AGE TRENDS IN STIMULUS OVERSELECTIVITY

LOUISE MCHUGH AND PHIL REED

SWANSEA UNIVERSITY

Stimulus overselectivity refers to the phenomenon whereby stimulus control over behavior is exertedonly by a limited subset of the total number of stimuli present during discrimination learning. It often isdisplayed by individuals with autistic spectrum disorders or learning disabilities, but is not exclusive tothose groups. The present studies investigated the impact of aging on stimulus control andoverselectivity. Three age groups—18–22, 47–55, and 70–80 year olds—were studied in twoexperiments. All participants were trained on a simple discrimination task, randomly assigned to oneof two conditions (either with or without a distractor task), and then tested for the emergence ofoverselectivity (Experiment 1). In Experiment 2 responding controlled by the overselected stimuluselements was reduced by introducing a verbal punisher. In subsequent tests, control of behavior by thepreviously underselected elements from Experiment 1 was enhanced across the two younger age groupsbut not the oldest group of participants. The results are discussed in relation to the attention-deficit andovershadowing accounts of overselectivity.

Key words: stimulus overselectivity, simple discrimination task, verbal reinforcer, verbal punisher,young adults, middle-aged adults, older adults

_______________________________________________________________________________

Stimulus overselectivity refers to the phe-nomenon whereby control over behavior isexerted only by a limited subset of the totalnumber of stimuli present during discrimina-tion learning (see Dube & McIlvane, 1999;Lovaas & Schriebman, 1971). It is oftendisplayed by individuals diagnosed with autis-tic spectrum disorders (ASD) or learningdisabilities, can result in limited learning interms of the range or number of stimulusfeatures that come to control behavior, andmay retard the acquisition of certain beha-viors. Although the majority of the research inthis area has been conducted with personsdiagnosed with ASD (Lovaas & Schriebman),a new body of research has begun to explorethis phenomenon in normally developingadults (Reed & Gibson, 2005).

There are two major reasons for the study ofoverselectivity: firstly, it gives insight into theprocesses of stimulus control of behavior; andsecondly, it examines the role overselectivityplays in impeding the formation of complexconcepts (Cumming & Berryman, 1965).According to Lovaas (1980), stimulus over-selectivity could cause many of the behavioralproblems commonly found in autism. Forexample, the understanding of speech involvesa number of cues, including hearing what the

other person is saying, watching lip move-ments and facial expressions, etc. In theextreme case, stimulus overselectivity maycause an individual to focus on only one ofthese cues. In turn, this may disrupt thedevelopment of an understanding of languageor, indeed, the ability to use language ina meaningful way (Lovaas, Schreibman, Koe-gel, & Rehm, 1971; Schreibman & Lovaas,1973).

The overselectivity effect bears on issuesfundamental to understanding the circum-stances under which stimulus control willoccur. Explanations proposed for stimulusoverselectivity have been based on the assump-tion that the problem is due to the failure ofstimulus elements to be learned during initialtraining. For example, an attention deficit canbe considered equivalent to the absence ofstimulus control and has been proposed toaccount for overselectivity in autistic individu-als (see Koegel & Schreibman, 1977). Analysisof the eye movements of retarded childrenwho display overselectivity when confrontedwith a complex stimulus appears to corrobo-rate this view. Such individuals do not appearto examine all the stimuli elements present inthe complex stimulus (Dube et al., 1999).Hence, these unattended elements do notcontrol behavior in subsequent test trials. (Acautionary note here is that eye movementsare imperfectly correlated with attention;Remington, 1980.)doi: 10.1901/jeab.2007.88-369

Correspondence concerning this article should beaddressed to Louise McHugh, Department of Psychology,Swansea University, email: l.mchugh@swansea.ac.uk.

JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 2007, 88, 369–380 NUMBER 3 (NOVEMBER)

369

An alternative argument is that all stimuluselements are attended to during training andlearned, but that only a subset of theseelements comes to control behavior. Whena stimulus is presented by itself, it may exertstrong control over behavior. However, pre-vious research indicates that, if that stimulus isaccompanied by another stimulus, then con-trol by the former stimulus may be reduced oreliminated by the presence of the latter (seeMackintosh, 1975). This effect is more com-monly known in the animal conditioningliterature as overshadowing (see Trabasso &Bower, 1968). Recent research has suggestedovershadowing as a model for overselectivity(Gibson & Reed, 2005).

