Sie sind auf Seite 1von 19

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/283789440

A Research Pioneer’s Wisdom: An Interview with Dr. Murray Sidman

Article  in  European Journal of Behavior Analysis · December 2010


DOI: 10.1080/15021149.2010.11434342

CITATIONS READS

6 113

1 author:

Per Holth
Oslo Metropolitan University
74 PUBLICATIONS   547 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Philosophy of Science: Category Mistakes and Explanations of Behavior View project

Improved procedure for the establishment of conditioned reinforcers View project

All content following this page was uploaded by Per Holth on 26 March 2018.

The user has requested enhancement of the downloaded file.


EUROPEAN JOURNAL OF BEHAVIOR ANALYSIS 2010, 12, 181 - 198 NUMBER 2 (WINTER 2010)
181

A Research Pioneer’s Wisdom: An Interview


with Dr. Murray Sidman
Per Holth
Akershus University College

One of the most outstanding contributors to modern behavior analysis answers questions regarding the
history of behavior analysis, behavior analysis and psychology, tactics of scientific research, philosophy
of science, behavior analysis and important social issues, research on coercion, and education. He also
reflects upon the usefulness of randomized controlled studies, his reservations against relational frame
theory, and on the future of behavior analysis.
Key words: Behavior analysis and psychology, tactics of scientific research, description vs.
explanation, social issues

Murray Sidman completed his Ph.D at Co- to its survival in a world of increasing competition
lumbia University in 1952. His principal advi- for ever more limited resources.
sors, Fred S. Keller and W. N. Schoenfeld, had Per Holth is Professor of Behavior Analysis
strong assists from Ralph Hefferline, Clarence at Akershus University College in Norway.
Graham, and a small group of fellow graduate Per received his Ph.D. from the University of
students. After that, he spent nine years in the Oslo. His doctoral research was conducted
exciting and productive interdisciplinary envi- in collaboration with Erik Arntzen and was
ronment of the Neuropsychiatry Division at the concerned with stimulus equivalence – very much
Walter Reed Army Institute of Research, work- inspired by the basic experimental work in that
ing with and learning from Joseph V. Brady and field by Dr. Sidman. At Akershus University Col-
many other innovative and productive research- lege, he teaches a variety of themes in behavior
ers. He then joined the Neurology Service of the analysis, including experimental and applied
Massachusetts General Hospital for another nine behavior analysis, and theoretical and conceptual
years. His human and nonhuman behavioral issues. Akershus University College now offers
research laboratories moved eventually to the E. education in behavior analysis at all levels from
K. Shriver Center and finally to Northeastern Bachelor through Master and Doctoral programs.
University, where he remained as Professor of
Psychology until he retired from Academe, con- Important contributions
tinuing his research at the New England Center
for Children. Although retired from there in Holth: You have made major contributions
2001, he continues research and writing. One within several different areas of a science of behav-
outcome of his lifetime of research is his convic- ior. You are one of very few behavior analysts who
tion that extending experimental results out of have had behavioral phenomena named after you.
the laboratory not only adds an intrinsically valu- We have “Sidman avoidance”, and people have
able dimension to basic research, but is essential occasionally used the term “Sidman equivalence”.
A third major area of contribution is your work
Notes: 1. I thank Iver H. Iversen for his helpful suggestions
on the interview questions. on “Tactics of scientific research” (e.g., Sidman,
2. Send correspondence Per Holth, Faculty of Behavioral Sci- 1960). Which one do you consider most impor-
ence, Akershus University College, PO box 423, 2001 Lillestrøm,
Norway; e-mail: per.holth@hiak.no tant of all your contributions?
181
182 Per Holth

Sidman: I think future developments will Entering psychology


prove or disprove judgments about the impor-
tance of any contributions. My own current Holth: What were some of the things that
opinion is that the future will judge “Tactics” led you to enter the field of psychology, and
to be my own most important contribution. It then, behavior analysis?
is not that the book introduced new research Sidman: At the time I was working for my
methodologies or principles. I have always undergraduate and graduate degrees, few young
thought of the book as descriptive rather than in- people thought of psychology as a profession
novative, as an account of the ways good research they might someday enter. Even most of those
has always been carried out. The illustrations, of already in the field had started in some other
course, came from behavior analytic research area before they became interested in one or
but I have always felt that the principles being more problems that turned out to be dealt with
illustrated were common to other sciences also. by psychologists. That was true of me, also. Like
(As behavior analysis has entered new research many idealistic young people, my ambition
areas, of course, new illustrations of research was to change the world but I felt that people
methodology become appropriate, and the lack made the world the way it was, so if it was to
of such new examples is possibly a major criti- be changed, people would have to be changed.
cism of the book as it now stands.) I became interested in the sciences because I
I am going to have more to say in a forth- thought it was there I could learn all about how
coming publication about what follows, but a to change people’s ways of dealing with their
number of respected colleagues have told me environments. I even took an extra undergradu-
that they wished I had said more in the book ate year of just science courses, looking for that
about the philosophy of science and had not science which would tell me what I had to know.
confined myself just to scientific methodology. I found it nowhere. Even in the lab courses, few
At first, I accepted this absence of philosophical of the experiments actually worked out the way
discussion as a justified criticism of the book. I the lab manuals said they would.
have always thought of myself as a reasonably And then I found myself in Keller and
capable experimenter but as only a novice in Schoenfeld’s pioneer introductory psychology
philosophy. I have long admired philosophers course that was based on B. F. Skinner’s work as
of science and at one time, was well read in summarized in his “The Behavior of Organisms
the area because I felt that philosophers have (1938).” There, in the laboratory sessions, my
much to tell me about my own place in the furry little subject did everything it was sup-
world of ideas and intellect. In the course of posed to do when I did what the lab manual
writing “Tactics,” however, I realized that most told me to do. It was not only my first science
philosophers of science had little notion of what course in which the lab results came out just
really went on in the design and performance as predicted, but it was the first time I found
of scientific experiments. Later, though, as my myself changing another organism’s behavior
own thinking matured, I came to realize that an not just by telling it in words what I wanted
understanding of and an acquaintance with the it to do but by telling it nonverbally, simply
methodological details of any experiment was by changing the effects its behavior had on the
critical to the understanding of the experiment’s environment, and by changing the environ-
validity, generality, and overall significance. I ment in which it could produce those effects.
now believe, therefore, that an understanding I had found what I was looking for and I kept
of research methods is a philosophic basic, a on with it, doing what interested and excited
requirement for any philosopher who wishes to me until I finally received a Ph.D., discovered
evaluate a particular science’s place in any more I had become a psychologist, and had to find a
general world view. That, then, is the outline of job somewhere in that field.
a major reason why I judge my “Tactics” book to As for entering the field of behavior analysis, I
be the most important of my own contributions. never did enter it because it did not exist at the time.
A Research Pioneer’s Wisdom: An Interview with Dr. Murray Sidman 183

