In respondent
conditioning we begin with an unconditioned stimulus (called UCS or US) that
elicits an unconditioned response (UCR). “Unconditioned” does not mean that the
UCS innately elicits the UCR, but only that at the present time the UCS elicits the UCR, either innately or because
of previous learning. In experiments with humans a popular UCS is a puff of air
to the eye with the resulting UCR of an eye-blink. A second stimulus is chosen,
the conditioned stimulus (CS), which does not elicit a response similar to the
UCR. The CS elicits some response, but the response is different than the UCR
and usually chosen to be a weaker response than the UCR. In the eye-blink example
a CS of a tone might be chosen. The response elicited by the tone is usually
just an orienting to the source of the tone.
Respondent
conditioning consists of pairing the CS and UCS, until the CS elicits a conditioned
response (CR) that is similar to the UCR. Generally, the most effective way
to do this pairing is to have the CS come on first and then the UCS about
one-half second later. For example, the tone would come on, followed by the
puff of air to the eye. This would be continued until the tone by itself was
capable of eliciting an eye-blink. Although the CR and UCR are similar responses,
they are generally not identical. They usually differ by magnitude, with the
UCR a stronger response than the CR.
Extinction in
respondent conditioning is accomplished by terminating the CS-UCS contingency,
that is, by stopping pairing the CS and UCS. Thus if we keep presenting the
tone to the person and no longer pairing it with the air puff, eventually the
CR will extinguish, the tone will no longer elicit an eye-blink.
Theories of respondent
conditioning emphasize different relationships that result from pairing the
CS and UCS. Most theories, however, center around one or both of the following
two possible relationships: First, there is the CS-UCS contiguity relationship,
the fact that the CS and UCS might become associated because they occur closely
together in time. Guthrie (1935) emphasized this association. Second, there
is a CR-UCS reinforcement relationship. That is, when the CR begins to occur
and when it precedes the UCS, the UCS might reinforce the CR. If the UCS is
a pleasant stimulus, such as wet food powder to a hungry dog, the onset of
the UCS might be reinforcing. If the UCS is an unpleasant stimulus, such as
the air-puff to the eye, it might be the offset of the UCS that is the reinforcement.
Spence (1956) was a theorist who emphasized the reinforcing role of the UCS.
Occasionally theorists also include the CS-UCR S-R relationship where the
stimulus CS simply by being paired with the response of the UCR results in
an association between the CS and UCR.
Some theories involve
a combination of relationships.
Joan E. Jones (1962) suggested that in early stages of some conditioning the
major variable is the CS-UCR relationship, while later in conditioning it
is the CR-UCS relationship.
In the first reading
Rescorla discusses different theoretical models of respondent conditioning
in the context of describing adequate control procedures. One theoretical
orientation emphasizes the number of times the CS and UCS have actually been
paired. That is, respondent conditioning basically reduces to the number of
times the CS has been closely followed by the UCS. An alternative orientation,
which Rescorla favors, emphasizes the contingency between the CS and UCS.
The subject learns not only what is paired with the CS, but also what is not
paired. Instead of just simple pairings, the subject is learning causal relationships
between stimulus events. From this latter orientation the ideal control group
for respondent conditioning is what Rescorla calls the “truly random” control
procedure. In this procedure the CS and UCS are presented randomly to each
other; that is, there is no contingency between the CS and UCS although there
may be some chance pairings.
An important theoretical
question is whether the subject is more disposed to learn the association
between one CS-UCS as opposed to another. The reading by Garcia and Koelling
suggests this might be the case. They found that rats readily learned the
association between audiovisual stimuli and electric shock and between gustatory
stimuli (taste) and toxins and x-rays that cause gastrointestinal problems,
but not conversely. The rats did not readily learn associations between gustatory
stimuli and electric shock or between the audiovisual stimuli and the nausea
producing stimuli. Similar types of findings are discussed in a later reading
by Seligman.
In human respondent
conditioning the CR might be considered undersirable. After being knocked
down a few times by a dog, a child might develop a conditioned fear of dogs
that carries into later life when the fear is no longer functional or desirable.
One way of changing an undesired CR is by means of counter- conditioning,
the respondent conditioning of a desired response in place of an undesired
response (cf. Mikulas, 1972, p. 32—38). The readings by Davison and by Bandura,
Blanchard, and Ritter show variations of the counterconditioning procedure.
Davidson describes the case of a college male who was sexually aroused by imagining sadistic scenes but not by “normal” sexual fantasies or activities. This situation was reversed by procedures that were primarily counterconditioning. To condition in the desired responses, heterosexual stimuli (CS) such as pictures from Playboy were paired with stimuli (UCS) from masturbation that produced a sexual response (UCR). Similarly to weaken the sexually arousing effects of the sadistic fantasies, sadistic images (CS) were paired with imagined scenes (UCS) that produced disgusting nauseous feelings (UCR). The idea of gradually conditioning in sexual responses appears to be one of the major variables underlying the procedures Masters and Johnson (1 970) outline for treatment of sexual inadequacies.
The reading by Bandura,
Blanchard, and Ritter describes a number of procedures for treating snake
phobias. The procedure of desensitization,
as developed by Wolpe (1958), generally requires the subject, while being
kept relaxed, to imagine scenes that gradually approximate the feared object.
Although the mechanisms underlying desensitization are greatly debated, the
most common explanation is that the relaxation responses gradually become
conditioned (counterconditioning) to the stimuli that previously elicited
anxiety. Another explanation (e.g., Wilson and Davison, 1971) is that desensitization
is basically respondent extinction, where the relaxation facilitates extinction
of the anxiety from the feared stimulus.
The second procedure
used in the Bandura study is modeling. Here it has been shown (Bandura,
1969, chapter 3) that simply having a subject with a snake phobia watch an
appropriate model approach and interact with the snake will tend to reduce
the subject’s fear of snakes. The third procedure, contact desensitization
(live modeling combined with guided participation), is essentially a combination
of parts of desensitization with modeling.
The general findings
of the Bandura study were that all the procedures were effective in reducing
anxiety and changing related attitudes, with contact desensitization the most
effective.
Its
Proper Control Procedures’
ROBERT
A. RESCORLA, Yale University
The traditional
control procedures for Pavlovian conditioning are examined and each is found wanting. Some procedures introduce nonassociative factors not present in the experimental procedure while others transform the excitatory,
experimental CS—US contingency into an inhibitory contingency. An alternative
control procedure is suggested in which there is no contingency whatsoever between
CS and US. This “truly random” control
procedure leads to a new conception of Pavlovian conditioning postulating that
the contingency between CS and US, rather than the pairing of CS and US, is the
important event in conditioning. The fruitfulness of this new conception of
Pavlovian conditioning is illustrated by 2 experimental results.
The operations
performed to establish Pavlovian conditioned reflexes require that the
presentation of an unconditioned stimulus
be contingent upon the occurrence of a conditioned stimulus. Students of
conditioning have regarded this contingency between CS and US as vital to the
definition of conditioning and have rejected changes in the organism not
dependent
_dir%5CtemF1B4seg241.jpg){kind=small}
upon this
contingency (such as sensitization or pseudoconditioning) as not being “true”
conditioning (i.e., associative). Therefore, in order to identify the effects
due uniquely to the contingency between CS and US, a variety of control
procedures have been developed. Each of these procedures attempts to retain
some features of the Pavlovian conditioning situation while eliminating the
CS—US contingency.
This paper argues
that, in fact, none of the conventional control procedures for nonassociative
effects is adequate, either taken alone or in combination; it further argues
that a new type of “random stimulus” control procedure does enable one to
identify the role of the CS—US contingency in Pavlovian conditioning.
TRADITIONAL
CONTROL PROCEDURES
The conventional control
procedures for Pavlovian conditioning are quite familiar, so they will be described only briefly. In
all of these descriptions, we assume that the conditioning or control treatment
is administered, and then all groups are tested with a single (unreinforced)
CS presentation. It is only the results of the test trial that are of interest.
(Similar descriptions could be given when anticipatory CRs rather than test
trial CRs are used as the index
of conditioning.)
The various control
treatments which are administered prior to the test trial in place of Pavlovian
conditioning are listed below together with examples of their use.
|
1.
|
CS-alone control. In this procedure a control subject (S)receives the same number of CS presentations as does an experimental 5; however, no US is administered. This control is designed to evaluate the effects of familiarity with the CS and any changes in the organism due solely to that familiarity (Rodnick, 1937; Thompson Sc McConnell, 1955). |
|
2.
|
Novel CS control. In this procedure, no CS is given prior to the test trial. The test trial gives an estimate
of the unconditioned effects
of the CS (Rodnick, 1937; Wickens & Wickens, 1940). |
|
3.
|
US-alone
control. Repeated presentations of the US alone are made in order to control
for sensitization by, or habituation to, the US (Notterman, Schoenfeld,
Sc Bersh, 1952; Wickens Sc Wickens, 1940). |
|
4.
|
Explicitly
unpaired control (sometimes called the random control). In this procedure,
S receives unpaired presentations of
CS and US. This can be done in a variety of ways, but the most typical
is presentation of both CS and US in the same session in random order
but never close together in time (Bitterman, 1964;
Harris, 1943). |
|
5.
|
Backward
conditioning. The CS and US are paired, but the US is always presented prior to the CS
(Kalish, 1954; Spence Sc Runquist, 1958). |
|
6.
|
Discriminative
conditioning. One stimulus (CS+) is paired with the US and the other (CS—) is not.
In this way CS— receives a treatment similar to that of CS+ except that
the contingency with the US is an “explicitly unpaired” one. Differences
between the reactions to CS+ and CS — are
taken to indicate Pavlovian conditioning (Solomon Sc Turner, 1962). |
The very variety
of control procedures which have been developed attests to the inadequacy
of any one. But it may be worthwhile to point briefly to the pitfalls of each
procedure because some of these have not been widely recognized. We take as
the logical criterion for an adequate control procedure that it retain as
many features as possible of the experimental procedure while excluding the
CS—US contingency. In general, each of the control procedures, although attempting
to eliminate the CS—US contingency, can be shown to do considerably more.
The result is that a variety of other differences, both associative and nonassociative,
between experimental and control procedures is confounded with the absence
of the CS—US contingency. Some of the confoundings are pointed out below.
|
1.
|
CS-alone
control. Quite obviously, an S treated in this way does not have the same
number of US experiences as the experimental S does; therefore, any differences
between Ss can be attributed to this difference in experience with the
US. But worse, repeated CS presentations in the absence of all USs may
not lead to the same rate of CS habituation as does repeated CS presentation
in a chamber in which the US also occurs. |
|
2.
|
Novel
CS control. It is useful to know the unconditioned properties of the CS,
but it is not clear what relevance this has for identifying “true” Pavlovian
conditioning. The experimental S has experienced the CS a large
number of times prior to the test trial and it is no longer novel to him.
Why compare him to an S for whom the CS is novel? Comparison with
a novel CS group allows one to assess the total change in reaction to the CS produced
by the conditioning procedure but does not permit isolation of those changes
due uniquely to the occurrence of Pavlovian conditioning. |
|
3.
|
US-alone
control. This procedure has faults similar to those of the novel CS procedure.
An S with this procedure receives a novel CS at the time of test, while
the experimental S receives a CS which it has experienced many times. |
|
4.
|
Explicitly
unpaired control. In many ways this procedure comes closest to being an
appropriate control, and it has become increasingly popular in recent
years. However, it contains flaws which cannot be overlooked. Although
it escapes the criticisms of Procedures 1, 2, and 3, it, too, does not
simply remove the contingency between CS and US; rather, it introduces
instead a new contingency, such that the US cannot follow
the CS for some minimum time interval. Instead of the CS being a signal
for the US, it can become a signal for the absence of the US. Although
this is an interesting procedure in itself, it does not allow a comparison
between two groups, one with a CS—US contingency and one without it. We
are, instead, in the position of having two different CS—US contingencies
which may yield different results. How can we know which group showed
Pavlovian conditioning? |
|
5.
|
Backward
conditioning. The relevance of this procedure rests upon the assumption
that in Pavlovian conditioning not only the CS—US contingency but also
their temporal order of presentation is important. It is not clear whether
this should be taken as part of the definition of Pavlovian conditioning
or as an empirical result. Nevertheless, some investigators have suggested
comparison with a backward conditioning group to evaluate the traditional
experimental group. For the purposes of analysis, let us assume that the
CS and US do not overlap in this procedure. We then have a sequence of
events: US—CS . . . US—CS... US—CS.. . in conditioning. This procedure produces the
same difficulty as does the explicitly unpaired procedure: The occurrence
of the CS predicts a period free from the US. Again, presentation
of the US is contingent upon CS occurrence but the contingency
is a negative one. Of course, if the CS begins during theUS in
this procedure, CS occurrence predicts the termination of the US,
which, in turn, introduces another contingency and further complications.
It is worth noting that Konorski (1948) considered the backward conditioning
paradigm as the prime example of an inhibitory conditioning procedure.
|
|
6.
|
Discriminative
conditioning. By now it should be clear that this control procedure falls
prey to the same criticisms as do Procedures 5 and 6. CS— is explicitly
unpaired with the US. In fact, the discriminative conditioning procedure
can be viewed as the Simultaneous administration to the same S of the
experimental procedure and Control Procedure 4. |
We can conclude that
each of the proposed control procedures either confounds some important nonassociative
change with the disruption of the CS—US contingency or changes the contingency
from a positive to a negative one. Furthermore, there is no obvious way in
which combined control procedures can be used to eliminate confoundings. Therefore,
we are in the unfortunate position of being unable to evaluate “true” Pavlovian
conditioning by the use of any or all of the conventional control procedures.
AN
ADEQUATE ALTERNATIVE
There is, however, a
control procedure which solves the problems raised above. We shall call this
procedure the “truly random” control procedure. In this procedure, both the CS and
the US are presented to S but there is no contingency whatsoever between
them. That is, the two events are programmed entirely randomly and
independently in such a way that some “pairings” of CS and US may occur by
chance alone. All CS and US occurrences for the control group are the same as
for the experimental group except that the regular temporal contingency between
CS and US is eliminated. The occurrence of the CS provides no information about
subsequent occurrences of the US. This procedure is similar in conception to
the explicitly unpaired procedure, (4), except that it eliminates the
contingency of that procedure which allows the CS to signal nonoccurrence of
the US.2
There are a variety
of ways of arranging a truly random control condition. Two major alternatives
are: (a) Present the CS asin the experimental group but randomly distribute
USs throughout the session; (b) conversely, present USs as in the experimental
group but randomly distribute CSs. Note that, in order for there to be no
contingency, the distributions must be such that CS occurrences do not
predict the occurrence of USs at any time in the remainder of the session.
If the CS predicts the occurrence of a US 30 minutes later in the session,
an appropriate random control condition has not been achieved.
Despite the apparent
adequacy of these alternatives, they actually add other confoundings. In the
usual Pavlovian conditioning procedure, several time intervals other than
the CS—US interval are kept relatively constant. Thus time intervals between
successive CSs and successive USs are of some (relatively large) minimum value.
