ArticlePDF Available

Derived relational responding as generalized operant behavior

October 2000
Journal of the Experimental Analysis of Behavior 74(2):207-27

DOI:10.1901/jeab.2000.74-207

Source
PubMed

Authors:

Olive Healy

Trinity College Dublin

Dermot Barnes-Holmes

University of Ulster

The major aim of the present study was to demonstrate that derived relational responding may be viewed as a form of generalized operant behavior. In Experiment 1, 4 subjects were divided into two conditions (2 in each condition). Using a two-comparison matching-to-sample procedure, all subjects were trained and tested for the formation of two combinatorially entailed relations. Subjects were trained and tested across multiple stimulus sets. Each set was composed of novel stimuli. Both Conditions 1 and 2 involved explicit performance-contingent feedback presented at the end of each block of test trials (i.e., delayed feedback). In Condition 1, feedback was accurate (consistent with the experimenter-designated relations) following exposure to the initial stimulus sets. When subjects' responding reached a predefined mastery criterion, the feedback then switched to inaccurate (not consistent with the experimenter-designated relations) until responding once again reached a predefined criterion. Condition 2 was similar to Condition 1, except that exposure to the initial stimulus sets was followed by inaccurate feedback and once the criterion was reached feedback switched to accurate. Once relational responding emerged and stabilized, response patterns on novel stimulus sets were controlled by the feedback delivered for previous stimulus sets. Experiment 2 replicated Experiment 1, except that during Conditions 3 and 4 four comparison stimuli were employed during training and testing. Experiment 3 was similar to Condition 1 of Experiment 1, except that after the mastery criterion was reached for class-consistent responding, feedback alternated from accurate to inaccurate across each successive stimulus set. Experiment 4 involved two types of feedback, one type following tests for mutual entailment and the other type following tests for combinatorial entailment. Results from this experiment demonstrated that mutual and combinatorial entailment may be controlled independently by accurate and inaccurate feedback. Overall, the data support the suggestion, made by relational frame theory, that derived relational responding is a form of generalized operant behavior.

Number of training trials across successive stimulus sets for each subject in Experiment 1.

…

Number of training trials across successive stimulus sets for each subject in Experiment 2.

…

Number of training trials across successive stimulus sets for each subject in Experiment 3.

…

Number of training trials across successive stimulus sets for each subject in Experiment 4.

…

Figures - uploaded by Olive Healy

Content may be subject to copyright.

Content uploaded by Olive Healy

Content may be subject to copyright.

207

JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR

2000, 74, 207–227

NUMBER

SEPTEMBER

)

DERIVED RELATIONAL RESPONDING AS

GENERALIZED OPERANT BEHAVIOR

LIVE

EALY

ERMOT

ARNES

-H

OLMES

AND

AUL

M. S

MEETS

UNIVERSITY COLLEGE CORK, IRELAND/CABAS

NATIONAL UNIVERSITY OF IRELAND, MAYNOOTH, AND

LEIDEN UNIVERSITY, THE NETHERLANDS

The major aim of the present study was to demonstrate that derived relational responding may be

viewed as a form of generalized operant behavior. In Experiment 1, 4 subjects were divided into two

conditions (2 in each condition). Using a two-comparison matching-to-sample procedure, all subjects

were trained and tested for the formation of two combinatorially entailed relations. Subjects were

trained and tested across multiple stimulus sets. Each set was composed of novel stimuli. Both Con-

ditions 1 and 2 involved explicit performance-contingent feedback presented at the end of each

block of test trials (i.e., delayed feedback). In Condition 1, feedback was accurate (consistent with

the experimenter-designated relations) following exposure to the initial stimulus sets. When subjects’

responding reached a predeﬁned mastery criterion, the feedback then switched to inaccurate (not

consistent with the experimenter-designated relations) until responding once again reached a pre-

deﬁned criterion. Condition 2 was similar to Condition 1, except that exposure to the initial stimulus

sets was followed by inaccurate feedback and once the criterion was reached feedback switched to

accurate. Once relational responding emerged and stabilized, response patterns on novel stimulus

sets were controlled by the feedback delivered for previous stimulus sets. Experiment 2 replicated

Experiment 1, except that during Conditions 3 and 4 four comparison stimuli were employed during

training and testing. Experiment 3 was similar to Condition 1 of Experiment 1, except that after the

mastery criterion was reached for class-consistent responding, feedback alternated from accurate to

inaccurate across each successive stimulus set. Experiment 4 involved two types of feedback, one type

following tests for mutual entailment and the other type following tests for combinatorial entailment.

Results from this experiment demonstrated that mutual and combinatorial entailment may be con-

trolled independently by accurate and inaccurate feedback. Overall, the data support the suggestion,

made by relational frame theory, that derived relational responding is a form of generalized operant

behavior.

Key words: generalized operant class, relational frame theory, mutual entailment, combinatorial

entailment, equivalence relation, matching to sample, humans

Derived relational responding has recently

stimulated much interest and research activ-

ity among behavior analysts. One of the main

reasons for focusing on such responding is

the fact that it cannot be readily accounted

for by the concept of conditional discrimi-

nation. In the typical equivalence experi-

ment, for example, if subjects are trained to

match Comparison Stimulus B1 to Sample

Stimulus A1, and Comparison Stimulus C1 to

Sample Stimulus B1, they will likely match

Comparison Stimulus A1 to Sample Stimulus

This research was conducted as part of Olive Healy’s

doctoral research program under the supervision of Der-

mot Barnes-Holmes. Portions of these data were present-

ed at the annual conference of the Experimental Analysis

of Behavior Group, London, April 1998, and at the an-

nual convention of the Association for Behavior Analysis,

Orlando, Florida, May 1998.

Requests for reprints may be sent to Dermot Barnes-

Holmes at the Department of Psychology, National Uni-

versity of Ireland, Maynooth, Maynooth, County Kildare,

Ireland (E-mail: Dermot.Barnes-Holmes@may.ie).

C1, demonstrating derived equivalence re-

sponding, without additional training. A con-

ditional discrimination, as normally deﬁned,

does not predict the emergence of this un-

trained performance. Neither A1 nor C1 has

a history of differential reinforcement as a

conditional discriminative stimulus with re-

gard to the other, and thus, neither stimulus

would be expected to reliably control selec-

tion of the other.

Relational frame theory (RFT) offers one

account of derived relational responding.

The important feature of RFT for the current

research is the fact that a relational frame, as

an analytic unit, is conceptualized as a three-

term contingency. For RFT, the contextual

cue is the third term, the relational response

(e.g., responding to Stimulus B in terms of A

and responding to A in terms of B) is the

second term, and a history of differential re-

inforcement correlated with the contextual

cue is the ﬁrst term in the contingency. From

208 OLIVE HEALY et al.

this perspective, therefore, responding to B

given A and to A given B may be considered

as a single response unit controlled by a rel-

evant contextual cue by virtue of its previous

correlation with differential reinforcement.

In effect, the RFT approach invokes a purely

functional concept of an operant, and the

term generalized operant class (e.g., Barnes,

1994, 1996; D. Barnes-Holmes & Barnes-

Holmes, 2000; Hayes, 1992; Hayes & Barnes,

1997) is used to emphasize this fact.

The RFT view of equivalence responding as

a form of generalized operant behavior may

be difﬁcult to appreciate if one typically con-

ceptualizes operants in structuralist terms

(see D. Barnes-Holmes & Barnes-Holmes,

2000, for a detailed discussion of this issue).

The concept of a response class with an inﬁ-

nite range of topographies is a deﬁning prop-

erty of operant behavior. Nonetheless, topo-

graphical and functional classes of behavior–

environment interactions quite often overlap,

and thus the two may become confused. Le-

ver pressing, for instance, may be deﬁned by

the effect of activity upon the lever, but al-

most all lever presses involve ‘‘pressing’’

movements. A sensitive lever may be activated

by coughing, but for most purposes such in-

stances can normally be ignored. Sometimes,

however, the independence between topo-

graphical and functional classes is made very

clear. The concept of generalized imitation

(e.g., Baer, Peterson, & Sherman, 1967; Ge-

wirtz & Stengle, 1968; Poulson, Kymissis,

Reeve, Andreatos, & Reeve, 1991) provides

one excellent example. After a generalized

imitative repertoire is established, an almost

inﬁnite variety of response topographies may

be substituted for the forms used in the ear-

lier training. The behavior of imitating is gen-

eralized because it is not limited to any par-

ticular response topography. In a similar vein,

some behavior analysts have argued that it is

possible to reinforce ‘‘generalized attending’’

(McIlvane, Dube, & Callahan, 1995; Mc-

Ilvane, Dube, Kledaras, Iennaco, & Stoddard,

1990), although what is being attended to will

change. Broadly similar arguments have also

been made with respect to many other phe-

nomena, such as generalized identity match-

ing and mismatching (e.g., Cumming & Ber-

ryman, 1965; Dube, McIlvane, & Green, 1992;

Saunders & Sherman, 1986), exclusion (e.g.,

Lipkens, Hayes, & Hayes, 1993; McIlvane et

al., 1987), arbitrary assignment (e.g., Saun-

ders, Saunders, Kirby, & Spradlin, 1988), and

one-trial learning (e.g., Catania, 1996; Dube

et al., 1992).

Although these and yet other examples

(see Neuringer, 1986; Pryor, Haag, &

O’Reilly, 1969; Stokes & Baer, 1977) consti-

tute a simple extension of the three-term con-

tingency as an analytic unit, speciﬁc qualiﬁers

are often included when operant classes are

not readily deﬁned topographically. Such

classes are referred to as generalized, higher or-

der, or overarching. These qualiﬁers are not

used in this instance as technical terms, and

they do not imply the existence of mediation-

al processes leading to the formation of op-

erants of this type. Instead, these qualiﬁers

emphasize that a speciﬁc functional class can-

not be deﬁned by its response forms or stim-

ulus forms, a fact that is true in principle for

all functional classes (D. Barnes-Holmes &

Barnes-Holmes, 2000). At the present time,

very little is known about the determinants of

operant class formation in those cases in

which there is minimal overlap between func-

tion and topography (see Pilgrim & Galizio,

in press). As a possible starting point for ad-

dressing this important issue, RFT argues that

derived relational responding, including

stimulus equivalence, can be explicitly inter-

preted as a type of discriminated operant be-

havior that is more usefully deﬁned by func-

tional rather than topographical properties

(D. Barnes-Holmes & Barnes-Holmes, 2000).

Consistent with earlier literature on the

topic, Hayes (1994) suggested that there are

four particularly important properties of dis-

criminated operant behavior: (a) operants

develop, (b) operants are ﬂexible and can be

shaped, (c) operants can come under stimu-

lus control, and (d) operants are controlled

by their consequences. Obviously, if deriving

stimulus relations is to be viewed as operant

behavior, all four of these properties should

apply. Supportive research has been provided

on all four points (e.g., Barnes, Browne,

Smeets, & Roche, 1995; Barnes & Hampson,

1993, 1997; Barnes, Hegarty, & Smeets, 1997;

Dymond & Barnes, 1995, 1996; Lipkens et al.,

1993; Roche & Barnes, 1996; 1997; Roche,

Barnes, & Smeets, 1997; Roche, Barnes-

Holmes, Smeets, Barnes-Holmes, & McGeady,

2000; Steele & Hayes, 1991; Wilson & Hayes,

209GENERALIZED OPERANT BEHAVIOR

1996). The least support, however, has been

obtained for the ﬁnal point.

