Hayes, S. C. & Hayes, L. J. (1992) - American Psychologist, 47, 1383-1395

Hayes, S. C. & Hayes, L. J. (1992). American Psychologist, 47, 1383-1395.
Verbal Relations and the Evolution of Behavior Analysis
Steven C. Hayes
Linda J. Hayes
University of Nevada
Running Head: VERBAL RELATIONS AND BEHAVIOR ANALYSIS
Key terms: behavior analysis, language, verbal behavior, verbal relations, cognition, stimulus
relations, behaviorism, Skinner

Verbal Relations and Behavior Analysis - Page 1
Abstract
For many years, much of the behavior analytic research done with humans assessed the
viability of the continuity assumption: the premise that research with nonhumans would yield
principles of direct applicability to human subjects. Although human subjects often showed
performances that differed in important ways from those of nonhumans, these differences were
regularly minimized and rationalized. Over the past decade many behavior analysts have
concluded that much of the difference between human and nonhuman performance is due to
human verbal behavior. Much of the human behavior analytic research has evolved into the study
of the nature of verbal events. New and exciting research methods, findings, and theories have
emerged that are quite unlike stereotypical views of behavior analysis. The present paper reviews
this literature, especially the work on stimulus equivalence, and describes relational frame theory --
a behavioral perspective that distinguishes direct and derived stimulus functions at the level of
psychological process. We argue that at least one new behavioral principle--verbal or relational
control--has been uncovered. We apply this analysis to issues of verbal meaning and rule
following, and claim that these developments demonstrate the vitality of contemporary behavior
analysis.
Verbal Relations, Cognition, and the Evolution of Behavior Analysis
The viability of research paradigms is not decided on entirely rational grounds. The history
of science is littered with approaches that died not as a result of their intellectual inadequacy, but
by virtue of their failure to develop new and exciting research questions (Lakatos, 1965). A
scientific area not busy being born is busy dying.
Behavior analysis has long been on many psychologists' list of dead or dying areas in
psychology. Few would deny that for the last 30 years basic behavioral perspectives have been
muscled off the center stage of psychology. Yet behind the scenes, out of view of all but the most
behaviorally-oriented, the Skinnerian tradition has evolved. It has identified new and exciting
research questions and has witnessed an explosion of research interest in these areas. While some
key assumptions have come into question, the core of the position seems remarkably intact. It
shows every sign of an area that is busy being born.
Early Behavior Analytic Research with Humans
and the Problem of Instructional Control
It wasn't long after the onset of behavior analysis as a distinct area that researchers began to
turn their attention to human subjects (e.g., Holland, 1958; Hutchinson & Azrin, 1961). Much of
this work was predicated on a core assumption of behavior analysis: the continuity assumption.
Behavior analysts assumed that principles identified with nonhumans would apply, more or less in
whole cloth, to the actions of human organisms. This assumption justified the research program of
behavior analysis, which relied heavily on studies with nonhuman organisms. Skinner made clear
from the beginning (1938) that he was not so much interested in animal behavior in its own right as
he was interested in it as a convenient and potentially worthwhile preparation for the identification
of basic processes common to human behavior.