Each of the two truly random controls would violate one of these relations
and thus introduce changes other than removal of the CS—US contingency. Fortunately,
this can be avoided if we depart from the traditional conditioning procedures
and use a wide variety of intertrial intervals for the experimental Ss. Then
it is possible to arrange truly random presentations of CS and US for the
control Ss while preserving the inter- US and inter-CS intervals of the experimental
condition. For instance, one could program CS—US pairings for the experimental
group with a random-interval programmer. Then a truly random control would
be arranged by using two independent random- interval programmers with the
same parameters as that of the experimental group—one to deliver CSs and one
to deliver USs.
We do not wish to
understate the importance of a variety of nonassociative factors which do
occur in Pavlovian conditioning. It is respect for their effects that leads
to the advocacy of the truly random control for contingency-produced effects.
One great advantage of the truly random control is that it holds constant
between the experimental and control procedures all of the factors
extraneous to the CS—US contingency without demanding that we be able to
specify in advance what factors might be operating. In contrast, the customary
control procedures have often been developed only to deal with one supposed
nonassociative factor.
It is also important
to realize that the actual results obtained with the truly random control
procedure are irrelevant to the present argument. It may be that in some conditioning
situations, Ss treated with the truly random control procedure will show strong
changes in behavior when the CS is presented. This simply means that important
changes not dependent upon a CS— US contingency occur in this situation; effects
due to that contingency still must be evaluated as deviations from the effects
produced by the truly random procedure.
Traditionally, the prime concern of American investigators has been the excitatory processes, and the inadequate conventional control procedures have reflected this concern. As noted above, many of these control procedures are biased toward the inhibitory side because of the explicit nonpairings of the CS and US. But the inhibitory effects of conditioning deserve attention in their own right. Clearly, we need an appropriate base condition against which to compare both the inhibitory and excitatory kinds of conditioning, relations. The truly random sequence of CSs and USs provides an unbiased control procedure for both positive and negative contingencies between CS and US. In fact, if we are going to retain the conceptual terms “conditioned excitatory” and “conditioned inhibitory” stimuli, the truly random control procedure will provide a base line against which to
define these
effects.
In addition to serving as a control condition for Pavlovian conditioning, the truly random presentation of CS and US provides an unbiased extinction procedure. To the degree that our concern in extinction of Pavlovian CRs is with how the animal loses its associative connection, simply removing the US from the situation is an inappropriate extinction procedure. Simple removal of the US eliminates not only the CS—US contingency but also whatever nonassociative effects the US might have. However, using the truly random presentation of CS and US as an extinction procedure permits examination of the loss of contingency-dependent learning independently of these other effects. Furthermore, the truly random procedure serves as an unbiased procedure for extinction of both excitation and inhibition. If inhibition can be acquired it seems reasonable that it can be extinguished. The truly random presentation of CS and US is the most natural extinction procedure for inhibitory as well as excitatory effects.
OBJECTIONS TO THE “TRULY
RANDOM” PROCEDURE: TWO
THEORETICAL VIEWS OF CONDITIONING
It seems certain
that our arguments will not be entirely convincing. All conventional control
procedures have a common feature: They never allow forward pairings of the
control CS and US. The reluctance which one might feel toward accepting
a truly random control procedure stems in part from the close temporal pairings
of CS and US which will occur by chance in that condition. One may
thus argue that the truly random control procedure itself allows Pavlovian
conditioning because of those few chance trials which pair CS and US; if
so, it can hardly be considered a “pure” control condition. According to
such an argument, the same processes may be operative in both the experimental
and control procedures, but to a lesser degree in the latter.
This objection
runs deep and is worthy of extensive examination. It rests upon an assumption,
often not made explicit, that the temporal pairing of CS and US is
the sufficient condition for “true” Pavlovian conditioning. It views Pavlovian
conditioning as a one-sided affair in which conditioning is either absent
or excitatory; the number of CS—US pairings determines the degree to which
conditioning is excitatory. It is this view which dominates American notions
of conditioning and which has been influential in preventing inhibitory
processes from playing a major role in our thinking. A good example of this
position is the Guthrian claim that the reinforcing event in Pavlovian conditioning
is simple contiguity between CS and US. From this point of view, a reasonable
control procedure for Pavlovian conditioning is one in which S “is not taught
that the US follows the CS.” This has been interpreted to include the possibly
quite different learning that “the CS is not followed by the US.”
With this type of bias, it might be reasonable to conclude that the “explicity
unpaired” and the discriminative conditioning procedure are appropriate
controls for Pavlovian conditioning.
An alternative
theoretical view of Pavlovian conditioning, and one which has not often
been distinguished from that in the previous paragraph, is that the temporal
contingency between CS and US is the relevant condition. The notion
of contingency differs from that of pairings in that the former
includes not only what is paired with the CS but also what is
not paired with the CS. Thus the truly random procedure contains no
contingency between the CS and US, even though it does contain some
chance CS—US pairings. From this point of view the appropriate control condition
for Pavlovian conditioning is one in which the animal is taught that “the
CS is irrelevant to the US.” Deviations from this base conditioning can
be either positive (CS is followed by US) or negative (CS is followed by
absence of US). This view of conditioning has the advantage of separating
out, from the simple absence of conditioning, a conceptualized inhibitory
process which has a status equal to that of excitatory processes. Intuitively
it seems clear that learning that the US does not follow the CS is different
from failing to learn that the US follows the CS or learning that the CS
is irrelevant to the US. In this sense, at least, the contingency view of
conditioning, and the truly random control procedure which it generates,
is more in the spirit of Pavlovian theory.3
The idea of contingency
used here needs explication. By it we mean the degree of dependency which
presentation of the US has upon prior presentation of the CS. This is clearly
a function of the relative proportion of US events which occur during or
at some specified time following the CS. Thus, in the truly random condition
no dependency exists, but in the standard Pavlovian conditioning situation
the dependence is complete. The control condition is brought closer to the
experimental condition as we increase the proportion of USs occurring in
the presence of the CS. When, at the other extreme, all USs occur in the
absence of the CS, the inhibitory end of the continuum is reached. These
proportions can be stated in terms of the probability of a US occurring
given the presence of a CS (or given that the CS occurred at some designated
prior time), and the probability of a US occurring given the absence of
the CS (cf. Prokasy, 1965). The dimension of contingency is then a function
of these two probabilities; if Pavlovian conditioning is dependent upon
the contingency between CS and US, it, too, will be a function of these
two probabilities. However, no attempt is made here to specify a particular
function which relates these two probabilities to a continuum of contingencies.4
If two conditioning
procedures have the same probability of reinforcement in the absence of
the CS, but have different probabilities in the presence of the CS, they
differ in what is usually called the degree of partial reinforcement. Whether
or not this affects the degree of contingency depends upon the function
of these two probabilities that we choose to describe degree of contingency.
We suggest that the contingency dimension, rather than the number of CS—US
pairings, is the theoretically fruitful dimension in Pavlovian conditioning.
As soon as one
admits a symmetry of inhibition and excitation in the Pavlovian conditioning
situation, the CS—US pairing view of conditioning begins to lose appeal.
Pavlovian conditioning consists of a sequence of CSs and USs arranged in
a particular temporal pattern. Suppose, now, that one is primarily interested
not in excitatory processes but in inhibitory processes, or in how an animal
learns that the CS signals a period free from the US. From the point of
view that the pairing of CS and US is the important Pavlovian event, the
truly random control procedure is inadequate for a reason that is exactly
the opposite to what it was for excitatory conditioning; now it contains
a number of non pairings of CS and US. Therefore, from such a view
we are forced to conclude that the symmetrical control procedure for the
study of inhibitory processes is to consistently pair CS and US.
This, it seems, is less than sensible.
It may also be
argued that the truly random control procedure does more than simply remove
the contingency of Pavlovian conditioning. It might, for instance, introduce
a new process of its own such as increasing the likelihood that S will ignore
or habituate to the CS since it bears no relation to the US. This is, of course, possible; but
it means that the arrangement of a contingency affects the rate at which
S comes to “ignore” a CS. Thus this ignoring of a CS is governed by its
associative relation to the US and is a proper part of the development of
a CR. From the point of view of this paper, then, the truly random control
procedure still provides the appropriate control.
Another objection
to the truly random control procedure rests again upon the notion that the
pairing of CS and US is the significant event for Pavlovian conditioning.
One can claim “what is random for the experimenter may not be random for
S.” Such an objection argues that if we use the truly random control, we
should arrange it so that the relation between the CS and US is phenomenally
random. One suspects that, at least in part, this objection is based upon
the notion of pairing of CS and
US. The statement implies that even though CS
and US are not related, S will behave as if the CS predicts the US. Those
who make this claim are rarely concerned that S will behave as if the CS
predicts no US!
It is, of course,
possible that some process which normally produces Pavlovian conditioning
when the US is made contingent upon the CS is operative even when the CS
and US are presented in random fashion. Such a process might fail to operate
only when there is a slight inhibitory, or only when there is a slight excitatory,
contingency between CS and US. In its most general form, this argument says
that the limits of our operational procedures do not necessarily define
the limits of psychological processes in the organism. It is difficult to
disagree. On the other hand, this is not an objection which applies uniquely
to the truly random control procedure. For instance, it applies also to
the traditional controls for Pavlovian conditioning: What is explicitly
unpaired for E may not be explicitly unpaired for S. A solution to this problem requires
an ability, which we do not yet have, to identify psychological processes;
until we do, there is little choice but to associate psychological processes
in Pavlovian conditioning with experimental operations.
A major advantage
of the contingency view of Pavlovian conditioning is that it provides a
continuum of CS—US contingencies along which a zero point can be located.
In the long run, the location of this zero with respect to process is not
crucial; if we discover that the assumed correspondence between experimental
contingency and psychological process is in error, it may be that results
can be brought into line by relocating the point of “zero contingency.”
TWO
EXPERIMENTAL PREDICTIONS
The truly random
control has led to the consideration of two theoretical views of Pavlovian
conditioning, the pairing view and the contingency view. The difference
between these two theoretical conceptions of Pavlovian conditioning is partly
semantic. From our present knowledge it is arbitrary whether we wish to
have a point of “zero” conditioning with deviations on both sides or a zero
point from which deviations can occur only in one direction. On the other
hand, the difference is also partly empirical, and in this framework the
question is whether the number of CS—US pairings, or the relative
probabilities of US in the presence and absence of CS, is the determinant
of Pavlovian conditioning. A comprehensive empirical answer to this question
requires an extensive program of research, but two specific predictions
can be extracted for illustrative purposes.
The area of most
blatant disagreement between the two conceptions of conditioning is the
notion of inhibition. (a) The pairing viewpoint fails to distinguish
between Ss failing to learn and Ss learning that the CS and US are explicitly
unpaired. Experimentally,
in accord with the pairing view, a CS which has been repeatedly presented
alone should not differ from one which has been explicitly unpaired with
the US. (This simple statement of the prediction neglects the operation
of such factors as sensitization which would produce more CRs in the explicitly
unpaired condition.) (b) From the viewpoint that CS—US contingencies
are the important determinants of Pavlovian conditioning, repeated CS
presentations may result in failure to condition; but, explicitly unpairing
CS and US should lead to the development of inhibitory phenomena. Thus,
under some circumstances, the contingency viewpoint predicts a difference
between the outcomes of these two treatments and the pairing view does not.
But it is important to note that the contingency approach only predicts
this difference when the CS is tested in the presence of some other excitatory
stimulus. Inhibitory effects can be measured only when there is some level
of excitation to be reduced. Again, at the risk of being pedantic, it is
important not to confuse the question of the presence or absence of inhibition
with the question of the ability to measure inhibitory effects.
The conditions
for testing the empirical fruitfulness of the contingency view were met
in an experiment by Rescorla and LoLordo (1965). In that experiment, two
groups of dogs were trained on a Sidman avoidance task. Both groups were
then confined and given Pavlovian conditioning treatments. While confined,
one group received repeated tone presentations without any shock USs, while
the other group received tones and shock sexplicitly unpaired in the manner
of Procedure 4 above. Later presentation of these tonal stimuli during the
Sidman avoidance performance led to a substantial reduction in avoidance
rate during the CS in the explicitly unpaired group and little change in
rate during the CS for the group that received only tones. Because previous
experiments supported the assumption that avoidance rate is in part a function
of the level of fear, these results were interpreted to indicate that explicitly
unpairing the CS and US led to the development of Pavlovian inhibitory processes
capable of reducing fear. Merely presenting the tones did not lead to this
result. The outcome of this experiment is consistent with the theoretical
view that CS—US contingency, rather than simply CS—US pairing determines
the outcome of Pavlovian conditioning procedures.
The two contrasting
views of Pavlovian conditioning also make differential pre4ictions for the
outcomes of excitatory conditioning procedures. Suppose that we condition
one group of Ss with a type of truly random conditioning procedure in which
USs are delivered on a variable interval schedule and CSs are randomly distributed
throughout the session. A second (experimental) group receives the identical
treatment except that the preprogrammed USs are allowed to reach S only
if they come in a 30-second period following a CS onset. Thus, for this
group a switch permits the delivery of the independently programmed USs
only for a period just after each CS. USs which are programmed for the truly
random Ss during other periods of the session never occur for the experimental
group. The Ss in this experimental group receive at least as many CS—US
pairings as do Ss in the truly random group, but USs can occur only following
CSs. If the number of CS—US pairings is important, then this procedure
should produce results similar to those of the truly random control. However,
if the CS—US contingencies are important, then a considerably greater
number of CRs should occur in the experimental group.
This conditioning
procedure was used in a paradigm like that of the Rescorla and LoLordo experiment
(Rescorla, 1966). All dogs were trained on a Sidman avoidance schedule.
Then, separately, half of the animals received the truly random control
treatment while the other half received the modified treatmentof the experimental
group described above. Shock was the US and tones served as CSs. After these
conditioning treatments, the tones were presented during performance of
the avoidance response. The CS of the truly random group had little effect
upon performance, while the CS of the experimental group showed marked fear-producing
properties, increasing the avoidance response rate. Again, this result supports
the view that the important dimension in Pavlovian conditioning is the CS—US
contingency rather than CS—US pairing.
These are but two
examples of the kinds of experiments which the contingency view of Pavlovian
conditioning generates. The fact that the results of these experiments support
the fruitfulness of the contingency view suggests a program of research
varying the relative probabilities which form the basis of the CS— US contingencies.
In this way we can explore the relations between CS—US contingencies and
Pavlovian conditioning.
In summary, we
have argued that the conventional control procedures for Pavlovian conditioning
are inadequate in a variety of ways. An alternative procedure, in which
the CS and US bear no relation to each other, was proposed, It was argued
that the failure previously to use this procedure stems from a particular,
and probably inadequate, conception of Pavlovian conditioning. Taking seriously
the truly random control procedure, we proposed an alternative theoretical
view of Pavlovian conditioning in which the CS—US contingency is important
rather than the CS—US pairing. The empirical usefulness of this alternative
view has been illustrated.
NOTES
|
1.
|
The
preparation of this paper and the experimental work related to it
were aided by United States Public Health Service Grant MH—04202 and
National Science Foundation Grant GB—2428 to Richard L. Solomon and
by a National Science Foundation predoctoral fellowship to the author.
The author would like to express his appreciation to Vincent M. LoLordo
and Richard L. Solomon for their advice and criticism of the ideas
presented in this paper. |
|
2.
|
A
similar control procedure has been suggested by Jensen (1961) and by
Prokasy (1965). |
|
3.
|
It
is worth pointing out that the argument advanced in this paper has direct
analogues for instrumental training. Whatever faults it might have,
the yoked-control procedure was introduced precisely to determine what
effects are uniquely due to instrumental reinforcement contingencies.