In the animal conditioning literature, appar-ently overshadowed stimuli, or stimuli that donot strongly control behavior, can be estab-lished as controlling stimuli by extinguishingother, more powerful or salient cues (e.g.,Kaufman & Bolles, 1981; Matzel, Schachtman,& Miller, 1985; Reed & Reilly, 1990). This effectwas demonstrated by Wilkie and Masson(1976). They exposed pigeons to compoundstimuli consisting of color and shape elements.Grain reinforcers were given in the presence ofa specific compound stimulus consisting of onecolor element and one shape element (the S+).Once pecking was exclusively directed to theS+, the color and shape elements were pre-sented individually without reinforcement.Each pigeon responded only to the colorelement and not to the shape element. In thenext phase of training, pecking was reinforcedwhen the shape (i.e., the overshadowed orunderselected) element alone was present.Pecking in the presence of the shape elementwas acquired more rapidly by the pigeons thatpreviously had been exposed to the shapeelement than by pigeons that previously hadnot been exposed to the shape element (i.e.,there were savings in the relearning task). Thisfinding suggests that, although the pigeonsexposed to the compound stimuli had learnedthe shape element during training, this ele-ment had not controlled behavior when it wascombined with the color element, that is, wasnot sufficient to control performance when themore salient stimulus was present simulta-neously.

To date there are few demonstrations thatan underselected stimulus demonstrates en-hanced stimulus control following extinction

of the overselected stimulus in humans (seeBroomfield, McHugh, & Reed, in press). Mostdemonstrations of the overselectivity effecthave been with clinical populations (e.g.,ASD), which may raise questions about thegenerality of the processes at work, that is, itmay be that clinical populations are influ-enced by different factors than nonclinicalpopulations. Previous research by Reed andGibson (2005) demonstrated that stimulusoverselectivity can be observed in humanparticipants without disabilities when theyconcurrently perform an additional task (e.g.,a concurrent memory task). In those studies,participants were presented with a simulta-neous discrimination task in which they had tolearn through trial and error to select one 2-element compound in preference to another2-element compound. Participants who weregiven a concurrent task subsequently re-sponded to one element of the compoundfar more than to the other, suggesting anoverselectivity effect. These findings led tofurther research on the basic processes in-volved in the phenomenon of overselectivityinvolving non-autistic adult participants.

The present study used a nonclinical sampleto examine the overselectivity phenomenonand also compared the effects of age onoverselectivity. Experiment 1 exposed youngadults, middle-aged adults, and elderly adultsto an overselectivity procedure, while manipu-lating the presence of a concurrent distractortask. If, for example, higher levels of over-selectivity emerged for the oldest age groupthan for the other two, this result woulddemonstrate a relation between age, distrac-tion, and overselectivity.

EXPERIMENT 1

A study by Reed and Gibson (2005) exam-ined whether overselectivity occurred in hu-man participants without disabilities in condi-tions where they were presented witha distracting task. As previously noted, partic-ipants were exposed to a simultaneous dis-crimination procedure in which they learnedthrough trial and error to select one 2-elementcompound over a second 2-element com-pound. Participants were randomly assignedto one of two groups. One group was requiredto concurrently memorize a stimulus grid (thedistracting task) while the other group was

370 LOUISE MCHUGH and PHIL REED

presented with the same grid but not requiredto memorize it. Participants with the distract-ing task subsequently responded to oneelement of the stimulus compound far morethan to the other, thus demonstrating theoverselectivity effect.

Experiment 1 extended the work carried outby Reed and Gibson (2005) by including threedifferent age groups to explore whether over-selectivity increased with age and, if so,whether that increase was related to thedistractor task. Research has documentedvarious declines in ability with age such asthe ability to process complex cues (Ardle,Ferrer-Caja, Hamagami & Woodcock, 2002).One potential explanation for such declinesmight be overselectivity. Employing the over-selectivity procedure previously used by Gib-son and Reed (2005) may provide an effectiveprocedure for determining whether overselec-tivity increases with age and is amplifiedfurther by a distractor task.

METHOD

Participants

Forty-eight participants, 16 in each of threeage groups (18–22, mean 19.5 6 1.2 – thestandard deviation; 47–55, mean 50.1 6 2.3;70–80, mean 73.1 6 2.7) participated. Allparticipants were volunteers recruited throughfaculty board announcements and from per-sonal acquaintances of the experimenters. Theparticipants in each age group were randomlyassigned to one of two experimental groups (n5 8). Participants did not receive any paymentfor participation.

