effects of stimulus variation on the reinforcing capability of nonpreferred stimuli

EFFECTS OF STIMULUS VARIATION ON THEREINFORCING CAPABILITY OF NONPREFERRED STIMULI

LEAH J. KOEHLER AND BRIAN A. IWATA

UNIVERSITY OF FLORIDA

EILEEN M. ROSCOE

NEW ENGLAND CENTER FOR CHILDREN

AND

NATALIE U. ROLIDER AND LAURA E. O’STEEN

UNIVERSITY OF FLORIDA

We examined the effects of stimulus (reinforcer) variation in several different contexts. In Study1, we identified high-quality (HQ) and low-quality (LQ) stimuli based on results of a paired-stimulus assessment and examined their effects when available under concurrent-reinforcementschedules for 8 participants. No participants showed preference for the LQ stimuli whencompared singly or in a varied arrangement to the HQ stimulus. In Study 2, we identifiednonpreferred (NP) stimuli based on results of a single-stimulus assessment and examined theireffects when available under single-reinforcement schedules for 3 participants. Results of Study 2were mixed. One participant’s data indicated that the varied presentation of NP stimuliproduced a modest improvement in performance over that observed when the stimuli werepresented singly. By contrast, a second participant’s data showed no facilitative effect for thevaried delivery of NP stimuli and that the inclusion of an HQ stimulus in the varied arrangementobscured the reinforcing effects of the HQ stimulus. The 3rd participant’s data showed no effectfor the varied delivery of NP stimuli but an apparent facilitative effect when an HQ stimulus wasincluded in the varied arrangement, which was attributable solely to the presence of the HQstimulus.

DESCRIPTORS: preference, reinforcer variation

_______________________________________________________________________________

The identification of reinforcers is animportant prerequisite to the development ofeffective behavioral contingencies, and a greatdeal of research has shown that empiricalmethods for assessing preference can be veryhelpful in selecting reinforcers for individualswith severe disabilities (Ivancic, 2000). Anumber of preference-assessment procedures

have been described in the literature; commonlyused methods include those in which approachresponses are observed to stimuli presentedsingly (Pace, Ivancic, Edwards, Iwata, & Page,1985), in pairs (Fisher et al., 1992), or inmultistimulus arrays (DeLeon & Iwata, 1996).

Empirical reinforcer identification does notnecessarily guarantee positive treatment out-comes because a number of factors maydetermine whether a reinforcement effect isproduced or maintained. One of these factors isrepeated exposure to the same reinforcer, whichmay result in performance decrements thatoften are attributed to satiation or habituation.Egel (1980, 1981) conducted a series of studiesshowing that this problem may be alleviated byvarying the stimuli that are delivered as re-inforcers. Egel (1980) observed that the pre-

This study was based on a thesis submitted to theUniversity of Florida Psychology Department in partialfulfillment of the requirements for the MS degree and wassupported in part by a grant from the Florida Departmentof Children and Families. We thank Sarah Bloom andNicole Hausman for their assistance in conducting thestudy.

Reprints may be obtained form Brian Iwata, PsychologyDepartment, University of Florida, Gainesville, Florida32611.

doi: 10.1901/jaba.2005.102-04

JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2005, 38, 469–484 NUMBER 4 (WINTER 2005)

469

sentation of varied reinforcers produced higherresponse rates than did the presentation of anyone of the reinforcers singly. He subsequentlyextended those findings to classroom perfor-mance (Egel, 1981).

An interesting implication of Egel’s (1980,1981) findings is that varied presentation of lesspreferred reinforcers might compete with a morepreferred reinforcer. Bowman, Piazza, Fisher,Hagopian, and Kogan (1997) addressed thisquestion by first establishing a hierarchy ofpreference for a number of stimuli. Theysubsequently compared, under a concurrent-schedule arrangement, the reinforcing effects ofthe highest ranked (high-quality or HQ)stimulus presented singly versus those rankedsecond, third, and fourth (slightly lower qualityor SLQ) presented in a varied format. Resultsshowed that 2 of the 7 participants preferred thesingle HQ reinforcer over the varied SLQreinforcers, 3 participants preferred the SLQreinforcers, and the remaining 2 participantsshowed switches in preference across sessions.Thus, the SLQ stimuli disrupted a strongpreference for the HQ stimulus in 5 of the 7participants (i.e., the 3 who preferred the SLQstimuli and the 2 who showed switches inpreference). In light of these somewhat mixedfindings and the fact that the rankings of theHQ and SLQ stimuli were very similar (in somecases, one or more of the SLQ stimuli hadpreference rankings identical to that of the HQstimulus), the extent to which variation pro-duces competition between less and morepreferred reinforcers remains unknown. Thepurpose of Study 1 was to replicate theBowman et al. procedures using HQ and low-quality (LQ) stimuli whose initial preferencerankings showed a greater degree of disparity.