Similarly, the distinction between pairing and contingency views has
recently been examined for operant conditioning by Premack (1965). |
|
4.
|
These
probabilities can be calculated whatever the number of CS and US events.
If, for instance, there is only one CS—US pairing, there is a high degree
of contingency since the probability of a US following a CS is one and
the probability of a US in the absence of the CS is zero. However, it
may turn out empirically that with only a few CS and US events the relative
importance of single pairings is greater. |
Consequence
in Avoidance Learning
JOHN
GARCIA and ROBERT A. KOELLING, Harvard Medical
School
and Massachusetts General Hospital
An audiovisual stimulus
was made contingent upon the rat’s licking at the water spout, thus making
it analogous with a gustatory stimulus. When the audiovisual’ stimulus and
the gustatory stimulus were paired with electric shock the avoidance reactions
transferred to the audiovisual stimulus, but not the gustatory stimulus.
Conversely, when both stimuli were paired with toxin or x-ray the avoidance
reactions transferred to the gustatory stimulus, but not the audiovisual
stimulus. Apparently stimuli are selected as cues dependent upon the nature
of the subsequent reinforcer.
A great deal of evidence stemming from diverse
sources suggests an inadequacy in the usual formulations concerning reinforcement.
Barnett (1963) has described
the “bait-shy” behavior of
wild rats which have survived a poisoning attempt. These animals utilizing
olfactory and gustatory cues, avoid the poison bait which previously made
them ill. However, there is no evidence that they avoid the “place” of the
poisoning.
In a recent volume
(Haley & Snyder, 1964) several authors have discussed studies in which
ionizing radiations were employed
METHOD
The apparatus was
a light and sound shielded box (7 in. x 7 in. x 7 in.) with a drinking
spout connected to an electronic drinkometer which counted each touch of
the rat’s tongue to the spout. “Bright-noisy” water was provided by connecting
an incandescent lamp (5 watts)
and a clicking relay into this circuit. “Tasty” water was provided by adding flavors to the drinking supply.
Each experimental
group consisted of 10 rats (90 day old Sprague-Dawley males) maintained
in individual cages without water, but with Purina Laboratory chow ad
libidum.
The procedure was:
A. One week of habituation to drinking in the apparatus without stimulation.
B. Pretests to measure in-take
of bright-noisy water and tasty water prior to training. C. Acquisition
training with: (1) reinforced trials where these stimuli were paired with
reinforcement during drinking, (2) nonreinforced trials where rats drank
water without stimuli or reinforcement. Training terminated when there was
a reliable difference between water intake scores on reinforced and nonreinforced trials. D. Post-tests to measure intake
of bright-noisy water and
tasty water after training.
In the x-ray study
an audiovisual group and a gustatory group were exposed to an identical radiation schedule. In the other studies
reinforcement was contingent upon the rat’s response. To insure that both
the audiovisual and the gustatory stimuli received equivalent reinforcement,
they were combined and simultaneously
paired with the reinforcer during acquisition training. Therefore, one group
serving as its own control and divided into equal subgroups, was tested
in balanced order with an audiovisual and a gustatory test before and after
training with these stimuli combined.
One 20-min. reinforced
trial was administered every three days in the x-ray and lithium chloride
studies. This prolonged intertrial
interval was designed to allow sufficient time for the rats to recover from
acute effects of treatment. On each interpolated day the animals received
a 20-min. nonreinforced trial. They were
post-tested two days after their last reinforced trial. The x-ray groups
received a total of three reinforced trials, each with 54 r of filtered
250 kv x-rays delivered in 20 min. Sweet water (1 gin saccharin per liter)
was the gustatory stimulus. The lithium chloride group had a total of five
reinforced trials with toxic salty water (.12 M lithium chloride). Nontoxic
salty water (.12 M sodium chloride) which rats cannot readily distinguish
from the toxic solution was used in the gustatory tests (Nachman, 1963).
The immediate shock
study was conducted on a more orthodox avoidance schedule. Tests and trials
were 2 min. long. Each day for four consecutive acquisition days, animals
were given two nonreinforced and two reinforced trials in an NRRN, RNNR
pattern. A shock, the minimal current required
to interrupt drinking (0.5 sec. at 0.08—0.20 ma), was delivered through
a floor grid 2 sec. after the first lick at the spout.
The delayed shock
study was conducted simultaneously with the lithium chloride on the same schedule. Non-toxic salty water was
the gustatory stimulus. Shock reinforcement was delayed during first trials
and gradually increased in intensity (.05 to .30 ma) in a schedule designed
to produce a drinking pattern during the 20-min. period which resembled
that of the corresponding animal drinking toxic salty water.
RESULTS
AND DISCUSSION
The results indicate
that all reinforcers were effective in producing discrimination learning
during the acquisition phase (see Fig.
1), but obvious differences
occurred in the post-tests. The avoidance
reactions produced by x-rays and lithium chloride are readily transferred
to the gustatory stimulus but not to the audiovisual stimulus. The effect
is more pronounced in the x-ray study, perhaps due to differences in dose.
The x-ray animals received a constant dose while the lithium chloride rats
drank a decreasing amount of the toxic solution during training. Nevertheless,
the difference between post-test scores is statistically significant in
both experiments (p < 0.01 by ranks test).
Apparently when
gustatory stimuli are paired with agents which produce nausea and gastric
upset, they acquire secondary reinforcing properties which might be described
as “conditioned nausea.” Auditory and visual stimulation do not readily
acquire similar properties even when they are contingent upon the licking
response.
In contrast, the
effect of both immediate and delayed shock to the paws is in the opposite
direction. The avoidance reactions produced by electric shock to the paws
transferred to the audiovisual stimulus but not to the gustatory stimulus.
As one might expect the effect of delayed shocks was not as effective as
shocks where the reinforcer immediately and consistently followed licking.
Again, the difference between post-test intake scores is statistically significant
in both studies (p < 0.01 by ranks test). Thus, when shock which produces
peripheral pain is the reinforcer, “conditioned fear” properties are more
readily acquired by auditory and visual stimuli then by gustatory stimuli.
It seems that given
reinforcers are not equally effective for all classes of discriminable stimuli.
The cues, which the animal selects from the welter of stimuli in the learning
situation, appear to be related to the consequences of the subsequent reinforcer.
Two speculations are offered: (1) Common elements in the time- intensity
patterns of stimulation may facilitate a cross modal generalization from
reinforcer to cue in one case and not in another. (2) More likely, natural
selection may have favored mechanisms which associate gustatory and olfactory
cues with internal discomfort since the chemical receptors sample the materials
soon to be incorporated into the internal environment. Krechevsky (1933)
postulated such a genetically coded hypothesis to account for the predispositions
of rats to respond systematically to specific cues in an insoluble maze.
The hypothesis of the sick rat, as for many of us under similar circumstances,
would be, “It must have been something I ate.”
NOTE
|
1.
|
This research
stems from doctoral research carried out at Long Beach |
V. A. Hospital and
supported by NIH No. RH00068. Thanks are extended to Professors B. F. Ritchie,
D. Krech and E. R. Dempster, U. C. Berkeley, California.
Fantasy
by a Client-Controlled
Counterconditioning
Technique:
A
Case Study1
GERALD
C. DAVISON, State University
of New York at Stony Brook
To the best of the
author’s knowledge, this is the 1st report of the elimination of a sadistic
fantasy by conditioning methods, as well as the 1st to describe a client-controlled
technique for counterconditioning sexual responses. The mainstay of the
therapy entailed client-controlled masturbation sessions, in which strong
sexual feelings were paired with pictures and images of females in nonsadistic
contexts. This presumed positive counterconditioning was supplemented in
the consulting room by imaginal aversive counterconditioning (“covert sensitization”),
whereby an extremely disgusting scene was paired in imagination with a typical
sadistic fantasy. Furthermore, therapeutic change seemed to be facilitated
through the client’s reconstruction of his problem in conditioning terms,
rather than in terms of mental illness and putative unconscious processes.
The modification
of deviant sexual behavior has been approached
largely through the contiguous pairing of a primary aversive stimulus with
a stimulus eliciting an undesirable response (the “symptom”), the goal being
to endow the inappropri
Of particular relevance
to the present study is the work of Thorpe et al. (1963). These writers
report therapeutic benefit following presumably counterconditioning sessions
during which efforts were made to pair female pictures with orgasm from
masturbation. It was assumed that this intensely pleasurable sexual response
counterconditioned the aversion to females which appeared to play a crucial
role in the behavior of the homosexuals. These authors recognized the importance
of a person’s fantasy life to his overt behavioral adjustment, and they
assumed that beneficial generalization would occur from pictorial to the
real- life situation, similar to the assumptions made for systematic desensitization
(Davison, in press; Wolpe, 1958). Although the therapeutic outcomes reported
by Thorpe and his co-workers are equivocal in respect to actual sexual behavior,
the procedures did have considerable effect on fantasies.
The possibility
of extending this kind of work to an outpatient setting presented itself
to the author during the course of his private practice. Various modifications
of procedures used by Thorpe et al. (1963) were employed, apparently to
good effect. In addition, other important issues became evident in the course
of therapy, which required fewer than 5 consulting-room hours over a span
of 10 wk., and it is for these heuristic reasons that the following is reported.
CASE
STUDY
The client was
a 21-year-old unmarried white male college senior majoring in history. The
university counseling center had received an anxious letter from his parents,
requesting help for their son in treating his introversion, procrastination,
and “masochism.” After working with the student for a few weeks on his tendency
to wait until the last minute in his academic work, the psychologist at
the center referred him to the author for help with his sexual difficulties.
Mr. M’s statement
of the problem was: “I’m a sadist.” There followed a rather troubled account
of a complete absence of “normal” sexual fantasies and activities since
age 11. Masturbating about five times a week, the client’s fantasies had
been exclusively sadistic ones, specifically, inflicting tortures on women.
He declared emphatically that he had never been sexually aroused by any
other kind of image. Although generally uninterested in dating girls, he
felt no aversion to them; on the contrary, he sometimes felt a “warm glow”
when near them, but did not describe this at all in sexual terms. Because
of his extreme concern over the content of his fantasies, however, he had
dated very little and expressed no interest in the co-eds at the college.
He recalled having kissed only two girls in his life, with no sexual arousal
accompanying these fleeting episodes. He had never engaged in any homosexual
activities or fantasies. Although expressing no guilt about his problem,
he was very much worried about it inasmuch as he felt it impossible to ever
contemplate marriage. This concern had recently been markedly increased
upon reading an account of a Freudian interpretation of “sado-masochism.”
He was especially perturbed about the poor prognosis for this “illness.”
Because his concern
over the gravity and implications of his problem seemed at least as disruptive
as the problem itself, the therapist spent most of the first session raising
arguments against a disease interpretation of unusual behavior. Psychoanalytic
notions were critically reviewed, and attention was directed especially
to the intestability of many Freudian concepts (Levy, 1963). Instances in
the therapist’s own clinical work were cited to illustrate the liberating
effects observed in many people when they interpret their maladaptive behavior
as determined by “normal” psychological processes rather than by insidious
disease processes (cf. Davison, 1966; Glaser, 1965; Maher, 1966; Mainford,
1962). Mr. M frequently expressed relief at these ideas, and
the therapist, indeed, took
full advantage of his prestigious position to reinforce these notions.
At the end of the
session, the counterconditioning orientation which would be followed was
explained (Davison, in press; Guthrie, 1935; Wolpe, 1958), as well as the
specific activities which he was to engage in during the coming week. When
assured of privacy in his dormitory room (primarily on the weekend), he
was first to obtain an erection by whatever means possible— undoubtedly
with a sadistic fantasy, as he indicated. He was then to begin to masturbate
while looking at a picture of a sexy, nude woman (the “target” sexual stimulus);
Playboy magazine was suggested to him as a good source. If he began
losing the erection, he was to switch back to his sadistic fantasy until
he could begin masturbating effectively again. Concentrating
again on the Playboy picture, he was to continue masturbating, using
the fantasy only to regain erection. As orgasm was approaching, he was at
all costs to focus on the Playboy picture, even if sadistic fantasies
began to intrude. It was impressed on him that gains would ensue only when
sexual arousal was associated with the picture, and that he need not worry
about indulging in sadistic fantasies at this point. The client appeared
enthusiastic and hopeful as he left the office. (Table 1 summarizes the
client-controlled masturbation assignments following this and succeeding
consulting-room sessions.)
At the second session
he reported success with the assignment: he had been able to masturbate
effectively and enjoyably three times over the weekend to a particular picture
from Playboy without once having to use a sadistic fantasy; however,
it did take significantly longer to climax with the Playboy photograph
than with the usual kind of sadistic fantasy. During the rest of the week,
when he had not had enough privacy for real- life visual stimulation, he
had “broken down” a few times and used his sadistic fantasies. Much of this
session was then spent in talking to him about some of the social-sexual
games which most males play in our culture, especially the “mental undressing”
of attractive women. The purpose was to engage him in the kind of “stud”
conversation which he had never experienced and which, it was felt, would help to change his orientation toward girls. The therapist reassured him that
the first direct contacts with girls are sometimes disappointing; he had
to admit, however, that his extreme sensitivity about the sadistic fantasies
had severely limited his experience.
During the coming
week he was, first of all, to ask out on a coffee date any girl whom he
felt he might find attractive, even for a sadistic fantasy. He was
also to spend some time between classes just looking at some of the co-eds
and noting some of their more remarkable attributes. Finally, his masturbation
sessions were to be structured as follows: The real-life pictorial stimuli
were to be girls either in bathing suits or lingerie, used in the same way
as the Playboy picture the preceding week; this latter stimulus was
to be used as “back-up” stimulus, replacing the sadistic fantasies in the
event that he was losing his erection. Attention was also to be directed
to imaginal sexual stimuli, and when masturbating in this way he was to use the Playboy image, with a sadistic fantasy as back-up.
The third session lasted half an hour. He had procrastinated so long in asking for a date that the girls he contacted
had already made other plans;
the therapist expressed his disappointment quite openly and urged him even
more strongly to follow through with this task. He had managed to spend
some time looking at girls but did not note significant sexual arousal,
except when a sadistic fantasy crept in occasionally. He had masturbated
only once to real-life stimuli, using some bathing-suit pictures from a
weekly national news magazine; this was successful, though it took longer
even than when the Playboy material was used previously. When masturbating
to imaginal sexual stimuli, he had relied almost exclusively on his sadistic
fantasies rather than utilizing the Playboy picture in imagination
as he had in real life 1 wk. earlier.
His reluctance
to give up the sadistic fantasies prompted the use of the following procedure,
the idea for which had been obtained from Lazarus (1958). With his eyes
closed, he was instructed to imagine a typical sadistic scene, a pretty
girl tied to stakes on the ground and struggling tearfully
to extricate herself. While
looking at the girl, he was told to imagine someone bring. ing a branding
iron toward his eyes, ultimately searing his eyebrows. A second image was
attempted when this proved abortive, namely, being kicked in the groin by
a ferocious-looking karate expert. ‘When he reported himself indifferent
to this image as well, the therapist (depicted to him a large bowl of “soup,”
composed of steaming urine with reeking fecal boli bobbing around on top.