The continuity assumption invited criticism from outside the field in part because it was not
clear to others that this assumption was strategic and not fundamental (Hayes, 1987). There is
nothing in evolutionary or behavioral theory that ensures continuity of this sort.
Evolutionary continuity might be expected in the sense that "new contains old" but it is not
necessarily true that "old contains new." Evolutionary continuity moves forward in time, not
backward. Nonhuman species are not older versions of humans. Investigations across phyla are
examinations of tips of branches, not rungs of a ladder. As such, these investigations always
involve connections both from new to old (from the present to common ancestors) and from old to
new (from common ancestors to the present). Similarity of psychological processes across species
thus cannot be assumed.
The continuity assumption was an issue of research strategy, however. Skinner and his
followers bet that the same basic processes would prevail across species, and that they would be
revealed more readily by studying relatively "simple" nonhuman behavior in relatively "simple"
environments, than by studying human complexity directly. We will term this strategy the
continuity strategy. If the continuity strategy had turned out to be fruitless behavior analysis could
still have continued more or less intact -- empirical examinations could simply have proceeded
with the human species as the focus. Behavior analysts were always aware that the strategy might
turn out to be useless in particular areas (Skinner, 1938, p. 442; cf. Skinner, 1957).
The continuity strategy turned out to be even more useful than expected. In many
situations humans respond very similarly to nonhumans. Perhaps most important at the time was
the great success of practical interventions based on basic behavioral principles. This success
buoyed the behavior analytic community, and strengthened their commitment to the continuity
strategy.
From the beginning, however, research in the basic laboratory with humans presented
problems for behavior analysts. Although humans often behaved similarly to nonhumans, it was
also true that human behavior often did not show the same effects of behavioral manipulations
(e.g., effects due to simple schedules of reinforcement) despite the fact that these effects were nearly
ubiquitous in the rest of the animal kingdom (e.g., Leander, Lippman & Meyer, 1968; Lippman &
Meyer, 1967; Weiner, 1964, 1965, 1969).
In addition, powerful effects were shown for a variable that could not be studied with
nonhumans in any obvious fashion: instructions. For example, Lippman and Meyer (1967)
examined the effects of instructions on performance on a fixed-interval (FI) 20 seconds schedule.
Some subjects were told that reinforcement depended merely on the passage of time; others were
told that the consequences were response produced. Subjects given the time instructions
responded at low rates. Subjects given the response instructions showed high rate performance.
Subjects given minimal instructions responded with one or the other of these two patterns, but not
the FI performance characteristic of nonhumans.
Weiner (1970) examined the effects of instructions during extinction. Some subjects were
told they could earn up to 700 pennies in the experiment, others that they could earn 999 pennies,
while were told nothing about the pennies they could earn. After 700 pennies had been earned all
subjects were exposed to a two-hour extinction period. Responding in extinction was greatest for
those subjects given no instructions, was less for those subjects told they could earn 999 pennies,
and was the least for those subjects told they could earn only 700 pennies.
Some behavioral researchers seemed surprised at the powerful effects of instructions,
especially when compared to direct contingency control. Early reports termed these effects
"remarkable" (Kaufman, Baron, & Kopp, 1966). Others explained instructional effects as simple
instances of stimulus control (e.g., Ayllon & Azrin, 1964). The dominant view seemed to be that the
effects of instructions were based on remote sources of control in the history of the individual that
were inaccessible to the experimenter. They were not perceived by behavior analysts to threaten
the continuity strategy. Most notable by its absence was the failure to consider instruction
following as an essentially verbal event, despite occasional references to its unique features (e.g.,
Schoenfeld & Cumming, 1963).
Skinner's Approach to the Listener
This approach to instructions was heavily influenced by Skinner's approach to verbal
behavior. Skinner formulated his analysis of verbal behavior from the point of view of the speaker
rather than the listener in a speaker-listener interchange. He defended this action on multiple
grounds, among which were that a complete account of the speaker implies an account of the
listener, that an analysis of the listener is unlikely to be empirically productive, and perhaps most
importantly, that "the behavior of a man as listener is not to be distinguished from other forms of
his behavior" (Skinner, 1957, p. 34). That is, in Skinner's view the behavior of the listener is
ordinary operant behavior under the discriminative control of speech (for a more complete review
of the plausibility of Skinner's three claims about the listener, see Hayes & Hayes, 1989). This idea
did not ascribe any special importance to the analysis of instructional control and it explicitly
rejected the idea that instructional control was a type of verbal event.
Skinner later drew a useful distinction between contingency-shaped behavior and rule-
governed behavior, but his analysis of the listener limited the utility and clarity of this distinction
(Parrott, 1987a, 1987b). Skinner defined rule-governed behavior as behavior controlled by
contingency-specifying stimuli (1966, 1969). He never defined what he meant by "specify."
Instead, he simply provided examples. Recently some behavior analysts have read "specify" to
mean "specify verbally" but Skinner himself did not say this until shortly before his death, and then
apparently not in writing.1 In his early presentations of the concept of rule-governed behavior,
Skinner did not distinguish between verbal rules and regularities drawn from other complex
antecedents, such as modeling stimuli (e.g., see Skinner, 1969, p. 163).
These characteristics made it difficult at first for behavior analysis to deal effectively with
the topic of verbal rules. Based on Skinner's analysis, the concept of "verbal rules" is
incomprehensible because he had no behaviorally sensible account of verbal stimuli. For Skinner, a
"verbal stimulus" was merely the stimulus product of verbal behavior (Skinner, 1957, p. 34). In this
usage, a "stimulus" is an object or entity, not a psychological function. In most behavior analytic
accounts, including those provided by Skinner, categories of stimuli (discriminative, eliciting,
reinforcing, etc.) are organized in accordance with the functional relations they sustain with
behavior. Given Skinner's analysis of verbal stimuli, it makes about as much sense to view control
by verbal rules as a distinguishable category of stimulus control as it would to claim that a pigeon
responding to an electric light was engaging in "Edison-governed behavior." Without a functional
category to examine, behavior analysis was delayed in its attempts to deal with instructional
control.
Verbal Control and Contemporary Behavior Analysis
Basic human operant research entered a period of stagnation in the late 1960's and most of
the 1970's. Much of the research continued to be driven by the continuity strategy, and discussion
sections focused largely on the similarities between human and nonhuman performances, not on
process accounts of the increasing number of identified differences. A basic analysis of
instructional effects was not pursued.
Non-behavioral authors often argued that instructional effects undermined the validity of
behavior analytic principles as applied to adult humans. In a well known chapter entitled "There is
no convincing evidence for operant or classical conditioning in adult humans," Brewer (1974)
attacked the idea of conditioning in adult humans when it is defined as an entirely automatic and
unconscious process -- that is, a process without significant verbal involvement. For example,
effects due to the self-formulated hypothesis "Every time the red light comes on I get shocked,"
were held by Brewer to be a function of cognitive processes (1974, p. 2), not conditioning. The
various behavior analytic studies showing the overriding effects of instructions were, in his eyes,
clear evidence for cognitive theory.
The behavior analysts' failure to provide a process account of the impact of verbal rules may
be partially explained, then, by two major sources. First, the continuity strategy mitigated against
it, especially when combined with the idea that verbal control was simply a complex form of
ordinary stimulus control. Based on Skinner's continuity strategy, if a researcher wanted to study
stimulus control, it would make little sense to study the impact of a complex stimulus (e.g.,
instructions) on a complex organism with an unknown history. Second, the rise of cognitive
psychology and its ready explanation of instructional effects via process concepts like awareness
and insight may have further dampened behavior analysts' enthusiasm for a basic analysis of
verbal control.
In the late 1970's things began to change. Some behavioral researchers began to conclude
that some of the differences between human and nonhuman performances could be traced to the
effects of verbal events on human action: The so-called "language hypothesis" (Lowe, 1979).
Skinner's concept of rule-governed behavior provided an avenue for this hypothesis, in part
because of his lack of clarity about it. Skinner himself had entertained the possibility that verbal
events involved new principles: "the only differences I expect to see revealed between the behavior
of rat and man (aside from enormous differences in complexity) lie in the field of verbal behavior"
(1938, p. 442).
Human research began to assume a new role within basic behavior analysis. No longer
simply a means of confirming the generality of findings with nonhumans, it has become a means
for constructing an empirically-based behavior analytic account of verbal behavior itself. This
change has had a profound effect on basic behavior analysis. For example, its flagship journal, The
Journal of the Experimental Analysis of Behavior, has gone from a small fraction of human studies in
the early 1970's, to nearly an even split in the current volume. The new human research tradition
has led to an increasing openness to new principles, preparations, and findings, and to more
diversity within the field on the question of the continuity strategy. It has also yielded an
interesting collection of empirical findings and theoretical suppositions that are both relevant to
other versions of human psychology (e.g., cognitive psychology) and virtually unknown outside of
behavior analysis.
Schedule Control
The role of verbal behavior is being studied, identified and explained in a variety of ways.
A popular preparation, early in the move toward more human research, was the FI schedule. This
schedule was of particular interest because it is a well established finding that human FI
performance differs significantly from that of other organisms (e.g., Leander, Lippman & Meyer,
1968; Lippman & Meyer, 1967; Weiner, 1964, 1965, 1969). There are ways, however, to reduce this
effect and they all seem to turn on the issue of verbal behavior.
For example, human temporal schedule performance is more like that of nonhumans when
steps are taken to reduce counting. Procedures have included concurrent verbal tasks such as
mental math or reading aloud (e.g., Laties & Weiss, 1963; Lowe, Harzem, & Hughes, 1978) or a
response-produced clock (Lowe, Harzem, & Bagshaw, 1978; Lowe, Harzem, & Hughes, 1978).
Similarly, if verbal events are involved in schedule control differences, then pre-verbal
humans should show patterns more characteristic of nonhumans. Lowe, Beasty, and Bentall (1983)
tested responding on FI schedules with two human infants, ages 9 and 10 months, respectively.
Their performances were more like those of nonhuman subjects than like those of adult humans.
Developmental studies have shown a gradual transition (especially from about age 2 to age 6 or 7)
in performance from the animal-like performances of infants to the pattern of adults, both in
schedule performance (e.g., Bentall, Lowe, & Beasty, 1985) and in the effects of self-instructions
(Bentall & Lowe, 1987).
Research on verbal rules has examined whether instructed subjects respond differently than
uninstructed subjects to changes in environmental contingencies. In general, instructions alter the
impact of contingency manipulations. For example, Shimoff, Catania, and Matthews (1981)
compared shaped to instructed performance with interval and ratio schedules, each with a
superimposed differential reinforcement of low rate (DRL) contingency. For both schedules when
the DRL contingency was relaxed, response rate increased for most subjects whose original
behavior was shaped but not for those who were instructed. Shimoff et al. concluded that this type
of insensitivity is a "defining property" of instructional control (p. 207). Other researchers have
since shown that this effect diminishes with time (Michael & Bernstein, 1991) and that instructions
can produce more control by programmed contingencies than no instructions under some
circumstances (Hayes, Brownstein, Zettle, Rosenfarb, & Korn, 1986; Joyce & Chase, 1990).
The changes in responding to programmed contingencies produced by verbal instructions
result from at least two different effects (Hayes, S. C., Zettle, & Rosenfarb, 1989): (a) instructions
alter the range of responding available to make contact with programmed contingencies, and (b)
instructions establish additional socially-mediated contingencies for responding in accord with
verbal rules. Several studies have shown these separate influences (e.g., Barrett, Deitz, Gaydos, &
Quinn, 1987; Hayes, Brownstein, Haas, & Greenway, 1986; Hayes et al, 1986; Joyce & Chase, 1990;
Shimoff, Matthews, & Catania, 1986; see Hayes, Zettle, & Rosenfarb, 1989 for a review). For
example, Hayes et al., (1986) had subjects press buttons to move a light through a matrix according
to a multiple fixed ratio (FR) 18 / DRL 6 s schedule to earn chances on money prizes. These
schedules were chosen because they control different rates of responding (high rate with the FR
and low with the DRL). An instruction to "Go Fast" was associated with one of the lights presented
to the subject and "Go Slow" with a different light. In one condition, only the Go Fast light was on;
in a second condition, only the Go Slow light was on; and in a third, the lights alternated within
each component of the multiple schedule. Thus, in all cases the instruction light in each condition
was accurate only half of the time, but the accuracy/inaccuracy was distributed differently across
the three groups. Those who always saw "Go Fast" earned points in the FR component, but not the
DRL component. In the "Go Slow" condition, the reverse was true. Those who had the instruction
lights rapidly alternated contacted the contingencies in both components.
While the instruction lights were lit the subjects behaved according to them regardless of the
actual schedule. For half the subjects the instruction lights were then unexpectedly turned off .
Subjects with alternating instructions immediately showed schedule-appropriate behavior.
Subjects with consistent instructions generally did not, since the restricted range of behaviors
produced by these instructions did not foster contact with the actual contingencies in both
components. Implicit social contingencies maintained instructional compliance while the
instructional lights were lit, but when they were turned off, only subjects with alternating
instructions emitted a range of behaviors that effectively contacted the underlying schedule. Thus,
instructions either increased or decreased schedule-appropriate behavior, depending upon the
specific instructions and specific schedules.
The general findings in behavior analytic studies of this kind are remarkably consistent (see
Hayes, S. C., 1989a for a review): On one hand, verbal rules or instructions interact with many
common behavioral procedures and preparations to produce behavioral patterns in humans not
commonly seen in nonverbal organisms. On the other, when unexplained differences between
human and nonhuman performances are studied carefully they often seem to resolve into
differences produced by the impact of verbal rules.
In effect, some of the conclusions reached by Brewer (1974) now make sense within a
behavior analytic account, but with a twist. Verbal processes often interact with other basic
behavioral processes in normal adults and it is indeed difficult to obtain unequivocal evidence for
operant or classical conditioning uncontaminated by verbal involvement. Rather than abandon
concepts like operant or classical conditioning, though, behavior analysts are busy examining the
nature of these verbal (or one might say, "cognitive") processes from a behavioral point of view.
Contemporary Behavior Analytic Research on Derived Stimulus Relations
It is one thing to document the involvement of verbal events in complex human behavior. It
is another to explain why verbal events have these effects, or even more basically, to explain what
verbal events are. While Skinner did not have a functional definition of verbal stimuli, behavior
analysts are identifying functional processes that seem to provide one.
This area of research is among the most rapidly developing in the behavioral analytic
tradition. Although there are several preparations of importance to the issues involved, by far the
most researched is that of stimulus equivalence.
Stimulus Equivalence
When humans are taught a series of related conditional discriminations, the stimuli that
enter into those discriminations can often become connected to each other in new ways, not
explicitly taught. The phenomena involved are typically investigated in a matching to sample
format. Suppose a person is taught, given a novel visual form (the sample - call it X1), to choose
another visual form from an array (the comparisons - call them Y1, Y2, and Y3). We could say that
the person learns "given X1 pick Y1, not Y2 nor Y3." The person is then taught to select another
unfamiliar visual form from another array of forms, given the same sample, or "given X1 pick Z1,
not Z2 nor Z3." So far we have two simple discriminations. In the presence of other samples,
however, what was correct is now incorrect and vice-versa. For example, the person would learn
"given X2 pick Y2, not Y1 nor Y3." In the typical study, all comparisons are correct and incorrect
equally often but the discriminations involved are conditional: a given comparison is correct only
given a certain sample (in the rest of the paper we will describe discriminations of this kind with
the shorter phrase "given X1, pick Y1" but it should be remembered that we are speaking of
conditional discriminations). With this kind of training, the person probably will select X1 from
an array of comparisons (not X2 nor X3), given Y1 or Z1 as samples, and Y1 (not Y2 nor Y3) from
an array given Z1 as a sample, and Z1 (not Z2 or Z3) given Y1 as a sample (e.g., Sidman, 1971;
Sidman, Cresson, & Willson-Morris, 1974). The same performances would be seen for the "2s" and
the "3s." While the relation between, say, X1 and Y1 is directly trained, that between Y1 and X1 is
derived. It has no direct history of training within the experiment and yet emerges only after the
trained relations have been established.
This set of phenomena is called "stimulus equivalence." The basic phenomenon was
acknowledged by the ancient Greeks, and has had several incarnations in modern psychology. S-R
learning theory struggled with the topic from the 1930's (e.g., Hull, 1934) until its disintegration in
the 1960's (e.g., Osgood, 1953; Jenkins & Palermo, 1964). Indeed, part of the reason for the collapse
of S-R psychology was the failure of S-R models of response and stimulus mediation (Jenkins &
Palermo, 1964). In its modern behavior analytic incarnation, however, the concept is associated
both with a more robust experimental methodology, and with behavioral theorizing unburdened
by S-R associationism and mediationalism.
In contemporary treatments, an equivalence class is usually said to exist if the trained
stimulus-stimulus relations show the three defining derived properties of reflexivity, symmetry,
transitivity, and combinations of these (Sidman & Tailby, 1982). In matching-to-sample
procedures, reflexivity is usually inferred from generalized identity matching. For example,
without explicit training, given X1 the person picks X1 and not X2 nor X3 from an array. Symmetry
refers to the functional reversibility of a trained relations: When "given X1 pick Y1" is trained,
"given Y1 pick X1" emerges in the absence of direct reinforcement (Sidman, Rauzin, Lazar,
Cunningham, Tailby & Carrigan, 1982). To evaluate transitivity, at least three stimuli in two linked
relations are required. If after "given X1 pick Y1" and "given Y1 pick Z1" have been taught, "given
X1 pick Z1" emerges without additional training, transitivity has been demonstrated. Although it
is somewhat confusing, the derived relation "given Z1 pick X1" is often termed simply an
equivalence relation (Fields, Verhave, & Fath, 1984) and is viewed as evidence that the trained
relations have the properties of symmetry and transitivity. These defining properties of stimulus
equivalence are shown in Figure 1.
-----------------------------------------------
Insert Figure 1 About Here
-----------------------------------------------
Stimulus Equivalence as a Working Model of Semantic Relations
A major source of interest in stimulus equivalence is the apparent correspondence between