It just grew around me and enveloped me Keller achieved the same results in an en-
within it as it grew. Now, mostly because the tirely different way. He was positive reinforce-
field has extended out of the laboratory and has ment personified. Instead of challenging us to
come to deal effectively with important practi- think in new ways, and making us justify our
cal problems, young people are coming to see thinking, he gradually shaped our thinking,
behavior analysis as an occupation worthy of using positive reinforcement all the way. His
becoming trained in and engaging in as a lifetime shaping technique, however, was not the simple
profession. one that we learned in the laboratory, in which
Holth: What were the roles played by Keller we gradually taught our subjects to perform
and Schoenfeld? the exact behavior we wanted from them.
Sidman: As I noted before, Keller and He shaped us not to say what he wanted us to
Schoenfeld’s introductory course was what really say, but rather, to come up with a more pro-
got me started. I also took other courses with ductive way of thinking that even he could not
them—verbal behavior, motivation, stimulus necessarily have known beforehand. We would
discrimination—and they are among my most go out of his office happy that he liked our idea
memorable school experiences. I was also greatly but not realizing that the idea we were going out
influenced by some of their earlier students. with was not the same one we had come in with.
Most noteworthy among those were Ralph Hef- I have tried at different times to imitate both
ferline, whose work led to my own dissertation of them, but I could never really match either
on avoidance behavior, and Jim Dinsmoor, who of their teaching techniques. Still, I think that
was always available with sensible advice for because of them, I have been a better teacher
the solution of experimental problems. Several than I would have been otherwise.
other students also were most important to my
own development, most notably Donald Cook, Research on coercion and avoidance
Michael Kaplan, and Carl Crocetti. The four of
us not only did experiments together and got Holth: In addition to your basic research
together outside for discussion and argument, on avoidance, you wrote a book about Coer-
but probably made life a bit hectic for Keller and cion and its fallout. Avoidance still seems to be
Schoenfeld. We were continually cornering them prevalent in human behavior and a huge part
in the corridors and in their offices with questions of psychopathology. Yet, there is very little basic
and requests for support of experiments. They research going on in this area. Do you think
were always available to us. that work in this area is still needed both with
Except for their interest in the same subject animals and with humans?
matter—the behavior of organisms—their teach- Sidman: I certainly do think that work on
ing styles were poles apart. Schoenfeld’s most coercive control should continue, with both
frequent question of us was, “How do you know human and nonhuman subjects. I am aware
that?” He continually questioned our most basic that because of the fairly generally accepted
assumptions, forcing us to think through and to conclusion that coercive control of behavior is
justify anything we said. He was always wondering undesirable, some have come to the conclusion
how we came to know things. I remember when that research on coercive control is no longer
he asked once, “How do you suppose people justified. Although my book strongly supports
learned that pregnancy was the result of sexual the notion that we should avoid coercive control
intercourse that had taken place nine months in everyday life, I strongly disagree with the
before?” His kinds of questions made life difficult position that research in the area is undesir-
for many students and many of them disliked his able. All research includes two activities: asking
courses intensely. I was enthralled with them, questions and attempts to answer questions.
exhiliarated not just by learning the answers to To proscribe the asking of questions—any
questions but by learning how to ask questions of questions—is, to me, an unthinkable option.
nature and of my own understanding of nature. Freedom of thought is essential to all progress.
184 Per Holth

And then, once asked, questions must be Sidman: I believe that both basic and ap-
answered, if possible. Some methods for an- plied behavior analysis will flourish for a long
swering questions may, of course, be undesir- time. About psychology, I am not so sure. I
able—for example, it is not permissible to find cannot believe that some areas of psychology
out the effects of torture by torturing people or will continue to exist.
animals—and good answers to some questions Clinical psychology, for example, is a sham
must await the discovery of allowable investiga- and a disgrace. Whatever success it has achieved
tive techniques, but the questions themselves has been accomplished by individuals who have
must always be allowed. themselves learned useful techniques from ex-
Holth: What do you think it would take to perience. They have no scientific background
get the proper research going? and produce little useful clinical data.
Sidman: For me, the proper research Physiological psychology should be a flour-
is any research in which good methods are ishing field but for the most part researchers
proposed to answer important questions. in that area ask the wrong questions about the
I do not think the criteria are different for behavior whose physiological and neurological
research on coercion than on any other set underpinnings they are trying to understand.
of problems. It is up to researchers to justify The field maintains itself because it has no
their research proposals to potential support- competitors. Behavior analysis, unfortunately,
ers—granting agencies, private foundations, has done little in that area but when it does
individuals or organizations with particular wake up to its potential contributions, it is
concerns—just as always. Research support likely to place traditional physiological psychol-
in general has become difficult to obtain. Re- ogy in the same disposable category as clinical
searchers must learn to argue persuasively for psychology.
their particular areas of concern. Educational psychology places no value on
scientific data which have shown that learn-
Behavior analysis and psychology ing is engendered by its consequences and
that learning can be errorless. It continues to
Holth: In the Coercion book, you wrote place its faith in trial-and-error learning and
that: in averaged continuous learning curves. When
educational psychology accepts the finding
“The science of behavior analysis had its roots in phi- that the true science of learning is the science
losophy, then separated itself off as a branch of the of teaching, that teachers rather than pupils
emerging discipline of psychology, and is now in the are responsible when learning takes place or
process of disengaging itself from psychology. The
parent has not yet let go (nor, for that matter, has the fails to take place, educational psychology will
grandparent), and struggles to maintain its adminis- then leave psychology and take its place within
trative dominance within Academe, but the lines of behavior analysis.
intellectual fracture are clear. Psychology as its name So yes, I do believe that psychology and
suggests, is the science of mind. Behavior analysis is behavior analysis are separate sciences, but I do
the science of behavior” (Sidman, 1989, pp. 4-5). not believe that all areas of each can flourish
side by side. They differ in philosophy, in the
In the Equivalence book, you mentioned kinds of questions they ask, in the kinds of data
that you decided not to publish equivalence they value, and in investigative methodology. A
research in traditional psychology journals (you main reason I did not publish my equivalence
realized that it was too late). You described research in traditional psychology journals was
psychology and behavior analysis as different that I did not think the audience would be
sciences – with different methodologies. Do receptive. The single-subject methodology and
you picture psychology and behavior analysis my inductive rather than deductive approach
as two sciences flourishing side by side in the were pretty much proscribed in the standard
future? psychology literature.
A Research Pioneer’s Wisdom: An Interview with Dr. Murray Sidman 185

