Comparing Different Classes of Reinforcement to Increase ...

Comparing Different Classes of Reinforcement to IncreaseExpressive Language for Individuals with Autism

Justin B. Leaf, Stephanie Dale, Alyne Kassardjian, Kathleen H. Tsuji,Mitchell Taubman, John J. McEachin, and Ronald B. Leaf

Autism Partnership Foundation

Misty L. Oppenheim-LeafBehavior Therapy and Learning Center

Abstract: One of the basic principles of applied behavior analysis is that behavior change is largely due to thatbehavior being reinforced. Therefore the use of positive reinforcement is a key component of most behavioralprograms for individuals diagnosed with autism. In this study we compared four different classes of reinforcers(i.e., food, praise, toys, and feedback) on the rate of skill acquisition for three individuals diagnosed withautistic disorder. Using a parallel treatments design, the results of this study showed that all four classes ofreinforcement were effective for teaching all of the participants. Results also indicated that food reinforecers weremore efficient and more preferred. Finally, participants performance during teaching was idiosyncratic acrossthe three participants.

One of the basic principles of Applied Behav-ior Analysis (ABA) is that behavior change canbe instigated through the use of positive rein-forcement (e.g., Catania, 1998). For individu-als diagnosed with an Autism Spectrum Disor-der (ASD), researchers have demonstratedreinforcement-based procedures to be effec-tive in decreasing self-injurious behaviors(e.g., Worsdell, Iwata, Hanley, Thompson, &Khang, 2000), aggression (e.g., Repp & Deitz,1974), and self-stimulatory behaviors (e.g.,Singh, Dawson, & Manning, 1981). Addition-ally, researchers have shown that reinforce-ment can be used to increase pro-social skillssuch as: social skills (e.g., Leaf, Dotson, Op-penheim, Sheldon, & Sherman, 2010), lan-guage (e.g., Petursdottir, Carr, Lechago, &Almason, 2008), and academic and pre-aca-demic tasks (e.g., Leaf, Sheldon, & Sherman,2010).

Positive reinforcers can take many formsincluding food (e.g., Schreibman, 1975), toys(e.g., Leaf, Sheldon, & Sherman, 2010), praise(e.g., Schreibman, 1975), tokens (e.g., Allyon& Azrin, 1968), and even allowing the learner

to engage in self-stimulatory behaviors (e.g.,Rincover & Newsom, 1985). Since reinforce-ment-based procedures have been highly ef-fective in changing behavior for people diag-nosed with ASD, researchers have conductednumerous studies investigating various aspectsof reinforcement. One area of interest hasbeen studying various methods for determin-ing learners preference of various stimuli andwhether these preference assessment resultsare predictive of reinforcer effectiveness. Re-searchers have evaluated single-item presenta-tion (e.g., Pace, Ivancic, Edwards, Iwata, &Page, 1985), paired preference assessment(e.g., Fisher et al., 1992), and multiple stimu-lus preference assessment (e.g., Carr, Nicol-son, & Higbee, 2000) for evaluating learnerpreference among stimuli and the correlationof preference level with reinforcer effective-ness. The research has indicated that all threeassessments are effective in determining bothpreference and reinforcement effectivenessand that these formal assessments are moreaccurate than asking teachers or parents(Green et al., 1988).

A second area that has been evaluated inthe reinforcement literature is determining ifchildren prefer contingent or non-contingentreinforcement (e.g., Luczynski & Hanley,

Correspondence concerning this article shouldbe addressed to Justin B. Leaf., 200 Marina Drive,Seal Beach CA 90740. E-mail: [email protected]

Education and Training in Autism and Developmental Disabilities, 2014, 49(4), 533–546© Division on Autism and Developmental Disabilities

Comparing Reinforcement Classes / 533

2009). Luczynski and Hanley (2009) evaluatedeight typically developing children’s relativepreference of contingent reinforcement aspart of a differential reinforcement of alterna-tive behavior (DRA) procedure versus non-contingent reinforcement to increase socialinteraction. Results indicated that 7 of the 8participants preferred contingent reinforce-ment. Although these findings are important,they were conducted with typically developingchildren; therefore, it is not known if findingswould be similar for children with ASD.

Additional areas of focus in the reinforce-ment literature include how the amount(magnitude) of reinforcement affects behav-ior change (e.g., Roscoe, Iwata, & Rand,2003), the effects of various reinforcementschedules (e.g., Worsdell et al., 2000) on be-havior change, the differences between posi-tive and negative reinforcement on behaviorchange (e.g., DeLeon, Neidert, Anders, &Rodriquez-Catter, 2001), and how learners’choice of reinforcement influences behaviorchange (e.g., Smith, Iwata, & Shore, 1995;Tiger, Hanley, & Hernandez, 2006). Despitethe amount of research on different compo-nents of reinforcement, very few studies haveevaluated how different classes of reinforcers(e.g., food, toys, or social) can affect skill ac-quisition.