the stimulus equivalence phenomenon and language phenomena. For example, in normal verbal
training, word-object relations are symmetrical. If a child of sufficient verbal abilities is taught to
point to a particular object among several given a particular written word, the child may, when
later given the object, point to the specific word among an array of words without specific training
to do so. Natural examples can readily be given for transitivity as well.
Such preparations have been used in several studies, and stimulus equivalence has readily
emerged (e.g., Dixon & Spradlin, 1976; Sidman, 1971; Sidman & Tailby, 1982; Sidman, Kirk, &
Willson-Morris, 1985; Spradlin & Dixon, 1976; Yamamoto, 1986). Some behavior analytic
researchers have established successful language training or reading programs using equivalence
and equivalence-like procedures exclusively (e.g., de Rose, de Souza, Rossito, & de Rose, 1992).
This kind of overlap suggests that the equivalence phenomenon may provide a new avenue for the
empirical investigation of verbal events. For some behavior analysts, it has become a kind of
working empirical model of semantic relations (e.g., Hayes & Hayes, 1989).
If stimulus equivalence provides a preliminary behavior analytic model of verbal
stimulation, one would expect to see it emerge readily in humans, but not so readily (or perhaps
not at all) in nonhumans. This expectation assumes only that humans have a relative facility with
verbal behavior and thus, if equivalence is in some way a model of verbal stimulation, humans
should also have a facility with equivalence. This has turned out to be the case. Stimulus
equivalence has been shown with a wide variety of human subjects using a wide variety of
stimulus materials (e.g., Dixon, 1977; Dixon & Spradlin, 1976; Gast, VanBiervliet, & Spradlin, 1979;
Hayes, L. J., Tilley, & Hayes, S. C., 1988; Mackay & Sidman, 1984; Sidman, 1971; Sidman et al.,
1974; Spradlin, Cotter, & Baxley, 1973; Sidman & Tailby, 1982; Spradlin & Dixon, 1976;
VanBiervliet, 1977; Wulfert & Hayes, S. C., 1988). Even children as young as two years old will
display such performances (Devany, Hayes, S. C., & Nelson, 1986). Stimulus equivalence has not
been shown with nonhuman organisms, however. While conditional discriminations are readily
demonstrated in a large variety of animals, including dolphins (e.g., Herman & Thompson, 1982),
rats (e.g., Lashley, 1938), and monkeys (e.g., Nissen, 1951), these have not resulted in stimulus
equivalence. To date, no unequivocal demonstration of stimulus equivalence in nonhumans has
been shown (D'Amato, Salmon, Loukas, & Tomie, 1985; Dugdale & Lowe, 1990; Kendall, 1983;
Sidman et al., 1982; Lipkens, Kop, Matthijs, 1988. As for McIntire, Cleary, & Thompson, 1987 and
Vaughan, 1988, see Hayes, S. C., 1989b and Saunders, 1989). Whether this is simply due to
methodological problems is not clear, but it is clear that nonhumans at least do not show
equivalence readily. In humans, performance on equivalence tasks seems to be correlated with
verbal performance, as would be expected if equivalence is a type of verbal relation (Barnes,
McCullagh, & Keenan, 1990; Devany, et al., 1986; Dugdale & Lowe, 1990).
Conditionality is also necessary for equivalence to serve as a preliminary model of verbal
stimulation, because it provides for the necessary precision seen in language. For example, the
meaning of the word "gay" could be either "happy" or "homosexual," depending in part on the
context in which it occurred. Conditional equivalence relations, in which elements are related to
others only in certain contexts, have been demonstrated (e.g., Bush, Sidman, & de Rose, 1989; Gatch
& Osborne, 1989; Kennedy & Laitinen, 1988; Wulfert & Hayes, S. C., 1988). Words themselves also
function as contexts in which equivalence class membership is reconstituted (Kohlenberg, Hayes, S.
C., and Hayes, L. J., 1991). This corresponds nicely to natural language. For example, compare the
different meanings of the word "bat" in the following sentences: "Babe Ruth held the bat" and
"Dracula held the bat." In this example, "bat" is in an equivalence relation with a piece of wood or a
flying mammal, depending upon other words ("Babe Ruth" or "Dracula," respectively) as
contextual stimuli.
The Challenge of Stimulus Equivalence
To understand why stimulus equivalence is having a major impact on behavioral
researchers one should note that it is not readily predicted from a direct contingency formulation.
If an organism selects Y1 over Y2 given X1, and selects Y2 over Y1 given X2. The probability of
reinforcement for selecting Y1 is greater in the presence of X1 than in the presence of, say, X2. We
would say that X1 is functioning as a conditional discriminative stimulus in the presence of which
Y1 is functioning as a discriminative stimulus for a selection response.
This does not mean, however, that the probability of reinforcement for selecting X1 is
greater in the presence of Y1 than in its absence. Consider a natural example. A monkey may look
for and find a snake (and not, say, a cloud) given a rustling sound, but not listen for a rustling
rather than some other sound given a snake. We could think of this as a trained relation: given
rustle find snake. Such a contingency arrangement provides few grounds to suppose that reversing
the functions of the stimuli will be reinforced: given snake find rustle.
In the natural environment of nonverbal organisms, trained sequential or other relations
between two events cannot be assumed to reverse. Most commonly, if the functions were reversed
the consequences would be neutral. Often they would be negative or even life-threatening. Thus it
is not clear why operant contingencies should be functionally reversible. Respondent conditioning
also provides a no clear process to account for the equivalence results, especially when we move
beyond simple symmetry (Hayes, S. C. & Hayes, L. J., 1989).
The development of theoretical accounts of stimulus equivalence is still in its infancy in
behavior analysis. Some researchers view stimulus equivalence as a primitive which cannot be
analyzed into component processes (e.g., Sidman, 1990). In behavior analysis, primitives such as
reinforcement or stimulus control can only be described in terms of the conditions in which they
are seen. They are not "explained" psychologically. If more of an explanation is demanded,
usually only an evolutionary one will be attempted. For example, it has been said that conditional
discriminations may lead to equivalence in humans (at least) simply because humans have evolved
that way (Sidman, 1990).
A second alternative has been to explain equivalence on the basis of verbal mediation (e.g.,
Dugdale & Lowe, 1990; McIntire et al., 1987), much as earlier S-R learning theorists did. These
analysis have so far either failed to be demonstrated empirically (e.g., see Hayes, 1989b) or appeal
to derived relations between stimuli and names that are of the same sort as the derived stimulus
relations they mean to explain.
Another alternative, the relational frame account, has appealed to the historical
development of relational responding (Hayes, S. C., 1991; Hayes, S. C. & Hayes, L. J., 1989). This
view will be discussed in some detail for three reasons. First, it shows how fairly traditional
behavior analytic concepts are leading to quite different views of verbal events. Second, it will be
important for other reasons later on to argue that equivalence might itself be learned, and relational
frame theory describes one plausible way this might happen. Finally, the analysis shares obvious
similarities with contemporary cognitive network theories of meaning, and thus might be of more
general interest to non-behavioral readers.
Arbitrarily Applicable Relational Responding
The relational conception of stimulus equivalence starts from a fairly conservative stance:
How can equivalence be explained via traditional contingency concepts? Relational frame theory
draws upon two central ideas. First, a variety of experimental findings (for example, transposition,
Reese, 1968) show that humans and nonhumans can respond to formal or non-arbitrary relations
between stimuli. Perhaps some organisms can take an additional step and bring relational
responding under the control of context rather than the form of the relata. Such abstracted
relational responding would then be arbitrarily applicable to any events, needing only contextual
cues to do so. Second, the behavior analytic account of generalized imitation (e.g., Gewirtz &
Stengle, 1968), grammatical frames (Skinner, 1957), and other phenomena draws upon the idea that
organisms can learn overarching behavioral classes containing virtually unlimited numbers of
members. Although equivalence and similar phenomena seem to show novel performances,
perhaps this performance is in actuality learned. That is, the action of relating two arbitrary stimuli
itself has a history.
Take the case of a young child learning the name of an object. A child is oriented toward
several objects. The parent points to one and asks "What's that?" At first parents do not expect a
response and after a pause, they say the name. If the child shows any sign of acknowledgement
(e.g., a smile when the name is given), the child is tickled, praised, or otherwise rewarded.
Sometimes incorrect names will be said deliberately, and the child will be expected to show that it
is not correct. Later only actual correct vocalizations by the child are rewarded. Incorrect
responses result in additional prompting of the name.
Eventually, the child learns "given object X, say name X, not name Y" (object → name.) But
the child is also asked "Where is (name)?" Picking the named object from among others is
rewarded with tickles, play, or verbal consequences ("That's right! That's the (name)"). We can say
that the child learns "given name X, point to object X not object Y" (name → object.) Bi-directional
training of this kind occurs only in certain contexts. There are a variety of cues to indicate that this
is a naming context, including the use of phrases such as "What's that?" and the juxtaposition of
objects and words.