Education scientific development but at the same time


has kept it from gaining understanding and
Holth: You have been quite critical of the popularity among the general public and other
field of education. In the equivalence book, you scientists. Let me explain.
wrote that: “I know of no service area except B. F. Skinner proposed Radical Behaviorism
education in which practitioners habitually dis- as the philosophical foundation for his way of
regard fundamental research” (Sidman, 1994, looking at the world of behavior. A basic feature
p. 532). If you were in the position of an om- of Radical Behaviorism is its avoidance of what
nipotent secretary of state for education, how Skinner called explanatory fictions. These are
would you go about to change this to make sure terms like “mind,” “representation,” “meaning,”
that students benefit optimally from a science “knowledge,” “symbol,” “referent,” “memory,”
of behavior? and many others. They are called explanatory
Sidman: If I really were omnipotent, even fictions not only because they are invisible but
more so than a secretary of education could because they explain nothing. Instead, they
possibly be, I would first wipe out most schools move the search for explanation from observ-
of education and then restart each of them able to unobservable variables, structures, and
when they could attract behavioral scientists processes. Many of the successes of experimental
who were capable of applying behavior analytic and applied behavior analysis stem from this
principles both in the basic learning laboratory philosophic background that calls for the rejec-
and in the classroom. That, of course, is never tion of explanatory fictions.
going to happen. I would, however, support Unfortunately, explanatory fictions remain a
the principle—one, incidentally, that I now see feature of our everyday vocabulary. For example,
being enunciated by the current Secretary of we say that someone has a brilliant mind, or that
Education in the United States—that teachers we get something done when we put our mind
should be evaluated by the performance of their to it, or that we store knowledge in our mind,
students. I would, however, also require that or that memory is an important feature of the
teachers receive training in how to increase the mind, and so on. When we say to nonbehavior-
level their pupils’ performances reach. If they ists that behavior analysis rejects the concept of
are going to be judged by their pupils’ successes, mind, the response is likely to be that behavior
then they have themselves to be taught how to analysts are interested only in mindless, or
teach successfully. Most teachers are justifiably trivial, behavior. Behaviorists have never popu-
afraid of the requirement that they teach suc- larized the notion that they are interested in all
cessfully because they have never been trained behavior but that they find explanations in other
to do that. Along with their pupils, teachers, places than in a mysterious thing called mind.
too, deserve better. Their own educational sys- In both popular speech and in nonbehavioral
tem has to be drastically revised. Basic teacher science, much of what we do is considered to
training, then, is the area in which I believe that be the result of memory processes: A person
improvement is most needed. learns rapidly because she has a good memory;
a person frequently loses his keys because he has
Past failure of behavior analysis to gain a poor memory; I have no memory for names;
popularity memory is said to deteriorate with age; and so
on. When behaviorists argue that there is no
Holth: Behavior analysis clearly has not such thing as memory, they are dismissed as
gained the ground and popularity that some proponents of an empty organism. Behaviorists
of us think it deserves. What do you think Be- have never made it clear that they are very much
havior Analysts could have done better in order interested in behavior that we call remembering
to promote their science over the last 50 years? or forgetting, and that we explain such behavior
Sidman: The language of Behavior Analysis in other ways than by reference to an invisible
has been one of its strengths with respect to its cause called memory.
186 Per Holth

Our insistence on purity of terminology I sometimes have the feeling, however, that
when we speak to each other has unfortunately sloppy “pilot’ studies are more common in ap-
carried over into our communication with oth- plied behavior analytic studies. The excuse there
ers, with the result that nonbehaviorists, both is often that the uncontrolled environments
nonscientific and scientific, have concluded that in which most applied studies must be carried
behaviorism has nothing of interest to say to out do not permit the experimental rigor that
them. Such a conclusion has not been unreason- basic experimentation demands. It is argued
able. After all, other sciences have recognized that uncontrolled studies must be carried out in
that popular language refers to events that are so order to get some idea of whether or not some
interesting and even important in everyday life hypothesis is worth the effort of applying the
that nonscientists have developed their own ter- kinds of controls that make valid data possible
minology to describe and explain them. And so, even in uncontrolled environments. Whatever
physicists do not scold us for talking about sun- the excuse, however, sloppy data are deceptive
sets but instead say, “Here is what is really going data. For the purpose of determining whether
on when we see the sun set.” Behavior analysts, to pursue some line of research, such data are
instead of telling people that there is no such just as likely to send one off on a wild goose
thing as a mind to be looked for, should instead chase as a wild guess would have done.
be telling them how to create a mind—which Holth: A very thought provoking statement
is really what behavior analysts are doing all the that you made in Tactics was that “It is simply
time. They should not be telling people that not true that the larger the group, the greater
there is no such thing as memory but should is the generality of the data” (Sidman, 1960, p.
instead show them what factors determine 47). To what extent do you feel that this point
whether we remember or forget something. got through to the researchers it was meant for?
Instead of insisting that representations do Sidman: I think that this point has be-
not exist, we should be pointing out how the come one of the basic foundations of both
formation of equivalence relations underlies experimental and applied behavior analysis.
representations and symbols. Ordinary language The substitution of systematic replication
develops from ordinary experience. No science with individuals for large group averages has
can ignore that experience and hope to survive. proven itself to produce data of great generality.
Unfortunately, the success of well-controlled
Tactics of Scientific Research individual studies in producing data of great
generality has led many to conclude that large
Holth: Let us turn to some issues raised in group studies have no value in the study of
your book Tactics of Scientific Research (Sidman, behavior. I do not believe this is correct (see
1960). You mentioned that, below).
Holth: What is your opinion on the move-
“Pilot studies that are not carried out in as rigorous a ment towards evidence-based treatment and its
fashion as possible have neither positive nor negative emphasis on the randomized controlled trials
value as indicators of subsequent results. A sloppy (RCT) with large groups of subjects as the gold
experiment is a poor experiment, and can never be standard?
justified by labeling it ‘pilot’” (p. 218). Sidman: Evidence-based treatment is, of
course, highly desirable. Treatment that is not
Do you think that the practice of designating based on evidence is so often just a scam that it
studies as “pilot” is still used in an attempt to should always at least be looked at with suspi-
justify sloppy experimenting? cion. What, however, constitutes evidence for
Sidman: No, I think that is a fairly rare the effectiveness of a treatment? As I noted be-
occurrence in experimental behavior analysis. fore, replicated studies with individuals rather
I would say, rather, that experimental rigor than groups have demonstrated themselves
is a hallmark of behavioral experimentation. again and again to produce generalizable data.
A Research Pioneer’s Wisdom: An Interview with Dr. Murray Sidman 187

If a treatment works with just one person, then Without a significant average improvement to
it will work for others also. When it is found start with, they will not be interested in finding
not to work with some others, then instead of out why the treatment works for some individu-
discarding the treatment on the basis of the als and not for others or even in knowing which
average number of successes and failures, the individual workers fall into each class. When
productive course is to investigate the reasons the aim is to produce large overall effects, the
for the differences; why does the treatment work effects of the treatment on any individual are
for some and not for others? of no concern. In such situations, statistical
Although behavior analysts usually work analysis is appropriate. I believe that one reason
with individual responses and individual sub- for the failure of behavior analysis to achieve
jects, people in general are interested in account- wider recognition is the absence of data indi-
ing for and changing large classes of behavior, cating how effective it is in large populations.
not restricted operants. They are also interested What about the use of randomized con-
in changing or accounting for large populations, trol trials to determine whether a treat-
not particular individuals. Therefore, group data ment is effective? The use of control
are more meaningful to them. groups in a statistical analysis of treat-
Behavior analysts who work with individu- ment effects raises a number of problems.
als do not always realize that some questions One is the difficulty in finding control groups
about the effectiveness of a treatment do require with the same relevant properties as the experi-
population statistics. For example, people who mental group. This is often impossible and as
are in the position of having to deal with practi- a result, many large-population studies receive
cal community and world problems—leaders unfavorable interpretations. A second set of
in government, the military, in diplomacy, in problems stems from ethical considerations.
education, and so on—do not want to be told Can one justify the failure to apply a treatment
that some treatment or procedure has been to those whom it would have helped if it were
shown to work with a few individuals. They are found to be effective?
not concerned with any particular individual Both of these kinds of problems can be
at all and are not disturbed if some individuals avoided by using the same population both
fail to show the desired behavior change when as the experimental and control group. Before
the treatment is applied. Rather, they want to applying the treatment, observe the groups
know how much of the population will show and take the relevant measurements. Then ap-
the desired effect, or how large the effect will be ply the treatment and observe whether or not
within the population or organization in which those measurements change. This of course is
the treatment is applied. a technique we have learned from our studies
For example, suppose we convince a com- with individuals. Establish baselines and then
pany to eliminate coercive control—punish- determine whether or not a particular treat-
ment and threats of punishment for undesirable ment changes those baselines.
behavior—and substitute consistent positive
reinforcement for desirable behavior. The Definition of punishment
company will be concerned to know about
any resulting changes in average work pro- Holth: Two different definitions of punish-
ductivity, in the overall rate of absenteeism, ment figure in the behavior-analytic literature.
in the total number of altercations between According to Skinner’s (1953) definition,
workers and supervisors, and in the profit and punishment is a procedure in which responses
loss statements reported by the accounting de- are followed by either (a) the removal of a
partment. If these overall measurements show positive reinforcer, or (b) the presentation
consistent improvements, the company may of a negative reinforcer (or aversive stimu-
or may not try to do something about individ- lus). The second definition of punishment
ual workers whose behavior failed to improve. was advocated by Azrin and Holz (1966).
188 Per Holth