Ferrari and Harris (1981) evaluated threedifferent classes of reinforcement: sensory re-inforcers (i.e., vibration, music, and strobelight), social reinforcement, and food rein-forcement on four children diagnosed withASD. The first task evaluated was how fre-quently participants were able to press aclown’s nose (e.g., bar pressing) during a 15min session. The second task evaluated howwell the participants were able to receptivelyidentify objects in a field of three. Results wereidiosyncratic across the five participants. Theresults did show that: (a) sensory reinforce-ment can be as effective as other classes ofreinforcement for children with ASD; (b) thatfood and praise were reinforcing for the fiveparticipants; and (c) that vibration appearedto be the most effective of the sensory stimuli.

In a later study, Rincover and Newsom(1985) compared the satiation effects of twoclasses of reinforcement: sensory (e.g., tickles,therapist finger tapping, and hand clapping)to food reinforcers on a two-choice visual dis-

crimination task. The authors selected fivesensory reinforcers and five food reinforcersfor each participant based on interviews anddirect observations. Results of the study indi-cated that sensory reinforcement was moreresistant to satiation when multiple reinforc-ers were presented during a teaching session.

In another study, Charlop, Kurtz, and Casey(1990) compared food reinforcers to usingparticipants’ self-stimulatory behaviors as rein-forcers on skill acquisition for children diag-nosed with ASD. Results of this study indicatedthat task performance was higher when theresearchers used a participant’s own self-stim-ulatory behaviors rather than food reinforc-ers. Thus, across the three studies it appearsthat sensory reinforcement can be more effec-tive than food reinforcement when teachingchildren a variety of tasks. Furthermore, sen-sory reinforcement might be more resistant tosatiation.

Despite these findings there are still severalquestions, which should be explored by re-searchers. For one, the research comparingreinforcement classes for children diagnosedwith ASD remains limited. Second, the previ-ous research did not utilize formal assess-ments of different potential reinforcers, so it isnot known if the researchers did indeedchoose the most preferred reinforcers withina reinforcement class. Third, none of the stud-ies incorporated participant choice, which hasbeen shown to influence the effectiveness ofreinforcement. Finally, none of the studieshave compared different reinforcementclasses to feedback (e.g., “Yes” or “No”) alone.

Thus, the purpose of this study was to eval-uate three reinforcement classes (i.e., foodreinforcement class, toy reinforcement class,social reinforcement class) and feedback-onlyon skill acquisition for three children diag-nosed with autistic disorder. This study evalu-ated the skill acquisition, maintenance, effi-ciency, and participant preference for each ofthe five participants.

Method

Participants

The participants in this study were three boyswho all: (a) had a formal diagnosis of autisticdisorder from a licensed psychologist from an

534 / Education and Training in Autism and Developmental Disabilities-December 2014

outside agency, (b) were between the ages of4 and 5 years old, (c) had a previous historywith discrete trial teaching, and (d) were re-ceiving intensive behavioral intervention ser-vices from a private agency for 10 to 21 hoursper week.

Perry was a four year old boy who was diag-nosed with autistic disorder. Perry had aWechsler Preschool and Primary Scale of In-telligence (WPPSI) full scale IQ score of 128,a Vineland Adaptive Behavior Scale (VABS)score of 94, and a Gilliam Autism Rating Quo-tient (GARS) of 98. At the time of the study,he was receiving 20 hours of behavioral inter-vention per week. Intervention focused on in-creasing attending and responding to adultsand peers, increasing language, initiating playwith peers, and using humor with peers.

J.D. was a four year old boy diagnosed withautistic disorder. J.D. had a WPPSII full scaleIQ score of 86, a VABS score of 83, and aGARS Autism Quotient of 80. At the time ofthe study, J.D. was receiving 10 hours of be-havioral intervention per week. Interventionfocused on following instructions the firsttime asked, tolerating frustrating situations,increasing language, and decreasing self-stim-ulatory behavior (i.e., fidgeting, silly/scriptedtalk, gazing).

Chris was a five year old boy diagnosed withautistic disorder. Chris had a WPPSII full scaleIQ score of 111, a VABS score of 85, and aGARS Autism Quotient of 83. At the time ofthe study, Chris was receiving 21 hours ofbehavioral intervention per week. Interven-tion focused on speaking clearly, sportsman-ship, increasing language, and staying ontopic in conversation with peers.

Setting

The researchers conducted all sessions in asmall research room that was a part of a pri-vate clinic that provided behavioral services tochildren and adolescents diagnosed with ASD.The research room measured 2 meters by 4meters. The room was furnished with a table,chairs, cabinets, and a couch.

Preference Assessment

The researchers conducted two paired prefer-ence assessments (Fisher et al., 1992) to deter-

mine each participant’s preference for 10 dif-ferent food items, two paired preferenceassessments to identify each participant’s pref-erence for 10 different toys, and two pairedpreference assessments to evaluate each par-ticipant’s preference for five different socialactivities prior to baseline. During the pairedpreference assessments the researchers di-rectly compared food items to food itemsonly, toys to toys only, and social activities tosocial activities only. Thus, no comparison wasmade of items across different reinforcementclasses (e.g., no food items were compared totangible items). The researchers selecteditems across each of the reinforcement classesby observing the participant in his naturalenvironment, interviewing the parents of theparticipant, and interviewing the participant’sclinical team (e.g., teachers and therapistswho work with the participant in the clinicalsetting). The top ten food and toy items iden-tified were used in the paired preference as-sessment and the top five social activities iden-tified were used in the paired preferenceassessment. The researchers conducted twopreference assessments on different days toensure that these preferences were consistentacross days. The top three items within eachreinforcement class that were selected mostfrequently, across the two preference assess-ments, were used throughout the course ofthe study.