As with generalized imitation, we might suppose that with enough instances of such
training the child will begin to learn to derive name → object given only object → name and vice
versa. In short, with enough instances of directly trained bi-directional stimulus relations in a
given context, bi-directional responding of a more general sort may arise in that context. The
process here might be similar to that discussed by Skinner in explaining the speaker's ability to
mand events never before manded (1957). On the basis of a history of reinforcement for specific
mands, the speaker is thought to learn the mand relation as a frame into which novel events may
be incorporated. The capacity to abstract the frame in the case of the mand, or with the kind of
stimulus relations we are discussing, may simply require a sufficient history of bi-directional
training or it may also require that a particular species be involved (i.e., humans). Beyond such an
historical or evolutionary appeal, behavior analytic theory does not ordinarily attempt an account
of such processes, either for the present topic or in other areas such as generalized imitation.
Once abstracted from a particular history, the derivation of stimulus relations is not based
solely on the formal properties of the relata. It is arbitrarily applicable in the sense that it may be
brought to bear on any stimuli in the presence of appropriate contextual cues. If, for example, a
person is told "X = Y > Z" the bi-directional relations between all possible pairs of X, Y, and Z (e.g.,
Z < X) can be stated, not based on the form of these related events, but on the arbitrarily specified
and derived relations among them brought to bear by cues such as = and <.
Arbitrarily applicable relational responding involves responding to one event in terms of
the other and vice versa based on contextual cues to do so. As such it is a pattern of the
transformation of stimulus functions. It has three characteristics (Hayes, S. C., 1991; Hayes, S. C. &
Hayes, L. J., 1989):
Mutual entailment. If X is related to Y in a given context, then Y is related to X in that

context. The specific relations involved need not be literally identical. If X is better than Y, then Y
is worse than X. The second relation is entailed by the first -- it is not possible to have a relation of
"better" without one of "worse." We may describe mutual entailment this way:
Crel { X rx Y ||| Y ry X }
That is: when a contextual event brings to bear upon the current situation a history of a particular
kind of relational responding ("Crel"), such that a relation of a particular kind (" rx ") is specified
between two events X and Y, then a second relation (" ry") is necessarily entailed ( "|||") between
Y and X.
Combinatorial entailment. If in a given relational context X is related in a particular way to
Y and Y is also related to Z, then some kind of relation must be entailed from X to Z and from Z to
X. More formally:
Crel { X rx Y and Y rz Z ||| X rp Z and Z rq X}
Combinatorial entailment differs from simple mutual entailment in more than mere complexity. In
mutual entailment, the specified relation between X and Y entails a relation between Y and X at the
same level of precision. In the formula above, however, the X-Z and Z-X relations ("rp" and "rq")
may not be specified to the same degree as that between X and Y, and Y and Z ("rx" and "rz" ). For
example, if X is different than Y and Y is different than Z, we cannot say what the relation is
between X and Z and between Z and X. This lack of precision, however, is specified by the nature
of the relation of "difference" that applies to X and Y, and to Y and Z. We can say that we cannot
say what the relation is. This is part of what distinguishes a psychological from a merely logical or
mathematical relation. A verbal relation of "no relation" is a relation psychologically speaking
because it requires responding to one event in terms of another. It is not a logical or mathematical
relation, however, which are more limited subsets of psychological stimulus relations.
Transfer of functions. It is this last characteristic of arbitrarily applicable relational
responding that provides the general psychological importance of such events. If an event X has a
psychological function, and that event is placed into a relation with another event Y, under certain
conditions Y may acquire a new psychological function based on the function of X and the relations
between X and Y. Because if all functions of one stimulus transferred to another and vice versa,
there would no longer be two separate psychological stimuli, which functions transfer must be
under contextual control. We can represent the transfer of functions this way:
Cfunc [ Crel { Xf1 ||| Yf2r and Zf3r } ]
That is, given the presence of contextual cues that select particular psychologically relevant, non-
relational stimulus functions ( "Cfunc"), that adhere in event X ("f1"), events Y and Z will show these
functions, transformed in terms of the derived relations between these events and event X, ("Yf2r
and Zf3r"). Because the transfer of functions is thought to be under contextual control, under some
conditions with some functions no transfer may be seen. In the relational frame conception,
however, all arbitrarily applicable stimulus relations enable some forms of such transfer.
Transfer of functions is a definitional aspect of relational frames in part to denote their
psychological nature, but it is more than that. Mutual entailment and combinatorial entailment are
themselves interpretable as transfer of functions in a limited sense. For example, in the traditional
equivalence study done using a conditional discrimination procedure, the trained conditional
function of a sample transfers to the related comparison if symmetry is shown--by definition. Is it
possible for only relational stimulus functions to transfer and for no non-relational stimulus
functions to transfer? It seems unlikely. For example, suppose a person learns a relation between
X and Y and is then asked to "picture what is related to Y." If the person has derived a strong
relation between Y and X, then presumably at least the perceptual functions of X will be available
to some degree with regard to Y. That is, the person will probably picture X. Transfer of these
perceptual functions is enabled by the property of mutual entailment, but it is distinct from it.
Thus, transfers of functions seem to be a definitional aspect of relational frames. Arbitrarily
applicable relations are patterns of the mutual transformation of both relational and non-relational
stimulus functions.
As a practical example of a transfer of functions, suppose a child is trained that the written
word DOG is called "dog" and that the word DOG goes with actual dogs. We may say that the
child has had two relations directly trained: DOG → dog and DOG → "dog." Later the child plays
with a dog for the first time and enjoys it. We may say that dogs have directly acquired various
functions by virtue of the play. Now, upon hearing his mother say "dogs" from another room the
child may smile and go to the other room even though a dog is not visible and the child has no
direct history of reinforcement for any of these activities in response to the word "dog." Based on
the relational frame account, we might say that dogs have become discriminative stimuli for
approach or eliciting stimuli for emotional responding directly through play. The word "dog" has
acquired similar functions not through a direct history, but indirectly through its participation in a
derived relation with dogs.
The literature on the transfer of functions across equivalence relations would support the
basic outlines of such an analysis. Transfer of discriminative functions have been shown across
simple equivalence relations(e.g., Hayes, S. C., Brownstein, Devany, Kohlenberg, & Shelby, 1987;
Kohlenberg et al., 1991), and conditional equivalence relations (Wulfert & Hayes, S. C., 1988), and
across symmetrically related stimuli (e.g., Catania, Horne, & Lowe, 1989; de Rose, McIlvane, Dube,
Galpin, & Stoddard, 1988; Lazar, 1977; Lazar & Kotlarchyk, 1986; Gatch & Osborne, 1989). In
much the same way, consequential functions transfer through equivalence relations (Hayes, S. C.,
et al., 1987; Hayes, S. C., Kohlenberg, & Hayes, L. J., 1991), as do conditioned emotional responses
(Augustson & Dougher, 1991).
Relational frame. The term "relational frame" was coined to designate particular kinds of
arbitrarily applicable relational responding (Hayes, S. C. & Hayes, L. J., 1989). A relational frame is
a type of responding that shows the contextually controlled qualities of mutual entailment,
combinatorial entailment, and transfer of functions; is due to a history of relational responding
relevant to the contextual cues involved; and is not based on direct non-relational training with
regard to the particular stimuli of interest. If a relation is a pattern of mutual transformation of
stimulus functions, a relational frame is the name for that particular pattern. It is a noun because
the structure of English demands it, but it is a situated action. A relational frame is always
"framing relationally."
In a relational frame view, stimulus equivalence is a special case of arbitrarily applicable
relational responding based on a "frame of coordination" or sameness (Hayes, S. C. & Hayes, L. J.,
1989). A variety of evidence fits with the relational frame view of equivalence (see Hayes, S. C.,
1991; Hayes, S. C. & Hayes, L. J., 1989), and it makes several predictions that differ from most other
accounts (e.g., derived relations should show developmental trends, equivalence is only one of
many such phenomena, equivalence relations can be reconstituted in ever more subtle ways based
on cues to do so, new kinds of relations can be trained, conditional discrimination procedures are
only one of many that should lead to derived stimulus relations). The strongest support for the
theory so far has been the demonstration that a variety of relations can be applied to arbitrary
matching-to-sample tasks, and give rise to equivalence or non-equivalence depending upon the
relations applied (Steele & Hayes, 1991). In the Steele and Hayes experiments subjects' responses to
three non-arbitrary stimulus relations (sameness, oppositeness, or difference) were first brought
under contextual control. For example, given a contextual cue (background color on a computer
screen) for "same," a subject shown a short line as a sample earned points for picking a short line
from a set of comparisons. Given a contextual cue for "opposite" and the same sample, picking a
long line was rewarded. Given the "different" context, selecting any comparison other than one
identical to the sample was reinforced. Subjects were then exposed in the presence of these same
contextual cues to matching-to-sample training in which the trained stimulus relations were
between arbitrary visual forms that reliably shared no formal properties. In two experiments it
was shown that derived performances were consistent with relational responding brought to bear
by the contextual cues. In contexts relevant to the relation of sameness, stimulus equivalence
emerged. Other kinds of relational networks emerged in the other contexts. For example, when
arbitrary stimuli were related in the presence of the opposite cue, derived relations of sameness
and oppositeness alternated with each interceding stimulus (e.g., an opposite of an opposite was
treated as the same, but an opposite of an opposite of an opposite was treated as opposite).
Why Equivalence May Involve a New Behavioral Principle
The behavior analytic work on the derivation of stimulus relations seems to be opening up
behavior analysis to a new behavioral principle. Behavioral principles are admitted into the
armamentarium only very slowly within behavior analysis, but both of the major interpretations of
stimulus equivalence and similar phenomena seem to lead to it.
On the one had, equivalence may be a primitive process which cannot be analyzed into
component processes (e.g., Sidman, 1990). If this is true, equivalence instantiates a new behavioral
principle by definition, since what is meant by a behavioral principle is a basic process. The
transfer of functions through equivalence classes would also be considered a new phenomenon, if
only because it is based on this new behavioral processes.