His grimaces, contortions, and groans indicated that an effective image
had been found, and the following 5 min. were spent portraying his drinking
from the bowl, with accompanying nausea, at all times while peering over
the floating debris at the struggling girl. After opening his eyes at the
end of the imaginal ordeal, he reported spontaneously that he felt quite
nauseated, and some time was spent in casual conversation in order to dispel
the mood.
His assignments
for masturbation during the coming week entailed increasing the frequency
of his real-life masturbatory exposures to bathing-suit pictures, along
with concerted efforts to use the Playboy stimuli in imagination
as he had in real life 2 wk. earlier, resorting to sadistic fantasies if
necessary.
The fourth session
lasted only 15 min. He had managed to arrange a date for the coming weekend
and found himself almost looking forward to it. Again, he had masturbated
several times to a real-life picture of a bathing beauty. In fantasy he
had managed to use the Playboy girl
exclusively two out of five times,
with no noticeable diminution in enjoyment.
He was to continue
using the bathing-suit pictures while masturbating to real-life stimuli,
but to avoid sadistic fantasies altogether,
the idea being that any frustration engendered by this deprivation would simply add to his general
sexual arousal and thereby
make it all the easier to use the Playboy stimuli in
imagination.
The fifth session,
also lasting only 15 min., opened with Mr. M animatedly praising the efficacy
of the therapy. He had masturbated several times, mostly to real-life bathing-suit
pictures, with no problems and, most importantly, had found himself unable
to obtain an erection to a sadistic fantasy. In fact, he even had difficulty
conjuring up an image. He had also spent considerable time with two girls,
finding himself at one point having to resist an urge to hug one of them—a
totally new experience for him. He enthusiastically spoke of how different
he felt about “normal dating,” and a 1-mo. period without interviews was
decided upon to let him follow his new inclinations.
The sixth session,
1 mo. later, revealed that his sadistic fantasies had not reappeared, and
that he had been masturbating effectively to both real-life and imaginal
appropriate sexual stimuli. He had not, however, been dating, and some time
was spent stressing the importance of seeking “normal” sexual outlets. He
felt strongly, however, that the sexual problem had been successfully handled
and requested that his procrastination problem be taken up. Two sessions
were subsequently devoted to following the same general strategy that had
been adopted, with some success, by the college counselor, that is, arranging
for various rewards to be made contingent upon certain academic task- performances.
Mr. M did report doing “an enormous amount of work” during 1 wk.—out of
fear of having to admit to the therapist
that he had been loafing. Practical considerations, how ever, made it clear that this handling of the problem, even if it should prove effective, was not as realistic
as his facing the reality that
there was no “magic pill” to eliminate his procrastination. Therapy, therefore,
was terminated, with no sadistic fantasies having occurred for over 1 mo.,
and with the problem of procrastination left more or less untouched.
A follow-up of
1. mo. was obtained by telephone. Mr. M reported that there was still no
sign of sadistic fantasies and that, indeed, he was no longer even thinking
about the issue. He had still not “gotten around” to asking any girl out
on a date, and the therapist urged him in no uncertain terms to tackle this
aspect of his procrastination problem with the vigor that he had shown in
regard to his studies (where significant improvement had been made). Extensive
and persistent questioning failed to evoke any reported aversion to girls
as the basis of his reluctance to ask them out.
DISCUSSION
As with every case
study, one must necessarily speculate, to a large extent, on the, “active
ingredients.” Hypotheses are not readily strengthened from such data. As
a demonstration of various strategies, however, the present report does
seem to be of heuristic value.
|
1.
|
The
first significant event in therapy was the author’s general reaction
to the client’s statement of the problem, “I’m a sadist.” After Mr.
M had recounted the horror with which he had read about his mysterious
“illness” in Freudian terms, the therapist countered with a logical
attack that made the hour take on more the characteristics of a graduate
seminar than a psychotherapy session, except perhaps for the warmth,
support, and acceptance which were deliberately conveyed. A key factor
in this initial phase was an attempt to change the client’s general
orientation to his problem. As this writer has usually found, the client
had been regarding himself as “sick,” qualitatively different from so-called
“normals.” Furthermore, the idea that much of his behavior was determined
by forces working in devious ways in his “unconscious” was quite troubling,
as was the poor prognosis. As reported in the case material, these issues
were dealt with immediately, and significant relief was afforded the
young man simply by reconstructing the problem for him in conditioning
terms. It would, indeed, have been interesting and valuable to attempt
some sort of assessment of improvement at this very point. |
|
2.
|
Inextricably
intertwined with the foregoing was the outlining of a therapeutic strategy:
his sadistic fantasies were to be attacked by procedures aimed at counterconditioning
the maladaptive emotional reactions to specific kinds of stimuli. The
client perceived the theoretical rationale as reasonable and was satisfied
with the actual techniques which would be employed. Furthermore, being
able to buttress the plan with both clinical and experimental data added
to its credibility. It must be emphasized that whether the data cited,
or the explanation offered, are valid is an irrelevant question in the
present situation. The important point is that the client’s enthusiastic
participation was enlisted in a therapeutic regime which, by all counts,
was to be highly unconventional. |
|
3.
|
A
third conceivably relevant variable was the “older brother” type of
relationship which the therapist established in talking with Mr. M about
conventional sex. Clearly the client had missed this part of the average
American male’s upbringing and, as has been reported, much time was
spent in deliberately provocative “locker-room talk,” not as an end
in itself, but rather as a means of exposing him to the kinds of heterosexual
ideations which seemed to the author useful in promoting nonsadistic
fantasies about girls. |
|
4.
|
It is likely that the two positive exposures to actual women contributed
to therapeutic improvement. Mr. M, having been goaded into direct social
contact with girls, was fortunately able to appreciate the enjoyment
that can come from a satisfactory relationship with a woman, albeit
on nonsexual terms. In addition, having felt a very strong urge to hug
one of them, in a nonsadistic fashion, was reported by the client as
a highly significant event and must surely have fostered some change
in his concept of himself as a sexual misfit. Furthermore, aside from
any alleged counterconditioning with respect to appropriate stimuli
(see below), it is also suggested that a favorable change in self-concept
developed as he saw himself able to respond sexually to imaginal and
pictorial stimuli that had previously left him unaroused. |
|
5.
|
It
is assumed that the most important variable in therapy was the masturbation
sessions which the client carried out privately. As discussed by Thorpe
et al. (1963), it was felt that more appropriate social-sexual behavior
would probably follow upon a change in sexual fantasies; in the present
case a focus on the fantasies seemed all the more reasonable in view
of the fact that they formed the basis of the referral. According
to the client, it was his fantasy life which had retarded his sexual
development, and it was this that he was most worried about. It was
assumed that generalization to real-life girls would be effected in
a fashion similar to the generalization which has been reported for
Wolpe’s technique of systematic desensitization (Davison, in press;
Lang & Lazovik, 1963; Lang, Lazovik, & Reynolds, 1965; Lazarus,
1961; Paul, 1966; Paul Sc Shannon, 1966; Rachman, 1966; Schubot, 1966;
Wolpin Sc Raines, 1966; Zeisset, 1966). Of course, whether Mr. M would
actually begin dating regularly, or at all, would seem to depend importantly
on factors other than those dealt with in this brief therapy, for example,
the client’s physical attractiveness, his conversational and sexual
techniques, the availability of women atLractive to him, and so forth.
The generalization spoken of here, then, is best restricted to the thoughts
and feelings which he had about women and about the prospects of relating
to them nonsadistically; the case-study data contain ample verification
for this. |
|
6.
|
Probably
very instrumental in changing the content of his fantasies was the intensive “imaginal aversive
counterconditioning” (or “covert sensitization,” viz, Cautela, 1966;
Lazarus, 1958) conducted by the therapist, in which extreme feelings
of disgust were generated by fantasy and then related to the sadistic
image. One can fruitfully compare this technique with the “emotive imagery”
procedure described by Lazarus and Abramovitz (1962), in which pleasant
images were generated in fearful children and then related by the therapist
to conditioned aversive stimuli. The procedure was resorted to in the
present case because the client appeared unable to give up the sadistic
fantasy solely on the basis of beginning to find the nonsadistic pictures
and images effective in maintaining erection and leading to orgasm.
|
The clinical data
on change in fantasy are self-reports, supplemented by the therapist’s inference
of the client’s credibility. The orderliness of response to therapy, along
with the enthusiasm which accompanied the progress reports, serves to bolster
the conclusion that Mr. M did, in fact, give up his sadistic fantasies of
10 years’ standing in favor of the kinds of fantasies which he felt were
a sine qua non for appropriate sociosexual behavior. Both preceding and
accompanying these changes was the radical difference in outlook. Simply
stated, Mr. M stopped worrying about himself as an “oddball,” doomed to
a solitary life, and did make some initial attempts to establish appropriate
relationships with girls. That he has not yet done so (as of this writing)
may, indeed, be due to a return of the original problem; however, this alternative
seems less likely than that verbalized by the client,
namely, that he has always
had trouble doing what he knows he ought to do, and that, above all, being
a so-called sexual deviate has ceased being an issue for him. Moreover,
as mentioned above, variables other than the content of fantasies would
seem to bear importantly on the matter of overt sexual behavior. Clearly,
if usual dating habits were to be used as a criterion for outcome, the therapy
must be considered a failure—although this would qualify many a young adult
as “maladjusted” or “abnormal.” Be that as it may, a relevant, well-established
class of behaviors was modified, setting the stage for a social adjustment
from which the client had initially seen himself utterly alienated.
Supplementary
Follow-Up Data
A follow-up report
was received by mail 16 mo. following termination. The client reported that,
since the therapy had so readily eliminated the arousal from sadistic fantasies,
and, most importantly, had altered his outlook for “normal” sexual behavior,
he allowed himself, “premeditatedly,” to return to the use of the sadistic
fantasies 6 mo. after termination, “...resolving to enjoy my fantasies until
June 1, and then to reform once more. This I did. On June 1 [1967], right
on schedule, I bought an issue of Playboy and proceeded to give myself
the treatment again. Once again, it worked like a charm. In two weeks, I
was back in my reformed state, where I am now [August 1967]. I have no need
for sadistic fantasies . .
. I have [also], been pursuing a vigorous (well,
vigorous for me) program
of dating. In this way, I
have gotten to know a lot of girls of whose existence I was previously only
peripherally aware. As you probably know, I was very shy with girls before;
well, now I am not one-fifth as shy as I used to be. In fact, by my old
standards, I have become a regular rake!”
A telephone call
was made to obtain more specific information about his return to the sadistic
fantasies. He reported that the return was “fairly immediate,” with a concomitant
withdrawal of interest in conventional sexual stimuli. His self-administered
therapy in June 1967 followed the gradual pattern of the original therapy,
although progress was much faster. The author advised him not to make any
more “premeditated” returns,
rather to consolidate his
gains in dating and other conventional heterosexual activities and interests.
The client indicated that this plan could and would be readily implemented.
Of the past 16
mo., then, the client has been free of the sadistic fantasies for 7 mo.,
the other 9 mo. involving what he terms a willful return for sexual stimulation
while masturbating. Constant throughout this follow-up period has been the
relief which he derived from finding himself able to respond sexually to
conventional sexual stimuli. Additional gains are his dating activities,
which, it will be recalled, were not in evidence while the writer was in
direct contact with him.
Still aware of
the limitations of these case-study data, it does seem noteworthy and possibly
quite important that the client’s self-initiated partial “relapse” took
place in a step-wise fashion, that is, without a gradual reorientation
to the sadistic fantasies:
he reported himself
almost immediately excited by them once he had made the decision to become
so. This sudden shift raises questions as to whether “aversive counterconditioning”
underlay the indifference to the fantasies which was effected during therapy.
This surprising finding also underlines the probable importance of other-than-conditioning
variables in the treatment.
NOTE
Relative
Efficacy of Desensitization
and
Modeling Approaches for
Inducing
Behavioral, Affective, and
Attitudinal
Changes’
ALBERT BANDURA, EDWARD B. BLANCHARD, and BRUNHILDE RIFLER, Stanford University
The present study
investigated basic change processes accompanying several social-learning
procedures from the perspective of a dual-process theory of avoidance behavior.
Snake-phobic subjects were administered either symbolic desensitization,
symbolic modeling, live modeling combined with guided participation (contact
desensitization), or they received no treatment. All three approaches produced
generalized and enduring reductions in fear arousal and avoidance behavior
as well as positive changes in attitudes. Of the three methods, modeling
with guided participation proved most powerful, achieving virtually complete
extinction of phobic behavior in every subject. Moreover, those who attained
only partial improvement through the other treatments displayed total extinction
of phobic behavior after a brief period of modeling with guided participation.
Consistent with social-learning theory, the favorable changes produced toward the phobic
object were accompanied by fear reduction toward threatening situations
beyond the specifically treated phobia, the decrements being roughly proportional to the potency of the treatments
employed. Moderately high positive correlations were found between behavioral
and attitudinal changes. Some evidence was obtained that modeling procedures
expedite behavioral changes through vicarious extinction of fear arousal
to aversive stimuli below the threshold for activating avoidance responses,
thus enabling persons to perform approach behaviors. Direct contact with
threats that are no longer objectively justified provides new experiences
that further extinguish residual anxiety and augment attitudinal changes.
Psychological approaches
to the modification of human behavior have relied heavily upon verbal influence
procedures. It would appear from the results of psychotherapy outcome studies
that the popularity of such methods is attributable more to their ease of
application than to their demonstrated effectiveness. Recent years have
witnessed a rapid growth of new treatment approaches that achieve psychological
changes mainly through guided learning experiences (Bandura, 1969a). The
present experiment was principally designed to assess the differential efficacy
of several of these approaches for inducing behavioral, affective, and attitudinal
changes in phobic subjects, and to investigate certain issues pertaining
to basic change processes.
The research reported
in this paper is guided by the dual- process theory of avoidance behavior.
According to this view, threatening stimuli evoke emotional arousal which
has both autonomic and central components. It is further assumed that these
arousal processes, operating primarily at the central level, exercise some
degree of control over instrumental avoidance responding.
The influential
role of arousal mediators in avoidance behavior is most clearly demonstrated
by Solomon and his colleagues (Rescorla & Solomon, 1967; Solomon &
Turner, 1962). In these studies, which use a three-stage paradigm, animals
first learn to make an avoidance response to a light stimulus. They are
then skeletally immobilized by curare to prevent avoidance responses from
being conditioned directly to external stimuli, and shock is paired with
one tone, while a contrasting tone is never associated with aversive stimulation.
In subsequent tests the animals display essentially the same degree of avoidance
in response to the negatively valenced tone and the light, both of which
evoke common arousal reactions, whereas avoidance responses rarely occur
to the neutral tone. Considering that the light and the tones were never
associated, and assuming that the curare blocked all skeletal activity (Black,
1967), thus precluding any differential conditioning of avoidance responses
to the tones, the controlling power of the aversive auditory stimulus must
be mediated either through events in central systems or through autonomic
feedback mechanisms.
There is evidence
that avoidance responses can be acquired and maintained in sympathectomized
animals (Wynne Sc Solomon, 1955), and that avoidance behavior persists long
after autonomic responses have been extinguished (Black, 1959; Notterman,
Schoenfeld, Sc Bersh, 1952). Moreover, the latency of autonomic reactions
is much longer than that of skeletal responding; consequently, avoidance
behavior is typically executed before autonomic reactions could possibly
be elicited. Findings of the preceding studies indicate that behavior is
in large part centrally regulated rather than under autonomic control, as
is commonly assumed in peripheral theories of anxiety.
