the effects of reinforcer pairing and fading on preschoolers' snack selections

THE EFFECTS OF REINFORCER PAIRING AND FADING ONPRESCHOOLERS’ SNACK SELECTIONS

KATHERINE M. SOLBERG, GREGORY P. HANLEY, STACY A. LAYER, AND

EINAR T. INGVARSSON

UNIVERSITY OF KANSAS

The effects of reinforcement pairing and fading on preschoolers’ snack selections were evaluatedin a multiple baseline design. Baseline preferences for snack options were assessed via repeatedpaired-item preference assessments. Edible, social, and activity-based reinforcers were thenexclusively paired with a less preferred snack option. Once the snack paired with reinforcementwas selected most frequently, the three types of reinforcement were systematically faded.Frequent selections of the previously less preferred snack option were produced with pairedreinforcement, but were disrupted for all children as the paired reinforcement was reduced to lowlevels. These data showed that paired reinforcement was initially effective in increasingpreference for the originally less preferred snack options, but more permanent changes in thevalue of the snack options were not achieved. Conditions for producing persistent changes inchildren’s snack choices are discussed.

DESCRIPTORS: conditioning, preschoolers, preference, reinforcement pairing and fading

_______________________________________________________________________________

Although reinforcement contingencies haveproven to be effective in increasing foodconsumption in preschool children (e.g., Ireton& Guthrie, 1972), the use of a reinforcementcontingency to promote consumption of non-preferred foods may not alter the value of thenonpreferred food; therefore, programmedcontingencies may be required to maintainconsumption. It may be possible to alter thevalue of a nonpreferred food and maintainconsumption of that food in the absence of theinitial intervention using conditioning proce-dures. Capaldi (1996) suggested that some foodpreferences may develop as a result of flavor-flavor conditioning, which occurs when oneflavor is paired with another flavor. Theassociation of one flavor with another is saidto result in the appetitive or aversive propertiesof one flavor transferring to the other (Capaldi).

Holman (1975) found that pairing a less pre-ferred flavor with a more preferred flavor canincrease preference for the originally less pre-ferred flavor in rats; similar findings have beenobserved with humans (Bayens, Eelen, Van denBergh, & Crombez, 1990; Zellner, Rozin,Aron, & Kulish, 1983).

Past studies concerned with increasing con-sumption of less preferred foods via pairinghave primarily involved two pairing procedures:(a) simultaneous presentation, which involvespairing a highly preferred food with a lesspreferred food so that the two foods areconsumed simultaneously, and (b) sequentialpresentation, which involves making the avail-ability of a highly preferred food contingent onconsumption of the less preferred food (i.e., thetwo foods are consumed separately). Kern andMarder (1996) showed that simultaneous pre-sentation resulted in a greater number of bitesof fruit accepted as compared to the number ofbites of vegetables accepted using sequentialpresentation, but because they associated fruitwith one presentation method and vegetableswith the other, it is not possible to determine ifthe presentation style was exclusively responsi-ble for the observed differences.

The research presented in this article was completed inpartial fulfillment of thesis requirements for the MAdegree by the first author.

Correspondence and requests for reprints should beaddressed to Gregory P. Hanley, Psychology Department,Western New England College, 1215 Wilbraham Road,Springfield, Massachusetts 01119.

doi: 10.1901/jaba.2007.633–644

JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2007, 40, 633–644 NUMBER 4 (WINTER 2007)

633

Piazza et al. (2002) directly comparedsequential and simultaneous presentation acrossall food groups. Increased food consumptionoccurred for 2 of 3 participants in thesimultaneous condition, in which a rarelyconsumed food was placed in or underneatha preferred food, but not in the sequentialcondition. Piazza et al. suggested that increasedconsumption of nonpreferred (rarely con-sumed) foods following intervention may bea result of flavor–flavor conditioning. Thisassertion was based on the fact that simulta-neous food presentation was more effective thansequential food presentation for increasingconsumption of nonpreferred foods in studiesby Kern and Marder (1996) and Piazza et al.However, as Ahearn (2003) pointed out, thenonpreferred foods were never presented aloneafter the simultaneous presentation procedure,so it is uncertain whether conditioning wasresponsible for the observed effects. Ahearnextended this research by conducting a postpair-ing condition after simultaneous food pre-sentation for increasing vegetable consumptionby a 14-year-old boy with autism. Resultsshowed that simultaneous presentation in-creased consumption of vegetables. However,when the highly preferred condiment wascompletely removed, consumption of all threevegetables returned to baseline levels. Theseresults suggest that it is unlikely that flavor–flavor conditioning was responsible for initialincreases in consumption of vegetables, and thatincreased consumption was more likely a directeffect of the reinforcing features of the condi-ments. It is possible that there were not enoughpairings for flavor–flavor conditioning to occur.However, there were more stimulus–stimuluspairings than in previous studies (Piazza et al.).