Conversely, equivalence and other derived stimulus relations may be learned. The
relational frame account is one such learning-based interpretation, but others seem possible. In this
case, neither equivalence per se nor higher-order concepts such as relational frames involve new
behavioral principles because the processes appealed to are known. At most they are new
behavioral outcomes, much as generalized imitation may be a new behavioral outcome.
If the derivation of stimulus relations is learned, however, the transfer of functions through
such relations apparently instantiates a new behavioral principle. Behavior analysts distinguish
between stimulus functions on the basis of history and current context. For example, conditioned
and unconditioned reinforcers are distinguished on the basis of the particular histories that give
rise to each. When, as in the example earlier, the child approaches upon hearing the word "dog,"
we cannot say that "dog" is a discriminative stimulus. A discriminative stimulus is a stimulus in
the presence of which there has been a greater probability of reinforcement for a given behavior
than in its absence. In the example (and in actual studies of such phenomena, e.g., Kohlenberg et
al., 1991), there is no such history for the word "dog." Its function is only discriminative-like, based
on a transfer of control through a derived stimulus relation. In behavior analytic accounts the
control exerted by any given discriminative stimulus is learned, but discriminative control per se is
not. If the discriminative-like functions of "dog" are based on a learned pattern of the derivation of
stimulus relations, then this seems to be a new kind of stimulus control in which the underlying
process responsible for the effect is itself learned. The same conclusion would be reached for all the
other functions that transfer through derived relations (reinforcing, eliciting, establishing, and
evocative).
It may still be possible to interpret equivalence and similar relational phenomena in a way
that does not suggest a new behavioral principle. If our logic above is correct, though, such an
interpretation would require both that equivalence is learned and that transfer of functions through
equivalence does not depend on the learned nature of equivalence. It is not obvious how that
could be done.
If there is a new behavioral principle, it is not clear what it should be called. Although it
may cause confusion because of its broad use both in lay language and psychology, we (Hayes, S.
C. & Hayes, L. J., 1989) have previously used the word "verbal" as a technical term to qualify
functions that depend upon derived stimulus relations. We have argued that verbal events are
simply those that involve relational frames. Thus, the new general principle might be termed
"verbal control" and the specific terms could include such things as "verbal reinforcers" (e.g.,
stimuli with reinforcing functions based on the transfer of functions through derived stimulus
relations), "verbal antecedents," "verbal motivation" and so on. Unfortunately, the term "verbal" is
often used to refer to events based on their form and not their function. For example, if the word
"good" increases behavior when it is used contingently as a consequence, some may call it a "verbal
reinforcer" simply because it is a "word," not because of any specific psychological processes that
led to its reinforcing functions. If behavior analysts concluded (as is suggested by our analysis)
that verbal functions are distinct at the level of psychological process, perhaps this topographical
use would cease, to be replaced by functional distinctions such as the ones we are making. If not,
perhaps "relational control," "relational reinforcers," "relational antecedents," and so on, could
provide an alternative.
Linking Control by Verbal Rules to Derived Stimulus Relations
Previous behavioral analyses of verbal meaning and of rule control have attempted to
reduce these phenomena to outcomes of simple and direct operant and classical conditioning
preparations. The analyses that have resulted seem cumbersome and have had a difficult time
accounting adequately for verbal effects.
The literature on semantic generalization and conditioned meaning provides a useful
illustration. Staats and Staats (1957), for example, directly paired nonsense words with emotionally
laden words and found a transfer of emotional functions to the nonsense words as measured by a
semantic differential even in subjects "unaware" of the stimulus relation. Applying this model of
verbal meaning to everyday language runs afoul of two difficulties. First, it is difficult to suppose a
history of repeated pairings in most or all such circumstances in the natural linguistic environment.
Secondly, Staats and Staats' conclusions do not explain why there was a correlation between
measures of verbal awareness of the stimulus relation and the emotional ratings given to the
nonsense words (e.g., Cohen, 1964). From a relational frame point of view, such a correlation
should be expected, not because verbal awareness is necessary to conditioning, but because when a
subject can describe a stimulus relation it is certain to have been derived (cf., Skinner, 1945; 1966).
Similarly, early behavioral accounts of rule-governance appealed to direct processes of
discriminative control (e.g., Galizio, 1979). This account too supposed a history that is difficult to
imagine, in which the discriminative effects of all verbal rules must be learned directly.
The behavior analytic literature on derived stimulus relations provides alternatives both for
the meaning of verbal formulae and for their effects as rules. A detailed analysis of these issues
would take us too far afield (see Hayes, S. C., & Hayes, L. J., 1989 for some of our efforts in these
areas), but a brief outline might help show how the research on derived stimulus relations opens
up new ways of approaching old problems and how it is reflecting back on the literature on
instructional control and rule-governance.
Verbal meaning. The literature on derived stimulus relations suggests that when a person
hears a word, some of the functions of that word may depend on the derived stimulus relations
sustained between it and other events. When a sentence is presented, a variety of new relations
may be derived, under the contextual control of the words themselves, their sequence, the
nonverbal context, paralinguistic cues, and other events.
Verbally understanding the meaning of a word or sentence might thus be a psychologically
distinct action that is based on the act of a listener deriving stimulus relations (e.g., forming
equivalence relations, relations of difference, if ... then relations, and so on). For example, I may
"understand" the word "fire" in the sense that its functions depend in part upon derived relations
between it and other events (e.g., words like "hot" or "oxidation"). This would represent a different
kind of "understanding" than that based on functions of the word that are directly trained (e.g., if
"fire" had reliably preceded being burned and thus elicited emotional reactions via classical
conditioning). These two types of meaning -- verbal meaning and direct meaning -- seem based on
different psychological processes.
The same kind of interpretation could be applied to the meaning of a sentence. Suppose a
person is told "when the bell rings, there is a fire." While the verbal meaning of "bell" or "fire" may
be based on pre-existing equivalence relations, the sentence provides cues for the listener to derive
an additional dependency relation (if ... then) between the bell and fire. This network of stimulus
relations and their resulting functions could be considered the verbal meaning of the sentence.
This view differs markedly from Skinner's view (1957) in which the verbal meaning and direct
meaning of a sentence are not distinguishable at the level of psychological process -- both are
simply forms of stimulus control. Skinner developed this analysis before equivalence had been
incorporated into behavior analysis, which may account for much of the difference between his
view and ours.
Rule-following. When defined this way, following a rule as a rule requires an