In line with this operant approach to rela-

tional responding, we recently reported a se-

ries of experiments that examined the effects

of differential consequences on derived rela-

tional responding (Healy, Barnes, & Smeets,

1998). Speciﬁcally, adult humans were

trained on four matching-to-sample (MTS)

trial types that established the following con-

ditional discriminations: A1-B1, A2-B2, B1-C1,

B2-C2 (the stimuli were three-letter nonsense

syllables but are labeled, for the purposes of

communication, with alphanumerics). Fol-

lowing training, the subjects were exposed to

a derived relations test that probed for tran-

sitivity (A1-C1 and A2-C2) and combined sym-

metry and transitivity (C1-A1 and C2-A2); the

generic term, combinatorial entailment, will be

used here to label these relational responses

(see Hayes, 1991). Using the same set of non-

sense syllables, each subject completed a total

of 11 cycles of training and testing. However,

unlike a traditional emergent relations exper-

iment, after each exposure to the test, a spe-

ciﬁc form of ‘‘test-performance feedback’’

was presented to the subject.

In Experiment 5 of Healy et al. (1998), for

example, following each of the 11 cycles of

training and testing, subjects were presented

with a point system that indicated the exact

number of class-consistent responses made

during the test phase (e.g., after training A1-

B1 and B1-C1, a class-consistent response

would be C1-A1). For some subjects the feed-

back was accurate following the ﬁrst ﬁve stim-

ulus sets and was inaccurate following the

next six (and vice versa for the remaining

subjects). During accurate feedback, if a sub-

ject emitted ﬁve class-consistent responses,

for example, ﬁve star shapes were presented

in a vertical column on the screen; if the feed-

back was inaccurate then 15 stars were pre-

sented (i.e., 20 stars minus 5 for the class-

consistent responses). Eight of the 10 subjects

responded in accordance with the accurate

and inaccurate feedback. Taken together, the

ﬁve experiments reported by Healy et al.

demonstrated that derived relational respond-

ing is often sensitive to differential conse-

quences provided at the end of consecutive

test phases. The authors argued that the data

supported the operant interpretation of such

responding.

However, the authors also pointed out that

the use of the same stimulus set across the

entire 11 sessions for each subject limits the

implications the results have for the interpre-

tation of relational responding as a general-

ized or overarching operant class. Consistent

with RFT, the term generalized is used here to

denote an operant that cannot be deﬁned to-

pographically. Because only one stimulus set

was used, the consistent changes in respond-

ing that occurred across stimulus sets could

have been due to the direct effects of the

feedback on responding to the same stimulus

set. The effects of the feedback do not, there-

fore, require an interpretation in terms of

emergence or in terms of generalized oper-

ant behavior. To constitute generalized op-

erant behavior, there should be no history of

differential consequences contingent on re-

sponding to a particular stimulus set. For ex-

ample, if each cycle of training and testing

involved a novel set of nonsense syllables, and

accurate performance-contingent feedback

on tests facilitated the emergence of relation-

al responding across sets, this would indicate

that generalized derived relational respond-

ing, as a response unit, was being shaped by

the differential feedback. This approach is

consistent with the following quotation from

Estes (1971):

In more mature human beings, much instru-

mental behavior and, more especially, a great

part of verbal behavior is organized into high-

er-order routines and is, in many instances,

better understood in terms of the operation

of rules, principles, strategies and the like

than in terms of successions of responses to

particular stimuli. . . . In these situations it is

the selection of strategies rather than the se-

lection of particular reactions to stimuli which

is modiﬁed by past experience with rewarding

or punishing consequences. (p. 23)

Although Estes used loosely deﬁned terms,

such as strategies, RFT adopts a similar posi-

tion but includes the term relational operant,

so that the ﬁnal sentence of the above quo-

tation would read, ‘‘In these situations it is

the selection of relational operants rather

than the selection of particular reactions to

stimuli which are modiﬁed by past experi-

ence with rewarding and punishing conse-

quences.’’

Another important issue that arose from

the Healy et al. (1998) study was the fact that

210 OLIVE HEALY et al.

some subjects showed relatively stable perfor-

mances across multiple exposures, whereas

others did not. One procedural reason for

this variability across subjects may be that no

mastery criterion was used before terminat-

ing the experiment or changing from one

type of feedback to another (i.e., all subjects

completed 11 stimulus sets irrespective of

performance). To demonstrate clearly the ef-

fects of differential feedback, a mastery cri-

terion should be employed that requires a

predetermined level of consistent responding

in accordance with the feedback before there

is any change in the type of feedback deliv-

ered, or before the experiment is terminated.

In Experiment 1 of the present series, all

subjects were ﬁrst exposed to conditional dis-

crimination training on four MTS trial types

(A1-B1, B1-C1, A2-B2, B2-C2). The subjects

then were tested for the formation of four

combinatorially entailed derived relations

(A1-C1, C1-A1, A2-C2, C2-A2). Following ex-

posure to this ﬁrst cycle of training and test-

ing, accurate or inaccurate feedback (the

conditions were counterbalanced) was deliv-

ered for the test performances. The next cy-

cle of training and testing then began, but

with a novel set of stimuli. This cycle of train-

ing and testing, using novel sets of stimuli for

each cycle, continued until a subject respond-

ed in accordance with the feedback across

three consecutive stimulus sets. Once this sta-

bility criterion was reached, the type of feed-

back switched (accurate to inaccurate or vice

versa) and the training and testing contin-

ued, using novel stimulus sets, until the test

performance again reached the stability cri-

terion. In the interests of clarity, the other

experiments will be introduced subsequently.

GENERAL METHOD

Subjects

Participants in the study were 13 students

pursuing both undergraduate and postgrad-

uate studies at University College Cork. Eight

were female, and 5 were male. Their ages

ranged from 18 to 25 years. Each subject was

randomly assigned to one of the experimen-

tal conditions. Subjects 1 through 4 partici-

pated in Experiment 1, Subjects 5 through 8

in Experiment 2, Subjects 9 through 11 in

Experiment 3, and Subjects 12 and 13 in Ex-

periment 4. None of the subjects had any

knowledge of stimulus equivalence or related

phenomena. All but 1 subject (Subject 1)

completed an entire experiment in one sit-

ting.

Apparatus

The experiment was conducted in a small

experimental room (4 square m) in which

the subject was seated at a table in front of

an Apple Macintosh LCIII

personal com-

puter. The computer was programmed in

BBC BASIC and it controlled stimulus pre-

sentations and recorded subjects’ responses.

During all conditions (except Conditions 3

and 4), the Z and M keys were designated as

response keys and were marked with yellow

paper squares. During Conditions 3 and 4,

the G, R, U, C, and N keys were designated

as response keys. The pool of stimuli consist-

ed of 168 nonsense syllables. Each of the stim-

ulus sets used for a particular subject was con-

structed by randomly selecting nonsense

syllables, without replacement, from this

pool, and then assigning them randomly to

their roles as sample and comparison stimuli.

Although the same pool of nonsense syllables

was used for each subject, the random selec-

tion and random assignment of these sylla-

bles allowed us to construct a completely nov-

el stimulus set for every cycle of training and

testing across all of the subjects in the study

(i.e., a particular set was only used with 1 sub-

ject).

General Procedure

All subjects were exposed to individual cy-

cles of training and testing, using novel stim-

ulus sets for every cycle. Each cycle of train-

ing and testing was as follows. All subjects

were trained in four MTS trial types using a

randomly generated set of six nonsense syl-

lables (each set is designated using the alpha-

numerics A1, A2, B1, B2, C1, C2; subjects nev-

er saw these labels). Subjects were seated in

the experimental room and the following in-

structions were presented on the computer

screen:

During this stage of the experiment you must

look at the nonsense syllable at the top and

then choose one of the two nonsense syllables

at the bottom by pressing one of the marked

keys on the keyboard. To choose the left non-

sense syllable press the marked key on the left.

211GENERALIZED OPERANT BEHAVIOR

To choose the right nonsense syllable press

the marked key on the right. Press the space-

bar twice to continue.

On each MTS trial, the sample and the two

comparison stimuli always differed in at least

two letters. The sample appeared centered in

the top half of the monitor screen, followed

1.5 s later by the comparison stimuli, which

were positioned to the left and right of the

sample, 5.00 cm from the bottom of the

screen. On each MTS trial, the position of the

comparison stimuli was varied randomly (i.e.,

the correct nonsense syllable could appear

on either the left or the right with equal

probability). Subjects were trained on the two

A-B and two B-C MTS trial types. The four

trial types were presented in a quasirandom

order, each trial type occurring twice every

eight trials until the subject produced eight

consecutive correct responses. For each stim-

ulus set, when A1 was the sample, B1 was cor-

rect; when B1 was the sample, C1 was correct;

when A2 was the sample, C2 was correct;

when B2 was the sample, C2 was correct.

The correct completion of an MTS trial re-

moved the stimulus display and produced

‘‘correct’’ in the center of the screen accom-

panied by a high-pitched beep for 1.5 s. The

incorrect completion of an MTS trial re-

moved the stimulus display and produced

‘‘wrong’’ in the center of the screen but with-

out auditory feedback, again for 1.5 s. A 1-s

intertrial interval followed both types of feed-

back, during which the computer screen re-

mained blank. When eight consecutive cor-

rect responses had been emitted, the

computer program progressed without inter-

ruption to an MTS test for combinatorial en-

tailment. The test consisted of four MTS trial

types presented in a quasirandom order, with

each of the four trial types occurring ﬁve

times across 20 trials (i.e., no intermixed

baseline trials). No feedback was given during

testing.

In summary, each training and testing cycle

employed a novel set of stimuli and consisted

of training to a mastery criterion (eight con-

secutive correct) followed by exposure to a

single block of 20 test trials. In effect, subjects

were required to learn a novel set of interre-

lated conditional discriminations, and were

tested for novel examples of combinatorial en-

tailment, across every cycle of training and

testing.

Posttest Feedback

Feedback was delivered after each test for

combinatorial entailment. This feedback pro-

vided differential consequences contingent

upon the exact number of class-consistent re-

sponses emitted during the previous derived

relations test. Once a test phase had been

completed, the words ‘‘Minimum Points

Earned’’ and ‘‘Maximum Points Earned’’ ap-

peared at the very bottom and top of the

screen, respectively. A computer-generated vi-

sual sequence presented a star shape for ev-

ery relational-consistent response (accurate

feedback) or nonrelational-consistent re-

sponse (inaccurate feedback) emitted during

the test. Each star appeared on the screen

accompanied by a high-pitched beep. The

ﬁrst star appeared directly above the words

‘‘Minimum Points Earned’’ and each subse-

quent star appeared above the other in a ver-

tical line. A broken line appeared midway on

the screen as a half-way mark. Once the stars

earned during a test exposure had been pre-

sented, they remained on the screen for 10 s

before the screen cleared, and the following

instructions appeared on the monitor screen:

‘‘That is the end of this part of the experi-

ment. Please report to the experimenter.’’