Pairing procedures also have been used toalter preferences for nonfood items (Hanley,Iwata, & Roscoe, 2006; Hanley, Iwata, Roscoe,Thompson, & Lindberg, 2003). In thesestudies, a less preferred activity was paired withreinforcement, and preference for the originally

less preferred activities emerged during pairing.Most important was that this shift in preferencewas maintained during evaluative probes fordays (Hanley et al., 2006) or weeks (Hanley etal., 2003) after pairing was terminated, suggest-ing that conditioning did indeed occur. Hanleyet al. (2003) provided intermittent pairingsbefore terminating the pairing procedures; thisaspect may have promoted a conditioningeffect. By contrast, paired reinforcement wassimply terminated in a single step by Ahearn(2003). Taken together, these data suggest thatpairing may be more likely to conditiona preference if the paired reinforcement is moregradually eliminated.

Stimulus fading has been used in combina-tion with other procedures to increase amountand type of food and fluid consumption(Freeman & Piazza, 1998; Patel, Piazza, Kelly,Ochsner, & Santana, 2001; Riordan, Iwata,Wohl, & Finney, 1980) and may be an effectivemeans for eliminating paired foods in condi-tioning procedures. In these studies, each timethe highly preferred food or drink was pre-sented, a small amount of a nonpreferred foodor drink was added in gradually increasingamounts, until the nonpreferred item waspresented alone. Mueller, Piazza, Patel, Kelley,and Pruett (2004) showed that consumption ofnonpreferred foods increased and was main-tained after being paired with a preferred foodwhen preferred foods were gradually eliminated.However, because the nonpreferred foods werepresented in a single-stimulus format, it is notpossible to determine if these procedures wouldresult in the child choosing a nonpreferred foodin the presence of more preferred foods. Thismay be important for less dependent children(e.g., typically developing school-aged chil-dren), who are typically faced with choosingbetween multiple foods, especially when choos-ing healthy over less healthy foods on themajority of occasions is the ideal outcome.However, before attempting to shift preferencesfor foods that may be highly discrepant in terms

634 KATHERINE M. SOLBERG et al.

of form and preference, such as chocolate caketo broccoli, it is important to apply preference-shifting procedures to foods that are formallysimilar, such as various cereal snacks. If in-ducing shifts in preferences between formallysimilar foods is difficult, it would very likely beeven more so if foods were formally discrepant.

Thus, the goal of the present study was tocombine and extend methods for increasingconsumption of less preferred foods demon-strated to be effective in different contexts withclinical populations to typically developingpreschool children. More specifically, we eval-uated the efficacy of pairing less preferred foodswith reinforcement (e.g., social interaction andchildren’s videos) to increase the selection andconsumption of less preferred snacks. Stimulusfading was then used to gradually eliminate thesuperimposed reinforcers until the target snackwas once again presented alone. For 2 partic-ipants, we assessed formally similar foods (i.e.,breakfast cereals), and we included moreformally discrepant foods ranging from vegeta-bles to candy for the 3rd participant.

METHOD

Participants and Setting

Three typically developing preschoolers,who were 4 years 5 months (Jessica), 4 years1 month (Tim), and 3 years 8 months (Susan)old at the beginning of the study, participated.All sessions took place in research roomscontaining child-sized chairs and a table, andwere conducted at roughly the same time eachday (after breakfast but before lunch) no soonerthan 30 min after breakfast.