Skills Taught

The researchers taught each participant toexpressively label pictures. Perry and Chriswere taught to expressively label pictures ofcartoon characters (e.g., Woody Woodpecker,Sylvester, Dennis the Menace); J.D. was taughtto expressively label pictures of different fooditems (e.g., bok choy, parsley, figs). The re-searchers selected these targets for two rea-sons. For one, the participants’ supervisorsindicated that these were skills that needed tobe worked upon. Second, the targets were notbeing intervened upon during clinical teach-ing, which minimized the chance of the par-ticipants inadvertently learning the skills out-side of the research sessions.

Each participant was taught a total of eightdifferent skills. The researchers randomly as-signed the eight targeted skills to the four


different reinforcement conditions (i.e., foodreinforcement, toy reinforcement, social rein-forcement, feedback-only). Thus, two skillswere taught at a time in each of the fourconditions. Table 1 provides a list of the skillstaught to each participant across the four re-inforcement conditions.

General Procedure

The researchers conducted sessions 3 to 5days a week with sessions running anywherefrom 15 to 40 minutes in total duration. Dur-ing each research session a participant re-ceived a probe to test skill acquisition (seebelow) and the various reinforcement condi-tions. The study was broken into three condi-tions: baseline, intervention, and mainte-nance.

Baseline and maintenance. The baseline andmaintenance condition consisted of the re-searchers implementing probe sessions (seebelow) to test if the participant already hadthe targets in his repertoire prior to interven-tion or if the participants were able to main-tain the skills directly taught to them followingintervention. In the baseline and mainte-nance condition, probe sessions consisted of32 probe trials (see below). After these probetrials the participant returned to his regularclinical setting.

Intervention condition. During the interven-tion condition, each research session startedwith the researcher implementing a probe ses-sion (8 to 32 probe trials) to test ongoinglearning and mastery of skills taught. Masterycriterion was set at three consecutive probesessions with 100% correct responding on

both targets being taught within a reinforce-ment condition. Probe sessions were followedby one of the four different reinforcementconditions. After the first reinforcement con-dition the researcher gave the participant a 2min break and then implemented the secondreinforcement condition. This was continueduntil all reinforcement conditions had oc-curred. After all reinforcement conditionswere implemented the participant returned tohis regular setting. The order of the reinforce-ment conditions was randomly determinedprior to the research session so that the re-searchers could minimize the potential of asequencing effect.

During each reinforcement condition theresearchers conducted a total of 10 teachingtrials. Once a participant reached mastery cri-terion during the probe sessions with one ofthe reinforcement conditions, the researchersno longer implemented that reinforcementcondition nor did they conduct probe trialsfor the skills that had reached mastery crite-rion. Therefore, the number of reinforcementconditions and the number of probe trials wasreduced as participants reached mastery crite-rion on each of the four conditions.

Probe Session

The researchers implemented probe sessionsprior to, during, and following intervention.Probe sessions were conducted to establish abaseline, to determine mastery, and to deter-mine maintenance for each of the five partic-ipants. A probe trial began with the researcherpresenting a picture to the participant, thenproviding an instruction to the participant

TABLE 1

Targeted Skills for All Participants

Participant Food Class Tangible Class Social Class Feedback

Perry Woody Woodpecker &Denis the Menace

Wilma Flintstone &Porky Pig

Sylvester the Cat &Willy thePenguin

Roger Rabbit &George Jetson

J.D. Olives & Enchiladas Grapefruit & Fig Passion Fruit &Pecan

Bok Choy &Parsley

Chris Woody Woodpecker &Denis the Menace

Wilma Flintstone &Porky Pig

Sylvester the Cat &Roger Rabbit

George Jetson &Willy thePenguin


(i.e., “Who is this?”, “What is his name?”,“What is the name of the food?”). The partic-ipant had 5 sec to respond to the instruction.Regardless of the participants response (i.e.,correct response or incorrect response) theresearcher provided neutral feedback (e.g.,“Thanks”, “Thank You”, or “Thanks for Try-ing”). If the participant did not respond (i.e.,remained silent without providing any expres-sive label) within 5 sec then the investigatorbegan the next predetermined trial withoutcomment.

During the baseline and maintenance con-dition there were always a total of 32 probetrials. During the intervention condition therecould be anywhere from 8 to 32 probe trials(dependent upon a participant reaching mas-tery criterion for skills in a certain condition).All targets were randomly sequenced prior tothe probe session. During intervention, thecriterion for learning a pair of stimulus itemsfor an individual reinforcement condition wasexpressively labeling each item of a pair cor-rectly on all probe trials (100%) across threeconsecutive sessions. Once participants metmastery criterion for a pair of stimuli, teach-ing on that pair stopped and the items wereno longer probed. Therefore, as participantsreached mastery criterion, the number ofprobe trials decreased. Following probe ses-sions during intervention, the researchersprovided the participant with a 2 min breakprior to beginning the first reinforcementcondition.