understanding of it (although more than that is involved: see Hayes, Zettle, & Rosenfarb, 1989). A
dog might be taught to get slippers when the master says "slippers" but the command is not
functioning as a rule for the dog, since it has no verbal meaning. Rather, the evocative functions of
the word depend solely on direct training. Conversely, consider a person who is told "when the
bell rings, there is a fire" and now runs at the sound of the bell. In this case, the action of running
could be interpreted as a form of rule-following that is dependent upon derived stimulus relations.
"Fire" might have some of the functions of actual fires via a transfer of such functions through of
relations of coordination, and now "bell" has acquired the function of an actual fire about to
happen. Because "bell" has these functions and is in a coordination relation with actual bells, when
the bell rings, the person may run, much as would have occurred had an actual fire been seen.
This interpretation of rule-following is speculative, but does not stray far from what is
already known. Even if relational frame theory is incorrect, the empirical phenomenon of
equivalence alone would support much of the analysis above.
The properties of derived stimulus functions and verbal rules. If rule-governance is
dependent upon derived stimulus relations and the non-relational functions that transfer through
them, then the findings in the instructional control literature make sense behaviorally. Derived
stimulus relations seem to be specific, indirect, and arbitrary, among other characteristics (Hayes, S.
C. & Hayes, L. J., 1989; Parrott, 1984), and the qualities of behavior guided by verbal rules seems to
correspond with these properties.
There seems to be a difference in the specificity of derived stimulus functions, one that may
come from their relative independence from form as a source of behavioral functions. A normal
discriminative stimulus cannot easily come to control a specific response and only that response for
two reasons. First, the form of the stimulus establishes some functions directly. A key light, for
example, is not only a stimulus in the presence of which pecking is reinforced, it is also a stimulus
that allows the bird to look at things in an otherwise darkened chamber; it may elicit an orienting
response; it may provide sensory stimulation for glances in its direction; it may make eye
movements more likely, and so on. It cannot "mean" just one specific thing. Second, all co-present
or proximal events are potentially relevant to a non-arbitrary stimulus function. Thus, it is difficult
to avoid multiple stimulus functions otherwise present in the environment from adhering to any
part of that environment.
When arbitrary stimuli gain their functions based on their relations to other stimuli and not
to their forms, these stimulus functions can be narrowed and broadened by the network of derived
relations, without significant contamination from functions based upon formal properties or co-
present events. Such verbal stimuli "mean" the transformed functions of events that participate in
derived relations with them. This also prevents intrusions in meaning from random event sources.
For example, if a person heard "I'll meet you next Wednesday at noon" and shortly thereafter
stepped on a tack, the phrase would be unlikely to become a powerful cue for careful walking.
Much of the impact of verbal rules is due to a change in the range of behaviors available to
contact with the world (Hayes et al, 1986; Joyce & Chase, 1990). If derived stimulus functions are
more specific than direct stimulus functions it would help to explain how such a change can be
produced and manipulated.
Because of the arbitrariness of derived stimulus relations the verbal community can alter the
derivation of stimulus relations via social convention. For example, if the verbal community agrees
that a given event is "good" or "bad," the derived relation between that event and these evaluations
may significantly alter the function of these events for members of the verbal community. This
may help explain the sensitivity of rule-governed behavior to social context (e.g., Barrett et al.,
1987).
We have already explained how the stimulus functions that transfer through equivalence
relations are indirect. This might help explain how rules can compete with control by programmed
contingencies. For example, a person might have been taught to relate sex to "immorality." The
direct effects of sexual contact may differ completely from the indirect effects that come from this
relation. Perhaps it is sensitivity to such indirect effects that helps produce the apparent
"insensitivity" to the direct effects of events that is seen in the rule-governance literature (Hayes, S.
C., 1989a).
Conclusion
Skinner rejected the notion that "certain basic linguistic processes [are] common to both
speaker and listener" (Skinner, 1957, p. 33) but this is precisely the position we are led to consider
when we try to extend the equivalence literature into the areas of verbal meaning and rule-
governance. Ideas present in the mind, however, are not the common ground between speaker and
listener. In the present view, derived stimulus relations sustained by social convention constitute
that ground.
The trend in behavior analysis to grapple with derived stimulus relations suggests a merger
of behavioral and cognitive research areas and interests. Although equivalence and relational
frames do not seem to have a place in the cognitive literature as yet, the phenomena involved seem
central to contemporary cognitive interests in such areas as stimulus class formation or semantic
meaning. Other types of derived stimulus relations, such as those involved in stimulus exclusion,
are being actively studied both by behavior analysts (e.g., McIlvane, Kledaras, Munson, King, de
Rose, & Stoddard, 1987) and by cognitive psychologists (e.g., Markman & Wachtel, 1988). The
network theory of meaning implied by relational frame theory overlaps considerably with
contemporary cognitive theories of meaning (e.g., Reese, 1991). If the underlying relational frame
conception is correct many of the cognitive findings shown with semantic networks (e.g., priming
effects) should apply to equivalence classes (or other relational networks) and vice versa. In short,
behavior analysis is now a field actively studying "cognition."
What, then, is the difference between the present kind of analysis and mainstream cognitive
psychology? In part, the answer is to be found in the most fundamental premises of the two
positions. While much of cognitive psychology -- especially that driven by computer metaphors --
seems mechanistic, the present analysis is self-consciously contextualistic (Hayes, Hayes, & Reese,
1988). One implication of this philosophical difference is that cognitive psychologists are interested
in developing models that correspond to the real world and are assessed by predictive verification,
while contextualistic behavior analysts focus on theories that help accomplish their pragmatic goals
of prediction and control. As a result, while cognitive theories embrace mediational accounts on
the grounds that they model mental machinery well, behavior analysts reject them on the grounds
that they are unhelpful to their goals (Hayes & Brownstein, 1986). Deriving stimulus relations, for
example, is conceptualized in relational frame theory as historically and contextually situated
actions of whole organisms, not as a mediational process or mechanism occurring in the brain or
mind. Some within the cognitive tradition have begun to argue for a similar approach (e.g.,
Watkins, 1990).
An explosion of research in stimulus equivalence and rule-governance has occurred since
the late 1970's. Too little time has passed for the assimilation of these new data into
characterizations of the behavioral perspective by those outside that perspective--indeed, the
assimilation within behavior analysis is still underway. It is clear, however, that behavior analysis
is actively studying many topics that were once considered outside its purview. It is also clear that
although the position is evolving and some of Skinner's theoretical suppositions are being
abandoned or transformed, it is still behavior analysis--alive and doing well.
References
Augustson, E. & Dougher, M. J. (November 1991). Transfer of a conditioned emotional response
via stimulus equivalence. Paper presented at the meeting of the Association for
Advancement of Behavior Therapy, New York.
Ayllon, T., & Azrin, N. H. (1964). Reinforcement and instructions with mental patients. Journal of
the Experimental Analysis of Behavior, 6, 327-331.
Barrett, D. M., Deitz, S. M. Gaydos, G. R., Quinn, P. C. (1987). The effects of programmed
contingencies and social conditions on response stereotypy with human subjects.
Psychological Record, 37, 489-505.
Bentall, R. P., & Lowe, C. F. (1987). The role of verbal behavior in human learning. III:
Instructional effects in children. Journal of the Experimental Analysis of Behavior, 47, 177-
190.
Bentall, R. P., Lowe, C. F., & Beasty, A. (1985). The role of verbal behavior in human learning. II:
Developmental differences. Journal of the Experimental Analysis of Behavior, 43, 165-181.
Brewer, W. F. (1974). There is no convincing evidence for operant or classical conditioning in adult
humans. In W. B. Weimer & D. S. Palermo (Eds.), Cognition and the symbolic processes.
Hillsdale, NJ: Erlbaum.
Bush, K. M., Sidman, M., & de Rose, T. (1989). Contextual control of emergent equivalence
relations. Journal of the Experimental Analysis of Behavior, 51, 29-45.
Catania, A. C., Horne, P., & Lowe, C. F. (1989). Transfer of function across members of an
equivalence class. The Analysis of Verbal Behavior, 7, 99-110.

Cohen, B. H. (1964). Role of awareness in meaning established by classical conditioning. Journal of
Experimental Psychology, 67, 373-378.
D'Amato, M. R., Salmon, D. P., Loukas, E., & Tomie, A. (1985). Symmetry and transitivity of
conditional relations in monkeys (Cebus apella) and pigeons (Columba livia). Journal of the
Experimental Analysis of Behavior, 44, 35-47.
de Rose, J. T., de Souza, D. G., Rossito, A. L., & de Rose, T. M. S. (1992). Stimulus equivalence and
generalization in reading after matching to sample by exclusion. In S. C. Hayes & L. J.
Hayes (Eds.), Understanding verbal relations (pp. 69-82). Reno, NV: Context Press.
de Rose, J. T., McIlvane, W. J., Dube, W. V., Galpin, V. C., & Stoddard, L. T. (1988). Emergent
simple discrimination established by indirect relation to differential consequences. Journal
of the Experimental Analysis of Behavior, 50, 1-20.
Devany, J. M., Hayes, S. C., & Nelson, R. O. (1986). Equivalence class formation in language-able
and language-disabled children. Journal of the Experimental Analysis of Behavior, 46, 243-
257.
Dixon, L. S. (1977). The nature of control by spoken words over visual stimulus selection. Journal of
Dixon, M. H. & Spradlin, J. E. (1976). Establishing stimulus equivalences among retarded
adolescents. Journal of Experimental Child Psychology, 21, 144-164.
Dugdale, N. & Lowe, C. F. (1990). Naming and stimulus equivalence. In D. E. Blackman & H.
Lejeune (Eds.), Behavior analysis in theory and practice: Contributions and controversies
(pp. 115-138). Hillsdale, NJ: Erlbaum.
Fields, L., Verhave, T. & Fath, S. (1984). Stimulus equivalence and transitive associations: A
methodological analysis. Journal of the Experimental Analysis of Behavior, 42, 143-157.