A further implication of findings fromresearch on reinforcer variation is that theinclusion of an HQ stimulus as one of theconsequences in a varied array that containsnonpreferred stimuli might actually strengthenthe reinforcing effectiveness of stimuli that

otherwise would not function as reinforcers.We examined this possibility in Study 2 byevaluating the effects of LQ stimuli when usedas reinforcers alone, under a varied arrange-ment, and under a varied arrangement thatincluded one HQ stimulus.

STUDY 1

METHOD

Participants and Setting

Eight individuals with developmental dis-abilities participated. Josh was a 31-year-oldman who had been diagnosed with mild mentalretardation and cerebral palsy. Matt was a 24-year-old man who had been diagnosed withmild mental retardation. Julie was a 41-year-oldwoman who had been diagnosed with profoundmental retardation and Down syndrome. Billwas a 24-year-old man who had been diagnosedwith mild mental retardation. Sam was a 25-year-old man who had been diagnosed withmoderate mental retardation, autism, andDown syndrome. Tom was a 37-year-old manwho had been diagnosed with mild mentalretardation and Klinefelter’s syndrome. Peterwas a 27-year-old man who had been diagnosedwith moderate mental retardation. Jack wasa 37-year-old man who had been diagnosedwith mental retardation and seizure disorder.All participants were ambulatory and couldfollow some instructions and communicatevocally, except for Julie, who could communi-cate with limited gestures only.

Sessions were conducted at a vocational dayprogram that all participants attended. Re-inforcer assessment sessions were 10 min induration and were conducted three to fourtimes daily, 4 to 5 days per week.

Response Measurement and Reliability

During the preference assessments, observersrecorded, on a preprinted data sheet, theoccurrence of approach responses to one oftwo edible items. An approach response wasdefined as the participant reaching for or

470 LEAH J. KOEHLER et al.

picking up one edible item within 5 s of theverbal prompt ‘‘pick one.’’ Preference for eachstimulus was calculated by dividing the numberof trials on which a stimulus was approached bythe number of trials on which it was presented.An independent observer collected data on 57%(range, 50% to 85%) of sessions. Reliability wascalculated on a trial-by-trial basis by dividingthe number of agreements by the total numberof trials and multiplying by 100%. Meanreliability for the preference assessments was99% (range, 97% to 100%).

During reinforcer assessment sessions, ob-servers using laptop computers recorded data onthe frequency of target behaviors and thedelivery of reinforcers. An independent observercollected data on 34% (range, 29% to 41%) ofall sessions. Session time was divided intocontinuous 10-s intervals, and observers’ rec-ords were compared on an interval-by intervalbasis. Interobserver agreement was calculated bydividing the smaller number of responses ineach interval by the larger number, averagingthese fractions, and multiplying by 100%.Mean interobserver agreement was 98% (range,94% to 100%) for all responses (HQ, LQ,SLQ, and control). Reliability was also calcu-lated for the delivery of reinforcement andaveraged 95% (range, 91% to 97%).

Preference Assessment

A paired-stimulus preference assessment(Fisher et al., 1992) was conducted for eachparticipant. Sixteen edible items were selectedbased on results of informal interviews with staffmembers familiar with the participant. Theparticipant was allowed to sample each edibleitem prior to the start of the assessment. Oneach trial, two items were presented, and theparticipant was allowed to select and consumeone of the items (simultaneous approach toboth items was blocked). Each item was pairedtwice with every other item during theassessment, and pairs were presented in a semi-random order (no pair was presented twiceconsecutively). An HQ stimulus was defined as

one selected on at least 85% of the trials, an LQstimulus was defined as one selected on 30% orfewer trials, and an SLQ stimulus was defined asone that was ranked second, third, or fourth inthe preference assessment.