Apparatus and Materials

The experiment was conducted in a roomfree of ordinary distractions. Participants werepresented with white cards that measured15 cm by 10 cm. Some of the cards containedtwo stimulus elements. These were the com-pound stimuli. The stimulus elements werecharacters obtained from various fonts availablein Microsoft Word 2000. The fonts wereWingdings, Wingdings 2, and Symbol. Verbalreinforcement was given for selecting one pairof characters (here designated by letters) overanother pair; for example, selecting A and Bover C and D, and E and F over G and H.Additional cards of the same size depicted onlyone of the elements of the compound stimulus.

The 4 3 4 grid with four different shapesthat was used in the distractor task is presentedin Figure 1 (see also Reed & Gibson, 2005).

Procedure

Distractor task. A distractor task was imple-mented as a between-subject variable. Half ofthe participants in each age group wererandomly selected and first required to mem-orize the stimulus grid (Figure 1), which waspresented for 20 s. Participants were presentedwith the distractor task individually. They wereinformed that they would be required toreplicate the grid by drawing it on paper atthe end of the experiment but that the gridwas not relevant to the next part of theexperiment. The other half of the participantsin each group were shown the grid for thesame length of time at the beginning of theexperiment, but no instruction to replicate (ormemorize) the grid was provided.

Training phase. Participants were trainedindividually. The experimenter sat directlyopposite the participant throughout the ex-periment. The stimulus cards were placed atthe center of the table between the participantand the experimenter. Participants were in-structed as follows:

You will be shown two cards containing two symbolson each. Please select a card by pointing to that card.Point to the card rather than an individual symbol.

Fig. 1. The grid used in the distractor condition inExperiment 1.

STIMULUS OVERSELECTIVITY AND AGE TRENDS 371

You will be given feedback of ‘‘yes’’ for some cardsand ‘‘no’’ for others. Your choices will be recorded.

Participants were presented with two cardssimultaneously. Each card contained twostimulus elements. On any given trial, partic-ipants were presented with one compoundstimulus card (e.g., here symbolized as AB orEF) that, if selected by the participant, resultedin verbal reinforcement in the form of theexperimenter saying ‘‘yes.’’ In other words,that card became an S+. Selecting the othercard (e.g., CD or GH) resulted in theexperimenter saying ‘‘no’’; in this case, thecard became an S2. The positions of the cardswere systematically randomized, that is, on 50percent of the trials the correct card waspresented on the left and otherwise on theright. Each trial lasted until a response wasmade and each intertrial interval was approx-imately 5 s.

During the training phase, a specific S+ wasalways paired with a specific S2. There weretwo S+ and two S2 stimuli. Participants wereconsidered to have acquired the discrimina-tion once they produced 10 consecutivelycorrect responses. The stimulus elements usedin the compounds were different for eachparticipant in order to prevent an intrinsicallymore salient stimulus from always having thesame role.

Test phase. The test phase followed thetraining phase. During the test phase of theexperiment the participants were presentedwith two cards simultaneously, each containingonly one character from the previous com-pound stimulus to which it belonged duringthe training phase. That is, one card containedan element of a previous S+ and the other anelement of a previous S2. Specifically, A or Belements were presented with C or D ele-ments, and E or F elements were presentedwith G or H elements, respectively. There werefive trials for each combination of elements fora total of 40 trials. No feedback was providedduring these trials. To ensure that the initialdiscriminations involving compound stimuliwere maintained throughout testing, each pairof compound stimulus cards (S+ and S2) alsowas presented 10 times during probe trials inthe test phase for a total of 20 probe trials. Nofeedback was provided during probe trials.The order of presentation of the compoundstimulus and single stimulus cards was ran-

domized over the total of 60 test trials. Thetraining and test trials all took place within thesame session.

RESULTS

Trials to Criterion in the Training Phase

The mean numbers of trials to criterion forthe three age groups in the no-distractorcondition were: 15.88 (s.d. 6 3.3), 17.38 (s.d.6 3.9), and 21.75 (s.d. 6 6.4) for the 18–22,40–55, and 70–80 year-old groups, respective-ly. The means for the three age groups in thedistractor condition were: 18.06 (s.d. 6 4.23),20.31 (s.d. 6 5.55), and 28.13 (s.d. 6 8.16),respectively. Thus, fewer trials were requiredwhen there was no distractor. Also, thenumber of trials to criterion increased asa function of age, irrespective of the presenceor absence of the distractor, with the fastestacquisition demonstrated by the youngestgroup and slowest acquisition by the oldestgroup.

