I. H. (1988). Tactics of graphic design: A
psychology. New York, NY: Appleton-Century-
review of Tufte’s The Visual Display of Quantitative
Crofts.
Information [Book review]. Journal of the
Experimental Analysis of Behavior, 49, 171–189. Kennedy, C. H. (2005). Single-case designs for educational
doi:10.1901/jeab.1988.49-171 research. New York, NY: Pearson Allyn & Bacon.

30
Single-Case Research Methods

Kimball, J. W. (2009). Comments on Spreckley and Boyd (Ed.), Single-subject research: Strategies for evaluating
(2009). Science in Autism Treatment, 6, 3–19. change (pp. 101–165). New York, NY: Academic
Press.
Kratochwill, T. R., & Brody, G. H. (1978). Single sub-
ject designs: A perspective on the controversy Pavlov, I. P. (1906). The scientific investigation of the
over employing statistical inference and impli- psychical faculties or processes in higher animals.
cations for research and training in behavior Science, 24, 613–619. doi:10.1126/science.24.620.613
modification. Behavior Modification, 2, 291–307.
Pavlov, I. P. (1927). Conditioned reflexes (G. V. Anrep,
doi:10.1177/014544557823001
Trans.). London, England: Oxford University Press.
Loeb, J. (1900). Comparative physiology of the brain and
Pavlov, I. P. (1928). Lectures on conditioned reflexes:
comparative psychology. New York, NY: Putnam.
Twenty-five years of objective study of the higher nervous
doi:10.5962/bhl.title.1896
activity (behaviour) of animals (Vol. 1). New York, NY:
Lovaas, O. I. (1987). Behavioral treatment and normal International Publishers. doi:10.1037/11081-000
educational and intellectual functioning in young Pavlov, I. P. (1955). Nobel speech delivered in Stockholm
autistic children. Journal of Consulting and Clinical on December 12, 1904. In K. S. Koshtoyants (Ed.),
Psychology, 55, 3–9. doi:10.1037/0022-006X.55.1.3
Copyright American Psychological Association. Not for further distribution.

I. P. Pavlov: Selected works (pp. 129–148). Honolulu,


Lowe, F. (2010, September). Can behavior analysis change HI: University Press of the Pacific.
the world? Paper presented at the Ninth International Schmidt, R. A., & Lee, T. D. (2005). Motor control and
Congress on Behavior Studies, Crete, Greece. learning: A behavioral emphasis. Champaign, IL:
Matson, J. L., Sevin, J. A., Fridley, D., & Love, S. R. Human Kinetics.
(1990). Increasing spontaneous language in three Shull, R. L. (1991). Mathematical description of operant
autistic children. Journal of Applied Behavior Analysis, behavior: An introduction. In I. H. Iversen & K. A.
23, 227–233. doi:10.1901/jaba.1990.23-227 Lattal (Eds.), Experimental analysis of behavior
McDougall, D., Hawkins, J., Brady, M., & Jenkins, A. (Vol. 2, pp. 243–282). New York, NY: Elsevier.
(2006). Recent innovations in the changing criterion Sidman, M. (1960). Tactics of scientific research:
design: Implications for research and practice in spe- Evaluating experimental data in psychology. New
cial education. Journal of Special Education, 40, 2–15. York, NY: Basic Books.
doi:10.1177/00224669060400010101
Sidman, M. (1971). Reading and auditory-visual equiva-
Montee, B. B., Miltenberger, R. G., & Wittrock, D. lences. Journal of Speech and Hearing Research, 14,
(1995). An experimental analysis of facilitated com- 5–13.
munication. Journal of Applied Behavior Analysis, 28,
189–200. doi:10.1901/jaba.1995.28-189 Sidman, M. (1981). Remarks. Behaviorism, 9, 127–129.
Morgan, D. L., & Morgan, R. K. (2009). Single-case Skinner, B. F. (1938). The behavior of organisms: An exper-
research methods for the behavioral and health sciences. imental analysis. New York, NY: Appleton-Century-
Los Angeles, CA: Sage. Crofts.
Morris, E. K. (2009). A case study in the misrepresenta- Skinner, B. F. (1953). Science and human behavior. New
tion of applied behavior analysis in autism: The York, NY: Macmillan.
Gernsbacher lectures. Behavior Analyst, 32, 205–240. Skinner, B. F. (1956). A case history in scientific method.
Neef, N. A., & Iwata, B. A. (1994). Current research on American Psychologist, 11, 221–233. doi:10.1037/
functional analysis methodologies: An introduction. h0047662
Journal of Applied Behavior Analysis, 27, 211–214. Skinner, B. F. (1966). Some responses to the stimulus
doi:10.1901/jaba.1994.27-211 “Pavlov.”Conditional Reflex, 1, 74–78.
O’Neill, R. E., McDonnell, J., Billingsly, F., & Jenson, W. Skinner, B. F. (1969). Contingencies of reinforcement. New
(2010). Single case designs in educational and commu- York, NY: Appleton-Century-Crofts.
nity settings. New York, NY: Merrill.
Skinner, B. F. (1978). The ethics of helping people. In
Onghena, P., & Edgington, E. S. (2005). Customization B. F. Skinner (Ed.), Reflections on behaviorism and
of pain treatment: Single-case design and society (pp. 33–47). Englewood Cliffs, NJ: Prentice
analysis. Clinical Journal of Pain, 21, 56–68. Hall.
doi:10.1097/00002508-200501000-00007
Skinner, B. F. (1979). The shaping of a behaviorist. New
Paré, W. P. (1990). Pavlov as a psychophysiological York, NY: Knopf.
scientist. Brain Research Bulletin, 24, 643–649.
doi:10.1016/0361-9230(90)90002-H Spreckley, M., & Boyd, R. (2009). Efficacy of applied
behavioral intervention in preschool children with
Parsonson, B. S., & Baer, D. M. (1978). The analysis and autism for improving cognitive, language, and adap-
presentation of graphic data. In T. R. Kratochwill tive behavior: A systematic review and meta analysis.