Experimental Conditions

Initial training. Each participant was taughtto perform a target task prior to the reinforcerassessment. Target behaviors were selected foreach participant based on three criteria: (a) Theresponse was discrete and easily measurable, (b)it consisted of or approximated a vocationaltask, and (c) it could be performed by theparticipant. If a vocational task that theparticipant could learn quickly was not identi-fied, a simple, arbitrary task (e.g., switchpressing) was selected. If the participant didnot emit the response following a modeledprompt, brief training (gestural and physicalprompting) was conducted until the participantdid so independently. During the first stage oftraining, occurrences of the target behavior werefollowed by the delivery of an edible item ona fixed-ratio (FR) 1 schedule. The item usedduring training had been selected on 40% to60% of the trials during the preferenceassessment and was different than those usedduring subsequent conditions. During Stage 2,a similar task (e.g., with materials of a differentcolor) was available concurrently but producedno consequences. The reinforcer to be deliveredwas placed on a plate behind one of the tasks,and its position was changed following everyfew responses. Training continued until theparticipant consistently performed the taskassociated with reinforcement. During subse-quent conditions (see below), three tasks(identical except for color) were concurrentlyavailable. Responses on either of two tasks (A orB) produced consequences (placed on platesbehind the tasks); responses on the third task(C) produced no consequences and served asthe control. The control task was included toidentify situations in which participants re-sponded randomly or failed to make discrimi-

STIMULUS VARIATION 471

nations. The response–stimulus pairings wererandomized across sessions. Responses on thecontrol task were rare, so these data are notpresented in the results but are available fromthe authors.

Single HQ versus single LQ stimuli (HQ vs.LQ1, LQ2, LQ3). The purpose of this condi-tion was to verify preference for the HQstimulus over each of the LQ stimuli. Thiscomparison was critical because if any of the LQstimuli competed with the HQ stimulus, anyvaried condition containing that LQ stimuluswould be unnecessary. In three sequentialcomparisons, the reinforcing efficacy of the HQstimulus was compared with that of each LQstimulus. The stimuli were available in a concur-rent FR1 FR1 arrangement: Each response to theHQ task produced access to the HQ stimulus;each response to the LQ task produced access tothe LQ stimulus, and each response to thecontrol task produced no consequences. Oneplate was situated behind each task and containedthe edible items associated with that task. Theparticipant was prompted to engage in the targetresponse associated with the HQ stimulus, theLQ stimulus, and the control prior to the start ofeach session.

Single HQ versus VLQ stimuli (HQ vs. VLQ).The purpose of this condition was to determinewhether the LQ stimuli, when presented ina varied format, would compete with the singleHQ stimulus. Procedures were similar to thosein the previous condition, except that eachresponse to the LQ task produced access to oneof the three LQ stimuli. All of the varied itemswere visible to the participant throughout thesession, and the order of stimulus delivery wasfixed within a given session but randomizedacross sessions. The participant was promptedto engage in each target response three timesprior to a session to establish contact with eachavailable reinforcer.

Single HQ versus varied SLQ stimuli (HQ vs.VSLQ). The purpose of this condition was todetermine whether the SLQ stimuli, when

presented in a varied format, would competewith the single HQ stimulus. Procedures weresimilar to those in the previous condition,except that each response to the SLQ taskproduced access to one of the SLQ stimuli.

Experimental Design

These conditions were arranged in sequentialorder. Within each condition, a concurrentschedule was used to assess preference for one ofthe reinforcer options.

RESULTS AND DISCUSSION

Figure 1 shows results of the preferenceassessments. The item selected as the HQstimulus was chosen most frequently by eachparticipant and on 100% of the trials by mostparticipants. The only exception to this wasTom: His highest ranked item (CuddlefishPeanut) was unavailable for purchase when webegan his reinforcer assessment, so we used hissecond-ranked stimulus as the HQ stimulus.The three items ranked just below the HQstimulus were selected as SLQ stimuli. Finally,the three lowest ranked items were selected asLQ stimuli.

Figure 2 shows results from the reinforcerassessments obtained for Josh (folding socks),Matt (copying words from note cards), Julie(operating a hole puncher), and Bill (copyingwords from note cards). When the HQ stimuluswas compared to each of the three LQ stimuliindividually, all 4 participants allocated moreresponding to the task associated with the HQstimulus. In most cases, participants never madeany responses to the task associated with the LQstimulus (Matt was the exception). When theHQ stimulus was compared with the three LQstimuli presented in a varied format, Joshshowed frequent switches between the HQand VLQ tasks. By contrast, Matt, Julie, andBill continued to show preference for the HQtask; they never responded on the VLQ task.When VSLQ stimuli were substituted for VLQstimuli, all participants allocated more respond-ing to the VSLQ task. Josh and Matt showed


exclusive preference for the VSLQ task; Julieand Bill switched periodically between the HQand VSLQ tasks but showed a general prefer-ence for the SLQ task.