To determine whether these trends werestatistically significant, a 233 mixed-modelanalysis of variance (ANOVA) was conducted,with distractor versus no distractor and agegroup (18–22, 47–55, 70–80) as the between-subject variables and trials to criterion as thedependent measure. The analysis revealed thatthere was a significant main effect for distrac-tion [F(1,45) 5 88.88, p , 0.0001], indicatingthat the use of the distractor increased thenumber of trials to criterion, a statisticallysignificant main effect for age group [F(2,45)5 9.76, p , 0.0001], indicating that the agegroups differed in the number of trials re-quired to meet the criterion, and a significantinteraction effect for distractor task by agegroup [F(2,45) 5 10.05, p , 0.0001].

To investigate where the significant effectsfor age group emerged for trials to criterion inthe distractor versus no-distractor conditions,simple-effects analyses were conducted. Theyshowed that there was a significant differencebetween the 18- to 22-year-old groups [F(1,45)5 9.65; p , .05], the 47- to 55-year-old groups[F(1,45) 5 96.52; p , .05], and the oldestgroups [F(1,45) 5 769.17; p , .05]. That is,the inclusion of the distractor task increasedthe number of trials to criterion in all threeage groups.

The differences between the numbers oftrials to criterion between the age groups alsowere compared using independent t-tests. The

372 LOUISE MCHUGH and PHIL REED

results revealed that the youngest age grouprequired significantly fewer trials to criterionthan the oldest group in both the no-distractorcondition [t(30) 5 3.26, p , .01] anddistractor condition [t(30) 5 4.38, p , .001].The same pattern also emerged between themiddle-aged group and the older-age group inboth the no-distractor condition [t(30) 5 2.33,p , .05] and the distractor condition [t(30) 53.17, p , .01]. There was no significantdifference between the youngest and middle-aged groups in number of trials to criterion foreither the distractor (p . .05) or no-distractorconditions (p . .05). The significant differ-ences that emerged between the groupssuggested a developmental trend in terms oftrials to criterion, with the two younger groupsof participants meeting the criterion morerapidly than the oldest group in both thedistractor and nondistractor conditions.

More-Often- versus Less-Often-Selected Elements inthe Test Phase

Correct responding on probe trials in thetest phase was consistently above 80%, in-dicating that the initial discriminations weremaintained throughout testing. The meannumber of times that each of the elements ofthe S+ stimuli was selected during the testphase was calculated in order that the pair ofmore-often-selected elements and the pair ofless-often-selected elements could be identi-fied. The results from the distractor and no-distractor conditions are shown in Figure 2 foreach group.

Inspection of these data indicates that in thedistractor condition, there was a larger differ-ence between the percentage of times themore-often- and less-often-selected elementswere chosen than in the no-distractor condi-tion. A 23233 mixed-model ANOVA wasperformed with element type (more-often-selected versus less-often-selected) as the with-in-subjects variable, distraction (distractor ver-sus no distractor) and age group (18–22, 47–55, and 70–80) as the between-subjects vari-ables, and percentage of times the more-often-or less-often-selected elements were selected asthe dependent measure. This analysis showedthat there was a significant main effect formore-often-selected versus less-often-selectedelements [F(1,45) 5 124, p , 0.0001],a significant main effect for distraction[F(1,45) 5 90.74, p , 0.0001], and a significant

interaction effect for the two variables [F(1,45)5 33.29, p , 0.0001]. There was also a signif-icant effect for age group [F(2,45) 5 48.97, p, 0.0001], a significant interaction betweenage group and the selected elements [F(2,45)5 11.63, p , 0.0001], and a significant three-way interaction between the three variables[F(2,45) 5 3.6, p , 0.05].

In order to determine where the significantdifferences emerged between the more-often-and less-often-selected elements across thethree age groups, a 232 repeated measuresANOVA was conducted for each age groupwith distraction as the between-subjects vari-able, element type as the within-subjects vari-able, and percentage of times the elementswere selected as the dependent measure. Theanalysis for the 18- to 22-year-old grouprevealed a significant effect for distraction[F(1,15) 5 39.19; p , .0001] and stimuluselement type [F(1,15) 5 23.6; p , .0001] anda significant interaction between the twovariables [F(1,15) 5 4.36; p , .05]. For the40- to 55-year-old group, the analysis revealeda significant effect for distraction [F(1,15) 534.71; p , .0001] and for element type[F(1,15) 5 40; p , .0001] and a significantinteraction [F(1,15) 5 15.64; p , .001]. Forthe 70- to 80-year-olds, distraction was signifi-cant [F(1,15) 5 29.39; p , .0001], as waselement type [F(1,15) 5 63.95; p , .0001], andthere was a significant interaction between thetwo [F(1,15) 5 18.46; p , .001].

