143

JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 1997, 68, 143–160 NUMBER 2 (SEPTEMBER)

INCONGRUOUS STIMULUS PAIRING AND CONDITIONALDISCRIMINATION TRAINING: EFFECTS ON

RELATIONAL RESPONDING

BRYAN ROCHE, DERMOT BARNES,AND PAUL M. SMEETS

UNIVERSITY COLLEGE CORK, IRELAND, ANDUNIVERSITY OF LEIDEN, THE NETHERLANDS

In Experiment 1, 5 subjects were exposed to a stimulus-pairing procedure in which two nonsensesyllables, identified by a letter-number code as A1 and C2, each predicted the onset of a sexual filmclip, and the nonsense syllables A2 and C1 each predicted the onset of a nonsexual film clip. Subjectswere then exposed to a matching-to-sample test in which the nonsense syllables A1 and A2 werepresented as sample stimuli and C1 and C2 were presented as comparison stimuli and vice versa(i.e., C stimuli as samples and A stimuli as comparisons). All subjects matched A1 with C2 and A2with C1. Subjects were then trained on the conditional discriminations A1-B1, A2-B2, B1-C1, B2-C2,after which the matching-to-sample test was again administered. All subjects continued to match A1with C2 and A2 with C1 in accordance with the earlier stimulus-pairing contingencies. An additional5 subjects were exposed first to conditional discrimination training and testing before being exposedto the incongruous stimulus pairing and matching-to-sample testing. Under these conditions, 4 ofthe 5 subjects always matched A1 with C1 and A2 with C2. Experiment 2 replicated Experiment 1,except that a matching-to-sample test was not administered following the initial training procedure.Under these conditions, matching-to-sample test performances were controlled by the contingenciesthat had immediately preceded the test. Experiment 3 indicated that initial matching-to-sample testperformances were unlikely to change, even after repeated exposure to incongruous training andtesting. Experiment 4 demonstrated that pretraining with unrelated stimulus sets increased the sen-sitivity of matching-to-sample test performances to incongruous contingencies when they were similarin format to those arranged during pretraining. These data may have implications for a behavior-analytic interpretation of attitude formation and change.

Key words: conditional discrimination, stimulus pairing, incongruous contingencies, stimulus equiv-alence, key press, humans

When a subject is taught to select StimulusB in the presence of Stimulus A and to selectStimulus C in the presence of B, it is likelythat the subject also will select A in the pres-ence of B (symmetry), B in the presence ofC (symmetry), B in the presence of B (i.e., apartial test for reflexivity), C in the presenceof A (transitivity), and A in the presence ofC (combined symmetry and transitivity, orequivalence) without further training. Whenthis occurs, the stimuli are said to participatein an equivalence relation (Sidman, 1990,1992; see also Barnes, 1994; Fields, Adams,

This material was presented at the annual meeting ofthe Experimental Analysis of Behaviour Group, London,England, April 1995. The research was conducted as apart of Bryan Roche’s doctoral research program, underthe supervision of Dermot Barnes. The research wasfunded by the Department of Applied Psychology, Uni-versity College, Cork, Ireland.

Correspondence and requests for reprints should besent to Dermot Barnes at the Behavior Analysis and Cog-nitive Science Unit, Department of Applied Psychology,University College Cork, Cork, Ireland (E-mail: lemmy@ucc.ie).

Verhave, & Newman, 1990). Numerous stud-ies have examined the formation of equiva-lence relations (e.g., Barnes, Browne, Smeets,& Roche, 1995; Barnes & Keenan, 1993;Barnes, Lawlor, Smeets, & Roche, 1995;Barnes, McCullagh, & Keenan, 1990; Bush,Sidman, & de Rose, 1989; Cullinan, Barnes,Hampson, & Lyddy, 1994; Dube, McIlvane,Mackay, & Stoddard, 1987; Dymond &Barnes, 1994; Wulfert & Hayes, 1988), butfew have examined the disruption or loss ofequivalence relations following new learningexperiences. In one such study, Pilgrim andGalizio (1990) trained adult subjects on a se-ries of conditional discriminations (i.e.,A1-B1, A2-B2, A1-C1, A2-C2) that led to theemergence of two three-member equivalencerelations during testing (i.e., A1-B1-C1,A2-B2-C2). Following equivalence testing,subjects received further training in whichthe original A-C relations were reversed (i.e.,A1-B1, A2-B2, A1-C2, A2-C1). This led to al-terations in the symmetry responding of 3 of

144 BRYAN ROCHE et al.

4 subjects, but none of the subjects respond-ed in accordance with the transitive relationsthat would be expected to follow from thenew (reversed) conditional discriminations.In effect, equivalence test performances werenot controlled by the modified conditionaldiscrimination contingencies that were in ef-fect, despite the fact that performances onsymmetry probes were sensitive to the novelreinforcement contingencies. This finding isconsistent with the results of both earlier andmore recent research with adult subjects thatshow the resistance of equivalence relationsto modification via the manipulation of base-line conditional discriminations (Pilgrim &Galizio, 1995; Saunders, Saunders, Kirby, &Spradlin, 1988; but see Pilgrim, Chambers, &Galizio, 1995, and Spradlin, Saunders, &Saunders, 1992, for evidence that equivalenceresponding is less resistant to change whenchildren, rather than adults, are used as sub-jects).

There is yet further evidence to suggestthat stimulus relations may be resistant tochange by novel reinforcement contingen-cies, at least for adult subjects. Specifically,Watt, Keenan, Barnes, and Cairns (1991)used standard equivalence procedures tostudy social categorization in Northern Ire-land, where many people respond to eachothers’ names as discriminative for their re-ligious backgrounds. In this study, NorthernIrish and English subjects were trained tomatch three Catholic names to three non-sense syllables, and subsequently to match thethree nonsense syllables to three traditionallyProtestant symbols. During the equivalencetest, subjects were presented with Protestantsymbols as sample stimuli; the comparisonstimuli were two of the Catholic names usedduring training and a novel Protestant name.All of the English subjects chose the Catholicnames (related through equivalence to theProtestant symbols), but 12 of the 19 North-ern Irish subjects chose the novel Protestantname in the presence of the Protestant sym-bols, thus failing to demonstrate laboratory-induced equivalence relations. On the basisof these findings, the researchers argued thatthe extraexperimental social contingenciesthat operate in Northern Ireland were re-sponsible for the responding of the 12 North-ern Irish subjects. In effect, the preexperi-mentally established social and emotional

functions were incongruous with the experi-mental conditional discriminations and mayhave controlled subjects’ performances onthe equivalence test. Similar findings havealso been reported in the context of sexualstereotyping (Moxon, Keenan, & Hine, 1993)and academic self-concept (Barnes, Lawlor,Smeets, & Roche, 1995).

Although Watt et al. (1991) offered an ac-count of their data in terms of preexperi-mentally established incongruous stimulusfunctions, these functions were only inferred;that is, it was assumed that social functions ofparticular words had been established preex-perimentally by the verbal community. Incontrast, Pilgrim and Galizio (1990) estab-lished incongruous conditional discrimina-tions in the laboratory. This was achieved byfirst training and testing subjects on a seriesof conditional discriminations and then ex-posing them to further training trials inwhich the original conditional discrimina-tions were reversed. An interesting avenue ofresearch, therefore, might be to extend theWatt et al. study by combining both of theforegoing research strategies. Rather thansimply inferring that particular stimuli al-ready have specific social and emotional func-tions, such functions could be established inthe laboratory before examining their inter-action with incongruous conditional discrim-ination training. For example, social or emo-tional functions might be established for twoarbitrary stimuli, A1 and C2, by repeatedlypairing them with powerful eliciting stimuli.Emotionally neutral functions could also beestablished for two other stimuli, A2 and C1,by pairing them with visual material that typ-ically elicits little or no response. Subsequent-ly, subjects would be provided with incongru-ous conditional discrimination training, inwhich choosing B1 in the presence of A1 andC1 in the presence of B1 is reinforced, andchoosing B2 in the presence of A2 and C2 inthe presence of B2 is reinforced. Given thistraining history, would subjects match A1with C2 and A2 with C1 consistent with thestimulus-pairing procedure, or would theymatch A1 with C1 and A2 with C2 based onthe conditional discrimination training? Thecurrent series of experiments were designedto address this question.