Figure 3 shows results from the reinforcerassessments obtained for Sam (pressing a but-ton), Tom (stuffing envelopes), Peter (foldingsocks), and Jack (punching holes in cards and

Figure 1. Results of paired-stimulus preference assessments. Items designated as high quality (HQ), slightly lowerquality (SLQ), and low quality (LQ) were used in Study 1.


Figure 2. Results of reinforcer assessments for Josh, Matt, Julie, and Bill.


Figure 3. Results of reinforcer assessments for Sam, Tom, Peter, and Jack.


putting them in envelopes). All 4 participants

showed preference for the HQ task over the LQ

task when any one of the LQ stimuli was

presented alone and when all three LQ stimuli

were presented in a varied format. During the

final condition, when the HQ stimulus was

compared with VSLQ stimuli, all participants

showed switching between the two response

options, although Jack’s performance tended to

favor the HQ option. On 16 of 20 sessions in

Jack’s HQ versus VSLQ condition, he allocated

more responding to the HQ task.In summary, all participants showed prefer-

ence for the HQ stimulus over each of the singleLQ stimuli when both were concurrentlyavailable, which was a prerequisite for sub-sequent comparisons. When the reinforcingeffects of the HQ stimulus were compared tothose of VLQ stimuli, none of the participantsshowed preference for the VLQ consequences.Josh was the only participant whose perfor-mance showed some switching between the HQand VLQ conditions. The consistency of theseresults indicates that stimulus variation does notappear to strengthen the reinforcing effects oflow-preference stimuli when highly preferredstimuli are available concurrently.

By contrast, when the reinforcing effects ofthe HQ stimulus were compared to those ofvaried stimuli whose initial preference rankingswere slightly lower than that of the HQstimulus (VSLQ), half of the participants (Josh,Matt, Julie, and Bill) showed preference for thevaried condition; the remaining participants(Sam, Tom, Peter, and Jack) switched betweenreinforcement options (i.e., they showed noclear preference). These results were similar tothose reported by Bowman et al. (1997), whoobserved that 3 of 7 participants preferred theVSLQ condition. It is unclear, however,whether the shift in preference observed forsome of the participants in our study, as well asin the Bowman et al. study, was a function ofstimulus variation per se. Because neither studyincluded a condition in which the HQ stimulus

was compared to single SLQ stimuli (i.e.,analogous to the first three conditions of ourreinforcer assessment), the relative reinforcingeffects of stimuli with similar preference rank-ings were unknown. Thus, it is possible that,under a concurrent arrangement, some partic-ipants would have shown preference for a singleSLQ stimulus over the HQ stimulus or, at thevery least, some switching across sessions.

Two additional features of the data warrantdiscussion. First, the two general patterns ofresponding (preference for HQ during the HQvs. VLQ condition; switching during the HQvs. VSLQ condition) were not observed forevery participant (see Josh and Matt), and it islikely that these patterns were influenced byeither the specific stimuli or the number ofstimuli included in the initial preferenceassessments. In other words, although it seemedreasonable to define preference (HQ, SLQ, LQ)based on rankings, a given ranking may not bepredictive of a reinforcement effect (Roscoe,Iwata, & Kahng, 1999). Depending on whichand how many stimuli are included in a prefer-ence assessment, it is possible that few stimuli,all stimuli, or all stimuli above a given rankwould function as reinforcers. Thus, thedifference in results observed between theVLQ stimuli and VSLQ stimuli (when com-pared to the HQ stimulus) must be interpretedwith caution.

Second, although the HQ versus VSLQcondition yielded switches in preference for 5participants, most of the switching occurredacross sessions; by contrast, most participantsconsistently chose either the HQ stimulus orthe VSLQ stimuli (i.e., they did not switch)within a given session. Although positioning ofthe reinforcers changed across sessions, none ofthe participants was observed to show a positionbias. The effects of satiation do not seem toaccount for this effect unless satiation to onereinforcer (but not another) occurred acrosssessions but not within a session. The findingthat subjects often respond exclusively toward


one option under concurrent FR schedules(Herrnstein & Loveland, 1975) also seemsunlikely as an explanation because it assumesthat one of the schedules was more favorable,and it is not clear how one schedule would bemore favorable on some sessions but not onothers. It is, of course, possible that othervariables, such as the type or amount of foodconsumed outside the session, may influenceeither satiation or the favorability of a givenreinforcer on a given day (e.g., North & Iwata,2005). In the present study, access to the stimuliused as reinforcers was controlled outside thesessions, and sessions were conducted atconsistent times relative to meals, but typeand amount of food were otherwise uncon-trolled.