When the subject reported to the experi-

menter, he or she was asked to ‘‘Please wait

here [i.e., outside the experimental room],

until I ask you to continue with the experi-

ment again.’’ When the experimenter had ex-

amined and saved the data to a diskette and

reset the computer for the next cycle of train-

ing and testing (with a novel stimulus set),

she returned to the subject and asked him or

her to ‘‘Please go back to the computer and

continue with the experiment.’’ No other in-

structions were provided, and no other inter-

actions occurred between the experimenter

and subject during the course of the experi-

ment.

EXPERIMENT 1

Procedure

Condition 1 (Subjects 1 and 2). In Condition

1, accurate feedback was delivered following

each cycle of training and testing until re-

sponding reached a predeﬁned mastery cri-

212 OLIVE HEALY et al.

terion. This criterion determined when a

switch in feedback would occur. The criterion

required that a subject produce a pattern of

responding consistent with the feedback

across three consecutive test exposures. Thus,

for example, during accurate feedback a sub-

ject was required to emit the conditional re-

sponses A1-C1, A2-C2, C1-A1, and C2-A2 at

least four times across each of the ﬁve expo-

sures to each trial type, across three consec-

utive stimulus sets. Once a subject had

achieved this mastery criterion, the experi-

menter altered the computer program and

the feedback switched to inaccurate. During

inaccurate feedback the subject was required

to emit the conditional responses A1-C2, A2-

C1, C1-A2, and C2-A1 at least four times

across each of the ﬁve exposures to each trial

type, across three consecutive stimulus sets,

before the experiment was terminated. If a

response pattern fell outside either of the

above categories, it was labeled as interme-

diate responding.

Condition 2 (Subjects 3 and 4). Condition 2

was identical to Condition 1 except that in-

accurate feedback was delivered following

each of the initial cycles of training and test-

ing, until the criterion was reached. Feedback

then switched to accurate, and exposure to

test conditions continued until performance

once again reached the stability criterion.

Results and Discussion

The number of training trials required to

meet the mastery criterion, per stimulus set,

by each subject is presented in Table 1. The

minimum number of trials required was 10

(Subject 1) and the maximum was 308 (Sub-

ject 3). The number of training trials per ex-

posure varied unsystematically for each sub-

ject. In general, however, subjects’ earlier

exposures to the training involved larger

numbers of trials than many of their later ex-

posures, but there were some notable excep-

tions (e.g., Subjects 1 and 2, Exposures 14

and 6, respectively).

Performances on each of the tests for com-

binatorial entailment are shown in Figure 1.

Subject 1 required a total of 15 exposures be-

fore achieving the stability criterion during

accurate feedback. This was followed by a rap-

id shift from high levels of combinatorial en-

tailment to three consecutive performances

in accordance with low levels with the intro-

duction of inaccurate feedback. For ease of

communication, the descriptor low level of

combinatorial entailment is used to label the fol-

lowing relational responses: A1-C2, A2-C1,

C1-A2, C2-A1. We recognize, however, that

this pattern may itself constitute combinato-

rial entailment rather than the absence of en-

tailed relations. For instance, the inaccurate

feedback might establish the relational frame

of opposition (e.g., A1 opposite to B1 and B1

opposite to C1 derives A1 not opposite to C1)

(e.g., Steele & Hayes, 1991). Alternatively, the

feedback might establish a frame of coordi-

nation but with negative stimulus control

emerging during the test phases (see Exper-

iment 2). In either case, the A and C stimuli

participate in a type of combinatorially en-

tailed relation. In the current article, there-

fore, the distinction between high and low

levels of entailment is used simply to reﬂect

the position of the data points on the graphs.

Subject 2 reached the criterion on Exposure

5, and switched responding promptly during

inaccurate feedback. In Condition 2, Subject

3 produced intermediate responding fol-

lowed by a variable pattern during inaccurate

feedback before achieving the criterion, and

when the feedback switched so did the re-

sponding to high levels of combinatorial en-

tailment. Subject 4 reached the criterion rap-

idly during both types of feedback, requiring

a total of eight exposures.

These data suggest that the feedback for

test performances, combined with the use of

a stability criterion, produced consistent pat-

terns of combinatorial entailment across all 4

subjects. Furthermore, these consistent pat-

terns were established across novel stimulus

sets. These data thus provide support for the

argument that derived relational responding

may be conceptualized as generalized oper-

ant behavior.

EXPERIMENT 2:

INCREASING THE

NUMBER OF CLASSES

Although the data from Experiment 1

showed that feedback affected derived rela-

tional responding across stimulus sets, an im-

portant interpretive problem arises. The use

of completely reversed feedback contingen-

cies, combined with only two comparisons on

each trial, may have facilitated control by

213GENERALIZED OPERANT BEHAVIOR

Table 1

Number of training trials across successive stimulus sets

for each subject in Experiment 1.

Stimulus

set

(exposure)

Training trials

S1 S2 S3 S4

148

104

116

112

132

176

158

308

272

196

114

232

143

124

128

194

167

102

148

186

104

114

what are called Type R and Type S relations

(see Carrigan & Sidman, 1992). Accordingly,

subjects may either select a particular com-

parison when presented with a sample (a

Type S or select relation) or they may reject

a particular comparison stimulus (a Type R

or reject relation). Both types of relation (se-

lect and reject) may generate consistent pat-

terns of combinatorial entailment. One way

to reduce the likelihood of Type R control in

MTS is to employ three or more comparisons,

because a greater number of reject, relative

to select, relations are then required (see Sid-

man, 1987). All of our research on the effects

of feedback on derived relational responding

has employed only two comparisons, and thus

by implication Type R relations may have

played an important role in generating the

observed effects. The possibility remains,

therefore, that similar results would not be

obtained if the number of comparisons were

increased. Experiment 2 was conducted to

address this possibility.

Procedure: Conditions 3 and 4

The procedures for Conditions 3 and 4 in

this experiment were similar to Conditions 1

and 2 in Experiment 1, respectively, except

that subjects were trained in eight rather than

four MTS trial types using sets of 12 rather

than six randomly selected nonsense syllables

(each set is designated using the alphanu-

merics A1, A2, A3, A4, B1, B2, B3, B4, C1,

C2, C3, C4; subjects never saw these labels).

Subjects 5 and 6 participated in Condition 3

(accurate-inaccurate feedback), and Subjects

7 and 8 participated in Condition 4 (inaccu-

rate-accurate feedback). On each MTS trial,

the sample appeared centered on the moni-

tor screen. Once the subject pressed the U

key on the keyboard (marked with colored

paper), the four comparison stimuli each ap-

peared on the four corners of the monitor

screen. On each MTS trial, the position of the

comparison stimuli was varied randomly (i.e.,

the correct nonsense syllable could appear

on any of the four corners with equal prob-

ability). In these two conditions, a one-to-

many training and testing design was em-

ployed (i.e., train A-B, A-C, and test B-C, C-B).

Furthermore, because of the increased num-

ber of MTS trial types, training was divided

into two blocks.

Subjects were trained on the four A-B MTS

trial types. The four trial types were presented

in a quasirandom order, each trial type oc-

curring twice across each block of eight trials,

until the subject produced eight correct con-

secutive responses. Subjects were then

trained on the four A-C MTS trial types.

These were also presented in a quasirandom

order, until the subject produced eight con-

secutive correct responses. The four A-B and

four A-C trial types were then presented qua-

sirandomly in blocks of eight trials (each trial

type occurring once in each block), until the

subject produced eight consecutive correct

responses. The A-B trial types always present-

ed the four B stimuli as comparisons, and the

A-C trial types always presented the four C

stimuli as comparisons. For each stimulus set,

when A1 was the sample, B1 was correct;

when A2 was the sample, B2 was correct;

when A3 was the sample, B3 was correct;

when A4 was the sample, B4 was correct;

when A1 was the sample, C1 was correct;

when A2 was the sample, C2 was correct;

when A3 was the sample, C3 was correct;

when A4 was the sample, C4 was correct.

When subjects had successfully completed

the mixed A-B and A-C training, the comput-

er program progressed without interruption

to an MTS test. The test consisted of eight

MTS trial types (B1-C1, B2-C2, B3-C3, B4-C4,

214 OLIVE HEALY et al.

Fig. 1. Class-consistent responding across successive test exposures for each subject in Experiment 1. Each data

point represents one exposure to a 20-trial test block involving a novel stimulus set. Subjects 1 and 2 were exposed

to accurate test-performance feedback followed by inaccurate feedback. Subjects 3 and 4 were exposed to inaccurate

feedback followed by accurate feedback.

215GENERALIZED OPERANT BEHAVIOR

Table 2

Number of training trials across successive stimulus sets

for each subject in Experiment 2.

Stimulus

set

(exposure)

Training trials

S5 S6 S7 S8

231

135

148

121

286

246

106

140

148

214

186

128

321

124

121

216

236

111

168

112

118

121

542

146

102

163

138

114

C1-B1, C2-B2, C3-B3, C4-B4) presented in a

quasirandom order, with each of the eight tri-

al types occurring twice across 16 trials. Sim-

ilar to the previous experiment, the mastery

criterion required that subjects respond in ac-

cordance with the feedback across three con-

secutive stimulus sets. However, because there

were only two exposures to each of the eight

trial types, the mastery criterion required that

all 16 responses be in accordance with the

feedback. During accurate feedback, one star

for each of the following relational responses

was delivered: B1-C1, B2-C2, B3-C3, B4-C4,

C1-B1, C2-B2, C3-B3, C4-B4. When the feed-

back was inaccurate, a star was delivered for

any three of the class-inconsistent responses

(i.e., B1-C2/C3/C4, B2-C1/C3/C4, B3-C1/

C2/C4, B4-C1/C2/C3, C1-B2/B3/B4, C2-

B1/B3/B4, C3-B1/B2/B4, C4-B1/B2/B3).

At the beginning of each exposure to a cy-

cle of training and testing (with a novel stim-

ulus set), each subject was seated in the ex-

perimental room and the following instructions

were presented on the computer screen (i.e.,

the same procedure as Experiment 1, but

with modiﬁed instructions to accommodate

the four-comparison MTS trial types):

You must look at the nonsense syllable in the

center of the screen, press the marked center

key, and then choose one of the 4 nonsense

syllables that appear at each corner of the

screen by pressing one of the four marked

keys on the keyboard. To choose the top-left

syllable press the marked key on the top-left.

To choose the top-right syllable press the

marked key on the top-right. To choose the

bottom-left syllable press the marked key on

the bottom-left. To choose the bottom-right

syllable press the marked key on the bottom-

right. Sometimes the computer will give you

feedback and sometimes it will not. Press the

space-bar twice to continue.

Results and Discussion

The number of training trials required by

each subject to meet the mastery criterion

during each stimulus set is presented in Table

2. The minimum number of trials required

was 42 (Subjects 5 and 6) and the maximum

was 542 (Subject 8). As in Experiment 1, the

number of training trials per exposure tend-

ed to vary unsystematically for each subject,

but again, in general, early training expo-

sures tended to involve larger numbers of tri-

als than many of the later exposures (see Sub-

ject 8, Exposure 7, for a notable exception).

Performances on each of the tests for com-

binatorial entailment are shown in Figure 2.

In Condition 3, Subject 5 reached the stability

criterion on Exposure 5 and maintained high

levels of combinatorial entailment following

a switch in feedback during Exposure 6. In-

termediate responding was shown during Ex-

posures 7 and 8, before low levels of combi-

natorial entailment on Exposures 9 to 11.