145INCONGRUOUS RELATIONAL RESPONDING

GENERAL METHOD

Subjects

Fifty subjects participated in the current re-search. All subjects were recruited as volun-teers through notice board advertisementsand personal contacts. Nine male and 11 fe-male subjects, aged between 18 and 23 years,completed Experiment 1. An additional 3male and 7 female subjects participated inExperiment 2. Five males and 3 females par-ticipated in Experiment 3. Finally, 4 malesand 8 females participated in Experiment 4.All subjects were undergraduate students.Twenty-five of the subjects were undergradu-ate psychology students who had not studiedstimulus equivalence as a part of their under-graduate training. The remaining subjectswere not psychology students or college grad-uates. Each of the 50 subjects completed thestudy in one session.

Apparatus

Conditional discrimination training and test-ing. Subjects were seated at a table in a smallexperimental room (2 m by 2 m) containinga microcomputer (Applet Model LC) thatdisplayed black characters on a white back-ground. Stimulus presentations and the re-cording of responses were controlled by thecomputer, which was programmed in BBCBASIC.

Six nonsense syllables were employed inExperiments 1, 2, and 3 (JOM, CUG, ZID,PAF, VEK, and BEH). These were randomlyassigned as sample and comparison stimulifor each subject and are labeled, in the inter-ests of clarity, using the alphanumerics A1,B1, C1, A2, B2, and C2. An additional sixnonsense syllables were employed in Experi-ment 4 (ROG, LER, YIM, NEP, MAU, andDAX). These were randomly assigned as sam-ple and comparison stimuli for each subjectand are labeled X1, Y1, Z1, X2, Y2, and Z2.

Stimulus pairing. Social or emotional func-tions were established using sexually arousingvisual materials. Sexual stimuli were used inthe current study for two reasons. First, thestudy formed part of an ongoing researchprogram on human sexual functioning at theCork laboratory (see Barnes & Roche, 1996;Grey & Barnes, 1996; Roche & Barnes, 1995,1996a, 1996b, 1997). Second, insofar as Wattet al. (1991) employed stimuli that were as-

sumed to possess relatively strong emotionalfunctions (e.g., sectarian symbols in a North-ern Irish context), we employed sexual stim-uli that had proven to be emotionally pow-erful in an experimental setting in theRepublic of Ireland (see Roche & Barnes,1995). Nonsexual funtions were establishedusing nonsexual visual materials (e.g., land-scape scenes).

Sexual and nonsexual visual stimuli werepaired with the nonsense syllables labeled A1,C1, A2, and C2 using 45- to 60-s film clipstaken from a popular sex-instruction videoand a geographic documentary. The fournonsense stimuli were quasirandomly pairedwith sexual or nonsexual material. In Exper-iment 4, sexual and nonsexual film clips alsowere paired with additional nonsense sylla-bles (X1, Z1, X2, and Z2).

The film clips differed on every trial butwere taken from the same sex-instruction vid-eo or nature documentary. Sexual film clipsdepicted heterosexual ‘‘necking,’’ heavy pet-ting, coitus, cunnilingus, and fellatio, whereasnonsexual clips depicted scenic landscapes ofmountainous or desert regions. (A detaileddescription of the film clips is available fromthe second author.) All film clips were playedon a Panasonict portable video player (Mod-el NV 80) and relayed to a Panasonict 14-in.television monitor located in an adjacentroom. The subject’s monitor was placed di-rectly beside a microcomputer. None of thefilm clips contained a soundtrack.

Procedure

Stimulus pairing. Before being exposed tothe stimulus-pairing procedure, subjects wererequired to read and sign a consent form(Appendix) acknowledging their awarenessof the sexually explicit nature of some of thefilm clips that they were about to see. Theywere also informed that they were free to ter-minate their participation at any time, andwere asked not to discuss the study with any-one. Subjects were seated comfortably, ap-proximately 1 m from the television monitor,and were asked to relax and watch the tele-vision. Subjects were then left in privacy.

Sexual and nonsexual film clips were eachpaired with two of four nonsense syllables(A1, C1, A2, and C2). The A1, C1, A2, andC2 stimuli (7 cm by 3 cm) were presented,one per trial, in the center of the subject’s

146 BRYAN ROCHE et al.

television screen. The stimulus remained onthe screen for 3 s and was followed by a 5-sinterval during which the screen went dark.At the end of the 5-s interval, a sexual or non-sexual film clip was presented. The stimulusthat appeared before the onset of the filmclip also flashed periodically during the filmclip in the top right corner of the screen.That is, once every 15 s the nonsense syllableappeared on the screen five times, with thestimulus on for 1 s, off for 0.2 s, on for 1 s,and so on. Because film clips varied from 45to 60 s in duration, the flashing stimulus pre-sentation appeared three to four times dur-ing each film clip (i.e., every 15 s). Intertrialintervals also varied from 45 to 60 s. Subjectswere exposed to 16 training trials (i.e., fourexposures to each of A1, C1, A2, and C2), theorder of which was randomized within andacross subjects, with the restriction that nostimulus could appear more than twice in suc-cession. The stimulus-pairing phase of the ex-periment lasted approximately 30 min.

Matching-to-sample training. To begin thisphase of the experiment, the subject was ori-ented towards the microcomputer on whichall conditional discrimination trials were pre-sented. The following instructions were thenpresented to the subject via the computermonitor.

During this stage of the Experiment you mustlook at the nonsense syllable at the top of thescreen, and then choose one of the nonsensesyllables at the bottom by pressing one of themarked keys on the keyboard. To choose theleft syllable press the marked key on the left.To choose the right syllable press the markedkey on the right. Press the space-bar twice tocontinue.

On all matching-to-sample tasks, the sam-ple stimulus appeared in the middle of thescreen, and after a 1-s delay two comparisonstimuli appeared to the left and right belowthe sample at the bottom edge of the screen.The screen position of the comparison stim-uli was counterbalanced across trials. Thesample and comparison stimuli remained onthe screen together until the subject pressedone of the marked keys (Z or M) on the com-puter keyboard (all other keys were deacti-vated). If the choice was defined as correct,the screen cleared and ‘‘CORRECT’’ ap-peared on the screen for 1.5 s, accompaniedby a beep from the computer. If the choice

was defined as incorrect, the screen clearedand the word ‘‘WRONG’’ appeared on thescreen for 1.5 s, and no beep was presented.The feedback (‘‘CORRECT’’ or ‘‘WRONG’’)followed all responses on all training trials.During the intertrial interval, the screen re-mained blank for 2.5 s.

There were four training-trial types:Choose B1 or B2 given A1 or A2, respectively,and choose C1 or C2 given B1 or B2, respec-tively. Subjects were exposed to the four trialtypes in a quasirandom order in blocks offour trials (i.e., each trial type was presentedonce every four trials) until they produced aminimum of eight consecutive correct re-sponses.

Matching-to-sample testing. The instructionsdelivered to subjects before testing were iden-tical to those delivered before training. Thetesting phase probed for (a) the transitive re-lations, A1-C1 and A2-C2, and (b) the com-bined symmetrical and transitive relations,C1-A1 and C2-A2. Testing occurred for 40 tri-als, with each of the four trial types presented10 times in a quasirandom order. During testtrials no feedback was provided (i.e., the sub-ject’s response was followed immediately bythe intertrial interval). A stable test perfor-mance was defined as choosing the samecomparison stimulus at least nine times outof every 10 exposures to each of the four trialtypes. The experimental protocol specifiedthat subjects who failed to produce a stableperformance be reexposed to the test phase,without training, up to a maximum of threetimes. In fact, no subject required more thantwo exposures to the test phase.