Reinforcement Sessions

Each of the four reinforcement conditionsconsisted of 10 teaching trials. The order oftargets was also randomly determined aheadof time. A teaching trial for each of the fourreinforcement conditions began with the re-searcher presenting the picture to the partic-ipant. Next, the researcher provided the par-ticipant with an instruction (e.g., “Who isthis?”). The researcher then gave the partici-pant 5 sec to respond to the instruction. If theparticipant responded correctly (e.g., said theword “figs” when showed a picture of figs), theresearcher made a general praise statement(e.g., “That’s it”) and provided the participantwith a choice between the three reinforcers,(i.e., the three food items, the three toys, or

the three social activities) depending on whichreinforcement condition was being imple-mented. After the participant had access to theitem for the specified duration (see below) theinvestigator started the next teaching trial. If theparticipant responded incorrectly (e.g., said theword “figs” when shown a picture of bok choy),or did not respond to the instruction, within 5sec the researcher said “No” and stated the cor-rect response (e.g., “It is figs.”).

Food condition. Three potential food rein-forcers were identified for each participantduring the paired preference assessments. ForPerry, the potential food reinforcers were an-imal crackers, marshmallows, and water (al-though not a food, was highly desirable). ForJ.D., the potential food reinforcers were fruitsnacks, pop corn, and gummy bears. ForChris, the potential food reinforcers were licksof a lollipop, gummy bears, and fruit snacks. Ifthe participant responded correctly, he had achoice of one of his three food items. Oncethe participant selected which food item hewanted, the researcher provided three smallpieces of that food item and did not start thenext teaching trial until the participant hadconsumed the food item.

Tangible condition. The assessments yieldedthree potential toy reinforcers for each partic-ipant. For Perry, the toy reinforcers were abook, a Superman action figure, and a carfrom the movie Cars. For J.D., the toy rein-forcers were a book, a light wand, and a ball.For Chris, the toy reinforcers were a car fromthe movie Cars, a ball, and a Superman actionfigure. If the participant responded correctly,he had a choice of one of his three toy items.Once the participant selected which toy itemhe wanted, the researcher provided 10 secaccess to that item and the researcher did notengage with the participant during this time.

Social condition. Three potential social re-inforcers were identified for each participantduring the paired preference assessments. ForPerry, the social reinforcers were the re-searcher and Perry engaging in a silly dance,the researcher giving Perry high fives at arapid pace, and the researcher taking Perry’sarms and moving them up and down rapidly(spaghetti arms). For J.D. and Chris, the socialreinforcers were the researcher and J.D./Chris engaging in a silly dance, the researchergiving J.D./Chris a foot high five, and the


researcher pretending to crack an egg onJ.D./Chris’ head. If the participant respondedcorrectly he had a choice of one of his threesocial reinforcers (which were pictured on anindex card) and the researcher provided 10sec access to the social activity that the partic-ipant selected.

Feedback-only condition. Teaching trials inthis condition were identical to the previousthree conditions with the exception of theconsequence that the researchers provided tothe participant. If the participant respondedcorrectly, the researchers simply said “Yes”and if the participant responded incorrectly,or did not respond at all, the researchers sim-ply said “No” and stated the correct response.The consequent event (e.g., saying “No” andproviding the correct response) was identicalto the consequent events used in the otherthree reinforcement conditions. It should alsobe noted, that the use of “Yes” and “No” asfeedback could also be classified as a condi-tioned social reinforcer, if it changed the be-havior of the participants. Nevertheless, thisconditioned social reinforcer is vastly differentfrom the reinforcement used in the social con-dition and is different from how social rein-forcement is used in best clinical practice.

Participant Preference among the FourReinforcement Cconditions

To assess participant preference among the fourreinforcement conditions, the researchers im-plemented a concurrent-chain arrangement(Hanley, Piazza, Fisher, Contrucci, & Maglieri,1997). Prior to baseline the researchers con-ducted a single paired preference assessmentacross ten different shirt colors. The four shirtsthat the participant selected at a near equivalentfrequency were randomly assigned to one of thefour reinforcement conditions. The shirt colorthat the participant selected most frequently wasused as a control shirt.

During every reinforcement session the re-searcher wore the color of shirt that corre-sponded to the reinforcement condition (e.g.,the researcher wore a yellow shirt during thefood condition if the yellow shirt was assignedto the food condition); additionally, a colormat of the same color was placed on the table.During every third research session, right afterthe probe session, the researcher placed all

five colors of shirt on the floor or on thecouch with the corresponding array of rein-forcers displayed on top of the shirt. The re-searcher asked the participant to touch whathe wanted to work for first. The reinforce-ment condition that corresponded with thecolored shirt and reinforcers that the partici-pant selected was then implemented first. Af-ter the first reinforcement session had con-cluded the researcher asked the participant tochoose the next color of shirt and reinforcersthat he wanted to work for and then imple-mented the corresponding condition; this oc-curred until all reinforcement conditionswere implemented. If, at any time, the partic-ipant selected the shirt color that corre-sponded to the control condition, the re-searcher concluded that the student was notmaking meaningful choices and the re-searcher selected the order of the remainingreinforcement conditions.