They defined punishment in the same way as ahead of time whether a particular behavioral
reinforcement, only with the opposite effect: consequence is a punisher; we first have to
Whereas reinforcement is defined by an in- observe whether it produces a decrease in the
creased response rate, punishment is defined by frequency of that behavior. If it does, then we
a decreased rate. The latter definition now seems explain the decrease as the effect of punishment.
to have become standard in most behavior- The same inelegant circularity exists in the iden-
analytic texts (cf., Holth, 2005). However, you tification of punishers as in the identification
have kept Skinner’s original definition of pun- of reinforcers.
ishment. In the Coercion book, for instance, Because Skinner confines the inelegance
you wrote that: of circularity to the concept of reinforce-
ment, whereas Azrin and Holz add the same
“Reinforcement differs in an important way from inelegance to the concept of punishment, I
punishment. We define reinforcers – positive or prefer Skinner’s more parsimonious definitions.
negative – by their special effect on conduct; they
increase the future likelihood of actions that they fol- What is the role of parsimony in science? A
low. But we define punishment without appealing to preference for parsimony derives originally
any behavioral effect; punishment occurs whenever from popular aesthetics. For example, Chi-
an action is followed either by a loss of positive or nese fortune cookies tell us, “There is beauty
a gain of negative reinforcers. This definition says in simplicity” and “The most valuable of all
nothing about the effect of a punisher on the action
talents is that of never using two words when
that produces it” (Sidman, 1989, p. 39).
one will do.” This wisdom made its way into
science with the principle of Occam’s razor:
What are your main reasons for preferring “Entities should not be multiplied unneces-
Skinner’s original definition of punishment? sarily.” Another form of this principle is, “The
Sidman: My own personal reasons for explanation requiring the fewest assumptions
preferring Skinner’s original definition of pun- is most likely to be correct.” It has always been
ishment have, first, to do with its parsimony. my impression that whenever two hypotheses,
As we all know, reinforcing events are identified theories, or methodologies seem equally
by observations of a particular consequence capable of explaining or predicting particular
of behavior—an increase in the future prob- data, then the more parsimonious alternative
ability of that behavior. But then, to explain is the one that eventually wins out. Of the two
a behavioral increase in behavior, we appeal defintions of punishment, Skinner’s is clearly
to reinforcement. The circularity here is evi- the more parsimonious.
dent; because we cannot know ahead of time My second reason for rejecting the definition
whether a particular behavioral consequence is of punishment as a behavioral consequence that
a reinforcer, we first have to observe whether it produces a decrease in the probability of some
does produce an increase in the frequency of particular behavior is that such a definition
that behavior. Skinner restricts that inelegant seems to apply also to instances that I find it dif-
circularity to the definition of reinforcement; it ficult to think of as punishment. For example,
does not apply to his definition of punishment. when pupils decrease their error frequency after
To identify a punisher he reqires no particular being exposed to additional instruction, do we
behavioral outcome. classify this as punishment? When one stops
The Azrin and Holz definition of punish- overeating after observing weight reductions on
ment, however, which seems to have become a scale, has one been punished? In general, when
generally adopted without much discussion, we get a person to stop one form of behavior
suffers from the same circularity as the defini- simply by providing positive reinforcement for
tion of reinforcement. Punishment is identi- some other behavior, have we engaged in pun-
fied by observations of a particular behavioral ishment? Such instances and many others like
consequence—a decrease in the future prob- them make me very uneasy about accepting the
ability of that behavior. Again, we cannot know Azrin-Holz definition of punishment.
A Research Pioneer’s Wisdom: An Interview with Dr. Murray Sidman 189

Stimulus equivalence equivalence relations take part. The second


development handles this common observa-
Holth: You once wrote, when asked to reply tion. That development was my own suggestion,
to Horne and Lowe’s (1996) criticisms of your at present confirmed by at least one published
equivalence position, that you did not hold a experiment, that equivalence classes include the
position, except the position that “data rather responses as well as the stimuli in a reinforcement
than debate will show the way” (Sidman, 1997, contingency. A corollary of this suggestion is that
p. 258). What do you think are the current differential responses will facilitate the develop-
most important research questions that you ment of equivalence relations. Such differential
would want to be able to answer with respect responses are often words or phrases. I do not
to stimulus equivalence? believe that this suggestion, if proven, would dis-
Sidman: When I referred here to data as prove the Horne and Lowe theory that language
opposed to debate, I was not thinking of any is necessary for equivalence relations. Instead, it
particular research questions that were out- would enfold that theory into my own.
standing at the time. I was referring just to my These considerations do not, of course,
preference for using research to decide which demonstrate either that equivalence relations
theory was the best rather than attending to are necessary for language or that language is
the seemingly endless verbal sparring about primary. I do not believe that either alternative is
which theory one should adopt. I did publish correct. Rather, I consider it a pseudo question.
a paper (Sidman, 2000), in which I suggested Although it is clear that language is not necessary
a number of experiments that would serve to for equivalence relations, it does play a role. It is
test the theory I had proposed, with perhaps the equally clear that language is much more compli-
most far reaching being experiments to test the cated a phenomenon than equivalence relations
prediction that in a reinforcement contingency, by themselves can explain. And yet, I do not
the resulting equivalence class would include the believe that language would be possible without
responses along with the stimuli. My general equivalence relations. Consider, for example, just
point, however, that “data rather than debate simple nouns, verbs, and adjectives. When people
will show the way” simply reflected my con- ask what such words mean (in this instance, what
viction that competing theories are eventually are the dictionary definitions?), the answers show
settled by the directions in which data lead the them to be equivalent to some thing, action, or
field rather than by verbal persuasion. property of things. Equivalence relations play a
Holth: There has been some speculation necessary role in language but that role, along
about relations between stimulus equivalence with the relevance of equivalence to other factors
and language, such as whether one is necessary essential to language, has yet to be fully explicated.
for the other. Could you explain how your
own thinking about these issues has evolved Limits to generality
over time?
Sidman: Two major developments have Holth: In the Coercion book, you wrote
influenced my thinking about equivalence and that“any experimental science must contain self-
language. The first is the series of convincing checking mechanisms for specifying the limits to
experiments from Schusterman’s laboratory in the generalizability of its methods and findings”
which sea lions formed equivalence relations (Sidman, 1989, p. 53). What do you think are
(e.g., Kastak, Schusterman, & Kastak, 2001). the most important things that need to be done
Complex as some of these experiments were, in order to specify the limits to the generalizability
language was clearly not involved. of stimulus equivalence?
On the other hand, everyday experience tells Sidman: One can view generalizability in
us that language is often involved in equivalence several contexts. For example, an early question
relations. Indeed, with humans, language is concerned the generalizability of equivalence
perhaps the most important area in which relations to the behavior of nonhuman animals.
190 Per Holth