The matching-to-sample test phase servedtwo purposes. First, it allowed the experi-menters to probe for equivalence relationsfollowing conditional discrimination training.Second, following stimulus pairing, thematching-to-sample test allowed the experi-menters to probe for control that was consis-tent with the stimulus pairing contingencies.

EXPERIMENT 1

Procedure

Nine male and 11 female subjects, aged be-tween 18 and 23 years, completed Experi-ment 1. This experiment consisted of fourconditions, with 5 subjects in each. Each of

147INCONGRUOUS RELATIONAL RESPONDING

the four conditions involved stimulus pairing,conditional discrimination training, andmatching-to-sample testing, as described inthe General Procedure. Conditions differed,however, in the order in which these phaseswere administered to subjects.

Condition 1. Subjects were first exposed tostimulus-pairing trials in which A1 and C2were paired with sexual film clips and A2 andC1 were paired with nonsexual film clips. Im-mediately following training, the matching-to-sample test was administered. Subjects werethen trained on the incongruous conditionaldiscriminations A1-B1, A2-B2, B1-C1, andB2-C2. Immediately following training, thematching-to-sample test was again adminis-tered to determine whether subjects wouldrepeat their initial matching-to-sample testperformance.

Condition 2. Subjects were exposed to thesame procedure as the subjects in Condition1, except that stimulus-pairing trials were con-gruent with conditional discrimination train-ing (i.e., A1 and C1 were paired with sexualstimuli, and A2 and C2 were paired with non-sexual stimuli).

Condition 3. Subjects were first trained onthe following conditional discriminations:A1-B1, A2-B2, B1-C1, and B2-C2. Immediatelyafter training, the matching-to-sample testwas administered. Subjects were then orient-ed towards the television monitor on whichincongruous stimulus-pairing trials were pre-sented (i.e., A1 and C2 were paired with sex-ual stimuli, and A2 and C1 were paired withnonsexual stimuli). Finally, subjects wereagain exposed to the matching-to-sample test.

Condition 4. Subjects were exposed to thesame procedure as in Condition 3, exceptthat stimulus-pairing trials were congruentwith conditional discrimination training trials(i.e., A1 and C1 were paired with sexual stim-uli, and A2 and C2 were paired with nonsex-ual stimuli).

Results and Discussion

Subjects’ response patterns from all con-ditions are presented in Table 1.

Condition 1. Following stimulus pairing, all5 subjects responded in accordance with thestimulus-pairing contingencies on their firstexposure to the matching-to-sample test (i.e.,they matched A1 with C2 and A2 with C1).Following incongruous conditional discrimi-

nation training (A1-B1, A2-B2, B1-C1, B2-C2),all subjects repeated their initial test perfor-mance on the second matching-to-sampletest.

Condition 2. Following stimulus pairing, all5 subjects responded on their first exposureto the matching-to-sample test by matchingA1 with C1 and A2 with C2. After congruentconditional discrimination training, all 5 sub-jects produced a matching-to-sample test per-formance that was consistent with both stim-ulus pairing and conditional discriminationtraining (i.e., they matched A1 with C1 andA2 with C2).

Condition 3. Subjects 11 and 13 respondedin accordance with the predicted equivalencerelations (i.e., they matched A1 with C1 andA2 with C2) on their first exposure to thematching-to-sample test phase. The remain-ing subjects (12, 14, and 15) required a sec-ond exposure to the test phase before theyproduced a stable performance in which theyalso matched A1 with C1 and A2 with C2. Fol-lowing exposure to incongruous stimuluspairing, 4 of the 5 subjects repeated their ini-tial performance on the matching-to-sampletest (i.e., they matched A1 with C1 and A2with C2). Subject 12, however, produced atest performance that was consistent with theincongruous stimulus-pairing contingencies(i.e., she matched A1 with C2 and A2 withC1).

Condition 4. All 5 subjects responded in ac-cordance with equivalence relations on theirfirst exposure to the matching-to-sample test.Following congruent stimulus pairing, all sub-jects produced a matching-to-sample test per-formance that was consistent with both con-ditional discrimination and stimulus-pairingcontingencies (i.e., they matched A1 with C1and A2 with C2).

In summary, Experiment 1 demonstratedthat test performances based on stimuluspairing and conditional discrimination wereresistant to alteration through the introduc-tion of incongruous conditional discrimina-tion and stimulus-pairing contingencies, re-spectively. Subjects in Condition 1 matchedA1 with C2 and A2 with C1 following stimuluspairing, and continued to do so following in-congruous conditional discrimination train-ing. When exposed to conditional discrimi-nation training first, 5 subjects in Condition3 behaved in a manner consistent with the

148 BRYAN ROCHE et al.

Table 1

Number of trials on which responses were consistent with initial experimental training inExperiment 1.

Subject no.and gender

No. of MTStraining

trialsTestno.

Test trial types

C1A2 A2C1 C2A1 A1C2 C1A1 A1C1 C2A2 A2C2

Condition 1 (incongruous pairing/MTS)1/F

2/F

3/M

4/M

5/F

16

16

20

108

44

1212121212

10101010101010109

10

10101010101010101010

10109

10101010101010

1010101010101099

10

Condition 2 (congruous pairing/MTS)6/F

7/F

8/M

9/M

10/F

60

36

28

36

16

1212121212

99

101099

109

1010

10101010101010101010

1010101010109

101010

10101010101010101010

Condition 3 (incongruous MTS/pairing)11/F

12/Fa

13/M

14/Fa

28

20

20

16

1212312123

9108

100

10106

1010

10109

100

10106

1010

10108

100

10106

1010

10108

101

1094

1010

15/Fa 400 123

101010

89

10

8109

101010

Condition 4 (congruous MTS/pairing)16/M

17/Ma

18/M

19/F

20/M

28

40

20

40

20

12123121212

101010101099

109

1010

1010101010101010101010

101098

10101010101010

109

10109

1010109

1010

a These subjects failed to produce stable performances (i.e., choose the same comparison at least nine times across10 exposures to each task) on the matching-to-sample test. The additional rows of data represent performances onsubsequent exposures to the test phase without additional training.

149INCONGRUOUS RELATIONAL RESPONDING

equivalence class (i.e., they matched A1 withC1 and A2 with C2), and 4 of these subjectssubsequently reproduced this performancefollowing exposure to incongruous stimulus-pairing contingencies (i.e., A1 and C2 pairedwith sexual stimuli and A2 and C1 paired withnonsexual stimuli).

EXPERIMENT 2

In Experiment 1, performances on the fi-nal matching-to-sample test were determinedby the training and testing to which subjectswere first exposed. One issue arises from thisfinding, however. Subjects were both trainedand tested before incongruous training trialswere administered. It is unclear, therefore, towhat extent training, testing, or both con-trolled the observed performances on the fi-nal matching-to-sample test (see also Spradlinet al., 1992). For example, what performancemight we expect on a matching-to-sample testfollowing conditional discrimination trainingif a subject was first exposed to incongruousstimulus pairing without a matching-to-sampletest? Experiment 2 was designed to addressthis issue.

Procedure

Three male and 7 female subjects partici-pated in Experiment 2. This experiment con-sisted of two conditions, with 5 subjects ineach. Both conditions involved stimulus pair-ing, conditional discrimination training, andmatching-to-sample testing, as described inthe General Procedure. The conditions dif-fered, however, in the order in which thesephases were administered to subjects.

Because all 10 subjects that were exposedto congruous conditional discrimination andstimulus pairing in Experiment 1 produced aperformance that was consistent with bothforms of training, Experiment 2 focused onlyon the effects of incongruity between thesetwo training phases.

Condition 1. Subjects were first exposed tostimulus pairing trials in which A1 and C2were paired with sexual stimuli and A2 andC1 were paired with nonsexual stimuli. Im-mediately following stimulus pairing, subjectswere trained on the incongruous conditionaldiscriminations A1-B1, A2-B2, B1-C1, andB2-C2. Following conditional discriminationtraining, subjects were exposed to the first

matching-to-sample test to determine wheth-er they would respond in accordance with thestimulus pairing (i.e., match A1 with C2 andA2 with C1) or with the emergent equiva-lence relations (i.e., match A1 with C1 andA2 with C2).