It would follow
from the dual-process theory that if the arousal capacity of subjectively
threatening events is extinguished, then both the motivation and one set
of controlling stimuli for avoidance behavior are removed. Black (1958)
has shown in experiments with curarized subjects that neutralization of
an aversive stimulus through repeated presentation without any accompanying
adverse experiences markedly facilitates elimination of avoidance behavior.
The psychological procedures investigated in the present study are likewise
predicated on the assumption that extinction of fear arousal will reduce
phobic behavior.
The method of systematic
desensitization (Wolpe, 1958) attempts to eliminate fear arousal through
repeated pairing of imaginal representations of threatening situations with
deep relaxation. Wolpe explains the effects of this form of treatment in
terms of reciprocally inhibitory processes occurring at the level of the
autonomic nervous system. These theoretical speculations about the mechanisms
governing the counterconditioning process are largely disputed by empirical
findings (Bandura, 1969a); nevertheless, numerous well-designed experiments
(Davison, 1968; Krapil 1967; Lang, Lazovik Sc Reynolds, 1965; Mealiea, 1967;
H. R. Miller, 1967; Moore, 1965; Paul, 1966; Paul Sc Shannon, 1966; Schubot,
1966) have shown that the systematic desensitization procedure produces
significant reduction in avoidance behavior.
Fear arousal can
also be eliminated on a vicarious basis. These vicarious extinction effects
are achieved by having persons observe models performing fear-provoking
behavior without any adverse consequences accruing to the performers (Bandura,
1968). The absence of anticipated negative consequences is a requisite condition
for fear extinction. Hence, the modeled displays most likely to have strong
effects on phobic observers are ones in which performances they regard as
hazardous are repeatedly shown to be safe under a variety of threatening
circumstances. However, presentation of modeled approach responses toward
the most aversive situations at the outset is apt to generate in observers
high levels of emotional arousal that can impede vicarious extinction.
Avoidance responses
can be extinguished with minimal distress if persons are exposed to a graduated
sequence of modeling activities beginning with displays that have low arousal
value (Bandura, Grusec, & Menlove, 1967; Bandura & Menlove, 1968).
After emotional reactions to attenuated threats have been extinguished,
progressively more aversive modeling events, which are weakened by generalization
of fear extinction from preceding displays, are gradually introduced and
neutralized. Stimulus graduation is not a necessary condition for vicarious
extinction, but it permits greater control over the change process and it
entails less fear elicitation than approaches involving repeated exposure
to modeling events having high threat value.
‘When fear arousal
is extinguished to symbolic representations of threats, as in systematic
desensitization and symbolic modeling, one would expect some loss in generalization
of fear reduction to actual events because they constitute more severe threats.
In instances where fear arousal is not reduced below the threshold for activating
avoidance behavior, persons will be unable to perform highly threatening
approach responses even though their fear has been extinguished to the symbolic
equivalents. One might expect less transfer loss when fear arousal is extinguished
through both vicarious and direct experiences with the actual threatening
events. However, when desired behavior is severely inhibited active response
guidance procedures may be required, in addition to graduated modeling,
to reinstate approach behavior.
The third approach
consists of a form of treatment combining graduated live modeling with guided
participation. The principal elements of this method were originally applied
by Ritter (1965) and further developed as contact desensitization (Ritter,
1968a, 1968b, 1969a). In the procedure employed in the present study, the
model initially demonstrates the desired behavior under secure observational
conditions, after which individuals are aided through further demonstration
and joint performance to execute progressively more difficult responses.
Whenever subjects are unable to perform a given behavior upon demonstration
alone they are assisted physically by enacting the activities concurrently
with the model. The physical guidance is then gradually reduced until all
subjects are able to perform the behavior without assistance.
In the present
experiment subjects who suffered from snake phobias received either systematic
desensitization, symbolic modeling, live modeling with guided participation,2
or were assigned to a nontreated control group. Prior to, and upon completion
of, their respective treatment programs subjects were administered a behavioral
avoidance test to measure the strength of their fear and avoidance of snakes.
In addition they completed a comprehensive fear inventory to determine whether
extinction of fear of snakes is associated with changes in other areas of
anxiety. Attitude measures were also included to furnish information regarding
the interesting but inadequately explored attitudinal effects of behavioral
changes induced through social-learning procedures.
It was predicted
that all three treatment approaches would extinguish both fear arousal and
avoidance behavior, but that live modeling combined with guided participation
would prove superior in this respect. No predictions were advanced, however,
concerning the relative efficacy of systematic desensitization and symbolic
modeling.
Psychodynamic theories
generally assume that anxieties are internally generated by arousal of repressed
impulses which are then displaced and projected onto environmental events.
External phobic objects are therefore regarded as pseudoevocative stimuli.
Thus, for example, snake phobias are believed to reflect phallic anxieties
(Fenichel, 1945). From this point of view, direct neutralization of a given
phobic stimulus should either have no lasting effect, or result in the emergence
of new phobic disorders because the underlying source of anxiety has not
been eliminated.
According to social-learning
theory, extinction of the arousal potential of a phobic stimulus should
produce not only stable decreases in avoidance behavior, but some reduction
in anxiety in other areas of functioning on the basis of stimulus generalization.
The transfer of extinction effects would be expected to vary as a function
of level of fear extinction achieved toward the treated phobic stimuli and
the degree of similarity between the neutralized and the other sources of
anxiety.
Considering both
the extensive research on attitude change and the powerful controlling functions
that are often conferred upon attitudes, there has been surprisingly little
investigation of the relationship between attitudinal and behavioral change.
A thorough search of the literature by Festinger (1964) yielded only a few
studies which disclose that changes in attitudes produced by persuasive
communications have little or no effect upon the performance of corresponding
actions. There is some reason to suppose that the degree of relationship
between attitudinal and behavioral change may be partly determined by the
affective and social consequences of the behavior being modified and by
the method of influence used to bring about the change.
One can distinguish
among three basic modes of attitude change. The cognitive-oriented approach
attempts to modify persons’ attitudes by altering their beliefs about the
attitude object through various forms of persuasive communications. As noted
above, this method can produce changes in attitudes but it often has little
effect upon overt actions. A second general strategy involves an affect-oriented
approach wherein both evaluations of, and behavior toward, particular
attitude objects are modified by altering their emotion-arousing properties,
usually through direct or vicarious conditioning procedures. The third approach
relies upon a behavior-oriented strategy. Results of the latter procedure
provide considerable evidence that attitudinal changes can be successfully
achieved by getting a person to engage in new behavior in relation to the
attitude object without untoward consequences.
The relative modifiability
of attitudes and actions, and the degree of correspondence obtained between
changes in these two sets of events, may vary with the affective consequences
accompanying the behavior. A given social influence might produce analogous
changes in both attitude and action when persons are indifferent to, or
favorably disposed toward, performing the advocated activities. Most attempts
to control consumer behavior through persuasive communications, for example,
would fall in this category. The process is much more complicated, however,
when persons resist advocated behavior that they can perform because it
results in self-devaluation, or when they are amenable to engaging in the
desired activities but are unable to do so because of strong fears and inhibitions.
In the latter instances, a weak method may alter responses that are readily
susceptible to change, such as verbal evaluations, but fail to modify overt
behavior which is rendered intractable by its adverse consequences. A relatively
powerful influence would be required to achieve correlative changes.
Two of the treatment
procedures investigated in the present experiment, namely systematic desensitization
and symbolic modeling, are designed to produce changes primarily by extinguishing
emotional arousal to symbolic representations of the phobic object. Modeling
with guided participation, on the other hand, eliminates emotional arousal
to actual threats and it also provides new direct experiences with the previously
avoided object that can serve as a further basis for modifying attitudes.
It was predicted that all three treatment procedures would produce favorable
changes in attitudes. On the assumption that a method operating through
behavior change furnishes an objective and genuine basis for new evaluations,
it was expected that modeling with participation would achieve the greater
modification in attitudes.
There is some evidence
from laboratory studies of counter- conditioning processes (Gale, Sturmfels,
Sc Gale, 1966; Poppen, 1968) that the aversive properties of threatening
stimuli can be extinguished more effectively when administered in conjunction
with anxiety-neutralizing events than when presented alone. In the present
experiment the symbolic modeling treatment was administered in conjunction
with relaxation procedures. Several investigators have also demonstrated
that relaxation can reduce physiological arousal to both imagined and external
threats (Grings Sc Uno, 1968; Paul, 1969), and that it increases tolerance
of psychologically aversive stimuli (Davison, 1968; Schubot, 1966). In order
to evaluate the contribution of relaxation to changes accompanying symbolic
modeling, after the posttreatment assessment was completed, subjects in
the control group received the symbolic modeling treatment except that they
did not utilize relaxation to counteract fear arousal. It was hypothesized
that symbolic modeling combined with relaxation would achieve more rapid
vicarious extinction of fear arousal, and greater changes in behavior, attitudes,
and emotional responsiveness, than symbolic modeling alone.
Under conditions
where a given influence procedure exercises weak behavioral control, other
variables (e.g., personality characteristics of change agents, attributes
of the recipients, and minor variations in the procedure) are likely to
emerge as influential determinants of change. However, if a method is sufficiently
powerful it should be able to override such influences. In order to demonstrate
that in cases exhibiting only partial improvement the major deficits may
reside in the method rather than in the recipient, all subjects who failed
to achieve terminal performances, including the treated controls, were subsequently
administered the treatment combining live modeling with guided participation.
Upon completion of this supplementary treatment program, each subject was
again administered the regular assessment procedures. Approximately 1 month
after subjects concluded their treatment they returned for a follow-up assessment
to evaluate the durability of the established changes.
METHOD
Subjects
Subjects were recruited
through an advertisement placed in community newspapers. The advertisement
requested volunteers to participate in an experiment testing procedures
designed to eliminate fear of snakes. The sample also contained a small
number of students recruited from an introductory course in psychology.
Of the total number
of 48 subjects who qualified for inclusion in the study on the basis of
a behavioral avoidance test, 5 were males and 43 were females. They varied
in age from 13 to 59 years, with a mean age of 27 years.
In virtually all
cases the phobia unnecessarily restricted subjects’ activities and adversely
affected their psychological functioning in various ways. Some of the people
were unable to perform their occupational duties in situations where there
was any remote possibility that they might come into contact with snakes;
others could not take part in recreational activities such as hunting, camping,
hiking, or gardening, because of their dread of snakes; and still others
avoided purchasing homes in rustic areas, or experienced marked distress
whenever they would be unexpectedly confronted with pet snakes in the course
of their social or occupational activities.
In the preliminary
assessment subjects completed a questionnaire in which they were asked to
describe any direct or vicarious aversive experiences that they or members
of their families had had in relation to snakes, to indicate the onset of
their fear, and to note any familial modeling of snake phobic behavior.
Pretreatment
Measurement of Attitudes
In the present
study attitudes are conceptualized as evaluative responses. Subjects’ attitudes
toward snakes were measured in two ways. First, they were administered six
attitude scales describing various encounters with snakes such as visiting
a reptile exhibit, being unexpectedly shown a documentary film on the habits
of snakes, encountering a snake on a hike, keeping snakes in the home, and
handling snakes and caring for them. Subjects were instructed to rate each
item on a 7-point scale which indicates strong enjoyment at one end, strong
dislike on the other, and indifference at the midpoint. The mean of the
six ratings constituted the attitude score.
Subjects’ attitudes
toward snakes were also assessed in terms of evaluative dimensions of the
semantic differential technique. The form used consisted of eight bipolar
adjective rating scales using the following pairs of contrasting adjectives:
good-bad, clean-dirty, ugly-beautiful, belligerent-peaceful, interesting-dull,
worthless-valuable, nice-awful, and pleasant-unpleasant. The pooled ratings
obtained from these scales were averaged to provide a summary evaluative
score.
Behavioral
Avoidance Test
The test of avoidance
behavior, which was similar to the one employed by Schubot (1966), consisted
of a graded series of 29 performance tasks involving increasingly more threatening
interactions with a 4-foot king snake. The tasks required subjects to approach
the snake in an enclosed glass cage, to look down at it, to touch and hold
the snake with gloved and then bare hands, to let it loose in the room and
then to replace it in the cage, to hold it within 5 inches of their faces,
and finally to tolerate the snake crawling in their laps while they held
their hands passively at their sides.
Prior to the test
of avoidance behavior subjects were given some factual information about
the characteristics of reptiles. They were told that snakes are dry rather
than slimy, that they feel cool to the touch because they are cold-blooded
and take on the temperature of their surroundings, and that they often flick
their tongues because they have faulty vision and use the tongue to explore
their environment. This information was provided in order to exclude moderately
fearful subjects who might achieve performance gains on the basis of incidental
information alone derived from testing and treatment experiences. Moreover,
by introducing informational factors prior to measurement of the behavioral
base line their potential influence was eliminated from the effects produced
by the treatment operations.
Subjects were tested
individually by a female experimenter. Those who were unable to enter the
room containing the snake were given a score of zero; subjects who could
go in were asked to perform the items in the graded series of tasks. To
control for any possible influence of expressive and postural cues from
the experimenter, she stood behind the subject and read aloud the approach
responses to be performed. She also rated the snake’s activity level and
recorded whether or not the subject successfully completed each test item.
In order to evaluate scorer reliability17 of the behavioral tests, randomly
selected from pretreatment, posttreatment, and follow-up phases of the experiment,
were scored simultaneously but independently by another rater who observed
the test sessions through a one-way mirror. The inter- rater agreement was
100 percent for approach responses and 92 percent for snake activity level.
The subject’s score
on the behavioral test was the number of approach tasks he was able to perform.
Those who succeeded in lifting the snake inside the cage with a gloved hand
for 5 seconds or more were eliminated from the experiment. On the basis
of this selection criterion, 38 percent of the subjects who had defined
themselves as snake phobic proved, much to their surprise, relatively fearless
in the behavioral test. Subjects were excluded only on the basis of approach
behavior without regard to any other psychological characteristics so as
to increase generality of the findings.
Fear Arousal
Accompanying Approach Responses
In addition to
measuring the attitudinal and behavioral effects of the different treatment
approaches, their efficacy in eliminating the fear arousal potential of
phobic objects was also assessed. During the behavioral test subjects were
asked to rate orally, in terms of a 10-interval scale, the intensity of
the fear they’ experienced when each snake approach response was described
to them and again while they were performing the corresponding behavior.
The scores, averaged across the responses that each subject was able to
complete, served as measures of anticipatory fear arousal and performance-related
fear.
Immediately after
the behavioral avoidance test was completed subjects were readministered
the attitude scales and the semantic differential to obtain a new attitudinal
base line reflecting any changes resulting from receipt of factual information
and exposure to an actual snake.
Appraisal
of Fear Proneness
As the final task
in the pretreatment assessment subjects completed a comprehensive fear inventory
to determine whether elimination of fears concerning snakes is associated
with concomitant changes in other areas of anxiety. The inventory, which
contained100 items, included 72 from the test developed by Wolpe and Lang
(1964) plus an additional 28 items designed to provide 20 items in each
of the following five fear categories: animals; social situations and interpersonal
behavior; physical afflictions and injuries; classical phobias; and a group
of miscellaneous fears.