STUDY 2

Study 1 examined the effects of stimulusvariation in the context of reinforcer competi-tion under a concurrent-schedule (HQ vs.VLQ) arrangement. In applied situations,however, therapists and educators may be moreconcerned with whether stimulus variationenhances the effects of reinforcement whenthey attempt to establish or maintain a par-ticular (single) performance, and the concurrentschedule used in Study 1 does not address thisissue. Using a single-schedule arrangement, Egel(1980, 1981) showed that stimulus variationproduced better results than those obtainedwith the same reinforcers delivered singly;however, the single-stimulus conditions initiallyshowed positive results. Thus, all the stimulifunctioned as reinforcers. As an attempt toextend Egel’s research, in Study 2 we sought todetermine whether stimulus variation mightstrengthen the reinforcing effects of stimuli thatdo not function initially as reinforcers. If so,therapists might actually be able to expand therange of reinforcers by varying the delivery ofa highly preferred consequence with the deliveryof consequences for which preference is negli-gible.

METHOD

Participants and SettingThree participants with developmental dis-

abilities participated. Ed was a 47-year-old manwho had been diagnosed with moderate mentalretardation and autism. Dan was a 27-year-oldman who had been diagnosed with moderatemental retardation. Brian was a 44-year-oldman who had been diagnosed with mild mentalretardation. All participants were ambulatory,could follow instructions, and communicatedvocally. The setting and session schedule werethe same as in Study 1.

Response Measurement and ReliabilityDuring the preference assessments, observers

recorded the occurrence of approach responsesas in Study 1. An approach response wasdefined as the participant reaching for orpicking one edible item within 5 s of the verbalprompt ‘‘pick one’’ (paired-stimulus assess-ment) or picking and then consuming theedible item being placed in front of theparticipant (single-stimulus assessment). Anindependent observer collected data on 33%(range, 0% to 100%) of trials during the paired-stimulus preference assessment and on 100% ofthe trials during the single-stimulus preferenceassessments. Reliability was calculated in thesame manner as in Study 1 and was 100%.

During reinforcer assessment sessions, ob-servers using laptop computers collected data onthe frequency of target behaviors and thedelivery of reinforcers. An independent observercollected data during 37% (range, 30% to 44%)of all sessions. Reliability was calculated in thesame manner as in Study 1. Mean interobserveragreement was 98% (range, 85% to 100%) forthe target response and 99% (range, 85% to100%) for the delivery of reinforcers.

Preference AssessmentA paired-stimulus assessment was conducted

to identify an HQ stimulus using the samecriterion that was used in Study 1. A single-stimulus preference assessment (Pace et al.,


1985) was conducted to identify nonpreferred(NP) stimuli. The rationale for using thisassessment was that stimuli not selected ina paired-stimulus format may neverthelessfunction adequately as reinforcers (Roscoe etal., 1999); as a result, we felt that nonapproachto a stimulus presented singly would yield morevalid information about lack of preference. Eachparticipant was allowed to sample each itemprior to the start of the assessment. Eachstimulus was placed singly in front of theparticipant for 5 s; contingent on an approachresponse, the participant was allowed toconsume the item. If no approach responseoccurred within 5 s, the item was removed anda new trial began. Five trials were conductedwith each stimulus, and three stimuli that werenever consumed were used as the NP items.

Experimental Conditions

Target behaviors during the reinforcer assess-ment were selected for each participant usingcriteria similar to those used in Study 1. Ifnecessary, brief training on the target responseoccurred until the participant emitted theresponse independently.

Baseline. The purpose of this condition wasto determine whether the participant wouldengage in the response in the absence of anycontingencies. No consequences were deliveredfor engaging in the target response.

Nonpreferred stimuli (NP). The purpose ofthis condition was to determine whether the NPitems would function as reinforcers. Contingenton the target response, one NP item wasdelivered on an FR 1 schedule. Each item wasassessed separately in a session, and sessionscontinued until a stable trend was established.Participants were prompted to engage in thetarget response prior to the start of the sessionto establish contact with the contingency.

Varied nonpreferred stimuli (VNP). If no orlittle responding was observed in the previouscondition, the three NP stimuli were presented ina varied manner throughout the session. One NPedible item was delivered contingent on the target

response on an FR 1 schedule; the order ofdelivering the three NP items was constant withina session but was randomized across sessions. Theparticipant was prompted to engage in the targetresponse three times prior to the session.

VNP + HQ. This condition was identical tothe previous one, except that the HQ stimuluswas added to the arrangement. One of the fouredible items (one HQ and three NP) wasdelivered contingent on the target response(again, order was constant within a session butwas randomized across sessions). The partici-pant was prompted to engage in the targetresponse four times prior to session.