Fig. 2. Mean percentage and standard error ofselection in the test phase of the less-often-selected andmore-often-selected stimulus elements in both the dis-tractor and no-distractor conditions for the three agegroups in Experiment 1.

STIMULUS OVERSELECTIVITY AND AGE TRENDS 373

A series of planned comparisons betweenthe more-often- and less-often-selected ele-ments was conducted comparing the distractorand no-distractor conditions in terms of themean percentages of selection of the two typesof stimulus element for each age group. Forthe youngest group of participants there wasa significant difference between the more-often-selected elements in the distractor con-dition and those in the no-distractor condition[t(15) 5 4.39; p , 0.001] and a significantdifference between the less-often-selected ele-ments in the two conditions [t(15) 5 4.74, p ,0.001]. The same pattern emerged for themiddle-aged group [t(15) 5 3.48, p , .01 andt(15) 5 6.48, p , 0.001]. However, there wasno significant difference between the more-often-selected elements for the older groupwhen compared across the two conditions[t(15) 5 1.95, p , 0.07]. This indicated thatthe elderly participants overselected in bothconditions. However, there was a significantdifference between the conditions for the less-often-selected elements at the same age level[t(15) 5 5.86, p , 0.001].

To summarize, there appeared to be a cleartrend in overselectivity: It increased with ageand with the use of the distractor task.Additionally, the effect of the distractor in-creased with age.

Individual Participants’ Data.

The individual participants’ data are pre-sented in Appendix 1. As can be seen in theno-distractor condition, there is a distinctdifference in percentage of selection betweenthe more-often-selected element and the less-often-selected element. This distinction be-comes more pronounced in the distractorcondition.

The numbers of participants who exhibiteda difference of 20% or greater in theirselection between the two types of stimuluselements were 2, 1, and 7, for the young,middle-aged, and elderly groups, respectively,in the no-distractor condition, and 5, 7, and14, respectively, in the distractor condition. Achi-squared analysis compared whether thenumber of participants meeting this criteriondiffered across the age groups. Overall, 21.9%of the participants in the 18- to 22-year-oldgroup met the criterion, 25% in the 47- to 55-year-old group did so, and 65.6% of the 70- to80-year-old group. The chi-squared analysis

indicated a significant difference between the18- to 22-year-olds and 70- to 80-year-old group[x2(1) 5 12.44, p , .001, W 5 .44] andbetween the 47- to 55-year-old and 70- to 80-year-old groups [x2(1) 5 10.65, p , .001, W 5.41], suggesting a significantly higher level ofoverselectivity in the oldest group.

To summarize, the distractor task increasedthe level of overselectivity, and the level ofoverselectivity increased as a function of age.Specifically, the participants in the oldest agegroup displayed higher levels of overselectivitythan those in the two younger groups. Theresults replicate the previous findings of Reedand Gibson (2005), who showed that adultparticipants were more likely to overselectwhen a distractor was used.

DISCUSSION

Previous work by Broomfield et al. (in press)explored whether underselected elementswould control behavior after a verbal punisherwas administered following the selection of theoverselected elements. In fact, responding tothe previously underselected elements wasenhanced. This was the first demonstrationof enhanced behavioral control by under-selected elements as a consequence of punish-ing the selection of the overselected elements.These results have implications for the re-mediation of overselectivity in populationswhere it may inhibit learning. To date,however, no study has attempted to enhancecontrol by underselected elements using par-ticipants from a normal adult population whodemonstrated higher levels of overselectivity.Experiment 2 was designed to explore whetherthe enhancement of stimulus control byunderselected elements could be facilitatedin participants who demonstrated the highestlevels of overselectivity in Experiment 1,namely, the older adults.