Measurement and Interobserver Agreement

The primary dependent variable measured inthis study was snack selections, and data werecollected using paper and pencil. Multiple-stimulus without replacement (MSWO) prefer-ence assessments as described by DeLeon andIwata (1996) and paired-stimulus preferenceassessments as described by Fisher et al. (1992)

were conducted to evaluate preferences amongsnacks, condiments, and activities. Selection wasscored during every preference assessment andconsumption was scored during each assessmentinvolving foods. Selection was defined as thechild removing a snack item or activity from thepair or array of items. Consumption wasdefined as the child placing the entire bit ofsnack in his or her mouth and having a cleanmouth (no visible snack in the child’s mouth)within 30 s of the selection. Consumption wasrecorded at the end of 30 s or after the childsaid that he or she was done and showed a cleanmouth. If 30 s had expired, the experimenterchecked for a clean mouth by giving the childa verbal prompt (e.g., ‘‘show me’’). If the childdid not comply after two prompts, theexperimenter moved to the next trial; however,all participants readily complied with thisrequest. Consumption of selected items rangedfrom 90% to 100% for all children; therefore,the percentage of selections is the measurereported.

Selection was scored by circling a numberthat corresponded to the item that was selected.Agreements for selection were scored when bothobservers circled the same number for the sametrial. The number of agreements was divided bythe total number of trials in the assessment.Consumption was scored at the end of each trialin all of the preference assessments involvingsnacks by circling ‘‘yes’’ or ‘‘no’’ next to thecorresponding trial number. Agreements forconsumption were scored when both observerscircled the same word. The number of agree-ments was divided by the total number of trialsin the assessment. Interobserver agreement wascollected during 69% of all sessions andaveraged 99% (range, 87% to 100%) forselection and consumption combined.

Procedure

Overview. First, MWSO assessments wereconducted with 15 snack foods. These largepreference assessments served not only asa method for identifying a range of preferred

REINFORCER PAIRING AND FADING 635

items but also as pre- and posttest measures ofpreference. To evaluate preference shifts ina smaller context, five snack foods were selectedbased on the results from the 15-item MWSOassessment (pretests). Foods that representeddifferent preference values (e.g., high, medium,low) were selected. Paired-choice preferenceassessments were then conducted with the fivesnack foods to establish a baseline for the pairingand fading assessment. Paired-choice assessmentsbegan either the same day as the 15-item MSWOassessments or the following school day. Next,potential reinforcers (activities and condiments)were selected based on the results of MSWOassessments. Once these items were identified,the pairing and fading intervention was imple-mented within the context of repeated paired-choice assessments. Finally, when the pairing andfading assessment was complete, the initial 15snacks were reevaluated in MSWO assessmentsto determine the effects of our procedures ina broader context. These measures were takeneither the same day as the pairing and fadingassessment or the following school day.

15-item MSWO assessments (pre- and posttestmeasures of preference). We selected 15 cereals asthe formally similar foods to be included in theinitial preference assessments for Tim andSusan. Cereals were reported to be a commonsnack for these children. For Jessica, a surveywas given to parents to identify more discrepantpreferred and nonpreferred snacks. Her motherwas instructed to circle the snacks that Jessicaliked and disliked. Snack items consisting ofdifferent cereals, crackers, chips, candy, fruits,and vegetables were included in the survey. Theresults were used to select 15 snacks of differingnutritional values to be included in her initialpreference assessments. For Jessica and Tim,five pre- and posttests were conducted. Fivepretests but only two posttests were conductedwith Susan because she left the preschool justprior to completion of this study.

Prior to the beginning of the initial MSWOassessment, children were allowed to sample

each item. A small, consistent portion of eachsnack (e.g., one slice of a carrot or one flake ofcereal) was placed on 15 small plates. The plateswere arranged in a semicircle on a table in frontof the child in an arbitrary order. Theexperimenter named each snack in the arrayprior to the first trial. The child was in-structed to select a snack by removing itfrom the table and consuming it. The selectedsnack was absent from the array in thesubsequent trials. Before the next trial began,the position of the remaining snacks was rotatedby taking the snack on the far left, placing it onthe far right, and adjusting the other snacks sothat they were equally spaced. This processcontinued until all of the snacks had beenchosen.

The results of these assessments provideda hierarchy of preferences. The snacks wereranked 1 through 15, with 1 being the mostpreferred and 15 being the least preferred. Amean rank for each snack was derived from thefive MSWO assessments. From this hierarchy,the snacks were partitioned into five groupswith three snacks in each group. The top three(mean rank of 1 to 3) were in the first group,the following three (mean rank of 4 to 6) werein the second group, and so on. One snack fromeach group was selected to provide a smallercontext from which to evaluate the effects ofpaired reinforcement and fading.