Subjects were asked
to rate their emotional responses to each object or situation in terms of
a 5-point scale describing increasing degrees of fearfulness. Two sets of
scores were derived from this test. One was the number of situations that
were rated as fear provoking and the other was a fear-intensity measure
obtained by assigning to each item numerical values ranging from 0 to 4
depending upon the level of fear checked. Number and intensity of fear scores
were determined separately for each of the five categories and summed across
all the items to provide an overall index of susceptibility to fear arousal.
It should be noted
in passing that attitudes, defined as evaluative responses, are differentiated
from fear arousal. These two sets of measures are distinguished on the grounds
that people can be attracted to things they fear, as evident in approach-avoidance
conflicts; conversely, it is not uncommon for people to dislike things they
do not fear.
Treatment
Conditions
Subjects were individually
matched on the basis of their pretreatment avoidance behavior and then randomly
assigned to one of four conditions. Each group contained 12 subjects. All
treatments were administered individually.
One group of subjects
received the standard form of systematic desensitization treatment originally
devised by Wolpe (1958). In this procedure deep relaxation was successively
paired with imaginal representations of snakes arranged in order of increasing
aversiveness. During the first of two sessions subjects received training
in muscular relaxation and in the use of positive imagery to diminish emotional
arousal. In subsequent sessions, after being deeply relaxed, subjects were
asked by the experimenter to visualize the least threatening item in the
hierarchy of emotion-arousing scenes involving snakes that they had previously
ranked from least to most aversive. This anxiety hierarchy contained a total
of 84 scenes ranging from relatively innocuous activities such as looking
at pictures and toy replicas of snakes to handling live snakes in ways that
would be fear provoking. ‘Whenever subjects signaled anxiety to visualization
of a threatening scene it was promptly withdrawn, relaxation was reinstated,
and then the item was repeatedly presented until it ceased to evoke anxiety.
If relaxation remained unimpaired in the imagined presence of the threat,
subjects’ emotional responses to the next item in the hierarchy were extinguished
and so on throughout the graduated series until the most threatening events
were completely neutralized.
A second group
of subjects participated in a self-administered symbolic modeling treatment
in which they observed a graduated film depicting young children, adolescents,
and adults engaging in progressively more threatening interactions with
a snake. The colored film, which was approximately 85 minutes long, began
with scenes showing the fearless models handling plastic snakes and proceeded
through displays in which they touched and held a large king snake, draped
it around their necks, and let it crawl freely over their bodies.
To further increase
the efficacy of this method two other features were added: First, subjects
were taught to induce and to maintain anxiety-neutralizing relaxation throughout
the period of exposure. The second factor concerned the control of stimulus
presentation. A self-regulated modeling treatment would be expected to permit
greater control over extinction outcomes than one in which subjects were
exposed to a sequence of aversive modeling stimuli without regard to their
anxiety responses. Therefore, the rate of presentation of modeling stimuli
was regulated by subjects through a Kodak analyst projector equipped with
remote control starting and reversing devices. Subjects were in- structured
to stop the film whenever a particular model performance was anxiety provoking,
to reverse the film to the beginning of the aversive sequence, and to reinduce
deep relaxation. They then reviewed the threatening scene repeatedly in
this manner until it was completely neutralized before proceeding to the
next item in the graduated sequence. After subjects became skillful in handling
the projector controls and the self-induction of relaxation, the experimenter
absented himself from the situation so that the subjects themselves conducted
their own treatment until their anxieties to the depicted scenes were thoroughly
extinguished. Treatment was terminated when they could view the entire film
without experiencing any emotional arousal.
During the symbolic
modeling treatment subjects rated on a 10-point scale the intensity of their
emotional responses to each scene and to subsequent reexposures to the same
items. These data were collected to provide information on the course of
vicarious extinction of emotional arousal as a function of repeated observation
of modeled approach responses.
Subjects assigned
to the third group received the treatment combining graduated live modeling
with guided participation. After observing intimate snake-interaction behavior
repeatedly modeled by the experimenter, subjects were aided through demonstration
and joint participation to perform progressively more threatening approach
responses toward the king snake. In the initial procedure subjects observed
through a one-way mirror the experimenter perform a series of threatening
activities with the king snake that provided striking demonstrations that
close interaction with the snake does not have harmful consequences. During
this period, which lasted approximately 15 minutes, the experimenter held
the snake close to his face, allowed it to crawl over his body at will,
and let it loose to slither about the room. After returning the snake to
its glass cage, the experimenter invited the subject to join him in the
room and to be seated in one of four chairs placed at varying distances
from the experimenter’s chair. The experimenter then removed the snake from
the cage and commenced the treatment, beginning with relatively non- threatening
performance tasks and proceeding through increasingly fear-provoking activities.
This treatment was conducted without the use of relaxation procedures.
At each step the
experimenter himself performed fearless behavior and gradually led subjects
into touching, stroking, and then holding the midsection of the snake’s
body with gloved and then bare hands while the experimenter held the snake
securely by the head and tail. Whenever a subject was unable to perform
the behavior upon demonstration alone she was asked to place her hand on
the experimenter’s and to move her hand down gradually until it touched
the snake’s body. After subjects no longerfelt any apprehension about touching
the snake under these secure conditions, anxieties about contact with the
snake’s head area and entwining tail were extinguished. The experimenter
again performed the tasks fearlessly, and then he and the subject enacted
the responses jointly; as subjects became less fearful the experimenter
gradually reduced his participation and control over the snake until eventually
subjects were able to hold the snake in their laps without assistance, to
let the snake loose in the room and to retrieve it, and to let it crawl
freely over their bodies. Progress through the graded approach tasks was
paced according to the subjects’ apprehensiveness. The threat value of the
activities for each subject determined the particular order in which they
were performed. When they reported being able to perform one activity with
little or no fear, they were eased into a more difficult interaction. Treatment
was terminated when subjects were able to execute all the snake interaction
tasks independently.
Subjects assigned
to the control condition participated in all of the assessment procedures
without receiving any intervening treatment. This group primarily furnished
a control for changes resulting from repeated measurements. A pseudotherapy
was not employed because several previous investigations (Davison, 1968;
Krapfl, 1967; Lang et al., 1965) have shown that snake avoidance behavior
is unaffected by such experiences. In addition, the controls were subsequently
used to test the efficacy of symbolic modeling without relaxation.
To evaluate the
reliability of treatment outcomes the procedures were administered by two
experimenters, one female and one male. Each experimenter applied each of
the three treatments to half the subjects. The experimenters received no
information about subjects’ pretreatment performances on any of the measures
so as not to alter the manner in which they administered the procedures.
The treatments,
which were typically scheduled twice a week, continued until subjects either
achieved the terminal criterion specified for each condition or the maximum
time allotment for 5.25 hours of treatment (not counting relaxation training)
was completed. Maximum contact with snakes, either in live or symbolic form,
was thus equated across conditions. The average duration
of treatment required for the different methods was 2 hours, 10 minutes
for contact desensitization, 2 hours, 46 minutes for symbolic modeling,
and 4 hours, 32 minutes for systematic desensitization. The latter method
required more treatment time than the two modeling procedures, which did
not differ significantly from each other. These treatment durations do not
include the time devoted to relaxation training in the symbolic modeling
and systematic desensitization conditions.
Several subjects
who were originally selected for the experiment had to be replaced by
subjects with comparable avoidant tendencies because of various difficulties
that precluded their participation. A male subject who was making satisfactory
progress in the contact desensitization treatment had to discontinue the
program when an occupational change made it difficult for him to meet
the required schedule; and a female subject in this same condition had
to be replaced because she was afflicted with mononucleosis; two controls
moved to distant cities during the waiting period, and one dropped out
after the pretreatment assessment.
Posttreatment
Assessment
Following completion
of the treatment series the assessment procedures employed in the pretreatment
phase of the study were readministered to all subjects. As in the pretest,
the attitude measures were administered prior to and following the behavioral
avoidance test to permit evaluation of the reciprocal interaction between
attitudinal and behavioral changes.
In order to determine
the generality of extinction effects, half the subjects in each of the
conditions were tested initially with the familiar brown-striped king
snake and then with an unfamiliar crimson-splotched corn snake that was
strikingly different in coloration; the remaining subjects were tested
with the two snakes in the reverse order. Two groups of 12 students, drawn
from an introductory psychology course, were tested with either the king
snake or the corn snake to compare their aversive properties. Except for
a slight tendency for the corn snake to evoke more negative evaluation
on the semantic differential (t = 2.33,
p <.05), both snakes produced equivalent approachbehavior, fear arousal
accompanying specific approach responses, and negative attitudes toward
reptiles.
The same female
experimenter who conducted the pretreatment assessment administered the
posttreatment measures. To control for any possible bias, she was provided
with no information about the conditions to which the subjects were assigned.
RESULTS
Differences in
approach behavior toward the two snakes were evaluated separately for
each treatment condition. Although subjects in the contact desensitization
and systematic desensitization treatments performed more approach responses
toward the familiar king snake than toward the generalization snake, none
of the differences reached the .05 significance level. Nor did subjects
experience differential levels of fear arousal while performing specific
approach responses toward the two snakes. The two sets of scores were,
therefore, averaged across snakes for evaluating the results of the experiment.
The data were
analyzed using analysis of covariance in which pretreatment measurements
served as the covariates. Separate three-way analyses were computed for
each independent variable with treatment conditions, experimenters, and
snake order representing the three factors.
Table 1 shows
the significance levels of the treatment effects, the differences between
pairs of treatment conditions, and the changes that occurred within each
group on each of 17 measures. The order in which the snake tests were
administered did not in itself produce any significant differences, and
except for one instance which will be discussed later, the two experimenters
achieved equivalent results. In the numerous analyses only three significant
interaction effects were obtained that will be discussed when results
are reported separately for each measure.
Approach
Behavior
The mean approach
responses performed by subjects in each of the four conditions at the
pretreatment and posttreatment phases of the experiment are presented
in Figure 1. As depicted in Table 1, subjects in the control group showed
no change in
The rates were
92 percent for modeling with participation, 33 percent for symbolic modeling,
25 percent for systematic desensitization, and 0 percent for the controls.
These differential rates of terminal performances were highly significant
(x2 = 23.14, p < .001).
As noted earlier,
the behavior changes produced by the treatment procedures generalized
extensively to the unfamiliar snake that subjects encountered for the
first time in the posttreatment assessment. However, some of the subjects
whose avoidance behavior was thoroughly extinguished in relation to the
familiar king snake were nevertheless unwilling to perform the terminal
task with the corn snake. Among subjects who achieved terminal performances
with the king snake, the percentage showing complete transfer to the other
reptile was 55 percent, 100 percent, and 0 percent for contact desensitization,
symbolic modeling, and systematic desensitization, respectively. When
approach scores are considered for all subjects, regardless of whether
or not they achieved terminal performances with the experimental snake,
the contact desensitization treatment of course produced greater approach
behavior (M = 24.6)
toward the generalization snake than either symbolic modeling (18.0) or
systematic desensitization (15.8).
Fear Arousal
Accompanying Approach Responses
The degree of fear arousal evoked by approach responses is partly dependent on the threat value of the behavior being performed. That is, looking at a caged snake is a much less fear- provoking activity than holding a writhing snake close to one’s face. The degree of fear extinction was measured by comparing the mean level of fear arousal accompanying approach responses that subjects performed before treatment with the fear levels reported in the posttreatment period for the same subset of approach responses and for all of the approach behavior that sub. jects successfully executed. If, for example, a given subject successfully completed 5 approach responses in the pretest and 20 responses in the posttreatment phase, the fear decrement for initial approach responses was based on differences in mean fear scores for the same 5 responses in the two assessments; the fear decrement for total approach behavior was measured in terms of differences between mean fear scores for the 5 pretest responses and for the 20 responses performed in the posttreatment period.
The fear extinction
data are depicted graphically in Figure 2 Since anticipatory and performance
fear arousal did not differ
Fear evoked
by initial approach responses. With
regard to the pretreatment subset of approach responses, subjects in all
conditions, including the controls, experienced less fear the second time
they performed the same behavior. The treatment conditions, however, produced
more marked fear decrements. Further comparisons among means revealed
that the magnitude of fear reduction achieved by subjects who received
systematic desensitization was significantly less than that shown by subjects
in the two modeling conditions, which did not differ from each other.
The analysis
disclosed a Treatment x
Order interaction effect at a borderline level
of significance (F =
2.97, p < .05).
Subjects who received the modeling treatments achieved greatest decrements
in fear arousal when they were tested first with the unfamiliar snake,
systematic desensitization subjects displayed greatest fears reduction
when they were tested initially with the familiar snake, while the order
of the behavioral test had no differential effects on fear arousal in
control subjects. The results also yielded a significant triple interaction
effect (F = 5.96,
p <.01) that is not easily interpretable.
Fear evoked
by total approach responses.
Further evidence for the differential efficacy of the treatment procedures
in extinguishing fear arousal is provided in comparisons of fear level
experienced in relation to pretreatment approach responses with fear arousal
accompanying total approach behavior. On this measure, control subjects
showed no significant fear reduction even though their posttreatment performances
did not differ much from their initial approach behavior. Subjects in
the treatment conditions, on the other hand, experienced significantly
less fear in connection with far more threatening performances. Except
for the significant treatment effect none of the other main effects or
interactions between variables were significant.
As was the case
with fear reduction in relation to pretreatment level performances, the
two modeling conditions did not differ from each other, but both produced
greater fear extinction for total approach behavior than did systematic
desensitization. Considering, however, that the mean fear level for contact
desensitization is based upon more fear-provoking behavior in posttreatment
than the mean for symbolic modeling, the data for the latter two groups
are not entirely comparable. A supplementary analysis was therefore performed
which compared the level of fear associated with the last approach response
that the more timorous member of each matched pair was able to perform.
That is, if a pair of matched subjects in the contact desensitization
and symbolic modeling treatments completed 29 and 25 approach responses,
respectively, the comparison included the fear level accompanying the
twenty-fifth response they performed in common. In this analysis subjects
receiving symbolic modeling experienced far greater fear arousal (M
= 4.21) than their counterparts in contact desensitization
(M = .69), a difference that is highly significant
(t = 3.29, p < .01).
Attitudinal
Changes
The changes in
attitudes produced by the
various treatment procedures are shown graphically in Figure 3 and evaluated
statistically in Table 1. Both measures—based on the attitude scales and
the semantic differential—yielded comparable results. There are no differences
on either measures between the two attitude assessments conducted in the
pretreatment phase (Pre2 — Pre1). For the types of
subjects included in this experiment, apparently factual information about
snakes and exposure to a snake in the behavioral test did not alter their
negative evaluations of reptiles. The refractory quality of these negative
attitudes is further shown by the control subjects whose attitudes remained
unaltered in the posttreatment assessment as well.
Analysis of changes
in attitude scores obtained between the pretreatment behavioral test and
following the treatment phase (Post1 — Pre2) reveals a highly
significant treatment effect for both measures. Subjects in all three
experimental conditions displayed favorable changes in their evaluation
of reptiles. Individual comparisons among the means for the different
conditions show that modeling combined with participation produced the
greatest attitudinal changes, the two modeling procedures were superior
to systematic desensitization, and all three treatment conditions differed
significantly in this respect from the nontreated control group. It is
interesting to note that subjects showed no additional attitudinal changes
as measured immediately after the posttreatment behavioral test (Post2
— Post1).