Subject 6 reached the stability criterion on

Exposure 7, and maintained a high perfor-

mance on the test for combinatorial entail-

ment across Exposures 8 and 9, before rap-

idly shifting to low levels during Exposures

10, 11, and 12. In Condition 4, Subject 7

reached the criterion on Exposure 7 and pro-

duced low levels of combinatorial entailment

during Exposures 8, 9, and 10, despite the

switch to accurate feedback. On Exposure 11,

this subject produced intermediate respond-

ing, and then on Exposure 12 responding

shifted to high levels of entailment in accor-

dance with accurate feedback. Subject 8

reached the mastery criterion during Expo-

sures 3, 4, and 5. After accurate feedback was

introduced, this subject switched to high lev-

els of entailment during Exposures 8 to 10.

These data show that the effects of posttest

feedback are not restricted to two-choice

MTS trial types. Although these data extend

the basic feedback effect beyond the two-

choice MTS format, they do not exclude Type

216 OLIVE HEALY et al.

Fig. 2. Class-consistent responding across successive test exposures for subjects in Experiment 2. Each data point

represents one exposure to a 16-trial test block involving a novel stimulus set.

217GENERALIZED OPERANT BEHAVIOR

Table 3

Number of training trials across successive stimulus sets

for each subject in Experiment 3.

Stimulus

set

(exposure)

Training trials

S9 S10 S11

127

141

135

101

142

116

178

162

112

108

104

123

127

111

109

101

122

113

106

R relations from involvement in test perfor-

mance. The use of four comparisons may

have reduced the likelihood of Type R con-

trol during the conditional discrimination

training, but the introduction of inaccurate

feedback during the tests may have facilitated

Type R control. In fact, visual inspection of

the raw data (not presented) showed that all

4 subjects tended to distribute their responses

across the three class-inconsistent compari-

sons when they were responding in accor-

dance with the inaccurate feedback. Such a

pattern of responding suggests that subjects

were not choosing a particular comparison,

but were simply responding away from the

‘‘correct’’ choice (i.e., Type R control). Fu-

ture analysis might reveal this effect in great-

er detail.

EXPERIMENT 3:

REPEATED FEEDBACK REVERSALS

In the previous two experiments, and in

our previously published research (Healy et

al., 1998), we employed separate AB and BA

designs, in which A constitutes accurate feed-

back and B is inaccurate feedback. These de-

signs produced relatively systematic out-

comes. When feedback was accurate, the

subjects eventually produced stable, combi-

natorially entailed relational responses, and

when feedback was inaccurate subjects even-

tually produced stable responding that was

not in accordance with combinatorial entail-

ment. A traditional single-subject experimen-

tal design, however, would involve not only

AB or BA but also a return to baseline (e.g.,

ABA or BAB, respectively). In fact, a highly

robust design would involve repeated rever-

sals (e.g., ABABAB) (e.g., Sidman, 1960). In

Experiment 3, patterns of responding were

examined when the type of feedback was re-

versed repeatedly across cycles of training

and testing (again using a novel stimulus set

for each cycle).

Procedure

The procedure of Experiment 3 was similar

to Condition 1 in Experiment 1 (e.g., two-

choice MTS, with a novel stimulus set em-

ployed in each cycle of training and testing)

with two important differences. First, Subjects

9 through 11 were exposed to training and

testing cycles until they responded in accor-

dance with accurate feedback across two,

rather than three, consecutive cycles. Second,

once this two-cycle mastery criterion had

been reached, the feedback then alternated

from inaccurate to accurate across six (Sub-

jects 9 and 10) or 10 (Subject 11) cycles of

training and testing.

Results and Discussion

The number of training trials required per

stimulus set by each subject is presented in

Table 3. The minimum number of trials re-

quired was 32 (Subject 10) and the maximum

was 178 (Subject 10). As in Experiments 1

and 2, the number of training trials per ex-

posure varied unsystematically for each sub-

ject, but again, in general, early training ex-

posures tended to involve larger numbers of

trials than many of the later exposures. Sub-

ject 9, however, appeared to depart a little

from this pattern, with relatively high num-

bers of training trials required across the ﬁnal

two stimulus sets.

Performance on each of the tests for com-

binatorial entailment are shown in Figure 3.

Subject 9 reached the stability criterion on

Exposure 4 and with the introduction of in-

accurate feedback switched responding from

high levels of combinatorial entailment to in-

termediate responding. On remaining expo-

sures as the feedback alternated from accu-

rate to inaccurate, responding switched

218 OLIVE HEALY et al.

Fig. 3. Class-consistent responding across successive test exposures for each subject in Experiment 3. Each data

point represents one exposure to a 20-trial test block involving a novel stimulus set. The presentation of inaccurate

or accurate test-performance feedback after each test is indicated by the letters IN and AC, respectively (see text for

details).

219GENERALIZED OPERANT BEHAVIOR

abruptly from high levels of entailment to low

levels in accordance with the type of feedback

delivered. Subject 10 reached the criterion

on Exposure 6 and also produced a variable

pattern of responding across Exposures 7 to

12 when the feedback alternated from accu-

rate to inaccurate. For both of these subjects,

the alternating feedback clearly produced

dramatic changes in the patterns of respond-

ing. However, in both cases, neither subject

successfully adapted responding to the feed-

back contingencies. For example, when the

feedback changed from accurate to inaccu-

rate at the end of Exposure 5 for Subject 9,

a shift from a high to an intermediate level

of combinatorial entailment occurred. How-

ever, at the end of Exposure 6, accurate feed-

back was again presented, and the subject

shifted back to a high level of combinatorial

entailment. Thus after both Exposures 5 and

6, Subject 9 received relatively few points for

his performances during those exposures,

and this pattern continued for the remaining

stimulus sets. That is, he switched his pattern

of responding back and forth between high

and low levels of entailment across each stim-

ulus set, thereby losing almost all of the feed-

back points available. A similar pattern was

also observed for Subject 10. To determine

whether this pattern of consistent point loss

would eventually break down, Subject 11 was

exposed to the same procedure, but addition-

al exposures to the alternating feedback were

provided. With sufﬁcient exposure to this

form of feedback, would a pattern of re-

sponding eventually emerge that allows the

subject to obtain most of the feedback points

available across stimulus sets?

Subject 11 reached the mastery criterion

during Exposures 5 and 6, but maintained

high levels of combinatorial entailment when

inaccurate feedback was introduced following

Exposure 7. This pattern of responding was

maintained until inaccurate feedback was de-

livered following Exposure 9, and responding

switched from high to low levels of entail-

ment. Responding across Exposures 10, 11,

12, and 13 alternated from high levels to low

levels of entailment, and thus, this subject lost

most of the feedback points available during

these exposures. During Exposure 14, how-

ever, a high level of entailment was main-

tained despite the delivery of inaccurate feed-

back. Responding then switched to low levels

(Exposure 15) and reverted to high levels on

the ﬁnal exposure. Therefore, across Expo-

sures 14, 15, and 16, Subject 11 achieved the

maximum number of points obtainable.

These data indicate that not only does dif-

ferential feedback establish, maintain, and

produce rapid shifts in test performances, but

it also produces responding that comes in-

creasingly under the control of repeated re-

versals of the feedback contingencies. This

ﬁnding may be considered evidence of yet an-

other example of generalized operant behav-

ior. We return to this issue in the General

Discussion.

EXPERIMENT 4:

FEEDBACK ON THE

COMPONENT OPERANTS OF

DERIVED RELATIONAL RESPONDING

The experiments conducted thus far have

shown that the use of differential conse-

quences can bring speciﬁc patterns of de-

rived relational responding under relatively

reliable control. In each of the two-choice

MTS experiments (including those reported

by Healy et al., 1998) the feedback contin-

gencies were designed to reverse the pattern

of derived relational responding. For exam-

ple, if the A1-C1, A2-C2, C1-A1, and C2-A2

relational responses were class consistent, the

inaccurate feedback contingencies aimed to

generate a complete reversal in these rela-

tional responses (i.e., A1-C2, A2-C1, C1-A2,

and C2-A1). This constitutes one way to test

the operant nature of derived relational re-

sponding, but additional tests are possible.

For example, the current work focused only

on combinatorial entailment, which is only

one of the deﬁning properties of relational

framing. Another property is mutual entail-

ment, which we have not yet examined. Some

relevant work has been conducted in this area

(e.g., Pilgrim & Galizio, 1990, 1995; Saunders

et al., 1988; Spradlin, Saunders, & Saunders,

1992) showing that the component proper-

ties of the relational frame of equivalence

may dissociate under speciﬁc environmental

control.

Pilgrim and Galizio (1990), for example,

trained adult subjects on a series of condi-

tional discriminations (i.e., A1-B1, A2-B2, A1-

C1, A2-C2) that led to the emergence of two

three-member equivalence classes during

220 OLIVE HEALY et al.

testing (i.e., A1-B1-C1, A2-B2-C2). Following

equivalence testing, subjects received further

training in which the original A-C relations

were reversed (i.e., A1-B1, A2-B2, A1-C2, A2-

C1). This altered the symmetry responding of

3 of 4 subjects, but none of them responded

in accordance with the transitive relations

that would be expected to follow from the

new (reversed) conditional discriminations.

In effect, equivalence test performances were

not controlled by the modiﬁed conditional

discrimination contingencies that were in ef-

fect, despite the fact that performances on

symmetry probes were sensitive to the novel

reinforcement contingencies. This ﬁnding is

consistent with both earlier and more recent

research with adult subjects showing the re-

sistance of equivalence relations to modiﬁca-

tion via the manipulation of baseline condi-

tional discriminations (Pilgrim & Galizio,

1995; Saunders et al., 1988; but see Pilgrim,

Chambers, & Galizio, 1995; Spradlin et al.,

1992, for evidence that equivalence respond-

ing is less resistant to change when children,

rather than adults, are subjects).

These data may indicate that symmetry and

transitivity (as examples of mutual and com-

binatorial entailment) do not always appear

together but may instead represent indepen-

dent stimulus-control relations. In the words

of Pilgrim and Galizio (1996), ‘‘The dissoci-

ation between symmetry and transitivity pat-

terns raises questions about the integrity of

the equivalence phenomenon, as deﬁned by

cohesiveness among the properties of reﬂex-

ivity, symmetry, and transitivity’’ (p. 177).

This view provides a contrast to previous def-

initions of equivalence classes, that of congru-

ent patterns of responding on tests of sym-

metry and transitivity (Sidman & Tailby,

1982). For Pilgrim and Galizio, therefore, the

formation of equivalence classes may be due

to a correspondence between multiple, small-

er behavioral units. Similar arguments have

been made by relational frame theorists (e.g.,

Roche et al., 1997; Wilson & Hayes, 1996).

The feedback procedures employed in the

current study may prove to be useful in ana-

lyzing the component properties of relational

frames. That is, if relational frames are com-

posed of component operant classes, then it

should be possible to gain control over the

component operants by employing differen-

tial consequences. That was the purpose of

Experiment 4.