Galizio, M. (1979). Contingency-shaped and rule-governed behavior: Instructional control of
human loss avoidance. Journal of the Experimental Analysis of Behavior, 31, 53-70.
Gast, D., VanBiervliet, A. & Spradlin, J.E. (1979). Teaching number-word equivalences: A study of
transfer. American Journal of Mental Deficiency, 83, 524-527.
Gatch, M. B. & Osborne, J. G. (1989). Transfer of contextual stimulus function via equivalence class
development. Journal of the Experimental Analysis of Behavior, 51, 369-378.
Gewirtz, J. L. & Stengle, K. G. (1968). Learning of generalized imitation as the basis for
identification. Psychological Review, 5, 374-397.
Hayes, L. J., Thompson, S., & Hayes, S. C. (1989). Stimulus equivalence and rule following. Journal
of the Experimental Analysis of Behavior, 52, 275-291.
Hayes, L. J., Tilley, K., & Hayes, S. C. (1988). Extending equivalence class membership to gustatory
stimuli. The Psychological Record, 38, 473-482.
Hayes, S. C. (1987). Upward and downward continuity: It's time to change our strategic
assumptions. Behavior Analysis, 22, 3-6.
Hayes, S. C. (Ed.). (1989a). Rule-governed behavior: Cognition, contingencies, and instructional
control. New York: Plenum.
Hayes, S. C. (1989b). Nonhumans have not yet shown stimulus equivalence. Journal of the
Hayes, S. C. (1991). A relational control theory of stimulus equivalence. In L. J. Hayes & P. N.
Chase (Eds.), Dialogues on verbal behavior (pp. 19-40). Reno, NV: Context Press.
Hayes, S.C. & Brownstein, A. J. (1986). Mentalism, behavior-behavior relations and a behavior
analytic view of the purposes of science. The Behavior Analyst, 9, 175-190.
Hayes, S. C., Brownstein, A. J., Devany, J. M., Kohlenberg, B. S., & Shelby, J. (1987). Stimulus
equivalence and the symbolic control of behavior. Mexican Journal of Behavior Analysis,
13, 361-374.
Hayes, S. C., Brownstein, A. J., Haas, J. R., & Greenway, D. E. (1986). Instructions, multiple
schedules, and extinction: Distinguishing rule-governed from schedule controlled behavior.
Journal of the Experimental Analysis of Behavior, 46, 137-147.
Hayes, S. C., Brownstein, A. J., Zettle, R. D., Rosenfarb, I., & Korn, Z. (1986). Rule-governed
behavior and sensitivity to changing consequences of responding. Journal of the
Hayes, S. C. & Hayes, L. J. (1989). The verbal action of the listener as a basis for rule-governance.
In S. C. Hayes (Ed.), Rule-governed behavior: Cognition, contingencies, and instructional
control. (pp. 153-190). New York: Plenum.
Hayes, S. C., Hayes, L. J., & Reese, H. W. (1988). Finding the philosophical core: A review of
Stephen C. Pepper's World Hypotheses. Journal of the Experimental Analysis of Behavior, 50,
97-111.
Hayes, S. C., Kohlenberg, B. K., & Hayes, L. J. (1991). Transfer of consequential functions through
simple and conditional equivalence classes. Journal of the Experimental Analysis of
Behavior, 56, 119-137.
Hayes, S. C., Zettle, R. D., & Rosenfarb, I. (1989). Rule following. In S. C. Hayes (Ed.),
Rule-governed behavior: Cognition, contingencies, and instructional control. (pp. 191-220).
New York: Plenum.
Herman, L. M. & Thompson, R. K. (1982). Symbolic, identity and probe-delayed matching of
sounds in the bottle-nosed dolphin. Animal Learning and Behavior, 10, 22-34.
Holland, J. G. (1958). Counting by humans on a fixed-ratio schedule of reinforcement. Journal of

Hutchinson, R. R. & Azrin, N. H. (1961). Conditioning of mental-hospital patients to fixed-ratio
schedules of reinforcement. Journal of the Experimental Analysis of Behavior, 4, 87-95.
Hull, C. L. (1934). The concept of the habit-family hierarchy and maze learning: Part I.
Psychological Review, 41, 33-52.
Jenkins, J. J. & Palermo, D. S. (1964). Mediation processes and the acquisition of linguistic
structure. In U. Bellugi & R. Brown (Eds.), The acquisition of language. Monographs of the
Society for Research in Child Development, 29, (Serial No. 92).
Joyce, J. H. & Chase, P. N. (1990). Effects of response variability on the sensitivity of rule-governed
behavior. Journal of the Experimental Analysis of Behavior, 54, 251-262.
Kaufman, A., Baron, A., & Kopp, R. E. (1966). Some effects of instructions on human operant
behavior. Psychonomic Monograph Supplements, 1, 243-250.
Kendall, S. B. (1983). Tests for mediated transfer in pigeons. Psychological Record, 33, 245-256.
Kennedy, C. H. & Laitinen, R. (1988). Second-order conditional control of symmetric and transitive
stimulus relations: The influence of order effects. Psychological Record, 38, 437-446.
Kohlenberg, B. S., Hayes, S. C. & Hayes, L. J. (1991). The transfer of contextual control over
equivalence classes through equivalence classes: A possible model of social stereotyping.
Lashley, K.S. (1938). Conditional reactions in the rat. Journal of Psychology, 6, 311-324.
Laties, V. G. & Weiss, B. (1963). Effects of concurrent task on fixed-interval responding in humans,
Lazar, R. (1977). Extending sequence-class membership with matching-to-sample. Journal of the

Lazar, R. M. & Kotlarchyk, B. J. (1986). Second-order control of sequence-class equivalences in
children. Behavioural Processes, 13, 205-215.
Leander, J. D., Lippman, L. G., & Meyer, M. M. (1968). Fixed interval performance as related to
subjects' verbalizations of the reinforcement contingency. The Psychological Record, 18,
469-474.
Lipkens, R., Kop, P. F. M., & Matthijs, W. (1988). A test of symmetry and transitivity in the
conditional discrimination performances of pigeons. Journal of the Experimental Analysis
of Behavior, 49, 395-409.
Lippman, L. G., & Meyer, M. M. (1967). Fixed interval performance as related to instructions and
to subjects' verbalization of the contingency. Psychonomic Science, 8, 135-136.
Lowe, C. F. (1979). Determinants of human operant behavior. In M. D. Zeiler & P. Harzem (Eds.),
Advances in analysis of behavior: Vol. 1. Reinforcement and the organization of behavior.
(pp. 159-192). Chichester, England: Wiley.
Lowe, C. F., Beasty, A., & Bentall, R. P. (1983). The role of verbal behavior in human learning:
Infant performance on fixed interval schedules. Journal of the Experimental Analysis of
Behavior, 39, 157-164.
Lowe, C. F., Harzem, P. & Bagshaw, M. (1978). Species differences in temporal control of behavior.
II: Human performance. Journal of the Experimental Analysis of Behavior, 29, 351-361.
Lowe, C. F., Harzem, P., & Hughes, S. (1978). Determinants of operant behavior in humans: Some
differences from animals. Quarterly Journal of Experimental Psychology, 30, 373-386.
Mackay, H. A. & Sidman, M. (1984). Teaching new behaviors via equivalence relations (pp.
493-513). In P. Brooks, R. Sperber, & C. McCauley (Eds.), Learning and cognition in the
mentally retarded. Hillsdale, N.J.: Erlbaum.

Markman, E. M. & Wachtel, G. F. (1988). Children's use of mutual exclusivity to constrain the
meanings of words. Cognitive Psychology, 20, 121-157.
McIlvane, W. J., Kledaras, J. B., Munson, L. C., King, K. A., de Rose, J. C., & Stoddard, L. T. (1987).
Controlling relations in conditional discrimination and matching by exclusion. Journal of
McIntire, K. D., Cleary, J., & Thompson, T. (1987). Conditional relations by monkeys: Reflexivity,
symmetry, and transitivity. Journal of the Experimental Analysis of Behavior, 47, 279-285.
Michael, R. L. & Bernstein, D. J. (1991). Transient effects of acquisition history on generalization in
a matching-to-sample task. Journal of the Experimental Analysis of Behavior, 56, 155-166.
Nissen, H. (1951). Analysis of complex conditional reaction in the chimpanzee. Journal of
Comparative and Physiological Psychology, 7, 449-516.
Osgood, C. E. (1953). Method and theory in experimental psychology. New York: Oxford
University Press.
Parrott, L. J. (1984). Listening and understanding. The Behavior Analyst, 7, 29-39.
Parrott, L. J. (1986). On the differences between verbal and social behavior. In P. N. Chase & L. J.
Parrott (Eds.), Psychological aspects of language: The West Virginia Lectures. Springfield,
IL: Thomas.
Parrott, L. J. (1987a). Rule-governed behavior: An implicit analysis of reference. In S. Modgil & C.
Modgil (Eds.), B. F. Skinner: Consensus and Controversy (pp. 265-276). Sussex, England:
Falmer Press.
Parrott, L. J. (1987b). Reply to Vaughan. In S. Modgil & C. Modgil (Eds.), B. F. Skinner: Consensus
and Controversy (pp. 265-276). Sussex, England: Falmer Press.
Reese, H. W. (1968). The perception of stimulus relations: Discrimination learning and