HQ alone. Participants were exposed to oneof two types of conditions in which the HQstimulus was delivered alone, depending ontheir performance in previous conditions. First,if no increase in responding had been observedduring any previous condition (NP, VNP, VNP+ HQ), the HQ stimulus was delivered on anFR 1 schedule simply to determine whethera reinforcement effect, which was absent to thatpoint, could be obtained. Second, if respondingin the VNP + HQ condition was higher thanin the VNP condition, it was possible thatimproved performance was solely a function ofthe HQ stimulus. To control for this possibility,the HQ stimulus was delivered on an FR 4schedule, which matched the schedule of HQdelivery in the VNP + HQ condition. Theparticipant was prompted to engage in thetarget response either once (FR 1) or four times(FR 4) prior to session.

Experimental DesignThe conditions were presented in the

following order: baseline, NP, VNP, VNP +HQ, and HQ alone, although some modifica-tions were made in the sequence based onperformance. A reversal design was used todemonstrate experimental control.

RESULTS AND DISCUSSION

Figure 4 shows results from the twopreference assessments. The item approached


Figure 4. Results of paired-stimulus and single-stimulus preference assessments. Items designated as HQ and

nonpreferred (NP) were used in Study 2.


most frequently in the paired-stimulus assess-ment was designated as the HQ stimulus; threeitems that were never approached in the single-stimulus assessment were designated as NPstimuli.

Figure 5 shows results from the reinforcerassessments. Ed never engaged in the targetresponse during baseline or during any of thesubsequent conditions in which the NP stimuliwere present (NP, VNP, and VNP + HQ).Thus, it appeared that none of the stimuli,including the HQ item, were reinforcers. Wethen implemented a condition in which the HQstimulus was delivered on an FR 1 schedule andobserved an immediate increase in responding.A return to the VNP + HQ condition resultedin a return to zero responding. These resultssuggest that the NP items not only failed toserve as reinforcers but also obscured thereinforcing effects of the HQ item. In Ed’scase, variation of NP and HQ items had anoverall detrimental effect on performance.

Dan exhibited no responses during baseline,NP, and VNP conditions. His respondingincreased immediately during the subsequentVNP + HQ condition. We therefore imple-mented the HQ (FR 4) condition and observedno change in response rates. When the VNP +HQ condition was reinstated, Dan’s perfor-mance was variable initially. He exhibited noresponses during two sessions but subsequentlyresponded at rates similar to the two previousconditions. His responding was maintainedduring the final HQ (FR 4) condition. Dan’sresults differed from Ed’s in that the inclusionof a highly preferred item with nonpreferreditems in a varied arrangement (VNP + HQ)appeared to facilitate performance. However,when these elements were presented separately,Dan did not respond during the VNP condi-tion, whereas his response rates during the HQ(FR 4) condition were similar to those observedduring the VNP + HQ condition. In addition,he consumed only one of the NP items (capers)throughout the VNP + HQ condition. Thus, it

seems likely that Dan’s responding throughoutthe study can be attributed to the presence ofthe HQ item.

Brian responded sporadically during thebaseline and NP conditions. His respondingalso was low during the VNP condition;however, unlike his response pattern in previousconditions, he exhibited some responses duringevery session. During the VNP + HQ condi-tion, Brian’s responding increased noticeablybut remained somewhat variable, and continuedin a similar manner (although at a slightly lowerrate) during a return to the VNP condition.Given the modest improvement observed in hisperformance during the VNP + HQ conditionand the fact that his performance also showeda slight improvement during the VNP condi-tion relative to the NP condition, we chose toreexamine the NP and VNP conditions ratherthan to implement the HQ (FR 4) condition.Brian’s responding decreased to zero during thereturn to baseline, increased slightly during theNP condition, and increased further during thefinal VNP condition. His results may illustratetwo effects related to stimulus variation. First,the inclusion of an HQ stimulus may haveenhanced the reinforcing effects of nonpreferredstimuli in the VNP + HQ condition, althoughthe extent to which observed increases inresponding were solely a function of the HQstimulus was unknown because we did notimplement the HQ (FR 4) condition. Second,the VNP condition resulted in a modestimprovement in performance over baselineand NP conditions.