This is an area in which considerable research “Having applied the set theory definition of the
remains to be done. equivalence relation to our behavioral observations.
A second question concerns the generaliz- . ., we now find that the emergent conditional
discriminations in our original experiment are no
ability of equivalence to stimuli of different longer mysterious. All those performances are
sensory modalities. This question has been to be expected if the baseline contingencies have
resolved; equivalence relations do not depend established an equivalence relation.”
on any particular stimulus modalities.
A third question concerns the generalizabil- Hence, is there some way still in which you
ity of the methods for establishing equivalence consider equivalence to be explanatory rather
relations. Although the original experiments than just descriptive?
used four-term contingencies (conditional Sidman: It is not easy to draw distinctions
discriminations) to establish and demonstrate between description and explanation. Per-
equivalence relations, it has become fairly clear haps the most general example would be the
that three-term (simple discrimination) and comparison of theory and experimental data.
probably two-term (response-reinforcer and Theory is clearly a way to explain data. Data
stimulus-reinforcer) contingencies suffice to by themselves, however, provide no explana-
establish equivalence relations. Unequivocal tion of why they are predictable. In what sense
demonstrations that equivalence relations is the way my collaborators and I talk about
have been established still, however, do require equivalence relations an explanatory theory
conditional discriminations. Mere stimulus and in what sense does it simply provide a way
substitutability does not suffice because it is to describe the observations?
possible to show that presumed establishing Theory plays a role in two aspects of our
procedures bring about the interchangeability formulation. The first is the question of where
of some but not all stimuli in a reinforcement equivalence relations come from. I have
contingency. For example, symmetry may be presented a detailed answer to this question
demonstrated even in the absence of transitivity. elsewhere (Sidman, 2000) that is too lengthy
And so, we find that while the original condi- to be repeated here. Briefly, that theory holds
tional-discrimination technique for establishing that equivalence relations arise directly from
equivalence relations is generalizable, other reinforcement contingencies, that each ele-
techniques can accomplish the same effects. ment of a reinforcement contingency becomes
Conditional discriminations, however, remain a member of an equivalence class. I presented
the only technique for testing definitively a description of many experiments that have
whether or not any establishing procedure has already supported this theory and many oth-
succeeded. ers that would help support or disprove it. In
As for the generalizability of any particular what follows, I shall assume the validity of our
finding, this has to be determined in the usual theory of the source of equivalence relations.
ways: direct and systematic replication. The second theoretical aspect of our for-
mulation is the use my colleagues and I have
Description versus explanation made of the mathematical theory of sets; we
have adopted that theory as a model. The ques-
Holth: The issue of description versus expla- tion with which I have wrestled is whether the
nation is a difficult one. You have written that: “. mathematical theory of sets provides a theo-
. . although set theory informs us how to find out retical or a descriptive model of equivalence
whether any particular event pair belongs to an relations in behavior?
equivalence relation it is silent about the origin I have come to the conclusion that our
of equivalence relations themselves” (Sidman, formulation does involve theory but not theo-
1994, p. 553) and, in a later article (Sidman, retical explanation. The theory consists only
1997) that stimulus equivalence is descriptive of our assumption that the relations among
rather than explanatory. Yet, you added that: elements of a reinforcement contingency—
A Research Pioneer’s Wisdom: An Interview with Dr. Murray Sidman 191

stimuli and responses—fit mathematical set transformation of function, often presumed to


theory’s definition of the equivalence relation. denote a separate behavioral process. Given set
In set theory, equivalence relations must dis- union, we automatically come up with elements
play certain regularities, described as reflexiv- that belong to more than one class. Transfer of
ity, symmetry, and transitivity. If our theory function, rather than being conceptualized as
that the elements of a reinforcement contin- an explanatory behavioral process, turns out to
gency are related by equivalence is valid, then be simply an example of set union. Like equiva-
the emergent relations that make our data so lence relations, set union and set intersection
interesting must display the regularities that set are concepts that belong to mathematical set
theory specifies. Set theory itself is not a theory theory, not to behavioral theory. When transfer
of behavior; it does not explain the remarkable of function takes place, it is simply a behavioral
explosion of performances that our behavioral example of mathematically defined set union.
procedures produce. Although set theory itself The behavioral process that is involved here is
says nothing about behavior, however, it proves discrimination formation, in which contex-
useful in describing behavior. In particular, tual factors determine whether set union or set
the relations among the environmental and intersection will provide a description of the
behavioral components of reinforcement observed behavior.
contingencies turn out to be specific examples Perhaps because our way of talking about
of set theory’s more general definition of the our data has been remarkably productive—a
equivalence relation. major characteristic of a good theory—oth-
Given the validity of our assumption that ers have called our formulation theoretical. I
our data fit set theory’s definition of the equiva- myself, however, consider it to be a descriptive
lence relation, then the emergent performances model, a useful framework within which to
are to be expected; they are simply examples of describe our data.
set theory’s equivalence relations. No behav-
ioral theory has predicted them. Our theory Relational Frame Theory
has simply placed our observations within an
already existing mathematical theory. That is Holth: Relational frame theory seems,
why I say that our theory does not explain the basically, to be very much inspired by your
data but just provides a way to describe them. early work on equivalence relations. How-
The theory simply places the data within a non- ever, in RFT, equivalence is just one of a
behavioral theory that describes the emergent number of different derived relations that
performances that make our data so interesting. can emerge from multiple exemplar training.
Our data simply provide examples of set theory’s What do you consider to be the stronger and
equivalence relations. weaker parts of this theory?
The theoretical assumption that our data fit Sidman: The analytic technique we used in
set theory’s definition of equivalence relations is our studies of equivalence showed that behav-
strengthened by the finding that other aspects ioral equivalence relations exhibit reflexivity,
of the data are also compatible with set theory. symmetry, and transitivity. RFT generalizes
For example, the abstractions, set union and set our original type of analysis to other kinds of
intersection also contribute to the descriptive relations that may exhibit different properties—
framework. Set union describes instances in nonreflexivity, nonsymmetry, or nontransitivity
which two or more equivalence classes merge (e.g., Hayes, 1991).
because they possess elements in common. Set In RFT’s analysis, symmetry and transitivity
abstraction defines the opposite observation; are subsumed under mutual and combinatorial
classes remain independent even though they entailment. For example, in an equivalence
possess elements in common. relation, if A is related to B, then symmetry
Set union provides a parsimonious descrip- requires B to bear that same relation to A. Thus,
tion of what is commonly called transfer or “father” cannot be an equivalence relation.
192 Per Holth