The results revealed
two additional significant effects. Subjects treated by the female experimenter
displayed greater changes on the semantic differential than those treated
by the male (F =
7.75, p <.01). A significant Treatment
x Order interaction effect
(F = 3.08, p < .05) was obtained in the
analysis of data from the attitude scales. The behavioral test order in
which the familiar snake was presented first produced more favorable changes
in attitudes of subjects receiving the modeling treatments, but variation
in snake order had no differential effects on subjects in the systematic
desensitization and control groups.
Fear Inventory
The changes in
the number and intensity of fears in each of the five areas measured are
given in Figure 4. Results of the statistical evaluation of these scores
are summarized in Table 1. The analysis of covariance indicates no significant
difference between groups except for intensity of animal fears. Further
comparison of pairs of means shows that subjects in the two modeling conditions
experienced a greater reduction in the degree to which they feared animals
than did the controls.
Analysis of change
scores within groups reveals some degree of fear reduction toward situations
beyond the specifically treated phobia, the decrements being roughly proportional
to the potency of the treatments employed (Table 1). Nontreated controls
showed no changes in either the number or intensity of fears; systematic
desensitization produced a decrease only in severity of fears toward other
animals; and symbolic modeling was accompanied by reduction in the intensity
of fear of animals and social events. Contact desensitization, on the
other hand, effected the most widespread fear reductions in relation to
a variety of threats including animals, physical injury, interpersonal
situations, and miscellaneous events.
Relationship
between Attitudinal and Behavioral Changes
Pretest attitudes
and behavior were correlated to assess the degree of relationship that
ordinarily exists between these two responses classes. Attitudes were
highly positively related to approach behavior for the unselected group
of subjects who were used to measure the aversiveness of the snakes. Approach
scores correlated .73 with attitudes prior to the behavior test and .87
immediately after the behavior test. The corresponding relationships between
the attitude measures based on the semantic differ. ential and approach
behavior were .56 and .70, respectively. These correlations are all significant
beyond the .01 level.
Similar product-movement
correlations were calculated for the phobic subjects as well, although
coefficients based on these data are less informative because the range
of approach scores for this sample is considerably curtailed. Nevertheless,
approach behavior correlated positively with attitudes prior to (r
= .48) and after (r = .56)
the behavior test. The data also reveal moderately high positive relationships
between approach behavior and evaluative responses on the semantic differential
as measured before (r =
.44) and after (r = .60)
the behavior test. All of the preceding correlation coefficients exceed
the .01 level of significance.
In order to determine
whether the treatment procedures produced analogous changes in attitude
and behavior, correlations were computed on amount of change obtained
between pretest and posttreatment scores for these sets of measures. Since
the corresponding correlations computed separately for data from the different
treatment conditions were in the same direction and did not differ significantly,
they were averaged by means of an r to z transformation.
Behavior change
is positively correlated with attitude change (Post1_Pre2) as measured
by both the semantic differential (r =
.39, p < .05)
and the attitude scales (r =
.59, p <.01). Moderately
high positive relationships are likewise obtained between these measures
when the attitude change scores are based on differences between pretest
attitudes and those exhibited in post- treatment following the behavioral
avoidance test (Post2 —
Pre2).
The correlations
of changes in behavior with changes in semantic differential and attitude
scores are r = .55 (p < .01) and r = .58
(p < .01), respectively.
The correlational
analysis disclosed no relationship between degree of behavioral change
and either initial number or intensity of fears in other areas of functioning.
Thus, the effectiveness of the treatment procedures was not diminished
by the presence of generalized anxiety. Nor did subjects’ initial attitudes
toward snakes affect the degree of behavioral improvement achieved by
the different treatment methods.
Although attitudes
were not predictors of behavioral change, the initial severity of avoidance
behavior was a significant predictor of degree of attitude change as measured
by both the attitude scales (r = —.43; p < .01) and the semantic differential (r = —.42;
p < .01). The more avoidant subjects were to begin with, the less they
altered their evaluations of reptiles in the positive direction. Moreover, within the two modeling treatments, degree of attitude change
on the attitude scales correlated negatively with numbers of fears (r
= —
.40; p < .05), and anxiety about physical injury (r = —
.54; p < .01).
Treated
Controls
Following completion
of the posttreatment assessment, subjects in the control group received
the symbolic modeling treatment without the relaxation component. They
simply reviewed threatening scenes repeatedly until completely neutralized,
and recorded their level of fear arousal during each exposure. Except
for one subject who had to discontinue toward the end of the treatment
to undergo major surgery, all of the controls completed this second phase
of the experiment. They were then readministered the same sets of measures
used in the preceding assessments.
In evaluating
the efficacy of symbolic modeling without relaxation, t tests
for correlated means were computed on changes in the performances of control
subjects after they had received the treatment relative to their posttest
scores. As shown in Table 2, symbolic modeling alone increased subjects’
approach behavior. In fact, 45 percent of the subjects exhibited terminal
performances toward both snakes. This treatment also produced favorable
attitudinal changes, and it reduced fear arousal to both snake approach
behavior and a variety of other potentially threatening situations measured
by the fear inventory (Table 2).
Statistical comparisons
were made of the changes achieved by control subjects through symbolic
modeling alone and by experimental subjects who received symbolic modeling
with relaxation. No differences were found between the groups either in
approach behavior or in generalized anxiety (Table 2). However, subjects
who paired aversive modeling cues with relaxation subsequently experienced
significantly less fear arousal while performing snake-approach responses,
and they showed greater positive changes in their attitude toward snakes.
Vicarious
Extinction of Fear Arousal to Modeled Events
As was mentioned
earlier, subjects receiving film-mediated treatment rated the degree of
fear arousal evoked by the modeled scenes initially and by each subsequent
reexposure to the same scenes. These ratings were averaged across subjects
and scenes at each exposure to provide an index of the rate with which
fear arousal was extinguished in subjects who observed the modeled events
with and without the benefit of relaxation. The data are plotted in Figure
5 for the first six exposures since subjects rarely required more than
six presentations to neutralize any given scene. The vicarious extinction
data for the subject who had to discontinue before completing the final
portion of the treatment and the fear arousal ratings of the matched subject
for the same duration of treatment were included in the statistical analysis.
Both groups of
subjects showed a progressive decline in fear arousal with each successive
exposure to modeled approach behavior. Separate comparisons of scores
between adjacent points reveals that the fear decrements with each reexposure
are highly significant for both sets of data.
Although repeated
observation of nonreinforced approach behavior eliminated fear arousal,
the addition of relaxation did not have a strong facilitative effect on
the rate of vicarious fear extinction. The two groups did not differ significantly
in level of fear arousal on first exposure to each modeled scene, but
subects who combined modeling with relaxation experienced a greater reduction
in fear on the second exposure to the aversive scenes than their counterparts
who received modeling alone (t
= 1.80,
p < .05); on subsequent reexposures, however, the rate of fear extinction
was essentially the same. Subjects who paired modeling with relaxation
required fewer reexposures (M = 24) than the modeling-alone
group (M = 58) to achieve complete extinction of fear arousal,
but there was considerable variability and the difference is significant
only at a borderline level of significance (t = 1.66, .10 <p < .05).
Changes
Following Supplementary Treatment with Contact Desensitization
A total of 23
subjects from the symbolic modeling, systematic desensitization, and treated
control groups who failed to attain terminal performances received live
modeling with guided participation. Although there was some variability,
the average length of this supplementary treatment was approximately 1
hour and 20 minutes. After these subjects completed their treatment they
were administered the regular assessment procedures.
As shown in Figure 6 and Table 2, subjects in all three groups displayed further significant increases in approach behavior. The data of this subgroup combined and those who required no additional treatment reveals that 96 percent of all the subjects who participated in the study achieved terminal performances with the experimental snake, while 70 percent showed complete extinction of avoidance behavior toward the generalization snake as well. These behavioral changes were maintained at the same level over the follow-up period, with the terminal performances being 96 percent and 67 percent for the experimental and generalization snakes, respectively. At the various test periods subjects in the different treatment conditions who failed
to complete
the final approach response usually successfully performed the remaining
tasks, which explains why differences in approach scores based on the
two snakes do not reach statistical significance.
Subjects’ attitudes
and level of fear arousal at the different assessment periods are summarized
graphically in Figures 7
and 8. The significance of the changes on these
and other measures by the subgroup of subjects receiving the supplementary
treatment is presented in Table 2.
Of the three
groups of subjects, those who originally received symbolic modeling
paired with relaxation benefited most from live modeling with guided
participation. In addition to the increases in approach behavior previously
reported, they showed further improvements in attitude, additional fear
extinction, and generalized reduction of anxiety in all five areas of
functioning measured by the fear inventory (Table 2). Following the
supplementary treatment, control subjects, who previously received symbolic
modeling alone, displayed a significant reduction in avoidance behavior,
positive changes on the attitude measure, and a significant decrease
in the total number of fears.
The supplementary
treatment likewise produced behavioral and attitudinal improvements
in the systematic desensitization subgroup. In addition, subjects in
this condition, who originally achieved the smallest decrement in fear
arousal compared to the other methods, showed marked extinction of fear
arousal. However, the supplementary treatment did not produce any further
change in their fear of other potential threats. The fact that subjects
receiving the symbolic modeling treatment originally achieved greater
decrements in fear arousal than either the treated controls or the systematic
desensitization subjects might explain why the former group showed the
most generalized reduction in other fears following the treatment combining
modeling with participation.
Separate analyses of variance were computed on the
17 measures after subjects attained terminal performances either through
their regular treatment alone or combined with modeling and participation.
At this phase of the experiment there were no significant differences
between the groups on any of the measures except the semantic differential
administered before the behavioral
avoidance test (Table 2). Subjects who received modeling with participation
exhibited greater changes than their counterparts in either the control
(F = 6.10, p < .05) or the systematic desensitization (F = 5.05, p < .05) conditions. The symbolic modeling group also manifested more
positive evaluative responses than the controls (F = 4.64, p <.05), but it did not differ significantly from the systematic desensitization
group. No significant group differences were found, however, either
on attitude scores (F
= .54)
or semantic differential scores (F
= 1.45) obtained immediately after the behavioral
avoidance test.
Maintenance
of Psychological Changes
In order to
evaluate the durability of induced changes subjects returned for an
additional evaluation approximately 1 month after they had concluded
their final treatment, t
tests were calculated on differences between
performances for the total sample of subjects in their last posttreatment
test and in the follow- up assessment. Results of these statistical
analyses for the combined sample are given in Table 2.
Subjects not
only maintained the same level of bold approach behavior, but they experienced
significantly less fear arousal while performing the same approach responses
in the follow-up assessment. However, subjects showed a small but nevertheless
significant decrease in positive attitudes toward snakes over the follow-
up period. As can be seen from Figures 8 and 9, subjects in the systematic
desensitization (t = 2.34, p < .05) and control
(t = 1.93, p <.10) conditions accounted
mainly for the change in attitude scores; control subjects (t = 2.82, p < .02) also contributed largely to the change on the semantic differential.
Analysis of
fear inventory scores revealed that subjects either maintained their
gains or showed further reductions in their fear of potential threats
in other areas of functioning (Table 2). Specifically, they reported
fewer fears and a significant diminution in the intensity of their subjective
reactions to threats. The greatest fear reduction over the follow-up
period occurred in relation to animals and apprehension over physical
injury. In addition, subjects reported decrements in various miscellaneous
fears.
Separate analyses
of variance were also performed on the 17 measures at the follow-up
period to determine whether subjects in the different conditions retained
their comparable improvements. As in the final posttreatment assessment,
there were no significant differences between the groups on any of the
measures, except the semantic differential obtained prior to the behavioral
avoidance test (Table 2). Individual comparisons of pairs of means shows
the control group as having lower positive evaluations of snakes than
subjects in either the contact desensitization (F
= 7,
p < .05) or symbolic modeling (F = 8.54, p < .01) conditions.
However, control subjects significantly increased their valuations of
snakes after the behavioral avoidance test and no significant group
differences were found (F
= 1.70)
on the final semantic differential score.
Antecedents
of Snake-Phobic Behavior
As was previously
mentioned, at the beginning of the experiment subjects were administered
a questionnaire measuring both direct and vicarious aversive experiences
with snakes and the incidence of familial modeling of snake phobic behavior.
Although not a single subject had ever been physically injured by a
snake, they reported a variety of frightening experiences involving
reptiles. For example, 68 percent of the subjects were frightened in
childhood by surprise encounters with snakes on walks, coiled up under
rocks, under household furniture, in boats, in tents, and in other unexpected
places. Some subjects described revolting child experiences in which
they witnessed snakes viciously beaten to death. To a young child incidents
of this type would tend to convey the impression that snakes must be
exceedingly dangerous to warrant such extreme onslaughts.
In 62 percent
of the cases, fear of snakes was further reinforced through pranks involving
live or dead snakes and toy specimens. In their childhood years the
subjects were chased by other children brandishing snakes menacingly,
they had dead snakes thrown at them, or hidden in their lunch baskets,
in beds, in tents, in closets, and in grocery bags. Those who were most
apprehensive about snakes were selected as the favorite targets in such
pranks.
Traumatic vicarious
experiences that often resulted in recurrent nightmares were also reported
by 58 percent of the sample. The episodes that subjects found most shocking
were sequences in movies or television programs in which snakes were
shown stalking their prey, crawling menacingly toward sleeping people,
wrapping themselves around animals or people and slowly crushing them
to death, or where persons were thrown into a pit of writhing snakes.
Familial modeling
of snake-phobic behavior also occurred with relatively high frequency
(56 percent) in this sample. The vast majority of cases (85 percent)
reported having experienced two or more of these different forms of
fear arousal (i.e., direct, vicarious, and familial modeling influences).
Although these findings cannot be fully interpreted in the absence of
comparative data from a nonphobic sample, they nevertheless reveal that
subjects in the present study had undergone numerous frightening experiences
capable of endowing snakes with strong aversive properties.
At the conclusion
of the experiment subjects filled out a questionnaire that asked them,
among other things, to describe their reaction when they first learned
about the type of treatment that they would be receiving and their confidence
in the method; and to indicate whether the treatment experiences in
any way enhanced or hindered their functioning. These results are reported
next.
Therapeutic
Expectations
The treatment
procedures were presented to subjects as experimental approaches, without
any claims made for their efficacy. Questionnaire results disclose that
when subjects first learned about the type of treatment they were to
receive, 67 percent did not expect to benefit from the program, 16 percent
were uncertain about what to expect, while 16 percent believed that
they would achieve beneficial results. Subjects in the symbolic modeling
and systematic desensitization conditions were skeptical that even if
the treatment eliminated their fears toward symbolic representation
of snakes, the extinction effects would transfer to actual snakes.
I
felt totally unconfident that it would
work on me. I thought I could probably get used to seeing snakes, but
I never thought that I could be able to pick one up calmly. . . I did not see how just imagining
snakes could help me. I did not have much confidence in the method.
Most subjects
in the contact desensitization condition, on the other hand, had serious
doubts that they could ever perform the snake approach responses required
by the treatment (“I was appalled and determined I could never handle
a snake.”).
The skepticism
regarding these more direct treatment approaches also stemmed in part
from the widespread belief that anxiety conditions can be successfully
modified only through verbal interpretive means. This attitude is reflected
in the following comments of one of the subjects:
When I heard
that it would be all involvement with snakes, I didn’t think it would be successful in my case. I had expected and hoped for more discussion
about snakes. However, I now realize that this would not have solved
my problem. Rather, it
would have been a waste of time.