Dependent Measures and Response Definitions

There were several dependent measures in thisstudy. The main measure was participant’s skillacquisition and maintenance of the targetstaught for each of the four reinforcement con-ditions. Skill acquisition and maintenance weredetermined by participant responding duringprobe sessions. During probe trials, researchersrecorded a correct response if the participantcorrectly labeled the picture that was shown(e.g., saying “figs” when shown a picture of figs)within 5 s of the researcher’s instruction. Theresearchers also recorded whether the partici-pant responded incorrectly or if they did notrespond within 5 s of the instruction.

The second measure of this study was par-ticipant responding during each of the fourreinforcement conditions. During teachingtrials, researchers recorded a correct responseif a participant labeled the picture that wasshown within 5 s of the researcher’s instruc-tion. The researchers also recorded whetherthe response was incorrect, or if the partici-pant did not respond.

The third measure of the study was theefficiency with which the participants learnedeach of the targets under the four reinforce-ment conditions. To measure the efficiencythe researchers looked at the total number ofsessions and the total amount of teaching tri-


als it took the participant to reach masterycriterion under each of the four reinforce-ment conditions. The fourth measure was theparticipants’ preference for the four rein-forcement conditions as determined by theconcurrent chain operation described above(Hanley et al. 1997).

Interobserver Agreement and Treatment Fidelity

The instructor recorded all participant re-sponding in vivo and an independent ob-server recorded participant responding by vid-eotape during 37% (range, 33% to 45% acrossparticipants) of probe sessions, 35.7% (range,33% to 37.5% across participants) of interven-tion sessions within the food condition, 37.5%(range, 33% to 42.6% across participants) ofintervention sessions within the tangible con-dition, 35.8% (range, 33% to 36.4% acrossparticipants) of intervention sessions withinthe social condition, and 40% (range, 35.7%to 50% across participants) of interventionsessions within the feedback-only condition.Interobserver agreement was calculated by to-taling the number of agreements (i.e., trialsfor which both observers scored the same par-ticipant behavior) on each type of participantresponse divided by the number of agree-ments plus disagreements (i.e., trials in whichthe two observers scored a different partici-pant behavior) and converting this ratio to apercentage. Percentage agreement across allparticipant responses was 99.6% (range,93.8% to 100% across sessions) of probe trials,100% during the food condition, 100% dur-ing the tangible condition, 99.2% (range,90% to 100% across sessions) during the so-cial condition, and 100% during the feedback-only condition.

The final measure was correct instructorbehaviors during the probe session and eachof the four reinforcement conditions. Correctinstructor behaviors during probe sessionswere defined as the researcher: (a) showingthe correct stimuli or placing the correct stim-uli on the table; (b) delivering a correct in-struction (e.g., saying “Who is it?”) to beginthe teaching trial; (c) allowing the participant5 s to respond; (d) not providing a prompt;and (e) saying “thank you” or “okay” if theparticipant responded. Correct instructor be-haviors during reinforcement sessions were

defined as the researcher: (a) showing thecorrect stimuli to the participant; (b) deliver-ing a correct instruction (e.g., saying “Who isit?”) to begin the teaching trial, (c) allowingthe participant 5 s to respond; (d) providing ageneral praise statement if the participant re-sponded correctly; (e) giving the participant achoice between the three reinforcer items; (f)providing the participant with access to thecorrect reinforcer for the appropriate amountof time; and (g) providing corrective and in-formative feedback only when the participantresponded incorrectly.

To assess treatment fidelity, an independentobserver measured correct instructor behav-ior (as described above) via video recordingduring 37% (range, 33% to 45.4% across par-ticipants) of probe sessions, 35.7% (range,33.3% to 37.5% across participants) of inter-vention sessions within the food condition,37.5% (range, 33.3% to 42.8% across partici-pants) of intervention sessions of the tangiblecondition, 35.8% (range, 33.3% to 36.4%across participants) of intervention sessionswithin the social condition, and 40% (range,35.7% to 50% across participants) of interven-tion sessions within the feedback-only condi-tion. Correct instructor behavior was dis-played during 99.8% of probe trials, 100% offood teaching trials, 98.8% of tangible teach-ing trials, 100% of social teaching trials, and100% of feedback-only teaching trials.

Experimental Design

The researchers used a parallel treatments de-sign to evaluate the effectiveness of the fourreinforcement conditions on each partici-pant’s acquisition of different skills. A paralleltreatments design can be used when there aretwo or more sets of dependent variables andwhere there are two or more sets of indepen-dent variables being implemented. The de-pendent variables should be equivalent interms of difficulty. Typically, when using aparallel treatment design there is a probe con-dition that is implemented prior to the rein-forcement conditions. A baseline condition isnot required, but is strongly recommended. Itis important that the order of the indepen-dent variables is random and counterbalancedand that the participant is able to discriminatebetween the different independent variables.


Results

Skill Acquisition and Maintenance

Figure 1 displays the participants’ perfor-mance during probe trials prior to interven-

tion, during intervention, and once interven-tion was completed. Each panel of the graphrepresents a different participant. Sessions aredisplayed across the X-axis and percentage ofcorrect probe trials are displayed along the

Figure 1. Probe Data for Perry, J.D., and Chris


Y-axis. Closed circles represent the food con-dition, open circles represent the tangiblecondition, closed squares represent the socialcondition, and open squares represent thefeedback-only condition. On Figure 1, thedata points for the four conditions fall onthe same session (e.g., the food condition forsession one and the tangible condition forsession one are both displayed on sessionone); thus, the graph does not show the orderin which the independent variable was imple-mented each session. However, Figure 2,which represents participant responding dur-ing teaching (see below), does show the orderof the implementation of the independentvariable. It should also be noted that the firstday in which the intervention was imple-mented there was no probe session conduct-ed; thus, the first probe session of the inter-vention condition was implemented after thefirst reinforcement session.