Condiment and activity preference assessments.Additional MSWO assessments were conductedto identify a preferred condiment and activity topair with the target snack. These assessmentswere conducted after baseline but beforeimplementation of the intervention. Preferenceassessments were conducted with the target low-preference snack paired with multiple condi-ments. For example, Jessica’s least preferredtarget snack was green peppers; therefore,a preference assessment was conducted witha variety of condiments (e.g., salad dressings,Easy CheeseE, ketchup, peanut butter, etc.) onthe green peppers. A variety of condiments were


also assessed for Tim’s and Susan’s target snacks(cereals) such as honey, applesauce, sugar,peanut butter, jelly, and chocolate. The targetsnack without a condiment was also included inthe array to ensure that having a condiment onthe snack was more preferred than the snackalone. During this assessment, condiments wereplaced on top of the snacks. The condimentamounts were selected so that the condimentcovered the snack. Condiment preferenceassessments were conducted with each childuntil preferences were relatively stable; fourcondiment assessments were conducted withTim, three with Susan, and two with Jessica.The most preferred condiment was used in thepairing and fading assessment.

MSWO assessments conducted with activitiesinvolved activities that could eventually besystematically faded (i.e., audio stimulationcould be systematically dialed down). Activitiesincluded in the assessment included a variety ofmovies, music, and toys such as Play-dohE andLegosE. Assessments were conducted untilpreferences were relatively stable; three, five,and two activity preference assessments wereconducted for Tim, Susan, and Jessica, re-spectively. The three most highly preferredactivities were included in the pairing andfading assessment. Three activities were selectedso that the children could be given a choice ofactivities each day.

Reinforcement pairing and fading assessment.Baseline sessions consisted of paired-stimulusassessments with the five assessment snacksextracted from the MSWO assessment rankings.Pairs of stimuli were presented to the childrenon spoons that were placed on tiny plates. Thepairs were presented in a random order, and theright or left placement of each item was rotatedevery other session. At the start of each trial, thechild was instructed to select one of the twoitems and consume it. The trial ended after1 min expired or after the child said that he orshe was done and showed a clean mouth.Consumption was then scored and the next trial

began. When the preference hierarchy wasstable, the reinforcement pairing phase began.

Paired-choice assessments were used for theevaluation of reinforcement pairing and fadingbecause preliminary data (not included in thestudy) suggested that the MSWO assessmentsyielded variable preference patterns over time,and 1 of the children for whom this was the casestated that she liked to ‘‘save the best for last.’’This sort of rule was not likely to affect theresults of paired assessments because notchoosing a preferred item results in omissionof the item, whereas the consumption of thepreferred item is simply delayed in the MSWOassessment.

The snack with the lowest mean rank inbaseline was selected for pairing. Paired-stimu-lus assessments were conducted as in baseline,with the exception that social, edible (condi-ment), and activity reinforcers now accompa-nied the target snack. Specifically, when thetarget snack was presented in this condition, itwas paired with a specific amount of a preferredcondiment placed under the snack item. Thecondiment was placed under the snack item inthis assessment to increase the likelihood thatthe food item and not the condiment wouldinfluence selections towards the close of thefading procedure. Tim’s Shredded WheatE waspaired with 3.0 ml of maple syrup, Susan’s cornflake was paired with 2.8 ml of MarshmallowCremeE, and 2.8 ml of Easy CheeseE wasplaced under a green pepper for Jessica. Theseinitial amounts were selected based on theapproximate amounts that were placed on thesnacks during the condiment preference assess-ment. In addition, after selection of the targetsnack, the children were provided with contin-uous attention (for 1 min), and access to 1 minof a preferred movie (after entering the sessionroom each day, the child was given a choicebetween the three movies that had beenidentified as preferred in an earlier assessment).If the child selected any snack other than thetarget snack, the experimenter provided access


to the item and did not provide access to anyother social or material reinforcers. Each childwas provided with 15 to 30 s of attentionbetween trials.

When the target snack moved up to the firstrank for two to four consecutive sessions, fadingwas initiated. The condiment, attention, andactivity were simultaneously and graduallyeliminated according to a predetermined sched-ule (for an example of a fading schedule, seeTable 1 for that used with Tim). The preferredcondiments were eliminated by gradually re-ducing the amount paired with the target snackuntil the target snack was once again presentedalone (which occurred with Tim and Susanonly). The movies were eliminated in two ways.First, the volume on the television was graduallyturned down until there was no sound, andsecond, the television was gradually turned awayfrom the child until the back of the televisionwas facing the child. The 1-min trial waspartitioned into 20 3-s intervals, and attentionwas eliminated by deleting 3 s of attention ateach fading step. This was accomplished byrandomly eliminating an interval of attention at

each fading step. The fading schedule included20 fading steps, followed by a return tobaseline. Experimenters progressed throughfading steps according to the six fading rulesincluded in Table 2. Additional fading steps(e.g., 21 and 22) were added for Tim andSusan. These steps were added when the rank ofthe target snack dropped after returning tobaseline (see Rule 6 in Table 2).