Procedure

The MTS training and testing procedures

for Experiment 4 were broadly similar to Ex-

periments 1 and 3 (i.e., two-choice MTS), ex-

cept that Subjects 12 and 13 were trained and

tested for both mutual and combinatorial en-

tailment. Training was identical to Experi-

ments 1 and 3 (e.g., four two-choice MTS trial

types were used and eight consecutively cor-

rect responses were required to complete the

training). During the initial cycles of training

and testing, training was followed directly by

a test for mutual entailment, and then by a

test for combinatorial entailment. The tests

for mutual entailment consisted of four MTS

trial types involving B-A and C-B relations

(i.e., B1-A1/not A2; B2-A2/not A1; C1-B1/

not B2; C2-B2/not B1). These trial types were

presented in a quasirandom order, with each

of the four trial types occurring ﬁve times

across 20 trials. The tests for combinatorial

entailment were similar to those employed in

Experiments 1 and 3 (i.e., four trial types in-

volving the A-C and C-A relations, presented

ﬁve times across 20 trials). The computer

screen remained blank for 30 s between the

last trial of the mutual entailment test and the

ﬁrst trial of the combinatorial entailment test.

No test-performance feedback was delivered

at this point in the experiment. This cycle of

training and testing, using a novel stimulus

set for each cycle, continued until a subject

produced both mutual and combinatorial en-

tailment across three successive stimulus sets.

In effect, subjects had to produce at least four

out of ﬁve class-consistent responses on each

of the eight trial types, across three consecu-

tive exposures to the tests for mutual and

combinatorial entailment (i.e., the mastery

criterion employed in Experiment 1 was ap-

plied to both mutual and combinatorial en-

tailment tests). This provided a baseline of

derived relational responding (Condition A)

from which to determine the effects of inac-

curate test feedback on mutual and combi-

natorial entailment.

Once a baseline had been established, both

subjects continued to receive cycles of train-

ing and testing, with new stimulus sets em-

ployed for each cycle, but different types of

test-performance feedback were introduced

221GENERALIZED OPERANT BEHAVIOR

across the 2 subjects. Subject 12 received ac-

curate feedback following tests for mutual en-

tailment (see below for details) and inaccu-

rate feedback following tests for combinatorial

entailment (Condition B). Subject 13 re-

ceived inaccurate feedback following tests for

mutual entailment (see below for details) and

accurate feedback following tests for combi-

natorial entailment (Condition C). When a

subject produced at least four of ﬁve feed-

back-consistent responses on each of the

eight trial types across three consecutive ex-

posures to each test, the alternative condition

was introduced, again using novel stimulus

sets with each cycle (i.e., Subject 12 switched

to Condition C, and Subject 13 switched to

Condition B). When a subject once again

produced at least four of ﬁve feedback-con-

sistent responses on each of the eight trial

types across three consecutive exposures to

each test, the baseline condition (A) was re-

introduced. That is, the test-performance

feedback was omitted after each exposure to

the tests for mutual and combinatorial en-

tailment. Once again, novel stimulus sets

were used with each cycle of training and test-

ing. Because we were uncertain what re-

sponse patterns would emerge during this re-

turn to baseline, we simply exposed each

subject to six cycles and then examined the

data. If a stable performance had not

emerged across the three ﬁnal cycles, addi-

tional exposures were planned, but this

proved to be unnecessary.

The feedback delivered after each test for

mutual entailment and each test for combi-

natorial entailment provided differential con-

sequences contingent upon the exact num-

ber of class-consistent responses emitted

during the two tests for derived relational re-

sponding. The format for both types of feed-

back was identical to the format used in the

previous experiments (i.e., stars presented

vertically on the screen). Similar feedback

contingencies operated for mutual entail-

ment as had operated for combinatorial en-

tailment in the previous experiments. For ex-

ample, emitting the relational responses

B1-A1, B2-A2, C1-B1, and C2-A2 each pro-

duced a star under accurate feedback contin-

gencies, but produced no stars under inac-

curate feedback contingencies. Similarly,

emitting the relational responses B1-A2, B2-

A1, C1-B2, and C2-B1 each produced a star

under inaccurate feedback contingencies, but

produced no stars under accurate feedback

contingencies. Mutual entailment feedback

always followed the test for mutual entail-

ment, and combinatorial entailment feed-

back always followed the test for combinato-

rial entailment. Each cycle of training and

testing, including the two types of feedback,

was presented without interruption. In effect,

the computer asked the subject to report to

the experimenter only after the following

general sequence had been completed: con-

ditional discrimination training

→

mutual en-

tailment test

→

mutual entailment feedback

→

combinatorial entailment test

→

combi-

natorial entailment feedback. The computer

screen remained blank for 10 s between the

end of the mutual entailment feedback and

the presentation of the ﬁrst trial of the com-

binatorial entailment test.

In summary, Subject 12 was exposed to an

ABCA design, and Subject 13 was exposed to

an ACBA design. Condition A constituted no

test-performance feedback, B constituted ac-

curate feedback for mutual entailment and

inaccurate feedback for combinatorial entail-

ment, and C constituted inaccurate feedback

for mutual entailment and accurate feedback

for combinatorial entailment.

Results and Discussion

The number of training trials required per

stimulus set by each subject is presented in

Table 4. The minimum number of trials re-

quired was 26 (Subject 12) and the maximum

was 162 (Subject 13). As in the previous ex-

periments, the number of training trials per

exposure tended to vary unsystematically for

each subject, but once again early training ex-

posures tended to involve larger numbers of

trials than many of the later exposures.

Performances on each of the tests for mu-

tual and combinatorial entailment are shown

in Figure 4. Derived relational responding oc-

curred with Subject 12 in accordance with the

stability criterion across Exposures 3, 4, and

5 (no feedback). With the introduction of ac-

curate feedback following the test for mutual

entailment and inaccurate feedback following

the test for combinatorial entailment, inter-

mediate responding occurred on the test for

combinatorial entailment along with high lev-

els of mutual entailment. The stability crite-

rion was reached across Stimulus Sets 9, 10,

222 OLIVE HEALY et al.

Table 4

Number of training trials across successive stimulus sets

for each subject in Experiment 4.

Stimulus set

(exposure)

Training trials

S12 S13

126

121

162

116

104

138

122

and 11 with high levels of mutual entailment

and low levels of combinatorial entailment.

Following this, feedback switched and the test

performances were affected across Exposures

13, 14, and 15. The mastery criterion was

reached across Exposures 16, 17, and 18 with

high levels of combinatorial entailment and

low levels of mutual entailment. With a re-

turn to baseline (no feedback), performance

on the test for mutual entailment gradually

shifted back to class-consistent responding,

and high levels of combinatorial entailment

were maintained across the ﬁnal three expo-

sures (i.e., meeting the mastery criterion).

Derived relational responding occurred

with Subject 13 in accordance with the mas-

tery criterion across Exposures 2, 3, and 4

(no feedback). With the introduction of in-

accurate feedback following the test for mu-

tual entailment and accurate feedback follow-

ing the test for combinatorial entailment,

responding on the test for mutual entailment

switched rapidly to low levels, with high levels

of combinatorial entailment maintained. The

mastery criterion was reached across Stimulus

Sets 7, 8, and 9 with high levels of combina-

torial entailment and low levels of mutual en-

tailment. Following this, feedback switched

and the mastery criterion was reached once

again across Exposures 13, 14, and 15 with

high levels of mutual entailment and low lev-

els of combinatorial entailment. With a re-

turn to baseline (no feedback), performance

on the test for combinatorial entailment shift-

ed gradually towards class-consistent respond-

ing. Intermediate levels of mutual entailment

were produced during Exposure 17, but re-

sponding returned to high levels during Ex-

posures 19, 20, and 21 (i.e., meeting the sta-

bility criterion).

In summary, both subjects produced de-

rived relational responding across stimulus

sets without feedback. With the introduction

of differential feedback following both tests,

responding was either maintained or

changed by the type of feedback delivered.

When the test-performance feedback was re-

moved, response patterns similar to those ob-

served during the ﬁrst baseline occurred.

GENERAL DISCUSSION

The present study supports previous re-

search by Healy et al. (1998) demonstrating

that patterns of combinatorial entailment can

be manipulated using differential feedback,

at least in adult human subjects. The use of

multiple stimulus sets in the current study,

however, has shown that derived relational re-

sponding may be interpreted as a form of gen-

eralized operant behavior. The present data

also indicate that derived relational respond-

ing, in certain contexts, may ‘‘fracture’’ into

independent stimulus-control relations, thus

supporting the ﬁndings of a number of pre-

vious studies (e.g., Pilgrim & Galizio, 1990,

1995; Roche et al., 1997; Wilson & Hayes,

1996).

One of the interesting ﬁndings from the

current research is that subjects showed levels

of responding in accordance with the feed-

back similar to those reported by Healy et al.

(1998). One might expect that more subjects

should have failed to respond in accordance

with the feedback when a new stimulus set

was introduced for each exposure than when

the same set was used across all exposures.

Perhaps, however, the use of novel stimuli for

each successive stimulus set eliminated non-

arbitrary forms of stimulus control that may

223GENERALIZED OPERANT BEHAVIOR

Fig. 4. Class-consistent responding across successive test exposures for each subject in Experiment 4. Each data

point represents one exposure to a 20-trial test block. For each stimulus set two data points are shown, one for the

mutual entailment test and one for the combinatorial entailment test. During Condition A no test-performance

feedback was delivered; during Condition B accurate feedback was delivered for mutual entailment and inaccurate

feedback was delivered for combinatorial entailment; during Condition C inaccurate feedback was delivered for

mutual entailment and accurate feedback was delivered for combinatorial entailment.

have persisted when only one set was used.

For example, during the ﬁrst test (with Stim-

ulus Set 1) a subject might have related the

stimuli in terms of alphabetical order (e.g.,

ZID with VEK because Z and V are both to-

wards the end of the alphabet). Using the

same stimulus set again might continue this

pattern, but if a new set were used this re-

sponse pattern might be punished (e.g., with

the second set, choosing DAX with CUG is

punished). Thus, nonarbitrary forms of stim-

ulus control would likely be ﬁltered out

across multiple sets, in a way that would not

be possible with only a single set (see Hayes,

Gifford, & Wilson, 1996, p. 287).

Healy et al. (1998) did not bring derived

relational responding under the control of

repeated feedback reversals, as has been

shown with other types of operant response

classes (e.g., Vaughan, 1988). In Experiment

3, however, we repeatedly reversed the type

of feedback across successive exposures, and

all 3 subjects rapidly shifted their levels of en-

tailment. With sufﬁcient exposure to the pro-

cedure, the behavior of 1 subject (11) adapt-

ed perfectly to the repeated reversals, in that

he obtained the maximum number of points

across the ﬁnal three consecutive test expo-

sures. In effect, this subject appeared to

adopt a win-shift/lose-stay pattern of derived

224 OLIVE HEALY et al.

relational responding across successive stim-

ulus sets. As indicated earlier, this type of per-

formance could be interpreted as generalized

operant behavior. Insofar as this was the case,

it appears that the generalized operant of de-

rived relational responding became a com-

ponent operant of a win-shift/lose-stay gen-

eralized operant class. This ﬁnding is

consistent with the RFT suggestion that many

examples of complex human behavior may

be usefully treated as generalized operants

that are either composed of other general-

ized operants (see below) or function as the

components of yet other generalized oper-

ants. In this regard, Barnes (1996) used the

phrase fractal-like to characterize the way in

which the generalized operant of relational

framing could give rise to increasing orders

of behavioral complexity. The data from Ex-

periment 3 (and Experiment 4; see below)

appear to provide an empirical example of

this fractal-like effect (see D. Barnes-Holmes

& Barnes-Holmes, 2000).