Perry was able to reach mastery criterion onhis eight targets across the four reinforcementconditions. Perry first reached mastery crite-rion on the food condition, followed by thefeedback-only condition, and then reachedmastery criterion on the social and tangibleconditions during the same research session.During baseline probe sessions, Perry demon-strated 0% correct responding and duringmaintenance probe session demonstrated100% correct responding, across the four re-inforcement conditions.

J.D. was able to reach mastery criterion onhis eight targets across the four reinforcementconditions. J.D. first reached mastery criterionon the food condition, followed by the feed-back-only and tangible conditions during thesame research session, and then the socialcondition. During baseline probe session, J.D.demonstrated 0% correct responding acrossthe four conditions. During maintenanceprobe sessions, J.D. demonstrated 100% cor-rect responding on the food, tangible andsocial conditions and an average of 96% cor-rect responding for the feedback-only condi-tion.

Chris was able to reach mastery criterion onhis eight targets across the four reinforcementconditions. Chris first reached mastery crite-rion on the food condition, followed by thetangible condition, then the feedback-onlycondition, and then the social condition. Dur-

ing baseline probe sessions, Chris demon-strated 0% correct responding across the fourconditions. During maintenance probe ses-sions, Chris demonstrated an average of 84%,100%, 100%, and 9% correct responding forthe food, tangible, social, and feedback-onlyconditions, respectively.

Teaching Trial Data

Figure 2 displays the participants’ perfor-mance during teaching across the four rein-forcement conditions. Each panel of thegraph represents a different participant. Ses-sions for individual conditions are displayedacross the X-axis and percentage of correctresponding on teaching trials is displayedalong the Y-axis. Closed circles represent thefood condition, open circles represent thetangible condition, closed squares representthe social condition, and open squares repre-sent the feedback-only condition. The orderof reinforcement conditions is displayed onthe graph.

Across all four reinforcement conditions,Perry’s average correct responding duringteaching trials was above 80%. For the foodcondition, Perry’s average correct respondingwas 90%. For the tangible condition, Perry’saverage correct responding was 82%. For thesocial condition, Perry’s average correct re-sponding was 85%. For the feedback-only con-dition, Perry’s average correct responding was80%. Thus, Perry showed the best correct re-sponding in the food condition and the leastcorrect responding in the feedback-only con-dition.

J.D.’s average correct responding duringteaching trials was above 80% for three of thefour reinforcement conditions. For the foodcondition, J.D.’s average correct respondingwas 83%. For the tangible condition, J.D.’saverage correct responding was 90%. For thesocial condition, J.D.’s average correct re-sponding was 89%. For the feedback-only con-dition, Perry’s average correct responding was72%. Thus, JD showed the best correct re-sponding in the social condition and the leastcorrect responding in the feedback-only con-dition.

Chris’s average correct responding duringteaching trials was above 80% for three of thefour reinforcement conditions. For the food


condition, Chris’s average correct respondingwas 85%. For the tangible condition, Chris’saverage correct responding was 86%. For the

social condition, Chris’s average correct re-sponding was 70%. For the feedback-only con-dition, Chris’s average correct responding was

Figure 2. Teaching Data for Perry, J.D., and Chris


84%. Thus, Chris showed the best correct re-sponding in the tangible condition and theleast correct responding in the social condi-tion.

Efficiency

Table 2 represents the total number of teach-ing sessions and teaching trials for each of theparticipants. For all three participants thefood condition resulted in the fewest numberof sessions and trials. For Perry, the tangibleand social conditions took the most trials andsessions to reach mastery; for J.D. and Chris,the social condition took the most trials andsessions to reach mastery criterion.

Participant Preference

Across all participants the food condition wasselected the most frequently. Perry had twoopportunities to make a preference selectionand selected the food and toy condition inthese two selections. J.D. had three opportu-nities to make a selection and chose the foodcondition twice and the feedback-only condi-tion one time. Chris had seven opportunitiesto make a preference selection; Chris selectedthe food condition four times, and the socialcondition, toy condition, and the control shirtone time each.

Discussion

This study evaluated the effectiveness of fourdifferent reinforcement classes on teachingexpressive labeling for three participants diag-nosed with autistic disorder. The results of thestudy indicated that all three participants were

able to reach mastery criterion for skills taughtwith all four classes of reinforcement. Resultswere mixed as far as participants’ correct re-sponding during teaching trials; each partici-pant had a better average rate of correct re-sponding during different reinforcementconditions. For each participant, the foodcondition did result in the fewest number oftrials and sessions for participants to reachmastery criterion and the social condition re-sulted in the most trials and sessions for theparticipants to reach mastery criterion. Fi-nally, participants showed a strong preferencefor food reinforcers compared to the otherreinforcement classes.