RESULTS

The five assessment snacks selected for thereinforcement pairing and fading assessment(see pretest rankings in Figure 1) for Timwere, in mean rank from the most preferredto the least preferred, Cocoa PuffsE, ColossalSquaresE, honey graham squares, Bran FlakesE,and Shredded WheatE. The assessment snacksfor Susan were Cocoa PuffsE, Cinnamon ToastCrunchE, Frosted FlakesE, corn flakes, andCorn CrispE. The five assessment snacksselected for Jessica were Gummy BearsE,SkittlesE, M&MsE, potato chips, and greenpeppers.

Table 1

Fading Schedule for Tim

Fading step TV placement (degrees) TV volume Condiment (ml) Intervals of attention

Baseline 0.0 01 20 11 3.0 202 20 11 3.0 193 20 10 2.8 184 40 10 2.8 175 40 9 2.6 166 40 9 2.4 157 60 8 2.2 148 60 8 2.0 139 60 7 1.8 1210 80 7 1.6 1111 80 6 1.4 1012 80 6 1.2 913 100 5 1.0 814 100 5 0.8 715 120 4 0.6 616 120 4 0.6 517 140 3 0.4 418 140 3 0.4 319 160 2 0.2 220 160 2 0.2 121 180 0 0.175 022 180 0 0.15 0


The three most preferred movies used in thereinforcement pairing and fading assessmentswere The Lion King, A Bug’s Life, and The Foxand the Hound for Tim; Rhythm & Blue, A Bug’sLife, and Arthur for Susan; and The LittleMermaid, Toy Story, and Aristocats for Jessica.The most preferred condiments were maplesyrup (Tim), Marshmallow CremeE (Susan),and Easy CheeseE ( Jessica). Data for thecomplete activity and condiment preferenceassessments are available from the secondauthor.

Data for the least preferred snack (targetsnack) from the reinforcement pairing andfading assessment can be found in Figure 2.(Data from the other four snacks included inthe preference assessment are not shown.) Theleast preferred item for Tim during baseline wasShredded WheatE, with ranks ranging from 3to 5 and a mean of 4.6. Ranks of his other itemsduring baseline were consistent with the initialMSWO assessments; the mean ranks were 1.6for honey graham squares, 2 for Cocoa PuffsE,2.6 for Colossal SquaresE, and 4.2 for BranFlakesE. When Shredded WheatE was pairedwith reinforcement, this item immediatelymoved up to a rank of 1 and was selected firstfor 14 sessions. A slight disruption was observedat Fading Step 13, but Tim continued to selectthe Shredded WheatE first until the return tobaseline, in which no remnants of the paired

reinforcement were present. Selecting ShreddedWheatE instead of the other items was clearlyinfluenced by the availability of a small amountof additional reinforcement. Shredded WheatEwas ranked first at Step 20 when only 0.2 ml ofmaple syrup was placed under it, the televisionscreen was no longer visible, the volume on thetelevision was set at 1 (barely audible), and hereceived 3 s of attention. By contrast, ShreddedWheatE ranked much lower during the baselineprobes in which these minute amounts ofreinforcement were absent. Attempting to fadeusing more incremental steps was not successful.The assessment ended with Shredded WheatEoccupying either rank 4 or 5 over the last eightsessions.

The item most consistently ranked the lowestfor Susan was corn flakes, with ranks rangingfrom 4 to 5 and a mean of 4.7 (Figure 2).During baseline the mean rank of Susan’sremaining four items (not shown in Figure 2)were Cocoa PuffsE (ranked 1), CinnamonToast CrunchE (ranked 2), Frosted FlakesE(ranked 3.3), and Corn CrispE (ranked 3.8).When reinforcement was paired with cornflakes, this item immediately moved up toa rank of 2 and maintained a rank of 1 or 2 for28 sessions. Susan continued to select the cornflake at Step 20 when the reinforcers weregreatly marginalized. During the return tobaseline, corn flakes dropped to a rank of 5.