Positive
Transfer to Naturalistic Situations
During the follow-up
period 47 percent of the subjects reported encounters with snakes in
one form or another. In each case they reported that the reduction in
fear of snakes achieved in treatment generalized to snakes in naturalistic
situations. The subjects no longer experienced marked distress when
unexpectedly confronted with snakes in the course of their social or
occupational activities; they could visit reptile exhibits and look
at pictorial displays of snakes without trepidation; they were able
to handle harmless snakes, and a few even served as model therapists
for their own children and faint-hearted friends.
Other subjects,
though they had no contact with snakes, were nevertheless able to participate
in recreational activities such as hunting, camping, picnicking and
hiking, which they had formerly avoided because of their dread of snakes.
As one subject explained, “I am no longer harassed by walking through
grassy areas in fear of running across a snake.”
DISCUSSION
Results of
the present experiment provide further evidence that treatment approaches
based on social-learning principles can be highly efficacious in producing
generalized and enduring psychological changes. Of the three methods
investigated, modeling combined with guided participation was most successful
in eliminating phobic behavior, in extinguishing fear arousal, and in
creating favorable attitudes. The generality of these findings is increased
by the additional evidence that subjects who achieved only partial improvement
through other treatments displayed substantial changes after a brief
period of contact desensitization.
It would appear
from these laboratory findings that a powerful form of treatment is
one in which therapeutic agents them
selves model the desired behavior and arrange optimal conditions for
clients to engage in similar activities until they can perform the behavior
skillfully and fearlessly.
Comparison of
symbolic desensitization and symbolic modeling shows both procedures to be equally effective in extinguishing avoidance behavior; however, symbolic modeling produced greater reduction
in both fear arousal and negative attitudes, and the behavioral changes it achieved appear
to be more generalized.
Indeed, findings of the present study and those
reported by Blanchard
(1969) disclose that subjects who attain terminal performances through
modeling alone show almost complete transfer on behavioral generalization
tests.
Although the
foregoing results demonstrate that significant psychological changes
can be reliably achieved by extinguishing the arousal potential of aversive
stimuli presented in symbolic form, they also indicate that such treatment
approaches have certain limitations if used alone. Virtually all subjects
who received systematic desensitization or symbolic modeling displayed
behavioral improvements that surpassed either their pretreatment performances
(96 percent) or the changes exhibited by matched nontreated controls
(91 percent). Nevertheless, most subjects in these
treatment conditions were unable to perform terminal-level activities that had been thoroughly neutralized
in symbolic form. Other
investigators (Agras,
1967; Hoenig & Reed, 1966) have found a similar discrepancy
between symbolic desensitization and actual performance.
From knowledge
of stimulus generalization one would ordinarily expect some transfer
loss in symbolically oriented treatments. A major advantage of modeling
with participation is that fear is eliminated toward actual threats.
In clinical practice, of course, symbolic desensitization is typically
supplemented with graded performance tasks that are executed in real
life situations, with active positive reinforcement of approach behavior
to overcome initial reluctance of phobic persons to reexpose themselves
to feared situations, and with modeling procedures to further augment
change in behavior. In laboratory investigations, these various “extraneous”
influences are intentionally excluded.
The prediction
that relaxation would augment the effects of symbolic modeling was only partially corroborated. Modeling coupled with relaxation produced greater decrements
in fear accompanying approach responses, more favorable attitudes, and
more rapid vicarious extinction of fear arousal on initial reexposure
to the modeled events. The groups did not differ, however, in approach
behavior. These results, while interesting, must be accepted with reservation
because subjects who received modeling without the benefit of relaxation
required significantly more observational
trials to extinguish their fearful reactions to the modeled approach
responses. It is conceivable that if
these treatments were
limited to the same number of observational extinction trials, the obtained
differences would have been even larger and a difference in approach
behavior might also have emerged. This expectation receives some support
from a recent study (Spiegler, Liebert, McMains, Sc Fernandez, 1969)
demonstrating that relaxation facilitates vicarious fear extinction
under conditions where subjects receive only a single exposure to the
modeled approach behavior.
It is of interest
that the efficacy of the treatment procedures was in no way limited
by subjects’ general level of anxiety. The correlational data from the present study are somewhat at variance with
previous findings (Bandura Sc Menlove, 1968) that susceptibility to
emotional arousal in children is inversely related to degree of vicarious
extinction achieved through film-mediated modeling. Several factors
might have accounted for these divergent results. The two experiments
differ in the age of the subjects and
in the type of phobic behavior being modified. Another possible explanation is in terms of markedly different ways in which the
modeling treatments were conducted. The earlier study involved only a single exposure to modeling stimuli
without regard to subjects’ fear arousal, whereas the present experiment
utilized a self-regulated modeling procedure which permitted subjects
the opportunity to review threatening scenes repeatedly until thoroughly
neutralized. Under conditions where aversive modeling
stimuli are presented only once, anxiety proneness in observers is more
likely to serve as a determinant of vicarious extinction.
It is also noteworthy that the various treatments were equally effective when applied by experimenters differing widely in personality characteristics. These findings are consistent with those of Paul (1966) and Mann and Rosenthal (1969), showing that changes produced by systematic desensitizations are not differentially affected by variations in experimenter characteristics.
Further
evidence that socially conducted and self-administered systematic desensitization achieves equivalent results (Donner, 1967;
Krapfl, 1967; Melamed & Lang, 1967) suggests
that social variables
are not appreciable contributors to the measured outcomes. Ideally,
psychological treatment methods should be sufficiently powerful to achieve
consistent changes by different therapists, just as one would not be
content with medical procedures whose effects depended heavily upon
the bedside manner of physicians.
Further research
is needed to isolate the factors in modeling cues that govern fear reduction
in observers. There is some reason to expect that the affective expressions
accompanying a model’s behavior may exercise some degree of control
over vicarious extinction. It has been shown in studies of vicarious
emotional arousal in primates (R. E. Miller, 1967; Miller, Banks, &
Ogawa, 1962; Miller, Murphy, Sc Mirsky, 1959) and in human subjects
(Bandura Sc Rosenthal, 1966; Berger, 1962), that negative affective
expressions by others can serve as powerful cues for arousing fear and
avoidance in observers. In fact, Miller and his colleagues (Miller
et al., 1959) have shown that exposure to a subject reacting in an apprehensive
or fearful manner could reinstate avoidance responses in observers even
after such responses had been completely
extinguished.
The foregoing
research suggests that modeled approach responses accompanied by positive
affective expressions would engender less fear arousal in observers
and hence faster vicarious extinction, than if models manifested fearful
reactions while per- forming the same approach behavior. In the present
experiment the models frequently expressed pleasant emotional reactions
as they performed approach responses in a relaxed manner.
It is generally
assumed in theories of identification that similarity of the model to
the observer enhances response matching. However, it remains an open
question whether utilization of fearful models would facilitate or hinder
the reduction of phobic behavior. According to the theory of identification
presented by Bandura (1969b), response consequences to models generally
outweigh their characteristics in producing indentificatory behavior
in observers. Thus, for example, witnessing a similar model bitten by
a snake would in all likelihood increase snake avoidance behavior, whereas
seeing a dissimilar model handle a snake without any untoward consequences
would weaken avoidance responses. The treatment film included not only
fearless adult models but also several young children, on the assumption
that their lack of fear while performing responses that adult observers
regarded as hazardous would provide the most dramatic disconfirmation
of anticipated aversive consequences. It would be of considerable interest
to investigate systematically the degree to which model-subject similarity
on a relevant dimension (i.e., fearfulness) and also on irrelevant dimensions
(i.e., attitudes, interests, general background) affects the rate of
vicarious extinction of phobic behavior.
The results
show that applications of social-learning procedures have important
attitudinal consequences. Both symbolic modeling and systematic desensitization,
which operate primarily through extinction of negative affect aroused
by aversive stimuli, produced favorable changes in attitudes toward
snakes. Consistent with expectation, the treatment condition that reduced
the feararousing properties of snakes and enabled subjects to engage
in intimate interactions with snakes achieved the greatest attitudinal
changes. These findings are sufficiently promising to warrant more extensive
use of social-learning procedures for studying theoretical issues concerning
the development, modification, and functional role of attitudes.
It will be
recalled that previous research, though admittedly meager, found changes
in attitude and actions to be essentially unrelated. More recently,
Greenwald (1965a) reported a positive, but low, correlation between
these measures on an academic activity that does not have much affective
impact. However, for subjects who expressed their negative attitude
prior to the influence attempt, persuasive communications changed their
attitudes but not their behavior (Greenwald, 1965b). In contrast to
these results, desensitization and modeling treatments produced corresponding
changes in both attitudes and behavior even though all subjects initially
committed themselves to a strong loathing for snakes. In a study employing
similar procedures, Blanchard (1969) also found a high positive correlation
(r = .72) between changes in attitude and behavior
as induced through modeling influences.
The correlated
changes produced by social-learning procedures in different response
systems may be interpreted in several different ways. According to most
contemporary attitude theories (Abelson, Aronson, McGuire, Newcomb,
Rosenberg, & Tannenbaum, 1968), there exists a strong drive for
consistency among beliefs, feelings, and actions. A change in any one
of the components will, therefore, engender congruous modification in
the other constituents. In these consistency models, changes in attitudes
or behavior are treated, not simply as consequent events, but as causal
factors affecting other classes of behavior. An alternative interpretation
is that social influences have similar but independent effects on attitudes,
behavior, and emotional arousal. In this view, attitude-behavior consistencies
represent correlated coeffects rather than outcomes of a process in
which modification of one type of behavior forces changes in other forms
of responding to eliminate cognitive disequilibrium.
Definitive
tests of the parallel effects and consistency explanations of change
processes are precluded by the absence of a methodology that would permit
simultaneous measurement of attitudes, affect, and actions. If incongruity
creates an internal stimulus for psychological change then a sequential
testing procedure unavoidably confounds the effects of external influences
and the consistency drive. Conversely, a given environmental influence
could have analogous effects on different classes of response that would
be erroneously attributed to the operation of a consistency drive. These
alternative formulations perhaps should be regarded as complementary
rather than competing.
Under most
conditions, powerful social influences produce corresponding changes
in different modes of responding, and performance of new behavior is
likely to have additional cognitive and emotional consequences.
The findings
of the present study also have implications for different theoretical
formulations regarding the conditions governing phobic behavior. Contrary
to expectation from psychodynamic theory, extinction of emotional responses
toward the phobic object not only enduringly eliminated fear arousal
and phobic avoidance of snakes, but the treatments produced significant
reductions in anxiety in other areas of functioning not specifically
treated. These findings are in accord with numerous studies demonstrating
that direct extinction of phobic behavior is typically accompanied by
generalized anxiety reduction as measured by self-ratings (Lang, et
al., 1965; Mealiea, 1967; H. R. Miller, 1967; Paul, 1966, 1967, 1968;
Paul & Shannon, 1966) and behavioral avoidance tests (Mealiea, 1967).
The
positive transfer obtained in the present experiment probably reflects
the operation of at least two somewhat different processes. The first
involves generalization of fear extinction effects from stimuli that
were neutralized by the treatments to related anxiety sources. Analysis
of differences between groups and changes within treatment conditions
revealed that the greatest fear decrements occurred in relation to similar
phobic objects such as other animals, which would be expected from knowledge
of stimulus generalization. The second process entails positive reinforcement
of a sense of capability through success which can mitigate emotional
arousal to potentially threatening situations. Having successfully eliminated
a phobia that had plagued them for most of their lives, a number of
subjects reported increased confidence that they could cope effectively
with other fear-provoking events. As one subject explained it, “My success
in gradually overcoming this fear of snakes has contributed to a greater
feeling of confidence generally in my abilities to overcome any other
problem which may arise. I have more faith in myself.” Others stated
that their treatment experiences not only changed their views about
the modifiability of personality patterns, but provided them with a
means of eliminating other unwarranted fears.
Within the
treatment combining modeling and guided participation several factors
are operative that might contribute to the psychological changes accompanying
this method. These component influences include observation of fearless
approach behavior repeatedly modeled without any unfavorable consequences
to the performer, incidental information received about the feared objects,
and guided interaction with threatening objects that engender no adverse
effects. Results of experiments subsequently conducted by Blanchard
(1969) and Ritter (1969a, 1969b) throw some light on the relative influence
of these component variables.
In a comparative
study of the effects of modeling, informational factors, and guided
participation, Blanchard (1969) found that modeling accounted for approximately
60 percent of the behavior change and 80 percent of the change in attitudes
and fear arousal; guided participation contributed the remaining increment.
Informational influences, on the other hand, had no effect on any of
the three classes of responses.
As mentioned
earlier, the guided participation component of the procedure under discussion
involves both enactment of progressively more difficult responses and
physical assistance in performing the required behavior. In a study
designed to evaluate the influence of these elements, Ritter (1969c)
found that modeling accompanied by physically guided performance produced
greater changes in acrophobic subjects than modeling with verbally guided
enactment which, in turn, was superior to demonstration alone.
Ritter (1968a)
gave special emphasis to the anxiety-reducing effects of physical contact.
In addition, when persons are physically assisted in performing the
behavior required at each step in the graded sequence, their fears and
inhibitions may be reduced to some degree by the added protection that
the model’s behavior provides. An experiment is needed to determine
whether the facilitative effects of physical guidance derive from interpersonal
contact, from protection against potential injurious consequences, or,
as seems most likely, from both factors.
Further research
is needed to clarify the mechanisms through which modeling combined
with guided participation achieves such uniformly powerful extinction
effects. Results of modeling procedures, particularly those based on
a nonresponse extinction paradigm, are consistent with the dual-process
theory of avoidance behavior. Data from subjects in the symbolic modeling
condition demonstrate that emotional arousal can be effectively extinguished
on a vicarious basis simply by having observers witness models exhibit
approach responses toward feared objects without experiencing any adverse
consequences. It has been further shown by Blanchard (1969) that the
more thoroughly emotional arousal to aversive modeling stimuli is vicariously
extinguished the greater is the reduction in avoidance behavior and
the more generalized are the changes.
In accordance
with the above findings the change process associated with the powerful
procedure involving modeling with guided participation may be conceptualized
as follows: Repeated modeling of approach responses and the anxiety-mitigating
influence of physical contact and physical protection decrease the arousal
potential of aversive stimuli below the threshold for activating avoidance
responses, thus enabling persons to engage, albeit somewhat anxiously,
in approach behavior. The favorable outcomes resulting from direct contact
with threats that are no longer objectively justified further extinguish
any residual anxiety and avoidance tendencies. Without the benefit of
prior vicarious extinction, the reinstatement of severely inhibited
behavior generally requires a tedious and protracted program. After
approach behavior toward formerly avoided objects has been fully restored
the resultant new experiences give rise to substantial reorganization
of attitudes.
NOTES
|
1.
|
This
research was supported by Public Health Research Grant M-5162 from
the National Institute of Mental Health. The authors are indebted
to Antonette Raskoff, Patricia Baker, and Robert O’Connor for their
generous assistance with various aspects of this research. |
|
2.
|
The
terms modeling with participation and contact desensitization are
used interchangeably to refer to the treatment condition. |