Findings from this study are similar to pre-vious research in that a variety of reinforce-ment classes can be effective in teaching chil-dren with ASD a variety of skills. This studydiffers from the previous research in that foodreinforcers were found to be more effectivethan other classes of reinforcement. Thisstudy also adds to the literature in severalways. For one, it is the first study to look atparticipants’ preference among different rein-forcement classes; all three of the participantspreferred the reinforcement class that was themost efficient. Second, the study was the firstto provide participant choice within reinforce-ment classes, which may help increase theeffectiveness of a reinforcement class (e.g.,Tiger, Hanley, & Hernandez, 2006). Finally,the study was the first to compare differentreinforcement classes to instructional feed-back (i.e., “Yes” or “No”) alone and show thatinstructional feedback alone may be just aseffective as using different reinforcers forteaching fairly easy tasks to some children withautism. Therefore, these results have several

TABLE 2

Total Reinforcement Sessions and Teaching Trials

Food Condition Tangible Condition Social Condition Feedback Condition

Participant Sessions Trials Sessions Trials Sessions Trials Sessions Trials

Perry 3 30 6 60 6 60 4 40J.D 3 30 7 70 11 110 7 70Chris 8 80 11 110 22 220 14 140


implications for clinicians working with chil-dren diagnosed with ASD.

For one, clinicians may wish to analyzewhich reinforcement classes are the most ef-fective prior to teaching new skills, so that themost effective and efficient teaching proce-dures are being implemented. Second, clini-cians should not assume additional reinforc-ers (e.g., food, toys, social activity) are neededto teach new skills, as feedback alone might bejust as effective. This could be because thefeedback (e.g., yes or no) serve as a mildconditioned social reinforcer or punisher andthat additional conditioned reinforcers arenot warranted. Third, although social praisemay not be the most effective, it may still beimportant for clinicians to teach the value ofpraise to children with ASD. Clinicians maywish to condition social praise using similarmethods as described by Leaf et al. (2012).Finally, clinicians should use a variety of rein-forcers, as the results from this study indicatethat different reinforcement classes are effec-tive in teaching new skills.

Despite the positive findings of this studyand the potential implications for cliniciansworking with children diagnosed with an au-tism spectrum disorder there are several limi-tations of the current study. One limitation ofthis study is that the researchers did not mea-sure the motivating operations for each of theparticipants under the four reinforcementclasses; nor did we withhold the toys and fooditems from the participants throughout thecourse of the day. Therefore, it is possible thatthe participants were in a state of satiationprior to teaching and that this state of satia-tion may have decreased the reinforcers’ ef-fectiveness. Additionally, recent research hasdemonstrated that the motivational power ofreinforcers can be significantly altered by themanner in which a student is exposed to them(Leaf et al., 2012). Thus, although one stimulimight not be the most effective reinforcerinitially a teacher can condition that stimuli tobe a powerful reinforcer. A second limitationis that we did not assess preference of itemsacross different reinforcement classes; thus, itis not known what the participants’ relativepreferences were. This is important, as partic-ipants’ preference for items are typically indic-ative of their effectiveness as reinforcers.Thus, a less preferred item across classes may

be less reinforcing. Future researchers shouldimplement preference assessments across dif-ferent classes.

Third, we did not measure how long partic-ipants had access to the food item during thefood condition. Instead, we did not start a newtrial until the participant finished consumingthe food item; this could have resulted in alonger or shorter intertrial interval, as com-pared to the other conditions. Future re-searchers may wish to hold the time constantduring a food condition. A fourth limitation isthat experimental control is partially weak-ened due to the rapid acquisition of skillsacross the four conditions. Ideally, when im-plementing a parallel treatments design, oneindependent variable results in more rapidskill acquisition than other independent vari-ables. In this study there was rapid skill acqui-sition for most conditions across the threeparticipants. Although there was rapid skillacquisition across conditions, differences canstill be observed in the rate of skill mastery.The data still demonstrates that multiple typesof reinforcers can be effective in teaching chil-dren diagnosed with autistic disorder.

A fifth limitation is that there were a limitednumber of participants (three) learning afairly simple task (expressive labeling). Thus,the results are limited in terms of externalvalidity. However, this was an initial studylooking at the effectiveness of different rein-forcers across different reinforcer classes. Fu-ture researchers can expand on these resultsby evaluating different reinforcer classes witha wide range of participants on a wide range oftasks. Additionally, future researchers maywish to compare the effectiveness and effi-ciency of reinforcer classes using a largernumber of participants, perhaps evaluatingthe effects using a group design rather than asingle subject design. A final limitation of thestudy is that each of the reinforcement condi-tions (e.g., food condition) was relatively shortin duration; so it is not known how effectivereinforcers in each of the four reinforcementclasses would be during longer teaching ses-sions. Future researchers should examine howthe use of reinforcers over more extendedtime periods could lead to more pronounceddifferences in factors such as satiation and theimpact of reinforcer type on health and wellbeing of the learner. It may be that certain


reinforcers that are less motivationally power-ful (e.g., social reinforcers) could have fewernegative side effects and/or more positive sideeffects. Despite these limitations this studyshowed that a variety of reinforcement classeswere effective in teaching three children diag-nosed with autistic disorder expressive lan-guage skills.