GENERAL DISCUSSION

We examined several characteristics of stim-ulus variation in the context of reinforcementfacilitation under conditions of concurrent andsingle reinforcement. In Study 1, 8 participantswere exposed to conditions in which thereinforcing effects of a stimulus ranked at thetop of a preference hierarchy (HQ) werecompared with those of varied stimuli ranked


Figure 5. Results of reinforcer assessments for Ed, Dan, and Brian.


at the bottom (VLQ) under a concurrent-reinforcement arrangement. Seven of the 8participants showed exclusive preference for thesingle HQ stimulus; 1 participant (Josh)switched between reinforcement options. Whenthe HQ stimulus was compared with variedstimuli ranked second, third, and fourth in thepreference hierarchy (VSLQ), half the partici-pants showed preference for the VSLQ stimuli;the other half switched preference acrosssessions. These data suggest that variation doeslittle to enhance the effects of NP stimuli,although this conclusion requires several quali-fications.

First, it is not clear that competition betweenthe VSLQ stimuli and the HQ stimulus can beattributed to variation per se because, aspreviously noted, the single presentation ofany one (or maybe all) of the SLQ stimuli couldhave produced similar results. We did notconduct HQ versus single SLQ comparisonsbecause our main interest was in the effects ofLQ stimuli. When we observed that the VLQcondition produced negligible results, we in-cluded the VSLQ condition to more closelyapproximate conditions of the Bowman et al.(1997) study and observed similar effects. Thus,results of the concurrent-schedule comparisonsshowed that stimulus variation is unlikely toimprove competition between stimuli rankedhigh and low on a preference hierarchy; theeffects of variation using more similarly rankedstimuli is less clear. Results of the HQ versusVSLQ comparison do show, however, thatstimulus variation may allow SLQ stimuli tocompete effectively with one HQ stimuluswhen both options are simultaneously available.

Second, concurrent presentation of stimuliwhose preference is established via paired-stimulus assessments is likely to produce false-negative outcomes. For example, Roscoe et al.(1999) showed that, although stimuli neverselected in a paired-stimulus assessment did notcompete with highly preferred stimuli undera concurrent-schedule arrangement, the same

stimuli functioned as reinforcers under a single-schedule arrangement. Therefore, negative con-clusions about the facilitative effects of stimulusvariation with NP stimuli in Study 1 must belimited to situations that involve reinforcercompetition. In most applied settings, however,therapists would want to know whether stim-ulus variation facilitates any reinforcementeffect. Results of Egel’s (1980, 1981) studiesindicated that variation enhances the reinfor-cing effects of stimuli that already function asreinforcers; the unknown implication of theseresults, as well as those from Study 1, is whethervariation might produce a reinforcement effectwith stimuli that otherwise do not function asreinforcers.

We therefore used a different methodology toevaluate stimulus variation in Study 2. NPstimuli were selected based on the absence ofapproach during single-stimulus preferenceassessments, because results of previous research(Green et al., 1988; Pace et al., 1985) showedthat such stimuli do not function as reinforcers;this was verified during the single NP condi-tions, in which 2 participants (Ed and Dan)never responded, and 1 participant (Brian)responded at rates similar to those observedduring baseline. When the VNP condition wasimplemented, Ed and Dan continued to exhibitno responses, whereas Brian’s responding in-creased, although only slightly. Thus, as mightbe expected, varying the delivery of nonpre-ferred stimuli appeared to have negligible effectson behavior. We then evaluated the effects ofincluding the HQ stimulus as one of the itemsin the varied arrangement (VNP + HQ). Edcontinued to exhibit no responses, and hisperformance during the subsequent HQ condi-tion suggested that the inclusion of VNP itemsin the VNP + HQ condition made his behaviorless sensitive to the effects of the HQ stimulus.In other words, varying NP and HQ itemsproduced detrimental results. One interpreta-tion of Ed’s results is that the VNP + HQcondition was functionally the equivalent of


delivering the HQ stimulus on an FR 4schedule that did not maintain responding(i.e., extinction occurred). Another interpreta-tion is that delivery of the NP items served aspunishment for responding. By contrast, Dan’sresponding increased immediately under theVNP + HQ condition, suggesting a facilitativeeffect. However, it appeared that this effectcould be attributed solely to the presence of theHQ stimulus because Dan rarely consumed theNP items, and his responding was maintainedwhen the VNP items were eliminated in theHQ (FR 4) condition. Finally, Brian’s respond-ing during the VNP + HQ condition, althoughhigher than during previous conditions, seemedmodest at best and prompted us to reexaminethe VNP arrangement. When the VNP condi-tion was reimplemented, Brian’s rate of re-sponding was higher than during two previousbaselines and two NP conditions and wassimilar to that observed during the VNP +HQ condition. Thus, although it was anunusual finding, the varied delivery of NPstimuli seemed to increase the reinforcingeffects of those stimuli on Brian’s behavior.