Table 2

Fading Rules

Rule Description

1 If the target snack retains rank 1, move onto the next fading step.2 If the rank of the target food drops from rank 1 to ranks of 2 or 3 and remains at rank 2 or 3 for two

consecutive sessions, progress to the next fading step. If the rank of the target food drops from rank 1to ranks of 2 or 3 and intermittently occupies two of these ranks for four consecutive sessions, progressto the next fading step.

3 If the target food drops to rank 4 or 5, return to the previous fading step and remain there until the targetfood is ranked 1, 2, or 3 for three consecutive sessions.

4 If you return to the previous fading step three times, remain at that step until the rankings are stable andend the assessment.

5 During return to baseline, if the target food drops to rank 2 or 3, return to Step 20 and continuefollowing fading rules.

6 During the return to baseline, if the target food drops to rank 4 or lower, return to Step 20 and addadditional, more incremental fading steps.


We returned to Step 20, added additionalfading steps, and then returned to an earlierfading step. However, we were unable torecover a rank of 1 or 2 for corn flakes.

The least preferred item during the re-inforcement and pairing baseline for Jessicawas the green pepper, with ranks ranging from3 to 5 and a mean of 4.6 (Figure 2). Duringbaseline, the mean ranks of Jessica’s remainingfour items (not shown in Figure 2) wereGummy BearsE (ranked 1), SkittlesE (ranked

3), M&MsE (ranked 3.2), and potato chips(ranked 3.4). Green pepper immediately movedup to a rank of 1 when it was paired withreinforcement and maintained the rank of 1 or2 for 22 sessions. Variability in rankings wasobserved at Step 14. High rankings were notrecovered by returning to a previous step; theassessment ended with highly variable rankingsof the green pepper, which were just slightlyhigher than those observed during baseline.

Figure 1 depicts the results from the pre- andpost-MSWO assessments for each child. The 14snack items not paired with reinforcementmoved an average rank of 2.4 in either direction(range, 2.1 to 2.8). Therefore, a shift in rankabove 3 units might be noteworthy. The meanrank for Shredded WheatE was 13.0 in the firstassessment and 13.8 in the second assessmentfor Tim. The mean rank for corn flakes was 10in the first assessment and 13 in the secondassessment for Susan. The mean rank for thegreen pepper was 11.2 in the first assessmentand 9 in the second assessment for Jessica.

DISCUSSION

An immediate and sustained effect for pairedreinforcement was observed across children.These results are consistent with the results ofAhearn (2003), Hanley et al. (2003), Mueller etal. (2004), and Piazza et al. (2002) in thatpreference for an originally less preferred itemincreased when it was paired with a highlypreferred item. If strong preferences for thetarget snack were maintained in the absence ofthe paired reinforcement, we would haveconcluded that preference had been changedvia flavor–flavor conditioning (stimulus–stimu-lus pairings). However, the ranks of the targetsnacks decreased as reinforcement was elimi-nated for all children. The pre- and post-MSWO assessments also showed a lack ofpreference shift in a broader context (14 othersnack choices as opposed to the four optionsduring the paired-choice assessments). Althoughthe posttest ranking of one target item (green

Figure 1. Pre- and post-MSWO preference assessmentsof 15 snack foods for Tim, Susan, and Jessica. Filled circlesrepresent pretests, and open circles represent posttests.Asterisks above data points represent snacks that werepaired with reinforcement.


pepper for Jessica) increased compared to theinitial MSWO ranking, this change was notgreater than the mean change of other items inthe assessment. Because strong preferences forthe target snacks were not maintained asreinforcement was removed and preferenceshifts were not observed in subsequent assess-ments, we can safely conclude that we did notcondition a preference via stimulus–stimuluspairings, and the initial shifts in preference werea direct function of the added reinforcers.

It is likely that the initial shifts in preferencewere a result of differential reinforcement. That is,

the children received highly preferred condiments,movies, and attention only if the target food wasselected. The current procedures succeeded inaltering the momentary value of the items, asevident by the large shifts in preference followingthe introduction of the additional reinforcers.However, based on the fact that the target snacksresumed their former status when the addedreinforcers were eliminated, it is apparent that ourprocedures were not successful in altering thepermanent value of the items.