Condition 2. Subjects were first trained onthe following conditional discriminations:A1-B1, A2-B2, B1-C1, and B2-C2. Immediatelyafter conditional discrimination training, thesubject was oriented towards the televisionmonitor on which incongruous stimulus-pair-ing trials were presented (i.e., A1 and C2were paired with sexual material, and A2 andC1 were paired with nonsexual material).Subjects were subsequently exposed to thefirst matching-to-sample test to determinewhether they would respond in accordancewith the stimulus pairing (i.e., match A1 withC2 and A2 with C1) or the emergent equiv-alence relations (i.e., match A1 with C1 andA2 with C2).

Results and Discussion

Subjects’ response patterns from both exper-imental conditions are presented in Table 2.

Condition 1. Immediately following expo-sure to stimulus pairing, subjects were ex-posed to incongruous conditional discrimi-nation training (i.e., A1-B1, A2-B2, B1-C1,and B2-C2). Subsequently, 4 subjects dem-onstrated control by the conditional discrim-ination training on their first exposure to thematching-to-sample test (i.e., they matchedA1 with C1 and A2 with C2). One subject,however, showed stable responding in accor-dance with the stimulus-pairing contingencieson her second exposure to the matching-to-sample test.

Condition 2. Following conditional discrim-ination training (A1-B1, A2-B2, B1-C1, andB2-C2), subjects were immediately exposed tostimulus pairing and the matching-to-sampletest. During testing, 4 subjects responded inaccordance with the stimulus-pairing contin-gencies (i.e., they matched A1 with C2 andA2 with C1), and 1 subject demonstrated con-trol by the conditional discrimination contin-gencies (i.e., he matched A1 with C1 and A2with C2).

These data indicate that test performancescan be sensitive to modified contingencieswhen initial training phases have not beenfollowed by matching-to-sample tests.

150 BRYAN ROCHE et al.

Table 2

Number of trials on which responses were consistent with initial experimental training inExperiment 2.

Subjectno. andgender

No. ofMTS

trainingtrials

Testno.

Test trial types

C1A2 A2C1 C2A1 A1C2 C1A1 A1C1 C2A2 A2C2

Condition 1 (incongruous pairing/MTS)21/F22/Fa

23/F24/F25/M

1668

204420

112111

0109010

01010000

07

10000

08

10000

Condition 2 (incongruous MTS/pairing)26/F27/M28/F29/F30/M

2016208820

11111

010000

09000

010011

010100

a This subject failed to produce a stable performance (i.e., choose the same comparison at least nine times across10 exposures to each task) on the matching-to-sample test. The additional row of data represents performance on asubsequent exposure to the test phase without additional training.

EXPERIMENT 3

Experiments 1 and 2 established that ex-posure to the matching-to-sample test follow-ing initial training helped to determine thesensitivity of later test performances to incon-gruous contingencies. It is possible, however,that test performances can be made moresensitive to modified contingencies by meansother than the exclusion of a matching-to-sample test after initial training phases. Forinstance, it may be possible to weaken persis-tent test performances by repeatedly expos-ing subjects to incongruous reinforcement orstimulus-pairing contingencies. This is of par-ticular importance given one interpretiveproblem arising from Experiment 2. Specifi-cally, given that a matching-to-sample test wasnot administered following the first trainingphase, we cannot be certain that test perfor-mances consistent with that training wouldhave emerged. What is required, therefore, isan alternative experimental procedure thatallows us to control test performances with-out removing either of the test phases. Ex-periment 3 addressed this issue.

Subjects in Conditions 1 and 2 of Experi-ment 3 were exposed to experimental pro-cedures that were identical to those em-ployed in Conditions 1 and 3 of Experiment1, respectively. In addition, subjects in both

conditions were exposed repeatedly to incon-gruous training and testing.

Experiment 3 also addressed an additionalissue. Specifically, the matching-to-sampletests administered in Experiments 1 and 2probed only for transitivity and combinedsymmetry and transitivity. Therefore, in casesin which transitive and combined symmetri-cal and transitive performances failed toemerge, it was unclear whether basic sym-metry responding would have emerged if test-ed. Consider, for example, a subject who isexposed to stimulus-pairing trials followed bya matching-to-sample test. Given the datathus far, we would expect the initial test per-formance to be reproduced even after expo-sure to incongruous conditional discrimina-tion training. However, what performancewould be observed if the matching-to-sampletest that followed incongruous conditionaldiscrimination training included probes forsymmetry? Previous research suggests thatsymmetry test performances are sensitive tonovel (incongruous) reinforcement contin-gencies even when transitivity test perfor-mances are not (e.g., Pilgrim & Galizio, 1990,1995). In Experiment 3, therefore, subjectswere exposed to probes for symmetry ontheir final reexposure to the matching-to-sample test.

151INCONGRUOUS RELATIONAL RESPONDING

Procedure

Five males and 3 females participated inExperiment 3. Four subjects participated ineach of two conditions. Both conditions in-volved stimulus pairing, conditional discrim-ination training, and matching-to-sample test-ing, including probes for symmetricalrelations on the final matching-to-sample test.

During symmetry testing, both B1 and B2were presented as sample stimuli with A1 andA2 as comparison stimuli, and both C1 andC2 were presented as sample stimuli with B1and B2 as comparison stimuli. Symmetryprobe types were presented 10 times each,randomly interspersed with the matching-to-sample test trials used in the previous two ex-periments. Thus, matching-to-sample testing,including symmetry probes, spanned 80 trials(i.e., 10 exposures to each of the two transi-tive, two combined symmetrical and transi-tive, and four symmetrical trial types).

Condition 1. Subjects were first exposed tostimulus-pairing trials in which A1 and C2were paired with sexual stimuli and A2 andC1 were paired with nonsexual stimuli. Fol-lowing training, the matching-to-sample testwas administered. Subjects were then trainedon the incongruous conditional discrimina-tions A1-B1, A2-B2, B1-C1, and B2-C2. Thematching-to-sample test was then readminis-tered. Subjects were subsequently reexposedto cycles of incongruous conditional discrim-ination training and matching-to-sample test-ing, three times in succession (i.e., trainingand testing, followed by training and testing,followed by training and testing) to deter-mine whether persistent test performanceswould be repeated. Furthermore, during thefinal reexposure to the test, probes for sym-metrical relations were presented for the firsttime, as described previously (i.e., B1-A1,B2-A2, C1-B1, C2-B2).

Condition 2. Subjects were first trained onthe following two conditional discrimina-tions: A1-B1, A2-B2, B1-C1, and B2-C2. Im-mediately after training, the matching-to-sam-ple test was administered. The subject wasthen oriented towards the television monitoron which incongruous stimulus-pairing trialswere presented (i.e., A1 and C2 were pairedwith sexual stimuli, and A2 and C1 werepaired with nonsexual stimuli). Subjects werethen reexposed to the matching-to-sample

test. Finally, subjects were reexposed to eightstimulus pairing trials followed by matching-to-sample testing, three times in succession(i.e., 24 additional stimulus-pairing trials intoto). The eight stimulus-pairing trials werechosen quasirandomly from the 16 originalstimulus-pairing trials, and the trial typeswere balanced. Only eight of the 16 originaltrials were used in order to keep sessionlength to a minimum. In addition, a pilotstudy had shown that eight trials were suffi-cient to reinstate the effects of the stimulus-pairing contingencies. During the final reex-posure to the matching-to-sample test, probesfor symmetry responding were presented forthe first time (i.e., B1-A1, B2-A2, C1-B1, andC2-B2).

Results and Discussion

Subjects’ performances from both experi-mental conditions are presented in Tables 3and 4.

Condition 1. Following stimulus pairing, allsubjects matched A1 with C2 and A2 with C1on their first exposure to the matching-to-sample test. Subjects then completed incon-gruous conditional discrimination training,after which all 4 again responded in accor-dance with the stimulus pairings on thematching-to-sample test. All 4 subjects subse-quently maintained their originalmatching-to-sample performances across allthree reexposures to the matching-to-sampletest.