References

Allyon, T., & Azrin, N. H. (1968). Reinforcer sam-pling: a technique for increasing the behavior ofmental patients. Journal of Applied Behavior Analy-sis, 1, 13–20.

Carr, J. E., Nicolson, A. C., & Higbee, T. S. (2000).Evaluation of a brief multiple-stimulus preferenceassessment in a naturalistic context. Journal of Ap-plied Behavior Analysis, 33, 353–357.

Catania, C. A. (1998) Learning Fourth Edition. Up-per Saddle River, NJ: Prentice Hall. Charlop,M. H., Kurtz, P. F., & Casey, F. G. (1990). Usingaberrant behaviors as reinforcers for autistic chil-dren. Journal of Applied Behavior Analysis, 23, 161–181.

DeLeon, I. G., Neidert, P. L., Anders, B. M., &Rodriquez-Catter, V. (2001). Examination of rel-ative reinforcement effects of stimuli identifiedthrough pretreatment and daily brief preferenceassessments. Journal of Applied Behavior Analysis,34, 463–473.

Ferrari, M., & Harris, S. L. (1981). The limits andmotivating potential of sensory stimuli as rein-forcers for autistic children. Journal of Applied Be-havior Analysis, 14, 339–343.

Fisher, W., Piazza, C. C., Bownman, L. G., Hago-pian, L. P., Owens, J. C., & Slevin, I. (1992). Acomparison of two approaches for identifying re-inforcers for persons with severe and 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).Identifying reinforcers for persons with profoundhandicaps: Staff opinion versus systematic assess-ment of preferences. Journal of Applied BehaviorAnalysis, 21, 31–43.

Hanley, G. P., Piazza, C. C., Fisher, W. W.,Contrucci, S. A., & Maglieri, K. A. (1997). Evalu-ation of client preference for function-basedtreatment packages. Journal of Applied BehaviorAnalysis, 30, 459–473.

Leaf, J. B., Dotson, W., Oppenheim, M. L., Sheldon,J. B., & Sherman, J. A. (2010). The effectivenessof a group teaching interaction for young chil-

dren with a pervasive developmental disorder.Research in Autism Spectrum Disorders, 4, 186–198.

Leaf, J. B., Sheldon, J. A., & Sherman, J. B. (2010).Comparison of simultaneous prompting andno-no prompting in two choice discriminationlearning with children with autism. Journal of Ap-plied Behavior Analysis, 43, 215–228.

Leaf, J. B. Oppenheim-Leaf, M. L., Leaf, R., Courter-manche, A. B., Taubman, M., McEachin, J., Shel-don, J. B., & Sherman, J. A. (2012). Observationaleffects on the preferences of children with au-tism. Journal of Applied Behavior Analysis, 45, 473–483.

Luczynski, K. C., & Hanley, G. P. (2009). Do chil-dren prefer contingencies? An evaluation of theefficacy of and preference for contingent versusnoncontingent social reinforcement during play.Journal of Applied Behavior Analysis, 42, 511–525.

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

Petursdottir, A. I., Carr, J. E., Lechago, S. A., &Almason, S. M. (2008). An evaluation of intraver-bal training and listener training for teachingcategorization skills. Journal of Applied BehaviorAnalysis, 41, 53–68.

Rincover, A., & Newsom, C. D. (1985). The relativemotivational properties of sensory and edible re-inforcers in teaching autistic children. Journal ofApplied Behavior Analysis, 18, 237–248

Repp, A. C., & Deitz, S. M. (1974). Reducing aggres-sive and self-injurious behavior of institutional-ized retarded children through reinforcement ofother behaviors. Journal of Applied Behavior Analy-sis, 7, 313–325.

Roscoe, E. M., Iwata, B. A., & Rand, M. S. (2003).Effects of reinforcer consumption and magnitudeon response rates during noncontingent rein-forcement. Journal of Applied Behavior Analysis, 36,525–539.

Schreibman, L. (1975). Effects of within-stimulusand extra-stimulus prompting on discriminationlearning in autistic children. Journal of AppliedBehavior Analysis, 8, 95–112.

Smith, R. G., Iwata, B. A., & Shore, B. A. (1995).Effects of subject-versus experimenter-selected re-inforcers on the behavior of individuals with pro-found development disabilities. Journal of AppliedBehavior Analysis, 28, 61–71.

Singh, N. N., Dawson, M. J., & Manning, P. (1981).Effects of spaced responding DRL on the stereo-typed behavior of profoundly retarded persons.Journal of Applied Behavior Analysis, 14, 521–526.


Tiger, J. H., Hanley, G. P., & Hernandez, E. (2006).An evaluation of the value of choice with pre-school children. Journal of Applied Behavior Analy-sis, 39, 1–16.

Worsdell, A. S., Iwata B. A., Hanley, G. P., Thomp-son, R. H., & Kahng, S. (2000). Effects of contin-uous and intermittent reinforcement for problem

behavior during functional communication train-ing. Journal of Applied Behavior Analysis, 33, 167–179.

Received: 18 April 2013Initial Acceptance: 20 June 2013Final Acceptance: 6 August 2013


Comparing Different Classes of Reinforcement to Increase ...

Documents

Transcript of Comparing Different Classes of Reinforcement to Increase ...