B E H A V I O R T H E R A P Y 8, 6 5 9 - 6 6 5 (1977)

Reinforcing and Punishing Thoughts

H O W A R D RACHLIN

State University of New York at Stony Brook

Thoughts cannot be reinforced or punished as can overt behavior. Although

reinforcers or punishers brought into temporal relationship with verbal behavior
(or other behavior that expresses the thought) may increase or decrease that
behavior, the thought itself, if it is changed, will not simply change in frequency
but will reflect the new relationship between a group of behaviors and the envi-
ronment.

A recent trend in behavior therapy has been to use those techniques

effective in modifying overt behavior to modify covert behavior. In some
cases the relation of this enterprise to operant conditioning has been
emphasized by using the term "coverant" for the covert behavior to be
reinforced or punished (Homme, 1965). In other cases the relation to
cognitive psychology is emphasized by retaining the term "cognition" or
"thought" for the behavior to be modified (Meichenbaum, 1969, 1974a).
The techniques proposed to modify thoughts vary somewhat but the
following procedure would not be atypical: A client is diagnosed as
obsessed with a certain harmful thought. The client is asked to say the
thought out loud and is then punished. Then, the client is asked to say the
thought to himself and indicate when it has occurred and is then punished.
Finally, the client is asked to punish himself when the thought occurs.1
Individual procedures vary by leaving out one or another of the above
steps or by reinforcing alternative thoughts rather than punishing the
thought to be eliminated. They also differ in their mode of reinforcement
or punishment. Usually, secondary reinforcers or punishers are used
(e.g., Meichenbaum, 1969); at times, hierarchies are arranged so that

Preparation of this paper was supported by a grant from the National Science Foundation.
Requests for reprints should be addressed to Howard Rachlin, Department of Psychology,
State University of New York at Stony Brook, Stony Brook, NY 11794.
1 1 have argued previously (Rachlin, 1974) that self-punishment of this kind, while it may
be effective, is not proper punishment. However, it is commonly used as such in behavior-
therapy, and we will consider it here together with overt punishment and reinforcement.

659
Copyright @ 1977 by Association for Advancement of Behavior Therapy.
All rights of reproduction in any form reserved. ISSN 0005-7894
660 HOWARD RACHLIN

events of higher value reinforce the thought to be modified (Homme,

1965).
The purpose of this paper is to show, following the assumptions made
by behavior modifiers themselves, behaviorists and cognitivists alike, that
these procedures misdirect the application of reinforcement and may
change behavior in unexpected ways.

CATEGORY MISTAKES
The logical mistake usually involved in the notion that thoughts can be
reinforced or punished is a category mistake. A category mistake consists
of treating the name for a class of events as if it, itself, were a member of
that class. Gilbert Ryle illustrates category mistakes as follows.
A foreigner visiting Oxford or Cambridge for the first time is shown a number of colleges,
libraries, playing fields, museums, scientific departments and administrative offices. He then
asks, "But where is the University? I have seen where the members of the Colleges live,
where the Registrar works, where the scientists experiment and the rest. But I have not yet
seen the University." It has then to be explained to him that the University is not another
collateral institution, some ulterior counterpart to the colleges, laboratories and offices
which he has seen. The University is just the way in which all that he has already seen is
organized. When they are seen and when their co-ordination is understood, the University
has been seen. His mistake lay in his innocent assumption that, it was correct to speak of
Christ Church, The Bodleian Library, the Ashmolean Museum and the University, to speak,
that is, as if "the University' stood for an extra member of the class of which these other
units are members. He was mistakenly allocating the University to the same category as that
to which the other institutions belong (Ryle, 1949).

In the case of thoughts and reinforcements, a similar category mistake

may be made. Reinforcement and punishment are relationships between
behavior and environmental events. A "thought" is a name for a set of
such relationships. One reinforces or punishes behavior, not thought. A
thought may already encompass reinforcement or punishment within it.
As an example, consider the following proposition which, when acti-
vated in a person's nervous system, might be said to constitute a thought:
" I f I touch fire I will be burned." This proposition could, according to
current cognitive theory, be located in a person's long-term store and
might be activated by the sight of a fire. When activated it becomes a
thought. It might be a conscious thought or even subconscious. It would
have been acquired by the previous association of touching fires and
burning, perhaps because these were both vivid temporally related events
or perhaps through verbal mediation. Whatever the means of acquiring
the association, the role of reinforcement or punishment, if any, would be
limited to the action of the fire on the act of touching the fire. But this
action is already encompassed in the thought. It has never been part of
cognitive theory, or any other theory, that still another reinforcer could
affect the thought. What another such reinforcement could be is difficult
 REINFORCING AND PUNISHING THOUGHTS 661