Condition 2. Following conditional discrim-ination training, all 4 subjects responded inaccordance with the emergent equivalencerelations (i.e., they matched A1 with C1 andA2 with C2). This performance was main-tained by all subjects following their first ex-posure to incongruous stimulus pairing. Allsubjects then reproduced their initial match-ing-to-sample test performance after each ofthe additional incongruous stimulus-pairingphases.

Symmetry probes. Under typical equivalencetraining and testing conditions, emergence ofthe following symmetrical relations would bepredicted from the conditional discrimina-tion training: B1-A1, B2-A2, C1-B1, andC2-B2. During their final reexposure to thematching-to-sample test phase, 3 subjects inCondition 1 responded in accordance withthese predicted symmetrical relations (see Ta-

152 BRYAN ROCHE et al.

Table 3

Number of trials on which responses were consistent with initial experimental training inExperiment 3.

Subjectno. andgender

No. ofMTS

trainingtrials

Testno.

Test trial types

C1A2 A2C1 C2A1 A1C2 C1A1 A1C1 C2A2 A2C2

Condition 1 (incongruous pairing/MTS 3 4)31/M

32/M

33/M

34/F

16201620

36162016

20165236

44201616

12345123451234512345

10101010109

10101010101010101010109

1010

1010101010109

101010101010101010109

1010

1010101010101010109

10101010101010101010

10101010101010101010101010101010109

1010

Condition 2 (incongruous MTS/pairing 3 4)35/Fa

36/F

44

28

12345612345

101010101010101010109

810101010101010101010

101010101010101010109

1010101010101010101010

37/M

38/M

36

28

1234512345

10101010101010101010

10101010109

10101010

10101010109

10101010

9101010101010101010

a This subject failed to produce a stable performance (i.e., choose the same comparison at least nine times across10 exposures to each task) on her first exposure to the matching-to-sample test. The subsequent row of data representsperformance on a subsequent exposure to the test phase without additional training.

ble 4), although transitive responding wasconsistent with stimulus pairing. Subject 33,however, responded in accordance with onlytwo of these symmetrical relations (i.e.,B1-A2, B2-A1, C1-B1, and C2-B2). All 4 sub-jects in Condition 2, who repeatedly pro-

duced a conditional-discrimination-based testperformance, responded in accordance withthe emergent symmetrical relations duringthe final test phase.

In summary, Experiment 3 demonstratedthat once a matching-to-sample test perfor-

153INCONGRUOUS RELATIONAL RESPONDING

Table 4

Number of trials on which responses to the symmetryprobes were consistent with matching-to-sample trainingin Experiment 3.

Subjectno. andgender

Test trial types

B1A1 B2A2 C1B1 C2B2

Condition 131/M32/M33/M34/F

101019

101019

109

109

10101010

Condition 235/F36/F37/M38/M

10101010

10101010

9101010

10101010

mance had been established, it was highly re-sistant to change, even after repeated expo-sures to incongruous training and testing.Furthermore, in Condition 1, subjects consis-tently and repeatedly failed to respond in ac-cordance with conditional discriminationtraining on combined symmetrical and tran-sitive probes, even though responses on sym-metry probes were sensitive to the condition-al discrimination contingencies. This findingsupports previous research in which symme-try test performances were manipulated in-dependently of transitivity test performances(Pilgrim & Galizio, 1990, 1995).

EXPERIMENT 4

Experiment 3 demonstrated that repeatedexposure to incongruous contingencies doesnot enhance those contingencies as sourcesof control over test performances. One wayin which we might enhance the behavioralcontrol that is exerted by a particular set ofincongruous contingencies, however, is toprovide subjects with an early history of in-teraction with similar contingencies. In Ex-periment 4, therefore, some subjects were ex-posed to preliminary stimulus pairing andmatching-to-sample testing before being ex-posed to conditional discrimination trainingand testing with novel stimuli followed by in-congruous stimulus pairing and testing. Oth-er subjects were exposed to preliminary con-ditional discrimination training and testingbefore being exposed to stimulus pairing and

testing with novel stimuli followed by incon-gruous conditional discrimination trainingand testing.

Procedure

Conditions 1 and 2 of Experiment 4 cor-responded precisely with Conditions 1 and 3of Experiment 1 with the following differ-ence. Subjects in Condition 1 were first ex-posed to preliminary conditional discrimina-tion training and matching-to-sample testing.Six nonsense syllables, unique to this prelim-inary condition, were used as stimuli (X1, Y1,Z1, X2, Y2, and Z2). Subjects in Condition 2were first exposed to preliminary stimuluspairing and matching-to-sample testing, usingfour stimuli (X1, Z1, X2, and Z2) unique tothis preliminary phase. Preliminary stimuluspairing and conditional discrimination train-ing were administered in precisely the samemanner as subsequent training phases. Fourmales and 8 females participated in Experi-ment 4. The 12 subjects were distributedamong three conditions (Condition 1, Con-dition 2, and a control condition), with 4 sub-jects in each condition.

Condition 1. Before being exposed to ex-perimental training and testing, subjects werefirst trained on the conditional discrimina-tions X1-Y1, X2-Y2, Y1-Z1, and Y2-Z2, andwere subsequently exposed to matching-to-sample tests. Subjects were then exposed toexperimental stimulus pairing trials in whichA1 and C2 were paired with sexual stimuliand A2 and C1 were paired with nonsexualstimuli. Following training, the matching-to-sample test was administered. Subjects werethen trained on the incongruous conditionaldiscriminations A1-B1, A2-B2, B1-C1, andB2-C2, after which the matching-to-sampletest was readministered.

Condition 2. Before being exposed to ex-perimental training and testing, subjects werefirst exposed to 16 preliminary stimulus-pair-ing trials in which the nonsense syllables X1and Z1 were paired with sexual stimuli andthe nonsense syllables X2 and Z2 were pairedwith nonsexual stimuli. A matching-to-sampletest was then administered.

Subjects were then trained on the followingtwo conditional discriminations: A1-B1,A2-B2, B1-C1, and B2-C2. Following training,the matching-to-sample test was adminis-tered. Each subject was then oriented towards

154 BRYAN ROCHE et al.

Table 5

Number of trials on which responses were consistent with preliminary training and then initialexperimental training in Experiment 4.

Subjectno. andgender

No. ofMTStrain-ing

trialsTestno.

Test trial types

X1Z1 Z1X1 X2Z2 Z2X2 C1A2 A2C1 C2A1 A1C2 C1A1 A1C1 C2A2 A2C2

Condition 1 (preliminary MTS → incongruous pairing/MTS)39/F 20

36

123

10 10 10 9100

101

100

100

40/Fa 44

16

1234

89

79

710

89

100

100

100

100

41/M 56

20

123

10 10 10 10101

100

100

100

42/M 40

16

123

10 10 10 990

100

100

100

Condition 2 (preliminary pairing → incongruous MTS/pairing)43/F

24123

10 10 10 9100

100

90

100

44/F20

123

10 10 10 10100

100

100

100

45/F48

123

10 10 10 1090

100

100

100

46/M32

123

10 10 9 10101

100

100

100

a This subject failed to produce a stable performance (i.e., choose the same comparison at least nine times across10 exposures to each task) on the matching-to-sample test. The additional row of data represents performance on asubsequent exposure to the test phase without additional training.

the television monitor on which incongruousstimulus-pairing trials were presented (i.e.,A1 and C2 were paired with sexual stimuli,and A2 and C1 were paired with nonsexualstimuli). Finally, subjects were reexposed tothe matching-to-sample test.

Control subjects. As a control measure, thepreliminary training phases from Conditions1 and 2 were reversed for 4 additional sub-jects. Two subjects received preliminarymatching-to-sample training and testing be-fore being exposed to experimentalmatching-to-sample training and testing fol-lowed by incongruous stimulus pairing andtesting. Similarly, an additional 2 subjectswere first exposed to preliminary stimuluspairing and testing before being exposed toexperimental stimulus pairing and testing fol-

lowed by incongruous conditional discrimi-nation training and testing.