Procedural differences between the currentstudy and previous work may account for the

Figure 2. Pairing and fading assessments for Tim, Susan, and Jessica. Circles represent baseline, and triangles indicatepairing and fading sessions. The numbers above the triangles indicate fading steps.


absence of a conditioning effect. For example,Mueller et al. (2004) blended the less andhighly preferred foods so that the less preferredfoods were not readily discernible, at leastinitially. In the current study, the two foodswere not blended, and the children were able tosee the less preferred foods at all pairing steps.Strong control of preference by the pairedreinforcer may have developed during ourassessment but may have been successfullyavoided with the procedures described byMueller et al. In addition, previous studies(Kern & Marder, 1996; Mueller et al.; Piazza etal., 2002) included the use of escape extinctionin the form of representation of the food andnonremoval of the spoon for some of theparticipants. Trials continued regardless of thechild’s choices in the current study, and trialtime was not increased until the child consumedthe target food. If the children did not selectand consume the target snack, they were able toconsume an alternative snack. Thus, thechildren could easily avoid consuming thetarget food in our study; this was not the casein previous studies. Finally, it is plausible thatthe paired reinforcement may have been morereinforcing for the participants in the Hanley etal. (2003, 2006) studies than the pairedreinforcement used with the participants inthe current study or that the difference in thevalue of the target activities in the earlier studieswas small compared to the difference in thevalue of the high- and low-preference snacks inthe current study.

Our results are similar those of Ahearn(2003), in that preference for an originally lesspreferred item increased with reinforcementpairing, but the preference shift was notmaintained when the paired reinforcementwas removed. Our procedures differed fromAhearn’s in that we gradually and systematicallyeliminated the paired reinforcers, but the fadingprocedure was ultimately not effective in pro-ducing a conditioned preference. The resultsfrom the current study and Ahearn suggest that

flavor–flavor conditioning (Capaldi, 1996) maybe more elusive than previously thought, andmay not be the most effective method to alterfood preferences for typically developing pre-schoolers. However, it is possible that ourmeasurement system was insensitive to changesin the value of the target snacks. For instance,Jessica may not have chosen green peppers overother snack options, but she may have con-sumed more green peppers when presentedsingly following our intervention than before it(similar to the results reported by Mueller et al.,2004). It is also possible that a more gradualschedule for eliminating the paired reinforcers,with special attention paid to transitioning fromsmall magnitudes of reinforcement to noreinforcement, might have resulted in successfulconditioning.

Although paired reinforcement was effectivein altering preference with all children, and thiseffect persisted even under conditions involvingminute amounts of the paired reinforcement for2 children, it is not clear from these datawhether the activity, condiment, social interac-tion, or all of the paired items were responsiblefor the initial shift in preference. Because wewere able to strengthen preference for lesspreferred snacks with paired reinforcement, itmay be reasonable to continue to embed targetsnacks in these presumably more reinforcingcontexts to maintain consumption of thesesnacks. It would be difficult to justify thecontinued use of a condiment to increaseconsumption of healthy snacks, unless a verysmall amount or intermittent use of high-sugaror high-fat condiments were successful formaintaining consumption of healthy snacks.At this point, it seems important to determinewhether intermittent condiment pairings wouldbe successful as opposed to the gradual andcomplete elimination of the condiment. Itwould likely be acceptable to continue socialinteraction and possibly some auditory or visualstimulation during snack periods. Future re-search should determine whether desirable shifts


in food preferences can be achieved with onlysocial and activity-based reinforcers. Previouswork by Birch, Zimmerman, and Hind (1980)has suggested that pairing a food item withadult attention may increase preference for thatfood item. However, the measure of preferencewas very different from that of the currentstudy. Birch et al. relied on verbal report ofpreference from the children as opposed todirectly observed preferences. In addition, somestudies suggest that using preferred foods asrewards (Birch et al.) as well as peer modeling(Birch, 1980) may be effective alternativemethods for developing preferences for typicallynonpreferred foods in preschoolers.

Effective intervention for and prevention ofobesity are important goals, given the currentepidemic with children and adults in the UnitedStates and around the world (Gill, King, &Caterson, 2005). Interventions that target in-creased consumption of healthy foods arerelevant. Repeatedly exposing children tohealthy snacks and restricting access to lesshealthy snacks are clearly effective methods forincreasing children’s intake of healthy snacks.However, restricting access to unhealthy foodsmay become increasingly difficult as childrengrow older and more independent. Therefore,helping children to make healthy choices in thepresence of less healthy choices may benecessary. These challenging circumstances werearranged in the current study by presentingyoung children less preferred snack options inthe presence of highly preferred snack options.These challenging conditions were likely re-sponsible for the marginal outcomes of theseintensive behavioral interventions, which pre-viously have been shown to both alter theabsolute value of food (Mueller et al., 2004)and condition preferences for nonfood items(Hanley et al., 2003, 2006).