to conceive. First, the thought would have to be activated and then

reinforced. Imagine showing a 10-yr-old child a fire, presumably activat-
ing the thought, " I f I touch the fire 1 will be burned," and then giving him
a piece of candy. To what extent is the thought affected? If the candy is
presumed to have any effect on the thought at all, it would itself be
associated with fire, thus doing the exact opposite of reinforcing the
original thought. The child may now approach fires more readily. Rein-
forcers and punishers are, if nothing else, vivid stimuli. They are likely to
be paid attention. They (or, according to cognitive psychologists, their
representations) form parts of thoughts themselves. The intended mecha-
nism of manipulation itself becomes part of the thing to be changed.
Reinforcement and punishment may change the thought to reflect the
relationship of its elements with the reinforcer or punisher. They cannot
act in a superordinate fashion to strengthen or weaken the thought itself.
While it is not possible to reinforce thoughts, it is possible to reinforce
or punish verbal behavior. When a change in verbal behavior is the only
desired outcome of the process, it is not necessary to refer to thoughts or
cognitions at all.
THE RELATION OF REINFORCEMENT, PUNISHMENT,
AND THOUGHT
Current psychological theory tells us that thoughts are propositions
represented either in an organism's mind, in its nervous system, or in its
behavior. The behaviorist believes that the proposition is simply a de-
scription of organized behavior. The cognitivist believes that the proposi-
tion gives rise to the behavioral organization and that this proposition has
a real existence somewhere within the body (Anderson & Bower, 1973).
Whether the proposition is in the mind or the nervous system, it is
generally agreed that thoughts have been observed so far only in the
organization of overt behavior, including verbal behavior.
If thought is behavioral organization (or causes such organization), it
follows that any one overt behavior cannot be a thought. To observe a
thought must be to observe an organized sequence that occurs over time.
The thought must encompass the whole set of behaviors, including verbal
expression, by which we know that the thought occurs. It is not possible
for the thought to occur over a brief time span while the behavior that it
causes occurs over an extended time span. This notion, that a behavior
can run off in a ballistic manner and still be thoughtful, was rejected by
Lashley (1951) and by most psychologists following him. Again, the
behaviorist says that the thought is the organization, while the cognitivist
says that it causes the organization. In neither case is the thought the
expression of the proposition in a sentence.
It is important to stress the nonidentity of thoughts and verbal be-
662 H O W A R D RACHLIN

havior. If a thought were only the verbal expression of a proposition and

this verbal expression could cause overt thoughtful behavior, then it
would be possible to reinforce and punish thoughts. It makes little sense,
however, that verbal behavior should be the cause of other behavior. For
instance, in what sense could the sentence, "I think I will go to the
movies," cause the act of going to the movies? Neither current cognitive
theory nor behavioral theory would attribute such power to the sentence.
Behaviorists would say the cause of both the verbal expression, "I think I
will go to the movies," and the behavior of going to the movies is a set of
observable events (behaviors, neutral stimuli, reinforcers, punishers) in
the past. Cognitivists would say that the cause of both the verbal expres-
sion and the behavior is an event or organization of events currently
active in the nervous system. In fact, one cognitive therapy (Ellis, 1962)
rests on discovering unexpressed thoughts underlying overt maladaptive
behavior.