Results and Discussion

Subjects’ training and testing perfor-mances for both experimental conditions arepresented in Table 5. Control subjects’ per-formances are presented in Table 6.

Condition 1 and controls. All 4 subjects com-pleted preliminary matching-to-sample train-ing within 56 trials. Subsequent matching-to-sample test performances were consistentwith equivalence relations (i.e., subjectsmatched Z1 with X1 and Z2 with X2). Follow-ing stimulus pairing, matching-to-sample testperformances demonstrated control by thestimulus-pairing contingencies (i.e., subjectsmatched A1 with C2 and A2 with C1). Follow-

155INCONGRUOUS RELATIONAL RESPONDING

Table 6

Number of trials on which responses were consistent with preliminary training and then initialexperimental training for the control subjects in Experiment 4.

Subjectno. andgender

No. ofMTStrain-ing

trialsTestno.

Test trial types

X1Z1 Z1X1 X2Z2 Z2X2 C1A2 A2C1 C2A1 A1C2 C1A1 A1C1 C2A2 A2C2

Control Condition 1 (preliminary pairing → incongruous pairing/MTS)47/M

20

123

10 9 10 101010

1010

109

1010

48/F

52

123

10 10 10 109

101010

1010

1010

Control Condition 2 (preliminary MTS → incongruous MTS/pairing)49/F 28

60123

10 10 9 91010

1010

109

1010

50/F 4432

123

10 9 9 91010

109

109

1010

ing incongruous conditional discriminationtraining, the previous stimulus-pairing-basedtest performances were altered; subjectsmatched A1 with C1 and A2 with C2, thusdemonstrating control in accordance withequivalence relations. Control Subjects 47and 48 (see Table 6), who had been exposedto stimulus pairing as opposed to conditionaldiscrimination pretraining, showed no im-pact from the final conditional discrimina-tion training. Both control subjects repeatedtheir previous test performances based on ex-perimental stimulus pairing.

Condition 2 and controls. Following expo-sure to preliminary stimulus pairing, all sub-jects matched X1 with Z1 and X2 with Z2 onthe matching-to-sample test, thus demon-strating control by the stimulus-pairing con-tingencies. Following conditional discrimi-nation training, subjects responded inaccordance with equivalence relations onthe matching-to-sample test (i.e., theymatched A1 with C1 and A2 with C2). Final-ly, following exposure to incongruous stim-ulus pairing, all 4 subjects altered their pre-vious matching-to-sample test performancesby matching A1 with C2 and A2 with C1, thusdemonstrating control by the stimulus-pair-ing contingencies. Matching-to-sample per-formances of Control Subjects 49 and 50(Table 6), who had received preliminary

conditional discrimination training as op-posed to stimulus pairing, were not influ-enced by the incongruous stimulus pairing.Both control subjects repeated their previ-ous test performances based on experimen-tal conditional discrimination training.

In summary, Experiment 4 demonstratedthat preliminary training and testing ap-peared to sensitize novel matching-to-sampletest performances to conflicting contingen-cies when the contingencies are similar inkind to those of pretraining.

GENERAL DISCUSSION

Experiment 1 demonstrated that matching-to-sample test performances were deter-mined by the training and testing to whichsubjects were first exposed. Experiment 2demonstrated that a critical factor involved inproducing persistent control by the initialtraining phase was the administration of amatching-to-sample test following exposureto the initial training phase. Experiment 3 at-tempted to gain control over persistent testperformances without removing either of thetest phases. It was found that once a match-ing-to-sample test performance had beendemonstrated, it was highly resistant tochange, even following repeated exposures toincongruous contingencies. Furthermore,

156 BRYAN ROCHE et al.

Experiment 3, Condition 1, demonstratedthat even though symmetry relations basedon subsequent conditional discriminationtraining sometimes emerged, combined sym-metrical and transitive relations based on thesame conditional discriminations did not. Fi-nally, Experiment 4 indicated that preexpo-sure to conditional discrimination or stimu-lus-pairing contingencies increased thesensitivity of novel matching-to-sample testperformances to incongruous contingencies.

One possible criticism of the current studyis that the number of stimulus-pairing trials(i.e., 16) was relatively small and predeter-mined, whereas the number of conditionaldiscrimination training trials was sometimeslarge and was determined by the training per-formance of each subject. It could thereforebe argued, for example, that subjects fromExperiment 1, Condition 3, demonstrated aconditional-discrimination-based test perfor-mance following incongruous stimulus pair-ing simply because of the small number ofstimulus-pairing trials employed. It is impor-tant to note, however, that some of these sub-jects received as few as 16 conditional dis-crimination training trials. In actuality, thedifference between the number of trainingtrials across training phases (i.e., conditionaldiscrimination and stimulus pairing) was of-ten small. Furthermore, subjects from Exper-iment 1, Condition 1, maintained stimulus-pairing-based test performances after as manyas 108 incongruous conditional discrimina-tion training trials. This persistence suggeststhat the relative number of stimulus-pairingtrials versus conditional discrimination train-ing trials employed in the current study wasnot directly responsible for the observed per-formances.

It may also be tempting to account for thecurrent data in terms of the contextual con-trol exerted by the different experimental set-tings that were associated with conditionaldiscrimination and stimulus pairing. For ex-ample, subjects from Experiment 1, Condi-tion 3, were first exposed to conditional dis-crimination training and matching-to-sampletesting on a computer monitor. The subse-quent stimulus-pairing trials, however, weredelivered via a television monitor, and noovert response was required on the part ofthe subject. Thus, when the matching-to-sam-ple test was readministered on the computer

monitor following stimulus pairing, it is pos-sible that subjects may have repeated theirinitial test performance simply because theexperimental setting controlling it (i.e., thecomputer) had been reinstated.

Although the foregoing argument may ap-pear plausible at first, it cannot account forthe persistent stimulus-pairing-based test per-formances produced by subjects in Experi-ment 1, Condition 1. These subjects were firstexposed to stimulus pairing via a televisionmonitor, followed by matching-to-sample test-ing on a computer. Subjects were then ex-posed to incongruous conditional discrimi-nation training, again on a computer.Subsequent matching-to-sample test perfor-mances, however, were insensitive to thesenovel reinforcement contingencies. In effect,test performances on the computer were con-trolled by stimulus pairing that took place ina different context (i.e., the television moni-tor). Furthermore, similar results were alsoobtained in Experiment 3, Condition 1,across repeated exposures to incongruousconditional discrimination training. It ap-pears, therefore, that the television and com-puter monitors as contexts cannot, on theirown, account for the observed performances.

We should not, however, completely dis-miss the possibility of contextual control byspecific features of the experimental settingor procedure. Another source of contextualcontrol, for example, might be the matching-to-sample test arrangement itself. There arethree possibly important differences betweenthe matching-to-sample test and the otherphases of the experiment. First, during thematching-to-sample test, three stimuli werepresented in a trigram configuration on eachtrial, whereas during the stimulus-pairing pro-cedure film clips were shown with a singlenonsense syllable. Second, the matching-to-sample test required subjects to respond onthe computer keyboard, but during the stim-ulus-pairing procedure no keyboard responsewas required. Third, feedback was never pro-vided during the matching-to-sample test (orduring stimulus pairing), but matching-to-sample training always involved feedback.Perhaps one or a combination of these pro-cedural differences (i.e., the visual proper-ties, the response requirements, and the pres-ence or absence of feedback) functioned as

157INCONGRUOUS RELATIONAL RESPONDING

a contextual cue or cues that controlled finaltest performances in each experiment.

In Condition 1 of Experiment 1, for ex-ample, all 5 subjects responded in accor-dance with the initial stimulus-pairing proce-dure across both matching-to-sample tests. Inthis condition, the absence of feedback dur-ing the final exposure to the matching-to-sample test may have functioned as a contex-tual cue to repeat the performance that hadbeen produced during the previous exposureto this test during which feedback was alsoabsent. Note that the visual properties or re-sponse requirements could not have func-tioned as contextual cues in this conditionbecause the visual and response properties ofthe matching-to-sample training were identi-cal to those of the final test.