An additional important implication of ourresults is that although any stimulus may beconditioned to be a reinforcer (Miltenberger,1997), the specific conditions have not been

well documented. Questions regarding theamount of pairing required, the relevant timeframes for conditioning to occur, and thestrategy for removing the small amounts ofpaired reinforcement while maintaining theconditioned effect remain to be answered.Although researchers continue to evaluate theparameters of effective preference conditioning,additional research should also be directedtowards identifying alternative interventions forpromoting independent healthy snack selectionsby typically developing young children.

REFERENCES

Ahearn, W. H. (2003). Using simultaneous presentationto increase vegetable consumption in a mildlyselective child with autism. Journal of AppliedBehavior Analysis, 36, 361–365.

Bayens, F., Eelen, P., Van den Bergh, O., & Crombez, G.(1990). Flavor-flavor and color-flavor conditioning inhumans. Learning and Motivation, 21, 434–455.

Birch, L. L. (1980). Effects of peer models’ food choicesand eating behaviors on preschoolers’ food prefer-ences. Child Development, 51, 489–496.

Birch, L. L., Zimmerman, S. I., & Hind, H. (1980). Theinfluence of social-affective context on the formationof children’s food preferences. Child Development, 51,856–861.

Capaldi, E. D. (1996). Conditioned food preferences. InE. D. Capaldi (Ed.), Why we eat what we eat: Thepsychology of eating (pp. 53–80). Washington, DC:American Psychological Association.

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

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.

Freeman, K. A., & Piazza, C. C. (1998). Combingstimulus fading, reinforcement, and extinction to treatfood refusal. Journal of Applied Behavior Analysis, 4,691–694.

Gill, T., King, L., & Caterson, I. (2005). Obesityprevention: Necessary and possible. A structuredapproach for effective planning. Proceedings of theNutrition Society, 64, 255–261.

Hanley, G. P., Iwata, B. A., & Roscoe, E. M. (2006).Some determinants of changes in preference overtime. Journal of Applied Behavior Analysis, 39,189–202.


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.

Holman, E. W. (1975). Immediate and delayed re-inforcers for flavor preferences in rats. AnimalLearning & Behavior, 6, 91–100.

Ireton, C. L., & Guthrie, H. A. (1972). Modification ofvegetable-eating behavior in preschool children.Journal of Nutrition Education, 4, 100–102.

Kern, L., & Marder, T. J. (1996). A comparison ofsimultaneous and delayed reinforcement as treatmentfor food selectivity. Journal of Applied BehaviorAnalysis, 29, 243–246.

Miltenberger, R. G. (1997). Behavior modification:Principles and procedures. Pacific Grove, CA:Brookes/Cole.

Mueller, M. M., Piazza, C. C., Patel, M. R., Kelley, M. E.,& Pruett, A. (2004). Increasing variety of foodsconsumed by blending nonpreferred foods intopreferred foods. Journal of Applied Behavior Analysis,37, 159–170.

Patel, M. R., Piazza, C. C., Kelly, M. L., Ochsner, C. A.,& Santana, C. M. (2001). Using a stimulus fadingprocedure to increase fluid consumption in a childwith feeding problems. Journal of Applied BehaviorAnalysis, 34, 357–360.

Piazza, C. C., Patel, M. R., Santana, C. M., Goh, H., Delia,D. D., & Lancaster, B. M. (2002). An evaluation ofsimultaneous and sequential presentation of preferredand nonpreferred food to treat food selectivity. Journalof Applied Behavior Analysis, 35, 259–270.

Riordan, M. M., Iwata, B. A., Wohl, M. K., & Finney, J.W. (1980). Behavioral treatment of food refusal andselectivity in developmentally disabled children.Applied Research in Mental Retardation, 1, 95–115.

Zellner, D. A., Rozin, P., Aron, M., & Kulish, C. (1983).Conditioned enhancement of humans’ liking forflavor by pairing with sweetness. Learning andMotivation, 14, 338–350.

Received August 1, 2005Final acceptance April 9, 2007Action Editor, Iser DeLeon


the effects of reinforcer pairing and fading on preschoolers' snack selections

Documents