MODIFYING THOUGHTS
A proposition (hence, a thought) could refer wholly to environmental
events, such as, "The ball is green," or to interactions between be-
havioral and environmental events such as, " I threw the ball."
Consider ways of reducing such thoughts that might be used by be-
havior therapists. First, suppose we connect the client to an electric
shock apparatus and shock him whenever he indicated he had the thought
or instruct him to shock himself whenever he had the thought. Probably
the verbal expression of the sentence would be reduced. But such
punishment might leave unaffected the behaviors exhibited in various
tests to see if the thought were really there. How many times would we
have to shock a mother after she said, "Playing near open windows is
dangerous," before she allowed her 2-year-old child to play on the win-
dowsill? What we are likely to do is to destroy not the thought itself but
the connection between the thought and its verbal expression. The person
punished for the thought might now simply lie. The therapist will note a
reduction in verbal expression of the thought and perhaps, by generaliza-
tion, other verbal expressions similar in sound or in meaning to the
original.
One could now imagine additional treatment. Suppose we punish the
lady not only for saying the thought but also for pulling her child away
from the window, for closing the window, for nodding her head whenever
she hears someone else express the thought, etc. Suppose we persist in
these methods until one by one we erase every known behavioral man-
ifestation of the thought. We will have now replaced the thought, " A - B , "
with " I f l say A-B or do A-B or agree with A-B, I will be punished." Such
procedures are undoubtedly what most people understand as brainwash-
 REINFORCING AND PUNISHING THOUGHTS 663

ing. What may make them aversive is the continuance of the subject's
experience with A-B. But as long as the subject, regardless of his or her
behavior with respect to A-B, still behaves as previously with respect to
C-D and all other experienced associations, it would be more parsimoni-
ous to assume that the relevant propositions in storage are " C - D , "
" E - F , " " G - H , " etc., and " A - B " plus the additional thought, "Expres-
sion of A-B leads to punishment." Regardless of how thorough our
techniques might be, none would remove the thought as long as the
experience of A-B remained as before.
To put it more concretely, we may get a person to say "The sky is
green" easily enough and to act as though the sky were indeed green,
even while, in fact, the sky remained blue. But unless the person were
also confused about other color names (for instance, calling the grass
yellow), we would most parsimoniously attribute to him the thought,
"The sky is blue, but I must say it is green." This is the only thought
consistent with his other thoughts, i.e., with his other behavior and
experience. But let experience itself be changed, say by the repeated
presentation of a green sky, and the thought will change accordingly.

THOUGHTS AND BELIEFS

One may distinguish between belief and thought by saying that the
thought is merely verbal expression, while belief involves something else.
But such an argument represents another category mistake. A belief is not
something other than a thought; it is an attribute of a thought. Given a
thought, we believe it is true or believe it is false, or are neutral about its
truth or falsity, or its truth or falsity is irrelevant. There are not cognitive
elements labeled "beliefs" that can be set aside and compared with
"thoughts." When behavior is examined for signs of thoughtfulness, it
can also be examined for degree of belief. Belief in a proposition may be
ascertained by the tone of voice in which it is expressed and by the
accompanying behavior. Before we would want to say that a person
strongly believed a proposition to be true, we would want to see a fairly
large sample of his behavior. Verbal behavior may or may not be part of
this sample.
In most cases behavior therapy is concerned with thoughts that the
client strongly believes are true or false. Some therapies (e.g., Ellis, 1962)
are willing to assign strong belief values to certain thoughts that may
never be verbally expressed by the client. Such a therapist might say,
"Your behavior is consistent with the thought, 'I must be loved by
everyone for everything I do.' " The therapist concludes that this is
indeed a thought that the client believes strongly, even though he may
never have expressed it verbally.
664 HOWARD RACHLIN

WHAT ARE THOUGHT-MODIFICATION PROCEDURES

ACTUALLY DOING ?
It is important to distinguish between what those who claim to be
modifying thoughts claim to be doing and what they are actually doing.
For instance, Meichenbaum (1969) has effectively modified verbal be-
havior of schizophrenics by various conditioning and modeling tech-
niques. Such treatment may well be effective to the extent that verbal
behavior constitutes a large part of what is dysfunctional in schizophrenic
behavior. But it is a major leap from these treatments to the following
quotation from Carlos Castenada that both Meichenbaum (1974b) and
Mahoney (1974) cite in summary of their implications.

The world is such-and-such or so-and-so only because we tell ourselves that that is the way
it i s . . . You talk to yourself. You're not unique at that. We carry on internal t a l k . . . I'll tell
you what we talk to ourselves about. We talk about our world. In fact we maintain our world
with our internal talk (Castenada, 1972, pp. 218-219).