Consider now Condition 3. In this condi-tion, persistent responding in accordancewith the initial conditional discriminationtraining could not have been controlled sim-ply by the presence or absence of feedback.Feedback was absent not only in the initialtest following conditional discriminationtraining but also in the stimulus pairing andsubsequent matching-to-sample test. In Con-dition 3, therefore, contextual control wouldhave to have been provided by the visualproperties or response requirements thatwere common to both the matching-to-sam-ple training and testing procedures.

In summary, in Condition 1 the presenceor absence of feedback may have functionedas a contextual cue over final test perfor-mances, whereas in Condition 3 the visualproperties or response requirements mayhave fulfilled this role. At the present time,therefore, appealing to the context of the ex-perimental procedures as the basis for persis-tent performances raises an important ques-tion. Exactly what aspects of the differentexperimental conditions established specificfeatures of the procedure as contextual cues?In fact, this question is also raised by the re-sults of Experiment 4. Consider the follow-ing.

In contrast to Experiments 1 and 3, Ex-periment 4 showed that test performancescould be reversed when subjects had beenprovided with an early history of exposureto stimulus pairing or conditional discrimi-nation contingencies using different stimu-li. One possible explanation for this out-

come is that once a particular trainingprocedure (e.g., conditional discriminationtraining) had established control over testperformances, as in Experiments 1 and 3,those test performances were unlikely to bealtered by incongruous contingencies estab-lished by a formally different training pro-cedure (e.g., stimulus pairing). In Experi-ment 4, however, preliminary stimuluspairing, for example, may have establishedthe stimulus-pairing procedure itself as asource of contextual control for test perfor-mances. Consequently, the stimulus-pairingprocedure was then able to control later testperformances, even when the contingenciesestablished by that procedure were incon-gruous with other previously introducedcontingencies. In a similar manner, prelim-inary conditional discrimination trainingmay have established the conditional dis-crimination procedure itself as a source ofcontextual control for test performances.Again, however, it remains unclear exactlywhat features of the two training proce-dures may have been functioning as contex-tual cues (i.e., the visual properties, the re-sponse requirements, the presence orabsence of feedback, or some other featuresof the procedures). Clearly, therefore, fu-ture research will need to examine this is-sue in some detail.

Another issue arising from the presentstudy that deserves further attention relatesto the possibility that the stimulus-pairingprocedure actually functioned as a form ofequivalence training (see Barnes, Smeets, &Leader, 1996; Leader, Barnes, & Smeets,1996). During stimulus-pairing trials, non-sense syllables were presented before andduring the presentation of film clips; thus,the sexual film clips, for example, may haveacted as a node through which two non-sense syllables became related to one anoth-er.

Of course, one might dismiss the ideathat the stimulus-pairing procedure pro-duced equivalence relations as traditionallydefined, because the nonsense syllableswere flashed repeatedly on the screen dur-ing the film clips. In effect, the symmetricalrelations between the syllables and filmclips might be seen as directly trained (seeStromer, McIlvane, & Serna, 1993). At thepresent time, therefore, it remains unclear

158 BRYAN ROCHE et al.

exactly what implications the current studymight have for previous research that haseither found or failed to find sensitivity toreversed baseline conditional discrimina-tions using only matching-to-sample proce-dures (Dube, McIlvane, Maguire, Mackay, &Stoddard, 1989; Pilgrim & Galizio, 1990,1995; Pilgrim et al., 1995; Saunders et al.,1988; Spradlin, Cotter, & Baxley, 1973;Spradlin et al., 1992; Wilson & Hayes,1996). Parenthetically, however, future re-searchers might bear in mind that the com-mon practice of presenting sample andcomparison stimuli simultaneously duringconditional discrimination training may al-low subjects to scan back and forth betweensample and comparison stimuli, thus givingrise to directly trained symmetrical relations(for a more detailed analysis of this and re-lated issues, the reader is referred to Mark-ham & Dougher, 1993; Smeets, Schenk, &Barnes, 1994, 1995; Stromer et al., 1993).

Important theoretical issues are raised bythe results of the current study. Consider,for example, the fact that the symmetry testperformances of subjects in Experiment 3,Condition 1, were sensitive to the incongru-ous reinforcement contingencies, eventhough combined symmetry and transitivitytest performances were not. Consider alsothat the presence of symmetry probes in thefinal matching-to-sample test effected noobservable change in response patterns onprobes for combined symmetry and transi-tivity. These data support the view thatequivalence classes are not integrated units,but consist of more flexible relations (seediscussions by Hayes, Gifford, & Wilson,1996; Pilgrim & Galizio, 1995; Spradlin etal., 1992; Wilson & Hayes, 1996). In a sim-ilar vein, we and others have argued thatequivalence, and derived relational re-sponding in general, may be approachedusefully as a form of operant behavior(Barnes, 1994, 1996; Barnes & Roche, 1996,in press; Hayes et al., 1996; Wilson & Hayes,1996). In the words of Wilson and Hayes,‘‘As with other operant behavior, respond-ing relationally has properties of both flex-ibility and stability—properties that wouldbe expected if deriving stimulus relations isitself operant behavior’’ (p. 279). The re-sults of the present study provide furthersupport for an operant account, insofar as

it was shown that derived relational re-sponding may persist, or remain stable, un-der certain conditions (Experiments 1 and3), but can be modified, or made flexible,given other conditions (Experiment 4). Ofcourse, this complex theoretical issue is farfrom resolved, but a growing body of em-pirical findings appear to support an oper-ant analysis of derived relational respond-ing.

As a final point, the current data mayhave important implications for our under-standing of attitude formation and change.A number of researchers (e.g., Barnes &Holmes, 1991; Barnes, Lawlor, Smeets, &Roche, 1995; Barnes & Roche, in press;Grey & Barnes, 1996; Moxon et al., 1993;Roche & Barnes, 1996a; Watt et al., 1991)have argued that social attitudes may beconceptualized usefully in terms of derivedand explicitly reinforced stimulus relations(e.g., a negative attitude towards condomuse could be seen as responding in accor-dance with an equivalence relation betweenactual condoms and descriptive terms suchas ‘‘unromantic’’ and ‘‘disgusting’’). Insofaras this interpretation is correct, the datafrom Experiments 1 and 3 suggest that oncean attitude has been formed (e.g., an equiv-alence relation between ‘‘condoms’’ and‘‘unromantic’’) this relation may persist, de-spite the introduction of incongruous re-inforcement or stimulus-pairing contingen-cies (e.g., seeing an advertisement for anAIDS awareness campaign; see also Kelly etal., 1991, 1992; Wulfert & Biglan, 1994).The findings from Experiment 4, however,indicate that it is possible to increase thesensitivity of equivalence responding to in-congruous contingencies in a laboratorysetting. Perhaps further research in thisarea will provide a solid empirical base fordeveloping effective interventions forchanging dangerous or unhealthy attitudesin the wider community (see Barnes &Roche, in press, for a detailed discussion).

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

frame theory. The Psychological Record, 40, 91–124.Barnes, D. (1996). Naming as a technical term: Sacrific-

ing behavior analysis at the altar of popularity. Journalof the Experimental Analysis of Behavior, 65, 264–267.

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

159INCONGRUOUS RELATIONAL RESPONDING

(1995). A transfer of functions and a conditionaltransfer of functions through equivalence relations inthree to six year old children. The Psychological Record,45, 404–430.

Barnes, D., & Holmes, Y. (1991). Radical behaviorism,stimulus equivalence, and human cognition. The Psy-chological Record, 41, 19–31.

Barnes, D., & Keenan, M. (1993). A transfer of functionsthrough derived arbitrary and nonarbitrary stimulusrelations. Journal of the Experimental Analysis of Behavior,59, 61–81.

Barnes, D., Lawlor, H., Smeets, P. M., & Roche, B.(1995). Stimulus equivalence and academic self-con-cept among mildly mentally handicapped and non-handicapped children. The Psychological Record, 46, 87–107.