The current study provides evidence that

derived relational responding is sensitive to

differential consequences, and could thus be

considered a form of generalized operant be-

havior. Insofar as this interpretation is cor-

rect, a number of possibly important impli-

cations arise for the experimental and

conceptual analyses of derived relational re-

sponding. Consider, for example, the data

from Experiment 4 showing that response

patterns in accordance with mutual and com-

binatorial entailment could be fractured us-

ing the appropriate feedback contingencies.

These data indicate that relational frames, as

behavioral units, may be broken down into

component operants by the appropriate re-

inforcement contingencies. These ﬁndings

are consistent with previous relational-frame

research that showed that mutual entailment

developed before combinatorial entailment

in the behavioral repertoire of a young child

(Lipkens et al., 1993). In other words, it ap-

pears that these relational operants (i.e., re-

lational frames) do not develop in whole

cloth, but are established as component op-

erants that are then combined through inter-

action with the contingencies operating in

the verbal community. The current ﬁndings

indicate that even when the component op-

erants have combined into relational frames,

they may be broken down again in an exper-

imental context. This ﬁnding supports the

RFT account of derived relational responding

as a form of generalized operant behavior (D.

Barnes-Holmes & Barnes-Holmes, 2000).

Other aspects of the data from Experiment

4 are also consistent with RFT. Consider the

fact that both of the subjects in Experiment

4 showed resurgence of their earlier class-

consistent responding when feedback was

withdrawn during the return to baseline. Ac-

cording to RFT, the generalized operant of

equivalence responding probably has the lon-

gest and richest history of reinforcement of

all the relational operants (Barnes & Roche,

1996), and thus, in the absence of alternative

controlling stimuli, equivalence responding

becomes the most likely outcome (see Roche

& Barnes, 1996). From the RFT perspective,

this is particularly likely when using an MTS

procedure because it may readily function as

a cue for equivalence, based, for example, on

its use in educational settings to establish

word–symbol equivalences (D. Barnes-Holmes

& Barnes-Holmes, 2000). The resurgence of

equivalence responding during the return to

baseline, when consequential control for

nonequivalence responding was removed, is

therefore consistent with the RFT interpre-

tation of equivalence as a generalized operant

class with a rich and protracted history of re-

inforcement. The present ﬁndings are also

consistent with earlier RFT-based research re-

ported by Wilson and Hayes (1996) that

showed systematic behavioral resurgence of

previously observed equivalence classes in a

laboratory setting. Of course, the present

data were obtained with only 2 subjects in a

single experiment, and thus further research

will be required to determine how general

the resurgence effect might be.

Although the ﬁndings of the present study

are consistent with RFT, three important is-

sues should be addressed. First, although the

present research showed consequential con-

trol over derived relational responding, such

responding had almost certainly been estab-

lished in the behavioral repertoires of the

adult subjects during their preexperimental

histories (consistent with this suggestion, Sub-

jects 12 and 13 demonstrated class-consistent

responding before test-performance feedback

was introduced). From this perspective,

therefore, the feedback inﬂuenced preexist-

ing repertoires of generalized operant behav-

225GENERALIZED OPERANT BEHAVIOR

ior, and did not establish those repertoires ab

initio. Consequently, the current data do not

provide strong evidence for the RFT view that

derived relational responding is established,

in the ﬁrst instance, as generalized operant

behavior. Nevertheless, given the paucity of

research on the effects of consequences on

equivalence class formation, demonstration

studies such as this one constitute a ﬁrst step

in exploring the utility of the RFT interpre-

tation of derived relational responding. Fur-

thermore, preliminary research from the

Maynooth laboratory has recently estab-

lished, ab initio, patterns of derived relational

responding in young children using operant

techniques similar to those reported here (Y.

Barnes-Holmes, Barnes-Holmes, & Roche, in

press).

Second, the postsession feedback used as

consequences for generalized relational re-

sponding was unusual in this area of research.

Speciﬁcally, the operations necessary to estab-

lish the postsession stars as reinforcers re-

main unknown, but it seems likely that some

type of instructional or verbal control was in-

volved. Although some might argue that this

possibility in some way diminishes the impor-

tance of the feedback effects demonstrated in

the current study, we would argue that in-

structional control is difﬁcult to avoid in any

experiment using verbally able participants

(including standard equivalence experi-

ments). Due to their biological and personal

histories, such subjects engage in vast

amounts of verbal behavior, and any psychol-

ogy experiment is therefore teeming with

stimuli that may function as instructional

cues. In fact, one of the major purposes of

the current study, as an example of RFT re-

search, was to contribute to developing a

functional analysis of instructional or verbal

control itself (see the ﬁnal paragraph of this

discussion). As such, we do not see the ‘‘in-

trusion’’ of instructional or verbal control as

a ‘‘contaminant’’ to be removed from our ex-

periments, but as the sine qua non of our

research agenda (see Hayes & Barnes-

Holmes, in press).

Third, although the ﬁndings of the current

study are consistent with the RFT view that

derived relational responding may be usefully

approached as generalized operant behavior,

the data do not directly contradict alternative

theoretical positions, such as those developed

by Sidman (1994) and Horne and Lowe

(1996). It was not our purpose, however, to

conduct a deﬁnitive study that would render

one or more theoretical accounts invalid. In

fact, we do not believe that such a study will

ever be forthcoming (see Barnes, 1994;

Barnes & Roche, 1996). Nevertheless, in due

course one of the currently available theoret-

ical accounts may be found to pertain to a

broader array of data or to suggest a larger

number of new and useful empirical investi-

gations. In this regard, we simply note that

RFT set the occasion for the present study

and others like it (e.g., Y. Barnes-Holmes et

al., in press; Healy et al., 1998). Perhaps more

important, however, are the broader theoret-

ical issues arising from the current work, and

we will brieﬂy consider these in closing.

At a conceptual level, approaching rela-

tional responding as generalized operant be-

havior may provide new and useful ways of

conceptualizing complex human behavior

typically referred to as language and cogni-

tion. From the perspective of RFT, relational

activities are considered to be the functional-

analytic bedrock of these complex behavioral

repertoires. Relational frame theory avoids

the typical approach to language and cogni-

tion taken by cognitive psychology, which has

tended to emphasize ‘‘content’’ by the train-

ing of speciﬁc words or the acquisition of spe-

ciﬁc concepts applicable in the real world.

For RFT the key focus should be on the re-

lational activities per se, rather than on par-

ticular words or concepts. In the present

study, for example, large numbers of non-

sense syllables, with no clear reference to

real-world events, were used for training and

testing derived relational responding. Per-

haps a similar approach could be taken in

educational settings in which learners are

trained in both real-world concepts and in

various types of relational responding. Con-

sider a classroom setting in which games

could be designed to improve the ﬂexibility

of a child’s relational responding. Questions

could be asked such as: ‘‘If X is the same as

Y, and Y is the same as Z, do I like Z if I like

X?’’ Although there is currently little evi-

dence to support this type of approach, the

data reported in the present study suggest

that the concept of the generalized operant

may be central to the functional analysis of

226 OLIVE HEALY et al.

derived relational responding and human

language and cognition.

REFERENCES

Baer, D. M., Peterson, R. F., & Sherman, J. A. (1967).

The development of imitation by reinforcing behav-

ioral similarity to a model. Journal of the Experimental

Analysis of Behavior, 10, 405–416.

Barnes, D. (1994). Stimulus equivalence and relational

frame theory. The Psychological Record, 44, 91–124.

Barnes, D. (1996). Naming as a technical term: Sacriﬁc-

ing behavior analysis at the altar of popularity. Journal

of the Experimental Analysis of Behavior, 65, 264–267.

Barnes, D., Browne, M., Smeets, P. M., & Roche, B.

(1995). A transfer of functions and conditional trans-

fer of functions through equivalence relations in

three- to six-year-old children. The Psychological Record,

45, 405–430.

Barnes, D., & Hampson, P. (1993). Stimulus equivalence

and connectionism: The implications for behavior

analysis and cognitive science. The Psychological Record,

43, 617–638.

Barnes, D., & Hampson, P. (1997). Connectionist mod-

els of arbitrarily applicable relational responding: A

possible role for the hippocampal system. In J. Don-

ahoe & V. P. Dorsel (Eds.), Neural-network models of cog-

nition (pp. 496–521). Amsterdam: Elsevier.

Barnes, D., Hegarty, N., & Smeets, P. M. (1997). Relating

equivalence relations to equivalence relations: A re-

lational framing model of complex human function-

ing. The Analysis of Verbal Behavior, 14, 1–27.

Barnes, D., & Roche, B. (1996). Relational frame theory

and stimulus equivalence are fundamentally different:

A reply to Saunders’ commentary. The Psychological

Record, 46, 489–507.

Barnes-Holmes, D., & Barnes-Holmes, Y. (2000). Ex-

plaining complex behavior: Two perspectives on the

concept of generalized operant classes. The Psycholog-

ical Record, 50, 251–265.

Barnes-Holmes, Y., Barnes-Holmes, D., & Roche, B. (in

press). Multiple-exemplar training and the develop-

ment of relational framing in young children. In S. C.

Hayes & D. Barnes-Holmes (Eds.), Relational frame the-

ory: Creating an alternative behavioral agenda in language

and cognition. Reno, NV: Context Press.

Carrigan, P. F., Jr., & Sidman, M. (1992). Conditional

discrimination and equivalence relations: A theoreti-

cal analysis of control by negative stimuli. Journal of

the Experimental Analysis of Behavior, 58, 183–204.

Catania, A. C. (1996). On the origins of behavior struc-

ture. In T. R. Zentall & P. M. Smeets (Eds.), Stimulus

class formation in humans and animals (pp. 3–12). Am-

sterdam: Elsevier Science.

Cumming, W. W., & Berryman, R. (1965). The complex

discriminated operant: Studies of matching-to-sample

and related problems. In D. I. Mostofsky (Ed.), Stim-

ulus generalization (pp. 284–330). Stanford, CA: Stan-

ford University Press.

Dube, W. V., McIlvane, W. J., & Green, G. (1992). An

analysis of generalized identity matching-to-sample

test procedures. The Psychological Record, 42, 17–28.

Dymond, S., & Barnes, D. (1995). A transformation of

self-discrimination response functions in accordance

with the arbitrarily applicable relations of sameness,

more than, and less than. Journal of the Experimental

Analysis of Behavior, 64, 163–184.

Dymond, S., & Barnes, D. (1996). A transformation of

self-discrimination response functions in accordance

with the arbitrarily applicable relations of sameness

and opposition. The Psychological Record, 46, 271–300.

Estes, W. K. (1971). Reward in human learning: Theo-

retical issues and strategic choice points. In R. Glazer

(Ed.), The nature of reinforcement (pp. 16–36). New

York: Academic Press.

Gewirtz, J. L., & Stengle, K. G. (1968). Learning of gen-

eralized imitation as the basis for identiﬁcation. Psy-

chological Review, 5, 374–397.

Hayes, S. C. (1991). A relational control theory of stim-

ulus equivalence. In L. J. Hayes & P. N. Chase (Eds.),

Dialogues on verbal behavior (pp. 19–40). Reno, NV:

Context Press.

Hayes, S. C. (1992). Verbal relations, time and suicide.