The applied relevance of Brian’s data may belimited because it is unlikely that therapistswould deliver as reinforcers those stimuli towhich approach had never been observed.However, Ivancic (2000) has noted that situa-tions may arise in which individuals approachfew, if any, stimuli in a preference assessment.In such situations, varying the delivery of anyavailable stimuli, or including a single preferredstimulus in a varied arrangement with thosestimuli, expands the range of available re-inforcement options.

In summary, results of the present studiesindicate that stimulus variation has little effecton the reinforcing capability of NP stimulirelative to that of a highly preferred stimuluswhen both alternatives are available. Differentresults, however, might be obtained withstimulus variation in the context of a single-task arrangement. Brian’s data suggested that

the varied delivery of NP stimuli or theinclusion of a preferred stimulus in a variedarrangement with NP stimuli may producemodest improvements in performance over thatobserved when NP stimuli are delivered singly.Furthermore, several arrangements not exam-ined in Study 2 might yield even clearer results.For example, Brian’s VNP condition wasconducted for relatively few sessions; it ispossible that longer exposure to this conditionwould be associated with further increases inresponding. In addition, a more optimalarrangement for increasing the effectiveness ofNP stimuli might consist of pairing them singlywith a highly preferred stimulus in an attemptto establish the former as a reinforcer (e.g., seeHanley, Iwata, Roscoe, Thompson, & Lind-berg, 2003). Finally, a less optimistic conclusionabout the effects of stimulus variation wasrevealed in Ed’s data, which indicated thatvarying nonpreferred and preferred stimuli asconsequences may occasionally produce un-desirable outcomes.

REFERENCES

Bowman, L. G., Piazza, C. C., Fisher, W. W., Hagopian,L. P., & Kogan, J. S. (1997). Assessment of preferencefor varied versus constant reinforcers. Journal ofApplied Behavior Analysis, 30, 451–458.

DeLeon, I. G., & Iwata, B. A. (1996). Evaluation ofa multiple-stimulus presentation format for assessingreinforcer preferences. Journal of Applied BehaviorAnalysis, 29, 519–533.

Egel, A. L. (1980). The effects of constant vs. variedreinforcer presentation on responding by autisticchildren. Journal of Experimental Child Psychology,30, 455–463.

Egel, A. L. (1981). Reinforcer variation: Implications formotivating developmentally disabled children. Journalof Applied Behavior Analysis, 14, 345–350.

Fisher, W. W., Piazza, C. C., Bowman, L. G., Hagopian,L. P., Owens, J. C., & Slevin, I. (1992). Acomparison of two approaches for identifying re-inforcers for persons with severe to profounddisabilities. Journal of Applied Behavior Analysis, 25,491–498.

Green, C. W., Reid, D. H., White, L. K., Halford, R. C.,Brittain, D. P., & Gardner, S. M. (1988). Identifyingreinforcers for persons with profound handicaps: Staffopinion versus systematic assessment of preferences.Journal of Applied Behavior Analysis, 21, 31–43.


Hanley, G. P., Iwata, B. A., Roscoe, E. M., Thompson, R.H., & Lindberg, J. S. (2003). Response-restrictionanalysis: II. Alteration of activity preferences. Journalof Applied Behavior Analysis, 36, 59–76.

Herrnstein, R. J., & Loveland, D. H. (1975). Maximizingand matching on concurrent ratio schedules. Journalof the Experimental Analysis of Behavior, 24, 107–116.

Ivancic, M. T. (2000). Stimulus preference and reinforcerassessment applications. In J. Austin, & J. E. Carr(Eds.), Handbook of applied behavior analysis (pp. 19–38). Reno, NV: Context Press.

North, S. T., & Iwata, B. A. (2005). Motivationalinfluences on performance maintained by foodreinforcement. Journal of Applied Behavior Analysis,38, 317–333.

Pace, G. M., Ivancic, M. T., Edwards, G. L., Iwata, B. A.,& Page, T. J. (1985). Assessment of stimuluspreference and reinforcer value with profoundlyretarded individuals. Journal of Applied BehaviorAnalysis, 18, 249–255.

Roscoe, E. M., Iwata, B. A., & Kahng, S. (1999). Relativeversus absolute reinforcement effects: Implications forpreference assessments. Journal of Applied BehaviorAnalysis, 32, 479–493.

Received July 21, 2004Final acceptance June 30, 2005Action Editor, Iser DeLeon


effects of stimulus variation on the reinforcing capability of nonpreferred stimuli

Documents