If A is the father of B, B cannot be the father demonstrated instance of functional transforma-


of A; the relation is not symmetric. A relation tion has required the equivalence relation as the
does, however, exist in both directions; B must vehicle of transformation.
be a child of A. Therefore, the relational prop- This direct inclusion of equivalence relations
erty, mutual entailment, does apply. Even if the within the very definition of relational frames
relation were symmetric, mutual entailment suggests that the originators of RFT really did
would still apply. understand that the relational frame, “equivalence,”
Furthermore, if A is the father of B, and differs from all other frames in ways that require
B is the father of C, then “father” cannot be RFT to adopt it as an integral part rather than
transitive; A cannot be the father of C. Like merely subsume it as just another frame. Equiva-
nonreflexivity, nontransitivity shows the rela- lence, therefore, is not another relation that can
tion “father” not to be an equivalence relation. be derived from RFT; it is itself the heart of one
A is, however, still related to C; that relation of the basic properties of RFT.
is “grandfather.” Therefore, although the rela- I also have considerable difficulty with RFT’s
tion, “father,” is nontransitive and therefore is basic assumption that equivalence relations are
not an equivalence relation, it does display the learned operants that arise from a history of mul-
relational property, combinatorial entailment. tiple exemplar training. It is certainly true that
If the relation were transitive, combinatorial many relations do come about through learning.
entailment would still apply. For example, relations like those frequently used
Thus, the two types of entailment do provide as examples in RFT expositions—same, opposite,
a more general way to classify relations than just different—certainly are learned. They are learned,
the limited properties of symmetry and transitiv- however, on the basis of shared physical charac-
ity. In spite of the greater generality of mutual teristics. That is to say, they involve stimulus dis-
and combinatorial entailment, however, I ques- criminations, the abstracting of physical properties
tion the dismissal of the equivalence relation as that are shared by all the examples. Equivalence
“just one of a number of different derived rela- relations, however, are not based on the abstraction
tions that can emerge from multiple exemplar of common physical stimulus elements. RFT tries
training.” What has been gained here? Whether to take care of this difficulty by means of multiple
or not equivalence emerges from a training his- exemplar training, the discrimination process by
tory (discussion of that point will come later), which we can learn relations simply by being ex-
equivalence relations remain unique in several posed to them many times (with reinforcement,
ways. For example, no other type of relation pro- of course).
vides a basis for investigating and understanding To be effective, however, multiple exemplar
what both linguistic theory and everyday termi- training requires discrimination on the basis of
nology refer to as the representational power of shared physical properties of stimuli. In equiva-
language. In RFT, linguistic representation is lence relations, the related elements need share
subsumed under the third property of relational no physical characteristics. The only feature that
frames, transformation of function. Upon close elements related by equivalence share in common
examination, however, one can see that the is the relation itself. For equivalence relations to be
property, transformation of function, constitutes learned, multiple experiences cannot be sufficient.
a direct adoption of the equivalence relation as If the elements of an equivalence relation share no
an integral part of RFT. As I indicated before, physical characteristic, such learning could take
along with set (class) union, the formation of place only by linguistic mediation, but for such
equivalence classes accounts for observations of mediation to teach the relation, the “meaning” of
function transfer or transformation. RFT does equivalence relation must itself already have been
not subsume the equivalence relation within a learned. The effectiveness of such mediation presup-
more general classification but rather, adopts poses the existence of already established linguistic
equivalence as an essential ingredient. In support equivalence relations. A similar situation exists
of this assertion, I note that every experimentally with respect to many nonequivalence relations.
A Research Pioneer’s Wisdom: An Interview with Dr. Murray Sidman 193

For example, the relation, “father” cannot be Facts and theory


learned simply on the basis of multiple examples
because no common physical stimuli exist from Holth: In one of your Remarks in the jour-
example to example. The “father” relation, nal Behaviorism, you wrote that:
like equivalence relations, requires linguistic
mediation if it is to be learned. Successful lin- “Their own failure to appreciate the inferential nature
guistic mediation requires preexisting linguistic of stimulus control has led some behaviorists to assert
that they, unlike cognitive theorists, are concerned
equivalences. only with observable events. They have been guilty
Although this reasoning leads to the conclu- here of false pride, one effect of which has been to
sion that multiple exemplar training cannot generate similarly inappropriate emotional reactions
suffice to teach equivalence relations, let us sup- among cognitive psychologists. But perhaps a more
pose that it was found empirically that multiple regrettable consequence has been to prevent behav-
iorists from including some cognitive phenomena
exemplar training could teach an equivalence
within their avowed subject matter” (Sidman,
relation to a nonverbal human or nonhuman.
1979, pp.123-124).
Such a finding would demonstrate a brand
new type of learning process, not included in
any current behavioral account of learning. Would you say that other core behavioral
Until such a demonstration has been verified, phenomena, such as reinforcement and ex-
therefore, the validity of RFT’s requirement that tinction are of a similar inferential nature as
equivalence constitutes a learned operant would stimulus control?
depend on a process that has not yet been shown Sidman: Because I was concerned at the
to exist. The postulation of a currently unknown time with problems of stimulus control, that
learning process detracts considerably from the interesting question did not occur to me. My
parsimony and coherence of RFT. conclusion that stimulus control is inferential
A number of other features of RFT keep came from my recognition that a single observa-
me from espousing it, but I have already out- tion can never suffice to identify control by any
lined those features in my book on equivalence specific stimulus aspect. Every stimulus presents
relations (Sidman, 1994, pp. 554-561) and many aspects at the same time. Multiple obser-
I need not repeat them here. An additional vations are therefore necessary before one can
point relevant to the concept of the relational say that a particular stimulus aspect controls any
frame has, however, occurred to me recently. item of behavior. The question being asked here
RFT holds that particular occurrences of any makes me realize that the same reasoning ap-
relation are to be explained as instances of a plies to reinforcement and extinction; multiple
more general relational frame. This notion is observations are required to infer those behav-
reminiscent of a theory of discrimination and ioral phenomena also. Reinforcement requires
concept learning proposed by Levine and his an increase and extinction requires a decrease
collaborators that enjoyed great popularity in in probability; measurements of probability
the 1960’s (e.g., Levine, 1963). Without going require several observations.
into the relevant experiments, one may summa- The passage of time, however, usually
rize the theory by citing Levine’s conclusion that brings with it contextual changes. As a science
reinforcements do not affect particular instances progresses, one must often evaluate known
of a discriminative response or concept. More phenomena under previously unconsidered
generally, reinforcements affect hypotheses that conditions. For example, errorless learning, ac-
mediate many such instances. Hypotheses and complished by programmed instruction, yields
relational frames seem to me to be function- one-trial learning. One does not need several
ally equivalent. In their evaluation of RFT, I observations to determine whether response
wonder if its originators might do well to look probability has increased. Having been taught
more closely at Levine’s theory with a view to all the prerequisites for a particular perfor-
determining why it did not survive. mance, a student’s “learning curve” will then
194 Per Holth