Barnes, D., McCullagh, P. M., & Keenan, M. (1990).Equivalence class formation in non-hearing impairedchildren and hearing impaired children. The Analysisof Verbal Behavior, 8, 19–30.

Barnes, D., & Roche, B. (1996). Relational frame theoryand stimulus equivalence are fundamentally different:A reply to Saunders’ commentary. The PsychologicalRecord, 46, 489–507.

Barnes, D., & Roche, B. (in press). Relational frame the-ory and the experimental analysis of human sexuality.Applied and Preventive Psychology.

Barnes, D., Smeets, P. M., & Leader, G. (1996). New pro-cedures for establishing emergent matching perfor-mances in children and adults: Implications for stim-ulus equivalence. In T. R. Zentall & P. M. Smeets(Eds.), Stimulus class formation in humans and animals:Advances in psychology (pp. 153–171). Amsterdam: El-sevier.

Bush, K. M., Sidman, M., & de Rose, T. (1989). Contex-tual control of emergent equivalence relations. Jour-nal of the Experimental Analysis of Behavior, 51, 29–45.

Cullinan, V., Barnes, D., Hampson, P. J., & Lyddy, F.(1994). A transfer of explicitly and non-explicitlytrained sequence responses through equivalencerelations: An experimental demonstration and aconnectionist model. The Psychological Record, 44,559–585.

Dube, W. V., McIlvane, W. J., Mackay, H. A., & Stoddard,L. T. (1987). Stimulus class membership establishedvia stimulus-reinforcer relations. Journal of the Experi-mental Analysis of Behavior, 47, 159–175.

Dube, W. V., McIlvane, W. J., Maguire, R. W., Mackay, H.,& Stoddard, L. T. (1989). Stimulus class formationand stimulus-reinforcer relations. Journal of the Experi-mental Analysis of Behavior, 51, 65–76.

Dymond, S., & Barnes, D. (1994). A transfer of self-dis-crimination response functions through equivalencerelations. Journal of the Experimental Analysis of Behavior,62, 251–267.

Fields, L., Adams, B. J., Verhave, T., & Newman, S.(1990). The effects of nodality on the formation ofequivalence classes. Journal of the Experimental Analysisof Behavior, 53, 345–358.

Grey, I. M., & Barnes, D. (1996). Stimulus equivalenceand attitudes. The Psychological Record, 46, 243–270.

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.

Kelly, J. A., St. Lawrence, J. S., Diaz, Y. E., Stevenson, L.Y., Hauth, A. C., Brasfield, T. L., Kalichman, S. C.,Smith, J. E., & Andrew, M. E. (1991). HIV risk be-havior reduction following intervention with key opin-ion leaders of a population: An experimental com-munity-level analysis. American Journal of Public Health,81, 168–171.

Kelly, J. A., St. Lawrence, J. S., Stevenson, L. Y., Hauth,A. C., Kalichman, S. C., Diaz, Y. E., Brasfield, T. L.,Koob, J. J., & Morganet, M. G. (1992). CommunityAIDS/HIV risk reduction: The effects of endorse-ments by popular people in three cities. American Jour-nal of Public Health, 82, 1483–1489.

Leader, G., Barnes, D., & Smeets, P. M. (1996). Estab-lishing equivalence relations using a respondent-typetraining procedure. The Psychological Record, 46, 685–706.

Markham, M. R., & Dougher, M. J. (1993). Compoundstimuli in emergent stimulus relations: Extending thescope of stimulus equivalence. Journal of the Experimen-tal Analysis of Behavior, 60, 529–542.

Moxon, P. D., Keenan, M., & Hine, L. (1993). Gender-role stereotyping and stimulus equivalence. The Psy-chological Record, 43, 381–393.

Pilgrim, C., Chambers, L., & Galizio, M. (1995). Reversalof baseline relations and stimulus equivalence: II.Children. Journal of the Experimental Analysis of Behavior,63, 239–254.

Pilgrim, C., & Galizio, M. (1990). Relations betweenbaseline contingencies and equivalence probe perfor-mances. Journal of the Experimental Analysis of Behavior,54, 213–224.

Pilgrim, C., & Galizio, M. (1995). Reversal of baselinerelations and stimulus equivalence: I. Adults. Journalof the Experimental Analysis of Behavior, 63, 225–238.

Roche, B., & Barnes, D. (1995). The establishmentand electrodermal assessment of conditioned sex-ual responses. The Experimental Analysis of HumanBehavior Bulletin, 13, 26–29.

Roche, B., & Barnes, D. (1996a). Arbitrarily applica-ble relational responding and sexual categoriza-tion: A critical test of the derived difference rela-tion. The Psychological Record, 46, 451–475.

Roche, B., & Barnes, D. (1996b). Sexual fetishism: Amodern experimental analogue. Clinical BehaviorAnalysis, 1, 2–4.

Roche, B., & Barnes, D. (1997). A transformation of res-pondently conditioned stimulus function in accor-dance with arbitrarily applicable relations. Journal ofthe Experimental Analysis of Behavior, 67, 275–301.

Saunders, R. R., Saunders, K. J., Kirby, K. C., & Spradlin,J. E. (1988). The merger and development of equiv-alence classes by unreinforced conditional selectionof comparison stimuli. Journal of the Experimental Anal-ysis of Behavior, 50, 145–162.

Sidman, M. (1990). Equivalence relations: Where dothey come from? In D. E. Blackman & H. Lejeune(Eds.), Behaviour analysis in theory and in practice: Con-tributions and controversies (pp. 93–114). Hove, En-gland: Erlbaum.

Sidman, M. (1992). Equivalence relations: Some basicconsiderations. In S. C. Hayes & L. J. Hayes (Eds.),Understanding verbal relations (pp. 15–27). Reno, NV:Context Press.

160 BRYAN ROCHE et al.

Smeets, P. M., Schenk, J. J., & Barnes, D. (1994). Es-tablishing transfer from identity to arbitrary match-ing tasks via complex stimuli under testing condi-tions: A follow-up study. The Psychological Record, 44,521–536.

Smeets, P. M., Schenk, J. J., & Barnes, D. (1995). Estab-lishing arbitrary stimulus classes via identity matchingtraining and non-reinforced matching with complexstimuli. Quarterly Journal of Experimental Psychology, 48B,311–328.

Spradlin, J. E., Cotter, V. W., & Baxley, N. (1973). Estab-lishing a conditional discrimination without directtraining: A study of transfer with retarded adolescents.American Journal of Mental Deficiency, 77, 556–566.

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.29–43). Reno, NV: Context Press.

Stromer, R., McIlvane, W. J., & Serna, R. W. (1993).Complex stimulus control and stimulus equivalence.The Psychological Record, 43, 585–598.

Watt, A., Keenan, M., Barnes, D., & Cairns, E. (1991).Social categorization and stimulus equivalence. ThePsychological Record, 41, 33–50.

Wilson, K., & Hayes, S. C. (1996). Resurgence of derivedstimulus relations. Journal of the Experimental Analysis ofBehavior, 66, 267–281.

Wulfert, E., & Biglan, A. (1994). A contextual approachto research on AIDS prevention. The Behavior Analyst,17, 353–363.

Wulfert, E., & Hayes, S. C. (1988). Transfer of a condi-tional ordering response through conditional equiv-alence classes. Journal of the Experimental Analysis of Be-havior, 50, 25–141.

Received May 8, 1995Final acceptance May 13, 1997

APPENDIXI consent that I am willing to participate in

this study. I am aware that as a requirementof this study, I will be exposed to film clipscontaining sexually explicit scenes. The filmsfrom which these clips are taken are widelyavailable in leading department stores andrecord shops in the Irish Republic. I have notbeen coerced in any way to participate in thisstudy and I understand that I may terminatemy participation in this study at any stage if Iso wish. I understand that my participation inthis study has no bearing upon grades for ac-ademic work in the department of AppliedPsychology or in University College Cork, ingeneral.