31
Iver H. Iversen

Journal of Pediatrics, 154, 338–344. doi:10.1016/j. Tufte, E. R. (1990). Envisioning information. Cheshire,
jpeds.2008.09.012 CT: Graphics Press.
Steege, M. W., & Mace, F. C. (2007). Applied behavior Ullman, L. P., & Krasner, L. (1965). Case studies in behav-
analysis: Beyond discrete trial teaching. Psychology in ior modification. New York, NY: Holt, Rinehart &
the Schools, 44, 91–99. doi:10.1002/pits.20208 Winston.
Sulzer-Azaroff, B., & Mayer, G. R. (1991). Behavior Van Houten, R., Axelrod, S., Baily, J. S., Favell, J. F.,
analysis for lasting change. Fort Worth, TX: Harcourt Foxx, R. M., Iwata, B. A., & Lovaas, O. I. (1988).
Brace. The right to effective behavioral treatment. Journal of
Applied Behavior Analysis, 21, 381–384. doi:10.1901/
Taub, E., Uswatte, G., King, D. K., Morris, D., Crago, jaba.1988.21-381
J. E., & Chatterjee, A. (2006). A placebo-controlled
trial of constraint induced movement therapy for Watson, J. B., & Rayner, R. (1920). Conditioned emo-
upper extremity after stroke. Stroke, 37, 1045–1049. tional reactions. Journal of Experimental Psychology,
doi:10.1161/01.STR.0000206463.66461.97 3, 1–14. doi:10.1037/h0069608
Thorndike, E. L. (1911). Animal intelligence. New York, Wheeler, D. L., Jacobson, J. W., Paglieri, R. A., &
Copyright American Psychological Association. Not for further distribution.

NY: Macmillan. Schwartz, A. A. (1993). An experimental assessment


of facilitated communication. Mental Retardation, 31,
Thorndike, E. L. (1927). The law of effect. American Journal 49–59.
of Psychology, 39, 212–222. doi:10.2307/1415413
Wolpe, J. (1958). Psychotherapy by reciprocal inhibition.
Todes, D. P. (2002). Pavlov’s physiological factory: Stanford, CA: Stanford University Press.
Experiment, interpretation, laboratory enterprise.
Baltimore, MD: Johns Hopkins University Press. Wood, J. D. (2004). The first Nobel prize for inte-
grated systems physiology: Ivan Petrovich Pavlov,
Tufte, E. R. (1983). The visual display of quantitative infor- 1904. Physiology, 19, 326–330. doi:10.1152/
mation. Cheshire, CT: Graphics Press. physiol.00034.2004

32

You might also like