jump immediately from near zero to a perfect this behavior by appealing to “symbolization”
performance. Errorless learning brings in new or “representation.” Each of these explanatory
behavior in one trial. Of course, if one wants processes is invisible and can be endowed with
to identify the particular aspect of a response whatever properties will include the observed
that is being reinforced, one still needs several behavior. No independent measurement of the
trials to find out, even with errorless learning. inferred explanations is required. Just observe
Lever pressing, for example, has several aspects: interesting behavior and then invent an explana-
force, distance, acceleration, the part of the tory process that has the same properties as the
body that is involved, and so on. To identify observed behavior. Behavior analysis explicitly
which of these is being reinforced still requires rejects the kind of explanatory inferences that
several observations. psychology typically espouses.
On the other hand, errorless extinction has
not been demonstrated, so extinction on any par- Behavioral science and socially
ticular observation is always inferential. Errorless important issues
discrimination learning has been demonstrated,
but because every stimulus has multiple aspects, Holth: Ever since Watson, behaviorists
the development of control by any particular have been concerned with the application of
aspect is still inferential. Even with errorless behavioral science to solve important social and
discrimination learning, it takes many observa- societal problems. Among many other things,
tions to determine which aspect of the stimulus you have been concerned with diplomacy based
is being discriminated. Therefore, any conclusion on empirical data (Sidman, 1989, p. 239) and
after a single observation is still inferential. with problems of terrorism and how behavior
Holth: If the basic phenomena in behavior analysis principles may be relevant in these
analysis are of an inferential nature, is there a areas. Regarding the issue of effective diplo-
simple way to distinguish the inferential nature macy, do you envision a sort of experimental
of these phenomena from those typically treated approach?
in psychology? Sidman: If, by envision, you mean foresee or
Sidman: Although reinforcement, extinc- predict, then I must admit to a pessimistic view
tion, and stimulus control are inferential in of the likelihood of an experimental approach to
behavior analysis, inference refers here only to diplomacy. Most high level diplomats come to
the impossibility of identifying in just a single their positions because they have already shown
observation whether these basic processes have strong biases toward particular approaches to
occurred. When measured by means of several such matters as world peace, disarmament,
observations, they are observed directly and are economic cooperation, and so on. Experimen-
no longer inferential. Reinforcement, extinc- tation is a term that either they do not have in
tion, and stimulus control become names that their vocabularies or signifies something risky,
identify a set of behavioral observations but do too dangerous to attempt. The very concept of
not explain those observations. The names do data is generally not understood at all, let alone
not themselves represent inferred explanatory the concept of basing our own behavior on data.
processes. In psychology, however, the usual type What I have suggested is the establishment
of inference is of a different nature. Psychology within, say, the state department, of a group
infers unobservable properties, and even unob- composed of scientists of various kinds who
servable entities, to explain observable features of would examine current diplomatic problems
behavior. For example, when they see behavior and come up with suggestions for securing
that behavior analysts call “remembering,” they data on possible ways to solve those problems.
explain it as being the product of something Behavior analysts could recommend a number
called “memory.” When they see people respond- of techniques to apply first in noncritical areas.
ing to written numbers in the same way they Response shaping, stimulus shaping, establish-
would respond to actual quantities, they explain ment of conditioned and generalized reinforcers,
A Research Pioneer’s Wisdom: An Interview with Dr. Murray Sidman 195

or generating particular equivalence relations Sidman: Although behavior analysis has


could be tried out at first in noncritical areas much of value to offer the general public, it has
in which we would incur no great loss if our not succeeded in significantly influencing public
reinforcement and shaping techniques did policy of any sort. As I indicated before, behavior
not work. Ensuring that both parties benefit analysts have never taught themselves how to
from particular simple interactions could converse effectively with the general public. That
gradually build up a context in which trust must come first.
and cooperation played more prominent Second, it is always difficult for one group of
roles. Particular people and particular forms specialists to influence other groups because they
of behavior would be observed and measured, are viewed as outsiders trying to obtain power be-
with future interactions being determined by yond their own areas of competence. Therefore, be-
those data. havior analysts who want to influence other fields
Other kinds of scientists would have their would do well to take positions within those fields.
own kinds of suggestions for experimentation. In education, for example, they must become
Physicists might propose experimenting with teachers, school principals, faculty members in
methods for exchanging previously secret schools of education, and so on. By working
information that would benefit both parties within the field they will be in advantageous
to such exchanges. Social psychologists might positions to accomplish the kinds of changes that
come up with methods for persuading large are needed. That, of course, is not an easy thing
groups to entertain the possibility of friendly to do. A major source of one’s own sources of
cooperation instead of competition. Whatever reinforcement is one’s environment. It is difficult,
the field that generates a proposed experiment, therefore, to maintain the rigor of a behavior
all suggestions would include methods for ob- analytic approach when one is working within
taining data that could be used for evaluating an environment dominated by traditional edu-
the success of the methods being proposed. cational practices.
Holth: Regarding Spread of terrorism, Similarly, if one wants to influence public
you have been concerned with a presumably policy more generally, one should go into fields
inadvertent but important role that the me- where public policy is a primary concern—for
dia may play in their focus on major terror example, politics. One should run for election
events. In the Coercion book, you wrote that, to public office and become a representative,
“The excuse that all the news must be reported senator, governor, and so on. Again, this is a dif-
is patently false: it has never been possible to ficult area in which to maintain a consistent data
report everything, and editors have always had oriented approach because success in influencing
to choose what to publish” (Sidman, 1989, p. one’s colleagues often depends on compromises
248). Did media people ever respond to this? in which one has at least temporarily to give up
Sidman: No, I have had no response to this one’s reliance on objective data. Nevertheless, one
from media people, and little response even will more effectively influence public policy from
from behavior analysts. Even though news the inside than from the outside.
reports of events like shootings in schools and A real problem here comes from the fact that
other unusual acts of violence have predicted behavior analysts are rarely taught to apply their
that others would try to imitate those acts skills in areas beyond the treatment of behavior
just to obtain publicity, I have never seen a problems. Diplomacy, which I discussed briefly
suggestion that refraining from publicizing before, is a name for the behavior of resolving
such events might make them less likely to conflicts. I do not believe that any behavior analy-
be copied. sis program has ever included courses on conflict
Holth: What are your thoughts on addi- resolution. If they did, we would find behavior
tional things that behavior analysts might do in analysts being sought not just in international
order to get this and similar points effectively diplomacy but in all areas in which conflict reso-
spread to the general public? lution is the dominant problem—labor relations,
196 Per Holth