Some experiments have purported to change behavior via thoughts. In

one experiment (Mahoney, Thoresen, and Danaher, 1972) subjects used
various mnemonic techniques with which they were experienced to re-
member noun pairs (e.g., "pig-typewriter"). Then, subjects were re-
warded for indicating that they used one technique or the other. Their
performance on the memory task varied accordingly, improving with
reward of the better techniques and diminishing with reward of the worse
techniques. This is evidence that instructions (or set) may have a strong
influence on how a subject responds to a stimulus, but not that thoughts
have been reinforced. 2
Perhaps the clearest example of an attempt to modify thought by
reinforcement is that of Homme (1965). The client is advised to reinforce
positive "coverants" (i.e., positive thoughts such as, "I'11 be healthier by
not smoking") with relatively pleasant daily activities such as drinking
coffee. Even if such self-reinforcement could increase the frequency of
verbal behavior, "Empirical evidence for the effectiveness of Homme's
suggested coverant control strategy is meager" as Mahoney himself has
pointed out (1974, p. 87).
This is not to say that it is impossible to reinforce sequences of behavior
or behavioral organizations. The applause a pianist receives at the end of
2 In cognitive psychology laboratories it is c o m m o n to reinforce or punish expressions by
saying " r i g h t " or " w r o n g " after the subject utters them. The behaviorist would argue that
this operation is not equivalent to reward or punishment of a single behavior (still less of a
thought) but to the provision of a discriminative stimulus for various relations between the
subject's behavior and more powerful rewards and punishments, such as the experimenter's
approval or disapproval and the eventual pay the subject might receive.
 REINFORCING AND PUNISHING THOUGHTS 665

his performance and the stars on helmets of college football players attest
to the power of reinforcement to work on complex behavior. But, in these
instances, overt behavior is clearly marked for reinforcement. Also, many
current behavior modification techniques come in "packages" whereby
both thoughts and behavior are said to be modified (Mahoney, 1974). To
the extent that such techniques are successful they modify thoughtful
behavior (an organized sequence of behavior), not two separate things:
thought plus behavior.
It is no therapeutic short cut to reinforce thoughts or cognitions or
coverants directly since none of these is a behavior susceptible to be
reinforced. Attempts to reinforce or punish nonbehavioral events may
end by unintentionally reinforcing or punishing behaviors, and, by en-
couraging clients to lie about what they are thinking, these attempts do
more harm than good. The advantage of behavior therapy has always
been that it forced the therapist and the client to define the behavior to be
changed and did not allow the evasions and abstractions of previous
techniques which claimed to cure the disease while often leaving all the
symptoms intact. This type of dodge has, it seems, proved all too tempt-
ing for behavior therapists themselves who now claim to get rid of cogni-
tions by reducing their verbal expression, yet they leave other overt
behavior intact.

REFERENCES
Anderson, J. R., & Bower, G. H. Human associative memory. Washington, DC: Winston,
1973.
Castenada, C. A separate reality: Further conversations with Don Juan. New York: Pocket
Books, 1972.
Ellis, A. Reason and emotion in psychotherapy. New York: Stuart, 1962.
Homme, L. E. Perspectives in psychology XXIV. Control ofcoverants, the operants of the
mind. Psychological Record, 1965, 15, 501-511,
Lashley, K. S. The problem of serial order in behavior. In L. A..leffreys (Ed.), Cerebral
mechanisms in behavior. New York: Wiley, 1951.
Mahoney, M, J., Thoresen, C. E., & Danaher, B. G. Covert behavior modification: An
experimental analogue. Journal of Behavior Therapy and Experimental Psychiatry,
1972, 3, 7-14.
Mahoney, M. Cognition and behavior modification. Cambridge, MA: Ballinger, 1974.
Meichenbaum, D. The effects of instructions and reinforcement on thinking and language
behaviors of schizophrenics. Behaviour Research and Therapy, 1969, 7, 101-114.
Meichenbaum, D. Cognitive behavior modification. Morristown, NJ: General Learning
Press, 1974a.
Meichenbaum, D. Self-instructional methods. In F. H. Kanfer & A. P. Goldstein (Eds.),
Helping People Change. New York: Pergamon, 1974b.
Rachlin, H. Self-control. Behaviorism, 1974, 1, 94-108.
Ryle, G. The concept of mind. London: Hutchinson, 1949.

RECEIVED: February 28, 1976; REVISED: April I, 1976

FINAL ACCEPTANCE; April 20, 1976