In S. C. Hayes & L. J. Hayes (Eds.), Understanding ver-

bal relations (pp. 109–118). Reno, NV: Context Press.

Hayes, S. C. (1994). Relational frame theory: A func-

tional approach to verbal events. In S. C. Hayes, L. J.

Hayes, M. Sato, & K. Ono (Eds.), Behavior analysis of

language and cognition (pp. 9–30). Reno, NV: Context

Press.

Hayes, S. C., & Barnes, D. (1997). Analyzing derived

stimulus relations requires more than the concept of

stimulus class. Journal of the Experimental Analysis of Be-

havior, 68, 235–233.

Hayes, S. C., & Barnes-Holmes, D. (Eds.). (in press). Re-

lational frame theory: Creating an alternative behavioral

agenda in language and cognition. Reno, NV: Context

Press.

Hayes, S. C., Gifford, E. V., & Wilson, K. G. (1996). Stim-

ulus classes and stimulus relations: Arbitrarily appli-

cable relational responding as an operant. In T. R.

Zentall & P. M. Smeets (Eds.), Stimulus class formation

in humans and animals (pp. 279–299). Amsterdam: El-

sevier.

Healy, O., Barnes, D., & Smeets, P. M. (1998). Derived

relational responding as an operant: The effects of

between-session feedback. The Psychological Record, 48,

511–536.

Horne, P. J., & Lowe, C. F. (1996). On the origins of

naming and other symbolic behavior. Journal of the Ex-

perimental Analysis of Behavior, 65, 185–241.

Lipkens, G., Hayes, S. C., & Hayes, L. J. (1993). Longi-

tidunal study of the development of derived relations

in an infant. Journal of Experimental Child Psychology, 56,

201–239.

McIlvane, W. J., Dube, W. V., & Callahan, T. D. (1995).

Attention: A behavior analytic perspective. In G. R.

Lyon & N. A. Krasnegor (Eds.), Attention, memory and

executive function (pp. 97–117). Baltimore: Paul H.

Brookes.

McIlvane, W. J., Dube, W. V., Kledaras, J. B., Iennaco, F.

M., & Stoddard, L. T. (1990). Teaching relational dis-

criminations to individuals with mental retardation:

Some problems and possible solutions. American Jour-

nal on Mental Retardation, 95, 283–296.

McIlvane, W. J., Kledaras, J. B., Munson, L. C., King., K.

A., de Rose, J. C., & Stoddard, L. T. (1987). Con-

trolling relations in conditional discrimination and

matching by exclusion. Journal of the Experimental Anal-

ysis of Behavior, 48, 187–208.

227GENERALIZED OPERANT BEHAVIOR

Neuringer, A. (1986). Can people behave ‘‘randomly’’?:

The role of feedback. Journal of Experimental Psychology:

General, 115, 62–75.

Pilgrim, C., Chambers, L., & Galizio, M. (1995). Reversal

of baseline relations and stimulus equivalence: II.

Children. Journal of the Experimental Analysis of Behavior,

63, 239–254.

Pilgrim, C., & Galizio, M. (1990). Relations between

baseline contingencies and equivalence probe perfor-

mances. Journal of the Experimental Analysis of Behavior,

54, 213–224.

Pilgrim, C., & Galizio, M. (1995). Reversal of baseline

relations and stimulus equivalence: I. Adults. Journal

of the Experimental Analysis of Behavior, 63, 225–238.

Pilgrim, C., & Galizio, M. (1996). Stimulus equivalence:

A class of correlations or a correlation of classes. In

T. R. Zentall & P. M. Smeets (Eds.), Stimulus class for-

mation in humans and animals (pp. 173–197). Amster-

dam: Elsevier.

Pilgrim, C., & Galizio, M. (in press). Stimulus equiva-

lence and units of analysis. In J. C. Leslie & D. E.

Blackman (Eds.), Experimental and applied analyses of

human behavior. Reno, NV: Context Press.

Poulson, C. L., Kymissis, E., Reeve, K. F., Andreatos, M.,

& Reeve, L. (1991). Generalized vocal imitation in

infants. Journal of Experimental Child Psychology, 51,

267–279.

Pryor, K. W., Haag, R., & O’Reilly, J. (1969). The creative

porpoise: Training for novel behavior. Journal of the

Experimental Analysis of Behavior, 12, 653–661.

Roche, B., & Barnes, D. (1996). Arbitrarily applicable

relational responding and human sexual categoriza-

tion: A critical test of the derived difference relation.

The Psychological Record, 46, 451–475.

Roche, B., & Barnes, D. (1997). A transformation of res-

pondently conditioned stimulus function in accor-

dance with arbitrarily applicable relations. Journal of

the Experimental Analysis of Behavior, 67, 275–300.

Roche, B., Barnes, D., & Smeets, P. M. (1997). Incon-

gruous stimulus pairing and conditional discrimina-

tion training: Effects on relational responding. Journal

of the Experimental Analysis of Behavior, 68, 143–160.

Roche, B., Barnes-Holmes, D., Smeets, P. M., Barnes-

Holmes, Y., & McGeady, S. (2000). Contextual con-

trol over the derived transformation of discriminative

and sexual arousal functions. The Psychological Record,

50, 267–291.

Saunders, R. R., Saunders, K. J., Kirby, K. C., & Spradlin,

J. E. (1988). The merger of and development of

equivalence classes by unreinforced conditional selec-

tion of comparison stimuli. Journal of the Experimental

Analysis of Behavior, 50, 145–162.

Saunders, R. R., & Sherman, J. A. (1986). Analysis of the

‘‘discrimination-failure hypothesis’’ in generalized

matching and mismatching behavior. Analysis and In-

tervention in Developmental Disabilities, 6, 89–107.

Sidman, M. (1960). Tactics of scientiﬁc research: Evaluating

experimental data in psychology. New York: Basic Books.

Sidman, M. (1987). Two choices are not enough. Behav-

ior Analysis, 22, 11–18.

Sidman, M. (1994). Equivalence relations and behavior: A

research story. Boston: Authors Cooperative.

Sidman, M., & Tailby, W. (1982). Conditional discrimi-

nation vs. matching-to-sample: An expansion of the

testing paradigm. Journal of the Experimental Analysis of

Behavior, 37, 5–22.

Spradlin, J. E., Saunders, K. J., & Saunders, R. R. (1992).

The stability of equivalence classes. In S. C. Hayes &

L. J. Hayes (Eds.), Understanding verbal relations (pp.

29–42). Reno, NV: Context Press.

Steele, D. L., & Hayes, S. C. (1991). Stimulus equiva-

lence and arbitrarily applicable relational responding.

Journal of the Experimental Analysis of Behavior, 56, 519–

555.

Stokes, T. F., & Baer, D. M. (1977). An implicit technol-

ogy of generalization. Journal of Applied Behavior Anal-

ysis, 10, 349–367.

Vaughan, W., Jr. (1988). Formation of equivalence sets

in pigeons. Journal of Experimental Psychology: Animal

Behavior Processes, 14, 36–42.

Wilson, K. G., & Hayes, S. C. (1996). Resurgence of de-

rived stimulus relations. Journal of the Experimental

Analysis of Behavior, 66, 267–281.

Received April 28, 1999

Final acceptance June 6, 2000

Flexibilidad cognitiva y control contextual de clases de equivalencia de estímulos según la naturaleza de las claves contextualesCognitive flexibility and contextual control of stimulus equivalence classes according to the nature of contextual cues

Article

Sep 2018
RACC

Contextual control of stimuli equivalence classes (SEC) emerges from second order conditional discriminations (CD): in this procedure, a stimulus can belong to different SEC according to the context. This characteristic allows the experimental study of flexible behaviors. The Wisconsin Card Sorting Test (WCST) also assesses cognitive flexibility, though by non-arbitrary criteria. In this study, the performance of 52 participants in a second-order CD task was evaluated, using contextual cues of different nature (arbitrary / verbal), and compared with their performance in the WCST. The results showed that, although verbal keys acquired their contextual function more readily than the arbitrary ones, they did not have a differential impact on the contextual control of SEC. Strong associations between performances in both tasks were observed. The results are discussed in terms of differences in the arbitrariness of the stimuli and the convenience of using rigorous procedures in the evaluation of relational flexibility.

A Comprehensive Behavioral Model of Emotion Rooted in Relational Frame Theory and Contemporary Extensions

Article

Jul 2024

There exists a vast literature on affect and emotion in psychological disciplines, yet contemporary conceptualizations and technologies to predict and influence emotion have been slower to emerge in behavior analysis. The current article is an attempt to conceptualize emotional experiencing through a radical behavioral lens using relational frame theory (RFT) and contemporary extensions. RFT provides a behavioral approach to cognitive appraisal within existing models of human emotion by emphasizing derived relational responding, transformation of stimulus function, and generalized reinforcement learning. Relational density theory (RDT) and the hyperdimensional multilevel (HDML) framework both expand upon RFT and may allow for a more complete account of emotional experiencing within complex networks. Synthesizing these two approaches yields multiple testable predictions that are consistent with RDT across levels of the HDML. Moreover, the ROE-M (relating, orienting, and evoking functions within a motivational context) is a dynamical unit that may be readily evident within emotional experiencing as it is generally described within the psychological literature, and compatible with the synthesized model. Taken together, these approaches and emerging research on affective dynamics may provide a starting point to develop a robust and comprehensive analysis of human emotion that can strengthen behavior analysis and therapies

Observation, Language Learning, and Development: The Verbal Behavior Development Theory

Article

Feb 2024

A review of recent applied research in observation suggests researchers could profit from a new account of observational learning. Current research in the identification and establishment of verbal developmental cusps demonstrates the importance of the range of observational cusps necessary for the acquisition of language. These cusps encompass learning through imitation, duplication of outcomes, understanding consequences of observed behaviors, acquiring new reinforcers, incidental unidirectional and bidirectional naming, and more. This account offers solutions to bridge gaps in the literature and complements related research, providing a comprehensive understanding of observational learning processes. This updated account of observational learning is especially relevant when we consider its implications for human language acquisition. In this article, we emphasize that language acquisition is not solely an individual cognitive development, but a socially mediated process, where observation plays a fundamental role in linguistic growth and development.

Skinner and Relational Frame Theory: Integrating Units of Analysis on a Continuum of Complexity

Article

Full-text available

Sep 2023
PSYCHOL REC

William F. Perez

Skinner and relational frame theory (RFT) present different behavioral perspectives on language. Although some bridges can be found between Skinner's account to verbal behavior and RFT at the conceptual level, the units of analysis presented by each tradition are not fully integrated. The unit of analysis suggested by RFT allows the components that constitute the practices of a verbal community (arbitrary, conventional relations) to be investigated within an operant point of view. Moreover, understanding relational responses as generalized operants, defined functionally by RFT, highlights the reinforcement contingencies involved in the multiple-exemplar training that gives rise to the generativity that hallmarks language as a phenomenon. The present article aims to explore an integration between the Skinnerian text and RFT. Such integration is grounded in the concept of mediated reinforcement contingencies that shape the practices of the verbal community. A conceptual treatment is offered whereby these practices can be better understood on a continuum of complexity that shapes specific instances of verbal responses (echoic, tact, mand etc.) up to purely functional generalized operants. Possible advantages of such an integrative view, as well as some of its limitations and practical challenges are discussed.