civil and criminal legal proceedings, legislation, Regarding collaboration with other sciences,
financial dealings, marriage problems, business the best I can do is to review some past instances
contracts, religious differences, and many others. of such ventures and perhaps draw some lessons
from those examples. In this regard, I have had
Future of behavior analysis some most instructive personal experiences.
For example, I spent approximately nine years
Holth: What do you think Behavior Analy- as a member of the psychology department that
sis will develop into in the next 50 years, and Joe Brady started in the early 1950’s in the Neu-
does that differ from what do you think it ropsychiatry Division of the Walter Reed Army
should develop into? Institute of Research. This was perhaps the most
Sidman: I would not venture to predict what stimulating and productive interdisciplinary
behavior analysis will look like 50 years from now. organization in which behavior analysis has ever
In this interview I have made a number of participated. Following Joe Brady’s lead, I took
suggestions that may be summarized by, “Be- part in many collaborative projects. For example:
havior analysis, heal thyself.” I hope, though, with neuroanatomists in studies of relations be-
that more than just healing will be going on. tween schedule-controlled behavior and central
Today’s students will determine the particular nervous system structures; with endocrinologists
new directions behavior analysis is going to in studies that demonstrated two-way interac-
take. Future development implies that some- tions between behavior and pituitary-adrenal
thing new is going to happen, something that cortical activity; with pharmacologists in studies
nobody today can predict. Who is going to put that followed up Peter Dews’ discovery that re-
us on the path to radically new developments? inforcement schedules play a role in determining
When will the changes happen? Nobody today the behavioral effects of drugs; with neuroanato-
can answer those questions either. It may even mists and physiologists in demonstrating that
be that the seeds of major developments are reinforcement schedules and other behavioral
present now but that the field simply has not variables could affect the extent to which intracra-
yet recognized them. nial electrical brain stimulation would reinforce
Do I have some ideas about what behav- behavior. Others in our group made original
ior analysis should develop into? No. If I did contributions to various other basic scientific
have such ideas, I would already have tried to and applied areas. For example: Findley’s work on
implement them or to persuade someone else 24-hour controlled environments led to Brady’s
to do so. At the moment, our science and our participation in the preparation of monkeys as
practices have enough problems to keep us all test subjects during space flights; Schuster and
busy looking for solutions. I do not feel capable his collaborators came up with techniques for
of specifying new goals or new ways to reach teaching animals to self-administer drugs, in the
our present goals. process demonstrating training and measurement
Holth: You have written that “Our basic techniques that revolutionized screening for ad-
scientists need to reestablish productive col- dictive properties of drugs.
laborative ventures in other areas, and our ap- Clearly, this was an exciting environment. In
plied scientists need to turn more to their basic it, behavior analysis played an important role,
science for data, principles, and procedures.” sometimes a leading role. What were this environ-
(Sidman, 2002, p. 12). What do you think it ment’s main determining characteristics? What
will take for that to happen? lessons does it have for us today? How might one
Sidman: You are really asking two questions recreate such productive collaboration? Upon
here. First, how can we reestablish collabora- reflection, several features stand out.
tion between behavior analysis and other areas First, the group consisted of first rate scientists
of science? Second, how can we encourage from each field. None of them were required to
applied behavior analysts to make use of their do collaborative work. They were, however, able
basic science? to look beyond their own specialties and when
A Research Pioneer’s Wisdom: An Interview with Dr. Murray Sidman 197

they saw others whose expertise they could also know how to change his/her therapeutic
respect, they were open to collaborative ven- procedures on the basis of what the client does.
tures. That is the kind of group one would have Successful behavior analytic practice does not
to establish. We can no longer expect a scientist depend on a set of fixed rules but consists of op-
from outside behavior analysis to include behav- tions that the practitioner can apply in response
ior analysts in such a group; behavior analysts to what the client does.
have long since broken most of their connec- We have a situation now in which practitio-
tions with other sciences. It will, therefore, ners who use methods that have come directly
take a behavior analyst to recruit capable and from basic research are often unaware of the
intellectually curious scientists from other fields. source of those methods. For example, the most
Perhaps someone from behavioral pharma- basic applied technique, positive reinforcement,
cology could succeed in creating a group of originated in and continues to be refined by
scientists from other areas who were capable basic research. Acquaintance with this research
of recognizing techniques for establishing and would make applied workers more capable of
measuring kinds of behavior that would interest responding effectively to many seeming failures
them. Then they might be able to bring in other in their standard reinforcement procedures.
kinds of behavior analysts, too, who were also Then too, applied workers regularly use tech-
interested in scientific collaboration. niques designed to generate errorless learning
The second problem is how to encourage while remaining unaware that the elimination
applied behavior analysts to make more use of of errors is at the heart of what they are doing.
their basic science. Such use is to a great extent Such understanding would permit them to
missing from the original training of many modify their teaching procedures to eliminate
applied workers. It is also largely unknown to seeming failures of a client to learn some new
practitioners who have learned only enough and more adaptive behavior.
to pass qualifying exams. For example, single- What I am saying here is that the practice of
subject methodology is fundamental in basic behavior analysis would become more generally
behavior analytic research; that aspect of the effective if the training programs for applied
methodology makes the science immediately behavior analysts were made more rigorous
applicable to behavior therapy. Also, Skinner’s than many are at present. Also, the qualifying
recognition that experimental behavior analysis tests for practitioners should also be made more
consists of two-way interactions between subject rigorous than they are at present.
and experimenter makes the science compatible Even more generally, I would suggest, as
with applied practice. Many applied workers, I have before, that neglect of basic behavior
however, have never been made aware of these analytic science will eventually reduce and
methodological differences between what they even eliminate the public approval of behavior
are doing and what traditional clinical psy- analytic practice. The concept of empirically
chologists do. Ignorance of the rationale for grounded practice is becoming known and ap-
single-subject methodology leads to ignorance preciated by the general public. Any practice
of the special importance of steady-state base- that lacks scientific backing will sooner or later
lines and multiple baselines of various sorts lead to public rejection of that practice. Further-
and of the necessity for refining such baselines more, even though basic science has provided
to evaluate the success or failure of their treat- the underpinning for applied behavior analysis,
ments. Also, like a successful experimental if the practitioners themselves are unaware of
program, effective behavior therapy also re- that reality, then the public will assume that
quires two-way interactions between therapist no such relation exists. When that happens,
and client. The behavior of each changes as practitioners will lose their public acceptance.
a function of what the other does. Not only Holth: Thank you for all your time and
must the client’s behavior change in response patience in completing this interview. It has
to therapeutic measures but the therapist must really been marvelous to listen to your thoughtful
198 Per Holth

answers to all of my questions. You have speci- 76(2), 131-158.


fied very clearly a number of issues that current Levine, M. (1963). Mediating processes in hu-
and future behavior analysts will need to attend mans at the outset of discrimination learning.
to, and I also believe that your answers will serve Psychological Review, 70, 254-276.
as an important source of inspiration to carry Sidman, M. (1960). Tactics of scientific research.
on that work. Thank you so much! New York: Basic books.
Sidman, M. (1979). Remarks. Behaviorism, 7,
References 123-126.
Sidman, M. (1989). Coercion and its Fallout.
Azrin, N. H., & Holz, W. C. (1966). Punish- Boston, MA: Authors Cooperative.
ment. In H. W.K. (Ed.), Operant behavior: Ar- Sidman, M. (1994). Equivalence relations and be-
eas of research and applications (pp. 380-447). havior: A research story. Boston, MA: Authors
East Norwalk, CT: Appleton-Century-Crofts. Cooperative.
Hayes, S. C. (1991). A relational control theory Sidman, M. (1997). Equivalence: A theoretical
of stimulus equivalence. In L. J. Hayes & P. or a descriptive model? Mexican Journal of
N. Chase (Eds.), Dialogues on verbal behavior Behavior Analysis, 23, 125–145.
(pp. 19-40). Reno, NV: Context Press. Sidman, M. (2000). Equivalence relations and
Holth, P. (2005). Two Definitions of Punish- the reinforcement contingency. Journal of
ment. The Behavior Analyst Today, 6, 43-47. the Experimental Analysis of Behavior, 74,
Horne, P. J., & Lowe, C. F. (1996). On the origins 127- 146.
of naming and other symbolic behavior. Jour- Sidman, M. (2002). Notes from the beginning
nal of the Experimental Analysis of Behavior, of time. The Behavior Analyst, 25(1), 3-13.
65(1), 185-241. Skinner, B. F. (1938). The behavior of organ-
Kastak, C. R., Schusterman, R. J., & Kastak, D. isms: An experimental analysis. New York:
(2001). Equivalence classification by Califor- Appleton-Century-Crofts.
nia sea lions using class-specific reinforcers. Skinner, B. F. (1953). Science and human behav-
Journal of the experimental analysis of behavior, ior. New York: Macmillan.

View publication stats

Das könnte Ihnen auch gefallen