transposition. New York: Academic Press.
Reese, H. W. (1991). Mentalistic approaches to verbal behavior. In L. J. Hayes & P. N. Chase (Eds.),
Dialogues on verbal behavior. Reno, NV: Context Press.
Saunders, K. J. (1989). Naming in conditional discrimination and stimulus equivalence. Journal of
Schoenfeld, W. N. & Cumming, W. W. (1963). Behavior and perception. In S. Koch (Ed.),
Psychology: A study of a science, (Vol. 5). New York: McGraw-Hill.
Shimoff, E., Catania, A. C., & Matthews, B. A. (1981). Uninstructed human responding: Sensitivity
of low rate performance to schedule contingencies. Journal of the Experimental Analysis of
Behavior, 36, 207-220.
Shimoff, E., Matthews, B. A., & Catania, A. C. (1986). Human operant performance: Sensitivity
and pseudosensitivity to contingencies. Journal of the Experimental Analysis of Behavior,
46, 149-157.
Sidman, M. (1971). Reading and auditory-visual equivalences. Journal of Speech and Hearing
Research, 14, 5-13.
Sidman, M. (1990). Equivalence relations: Where do they come from? In D. E. Blackman & H.
Lejeune (Eds.), Behaviour analysis in theory and practice: Contributions and controversies
(pp. 93-114). Hillsdale, N.J.: Erlbaum.
Sidman, M., Cresson, O., & Willson-Morris, M. (1974). Acquisition of matching-to-sample via
mediated transfer. Journal of the Experimental Analysis of Behavior, 22, 261-273.
Sidman, M., Kirk, B.,& Willson-Morris, M. (1985). Six-member stimulus classes generated by
conditional-discrimination procedures. Journal of the Experimental Analysis of Behavior,
43, 21-42.
Sidman, M., Rauzin, R., Lazar, R., Cunningham, S., Tailby, W., & Carrigan, P. (1982). A search for
symmetry in the conditional discriminations of rhesus monkeys, baboons and children.
Sidman, M. & Tailby, W. (1982). Conditional discrimination versus matching to sample: An
expansion of the testing paradigm. Journal of the Experimental Analysis of Behavior, 37,
5-22.
Skinner, B. F. (1938). Behavior of organisms. New York: Appleton-Century-Crofts.
Skinner, B. F. (1957). Verbal behavior. N.Y.: Appleton-Century-Crofts.
Skinner, B. F. (1966). An operant analysis of problem solving (pp. 225-257). In B. Kleinmuntz (Ed.),
Problem-solving: Research, method, and theory. New York: Wiley.
Skinner, B. F. (1969). Contingencies of reinforcement: A theoretical analysis. New York:
Appleton-Century-Crofts.
Spradlin, J. E., Cotter, V. W. & Baxley, N. (1973). Establishing a conditional discrimination without
direct training: A study of transfer with retarded adolescents. American Journal of Mental
Deficiency, 77, 556-566.
Spradlin, J. E. & Dixon, M. (1976). Establishing a conditional discrimination without direct training:
Stimulus classes and labels. American Journal of Mental Deficiency, 80, 555-561.
Staats, C. K. & Staats, A. W. (1957). Meaning established by classical conditioning. Journal of
Experimental Psychology, 54, 74-80.
Steele, D. L. & Hayes, S. C. (1991). Stimulus equivalence and arbitrarily applicable relational
responding. Journal of the Experimental Analysis of Behavior, 56, 519-555.
VanBiervliet, A. (1977). Establishing words and objects as functionally equivalent through manual
sign training. American Journal of Mental Deficiency, 82, 178-186.

Vaughan, W. (1988). Formation of equivalence sets in pigeons. Journal of Experimental Psychology:
Animal Behavior Processes, 14, 36-42.
Watkins, M. J. (1990). Mediationism and the obfuscation of memory. American Psychologist, 45,
328-335.
Weiner, H. (1964). Conditioning history and human fixed-interval performance. Journal of the
Weiner, H. (1965). Conditioning history and maladaptive human operant behavior. Psychological
Reports, 17, 935-942.
Weiner, H. (1969). Human behavioral persistence. Psychological Record, 20, 445-456.
Weiner, H. (1970). Instructional control of human operant responding during extinction following
fixed ratio conditioning. Journal of the Experimental Analysis of Behavior, 13, 391-394.
Wulfert, E. & Hayes, S. C. (1988). The transfer of conditional sequencing through conditional
equivalence classes. Journal of the Experimental Analysis of Behavior, 50, 125-144.
Yamamoto, J. (1986). Establishing stimulus equivalence in autistic children (in Japanese with
English summary). Japanese Journal of Behavior Analysis, 1, 1-21.

Author Notes
Requests for reprints should be addressed to Steven C. Hayes, Department of Psychology,
University of Nevada, Reno, NV 89557-0062.

Footnotes
1. In a draft chapter Skinner wrote for a book edited by the first author (Hayes, 1989) he described
rule-governed behavior as "verbal stimulus controlled behavior." He later edited it out, but
its appearance shows that he was beginning to deal with how rules could be defined as
"verbal rules" within his system.

Figure Captions
Figure 1. A diagrammatic presentation of the trained and derived relations in a typical stimulus
equivalence network. The solid arrows indicate trained relations; the dotted arrows indicate
derived relations. For the sake of clarity, only the trained relations among one set of stimuli
are shown, but in the typical study the need for conditionality in the trained relations would
require trained relations among two or more sets of related stimuli.

Hayes, S. C. & Hayes, L. J. (1992) - American Psychologist, 47, 1383-1395

Uploaded by

Copyright:

Available Formats

Hayes, S. C. & Hayes, L. J. (1992) - American Psychologist, 47, 1383-1395

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Hayes, S. C. & Hayes, L. J. (1992) - American Psychologist, 47, 1383-1395

Uploaded by

Copyright:

Available Formats

Hayes, S. C. & Hayes, L. J. (1992). American Psychologist, 47, 1383-1395.

Verbal Relations and the Evolution of Behavior Analysis

Running Head: VERBAL RELATIONS AND BEHAVIOR ANALYSIS

relations, behaviorism, Skinner

Verbal Relations, Cognition, and the Evolution of Behavior Analysis

scientific area not busy being born is busy dying.

shows every sign of an area that is busy being born.

Early Behavior Analytic Research with Humans

and the Problem of Instructional Control

of basic processes common to human behavior.

nothing in evolutionary or behavioral theory that ensures continuity of this sort.

thus cannot be assumed.

Meyer, 1967; Weiner, 1964, 1965, 1969).

examined the effects of instructions on performance on a fixed-interval (FI) 20 seconds schedule.

the FI performance characteristic of nonhumans.

Some behavioral researchers seemed surprised at the powerful effects of instructions,

Schoenfeld & Cumming, 1963).

Skinner's Approach to the Listener

This approach to instructions was heavily influenced by Skinner's approach to verbal

(Parrott, 1987a, 1987b). Skinner defined rule-governed behavior as behavior controlled by

contingency-specifying stimuli (1966, 1969). He never defined what he meant by "specify."

antecedents, such as modeling stimuli (e.g., see Skinner, 1969, p. 163).

accounts, including those provided by Skinner, categories of stimuli (discriminative, eliciting,

responding to an electric light was engaging in "Edison-governed behavior." Without a functional

Verbal Control and Contemporary Behavior Analysis

process accounts of the increasing number of identified differences. A basic analysis of

instructional effects was not pursued.

clear evidence for cognitive theory.

(Bentall & Lowe, 1987).

uninstructed subjects to changes in environmental contingencies. In general, instructions alter the

The changes in responding to programmed contingencies produced by verbal instructions

instructions establish additional socially-mediated contingencies for responding in accord with

lights rapidly alternated contacted the contingencies in both components.

Subjects with alternating instructions immediately showed schedule-appropriate behavior.

components. Implicit social contingencies maintained instructional compliance while the

instructions either increased or decreased schedule-appropriate behavior, depending upon the

specific instructions and specific schedules.

differences produced by the impact of verbal rules.

operant or classical conditioning uncontaminated by verbal involvement. Rather than abandon

Contemporary Behavior Analytic Research on Derived Stimulus Relations

analysts are identifying functional processes that seem to provide one.

most researched is that of stimulus equivalence.

trained relations have been established.

by S-R associationism and mediationalism.

In contemporary treatments, an equivalence class is usually said to exist if the trained

transitivity, and combinations of these (Sidman & Tailby, 1982). In matching-to-sample

equivalence are shown in Figure 1.

Stimulus Equivalence as a Working Model of Semantic Relations

A major source of interest in stimulus equivalence is the apparent correspondence between

to do so. Natural examples can readily be given for transitivity as well.

If stimulus equivalence provides a preliminary behavior analytic model of verbal

equivalence. To date, no unequivocal demonstration of stimulus equivalence in nonhumans has

equivalence readily. In humans, performance on equivalence tasks seems to be correlated with

verbal performance, as would be expected if equivalence is a type of verbal relation (Barnes,

Conditionality is also necessary for equivalence to serve as a preliminary model of verbal

function as contexts in which equivalence class membership is reconstituted (Kohlenberg, Hayes, S.

The Challenge of Stimulus Equivalence

To understand why stimulus equivalence is having a major impact on behavioral