Deictic Framing Performance Fails to Generalize to Other Perspective-Taking Tasks

Article

Full-text available

Jun 2023

The behavioral processes underlying perspective taking have not been studied extensively. One approach to understanding and enhancing perspective taking, deictic framing, has been proposed. Proponents of this approach have suggested that deictic framing is a core property of perspective taking. A training protocol based on deictic framing has been developed and tested, but researchers generally evaluate the effectiveness of the protocol using tasks that have a similar format to the training protocol. Little research has examined the protocol’s effectiveness for improving performance in different perspective-taking tasks. We investigated generalization of the performance of three groups of university students trained with a deictic-framing protocol (or not) and tested with two other perspective-taking tasks: a visuospatial perspective taking using a cupboard containing a range of objects and a version of the implicit relational assessment procedure specifically designed to measure perspective taking. The first group was trained with the original verbal protocol with deictic expressions; the second group with the same protocol involving nondeictic words; and the third group was merely exposed to deictic expressions as a control condition. The results suggested that deictic framing is not fundamental to perspective taking, as the performance of the two experimental groups was not significantly different from the control group’s performance. Identification of specific stimulus functions involved in successful perspective taking and how those functions can be established should be addressed in future research.

Relational Frame Theory (RFT): An Empirical Approach to Language and Cognition

Article

Jan 2023

The effort of understanding human language and cognition, which distinguishes humans from other organisms and provides them with the opportunities they have on earth, dates back to ancient times. In line with this effort a theory of language and cognition, which is fairly comprehensive and has empirical support, has been developed based on a functional contextualistic view in the last century. This theory, called Relational Frame Theory (RFT), considers language as a verbal behavior and argues that the basis of human language and cognition is the behavior known as Arbitrarily Applicable Relational Responding (AARR). This behavior includes responding based on arbitrary relations independent of the non-arbitrary (physical) relations between stimuli, established by social whim and that do not make sense outside the social context in which they are located. AARR is characterized by three core properties such as mutual entailment, combinatorial mutual entailment and transformation of stimulus functions, and this behavior itself needs to be learned. This review aims to explain RFT, a very complex theory, in a comprehensive and understandable way. The historical process that brought about RFT will be mentioned, the basic principles and concepts of it will be explained, and the contributions of this theory to the understanding of complex human behaviors and its approach to psychopathology will be discussed.

Teaching Verbal Behavior to Children with Autism Spectrum Disorders

Chapter

Nov 2024

This chapter provides a comprehensive analysis of the strategies for teaching verbal behavior to children with autism spectrum disorders (ASD). It explores the fundamental role of language and communication in human development with a focus on the functionality of verbal behavior. By breaking down verbal behavior into teachable units, such as mands, tacts, echoics, and intraverbals, the chapter underscores the necessity of a tailored, functional communication approach in the remediation of language delays and the fostering of an adequate communicative repertoire in children with autism.

Teoria das Molduras Relacionais (RFT): Apresentação para analistas do comportamento

Article

Full-text available

Nov 2023

A Teoria das Molduras Relacionais (RFT) é uma proposta teórica para explicar um conjunto de evidências relacionadas ao fenômeno comportamental do responder relacional derivado, com implicações teóricas e aplicadas, especialmente no âmbito do comportamento verbal humano. Embora a RFT esteja tendo um impacto significativo e progressivamente maior na ciência comportamental, muitas de suas publicações estão em língua inglesa, prejudicando sua acessibilidade à comunidade analítico-comportamental brasileira. Portanto, o objetivo deste artigo é apresentar de modo introdutório os principais pilares conceituais da RFT para leitores familiarizados com princípios básicos de Análise do Comportamento. Isso inclui: as bases filosóficas do Contextualismo Funcional; os fenômenos comportamentais de responder relacional (não-arbitrário, arbitrário e derivado); os conceitos de Responder Relacional Arbitrariamente Aplicável (RRAA) e molduras relacionais, incluindo suas propriedades e a história de aprendizagem; e, por fim, as dimensões do RRAA.

Randomized-controlled trial examining differing content of ABA therapy delivered to autistic children

Article

Mar 2023

Training temporal relational framing in young children

Article

Mar 2023

Contextual Control over the Derived Transformation of Discriminative and Sexual Arousal Functions

Article

Full-text available

Mar 2000

A previous study by Roche and Barnes (1997) examined the transformation of conditioned sexual arousal in accordance with arbitrary relations. The current research replicated and extended that study by attempting to bring the derived transformation effect under contextual control. In Experiment 1, the functions of hand waving and clapping were first established for two nonsense syllables (called B1 and B2, respectively). Subjects were then exposed to relational pretraining, similar to that employed by Steele and Hayes (1991), in order to establish the contextual functions of Same and Opposite in two arbitrary stimuli. Subsequently, subjects were trained in the following relations; Same/A1-[B1-B2-N1], Same/A1-[C1-C2-N2] Opposite/A1-[B1-B2-N1], Opposite/A1-[C1-C2-N2] (underlined comparison stimuli indicate reinforced choices) from which the following relational responses emerged; Same/B1-C1; Same/B2-C2; Opposite/B1-C2; Opposite/B2-C1. During a testing phase, the stimulus functions established for B1 emerged for C1 in the presence of Same (i.e., the subjects waved) but those established for B2 emerged for C1 in the presence of Opposite (i.e., the subjects clapped). Similarly, the functions of B2 emerged for C2 in the presence of Same (i.e., the subjects clapped), but those established for B1 emerged for C2 in the presence of Opposite (i.e., the subjects waved). Experiment 2 established similar results using respondent eliciting functions in the place of hand clapping and waving.

Relational Frame Theory and Stimulus Equivalence are Fundamentally Different: A Reply to Saunders’ Commentary

Article

Jul 1996

Stimulus Equivalence and Connectionism: Implications for Behavior Analysis and Cognitive Science

Article

Oct 1993

Recent developments in behavior analysis and cognitive science can be used to improve the uneasy relationship between these two approaches to psychological inquiry. Stimulus equivalence phenomena demonstrate the power of behavior analytic procedures to induce complex generative performances that are typically studied by cognitive scientists, while connectionism challenges the previously dominant symbol-based accounts of cognition and emphasizes the importance of environmental constraints and fundamental principles of learning as accounts of cognitive processing. The nature and possible contribution of stimulus equivalence to cognitive science are outlined, and connectionist simulations of complex equivalence phenomena are described. These simulations provide a possible rapprochement between behavior analysis and cognitive science.

The complex discriminated operant: Studies of matching-to-sample

Article

Jan 1965

Stimulus equivalence and relational frame theory

Article

Jan 1994

D. Barnes

Verbal relations, time, and suicide

Article

Jan 1992

Steven C Hayes

Derived Relational Responding As An Operant: The Effects Of Between-Session Feedback

Article

Jun 1998

Relational Frame Theory (RFT) views derived relational responding as an overarching operant class of behavior. One approach to the analysis of derived relations from the RFT perspective, therefore, would be to examine systematically the effects of differential consequences on derived responding. In Experiment 1, 15 undergraduate subjects were divided into three conditions (5 subjects in each condition). In each condition subjects were exposed to 11 sessions, and in each session they were trained and tested for the formation of combinatorially entailed relations. In Condition 1, no differential consequences were delivered after any session. In Condition 2, response-independent, positive feedback was delivered after each of the first five sessions (i.e., the experimenter said to the subject "You are doing very well") and response-independent, negative feedback was delivered after Sessions 6 to 10 (i.e., the experimenter said to the subject "You are doing badly"). Condition 3 was similar to Condition 2, except that the first five sessions were followed by negative feedback and the remaining sessions were followed by positive feedback. The results showed that (a) delivering positive feedback before negative feedback attenuated relational responding relative to the negative before positive feedback condition, and (b) delivering differential feedback produced more frequent relational responding relative to the no-feedback condition. Experiment 2 replicated Conditions 2 and 3 of Experiment I, except that the positive and negative feedback was replaced by accurate and inaccurate feedback, respectively. During accurate feedback, subjects were told "You're doing well" if responding was above 50% correct and were told "You're doing badly" if responding was less than 50% correct. This was reversed for inaccurate feedback. Experiment 3 replicated Experiment 2 except that the feedback was delivered automatically via the computer. Experiment 4 involved a control whereby the test phase was unrelated to baseline training. The feedback delivered remained identical to that of the previous experiment. Experiment 5 replicated Experiment 3, but the feedback was precise in that, following the test phase, subjects received a point for every correct response made. The results showed that, in general, derived relational responding was highly sensitive to the response-contingent feedback and this responding was more frequent when precise feedback was used. Overall, the data are consistent with the suggestion that derived relational responding may be viewed as generalized operant behavior.

Arbitrarily Applicable Relational Responding and Sexual Categorization: A Critical Test of the Derived Difference Relation

Article

Jul 1996

The major aim of the current study was to demonstrate that preexperimentally established verbal functions can be examined, and transferred to arbitrary stimuli, using the procedures adopted by relational frame theory. Ten subjects were first exposed to relational pretraining, similar to that employed by Steele and Hayes (1991), in order to establish the relational functions of SAME, OPPOSITE, and DIFFERENT in three arbitrary stimuli. Subjects were then trained in the following relations: SAME/S1-DOMINATE, DIFFERENT/S1-FORGET, OPPOSITE/S1-SUBMIT, SAME/S1-X1, DIFFERENT/S1-X2, and OPPOSITE/S1-X3. Testing involved presenting the subjects with PENIS or VAGINA as samples and X1, X2, X3, and a question mark as comparisons (subjects were instructed to choose the question mark if they felt that none of the other comparisons were correct). All subjects consistently chose X1 given PENIS in the presence of SAME but X3 in the presence of OPPOSITE. Similarly, they chose X3 given VAGINA in the presence of the SAME contextual cue, but chose X1 in the presence of OPPOSITE. Four of the ten subjects chose X2 in the presence of DIFFERENT, given either PENIS or VAGINA as a sample. The six remaining subjects chose the question mark on these tasks, thus indicating that the relation between X2 and PENIS was unspecified. Five of the subjects were exposed to an additional test in which AMNESIA was presented as a sample. In the presence of SAME, three subjects consistently chose the question mark, whereas two subjects consistently chose X2. In the presence of DIFFERENT, subjects consistently chose either X1 or X3. Finally, in the presence of OPPOSITE, subjects consistently chose the question mark. These data support a relational frame account of sexual categorization.

Reward in human learning: Theoretical issues and strategic choice points

Article

Jan 1971

W.K. Estes

An Analysis of Generalized Identity Matching-to-Sample Test Procedures

Article

Jan 1992

This paper examines issues that must be considered when evaluating generalized identity matching to sample. We suggest that (a) stimuli displayed together on tests of generalized identity matching should have comparable experimental histories, and (b) identity matching procedures should be differentiated as either “conditional” or “nonconditional,” depending upon whether or not conditional discrimination is required. Although either type of procedure might be adequate for demonstrating a “matching concept,” we suggest that only generalized conditional identity matching provides an adequate test for the reflexive property of conditional relations.

Derived relational responding as generalized operant behavior

Abstract and Figures

Recommended publications

Synchronised and complementary coordination mechanisms in an asymmetric joint aiming task

Transactional Properties of Permissioned Blockchains

Poland and the Nordic States

Practical Approach to Prediction of Plant Technological Parameters with Missing Data