EXPERIMENT 2

As previously mentioned, possible causes ofoverselectivity have been suggested in theliterature. Dube et al. (1999) suggested thepossibility of an attentional deficit, that is, thatnot all the elements that compose a complexstimulus are learned. An alternative view isthat, although all of the stimulus elements arelearned, overshadowing may take place whenrelative stimulus control is tested with a simul-

374 LOUISE MCHUGH and PHIL REED

taneous discrimination procedure. Also, thepossibility that two separate processes affectstimulus control has been suggested by Whiteand Ruske (2002). Experiment 2 aimed toassess whether the higher levels of overselec-tivity observed in the older-adult participantsin Experiment 1 was an attentional deficit orthe result of overshadowing. To this end, thepreviously overselected elements were pun-ished verbally, and the previously underse-lected elements were then tested for anychange in the stimulus control that theyexerted when tested again against the ele-ments from the previously nonreinforcedcompound. Experiment 2 also addressed thequestion of whether the enhancement ofstimulus control by the underselected ele-ments would differ across the three age groups(18–22, 47–55, and 70–80 year olds).

METHOD

Participants, Apparatus, and Materials

Twenty-four of the participants in Experi-ment 1 also participated in Experiment 2.Specifically, they were the 8 participants fromeach age group for whom the distractorcondition was in place during the trainingphase of Experiment 1. The setting, apparatus,and materials were identical to those em-ployed in Experiment 1 with the exceptionthat novel stimulus elements were introducedin the training phase. No payment for partic-ipation was provided.

Procedure

Training phase. The elements that wereselected more often in Experiment 1 (i.e.,the overselected elements) were identified foreach participant. Training trials were con-ducted in which one of the more oftenselected elements and a novel element werepresented at the same time, each on a separatecard. There were four novel elements, one ofwhich was selected randomly for each of theparticipants. A verbal reinforcer (‘‘yes’’) fol-lowed selection of the novel element. A verbalpunisher (‘‘no’’) followed selection of thepreviously overselected element. Training con-tinued until the participants selected the novelelement in 10 consecutive trials.

Test phase. The test phase followed thetraining phase. The test procedure involvedthe same 40 trials consisting of individual

elements of S+ and S- that were included inthe test phase of Experiment 1. UnlikeExperiment 1, no probe trials containing theoriginal compound stimuli were presented.

RESULTS

The mean percentages of selection of thepunished elements (i.e., the originally more-often-selected elements) and the mean per-centages of selection of the originally less-often-selected elements from Experiment 1(unpunished elements) were calculated foreach age group and are presented in Figure 3.For purposes of comparison the percentagesof selection of the originally underselectedand overselected elements in Experiment 1also are presented in the Figure 3.

The results from the test phase of Experi-ment 2 were compared to those from the testphase of Experiment 1 using the percentagesof selection of the more-often-selected stimu-lus elements across the three age groups in thedistractor condition. To that end, the meanpercentage of selection of the originally more-selected stimulus elements from Experiment 1was subtracted from the mean percentage ofselection of these now punished (but pre-viously more often selected) stimulus elementsfrom Experiment 2. A negative differenceindicates a lower percentage of selection inExperiment 2. The smallest differenceemerged for the older-adult group [228.75(617.27)], compared to the two youngergroups of participants [18–22: 231.25(614.6); 47–55: 231.25 (69.9)]. The differ-ences clearly show that stimulus control by theoriginally underselected stimulus elementsemerged following punishment of the origi-nally overselected elements.

A 233 mixed ANOVA was performed.Selection of the punished elements versusselection of the originally less-often-selectedelements was the within-subjects variable, agegroup was the between-subjects variable, andmean percentages of selection of element typewas the dependent measure. The analysisrevealed a significant main effect for elementselected [F(1, 21) 5 15.59, p , 0.001],a significant main effect for age group [F(1,21) 5 17.24, p , 0.0001], and a significantinteraction effect [F(2, 21) 5 5.934, p , 0.01].These results indicate that the use of theverbal punisher may have reversed the selec-tion of the two types of stimulus element.

STIMULUS OVERSELECTIVITY AND AGE TRENDS 375

A series of planned comparisons was con-ducted to determine where the significantdifferences in element selection emerged withregard to age group. Simple-effects analysisrevealed a significant difference for the 18- to22-year-old-group [F(1,4) 5 26.71; p , 0.05],for the 40- to 55-year-old group [F(1,4) 5 8.7; p, 0.05], and for the 70- to 80-year-old group[F(1,4) 5 19.22; p , 0.05], but the differencefor the last group was in the opposite directionfrom that for the two other groups.

For the two younger age groups, theoriginally less-often-selected elements becamemore salient (that is, were selected moreoften) when the originally more-often-selectedelements were punished. This did not occur inthe older-adult group. That is, the enhance-ment of stimulus control by the originally less-often-selected elements occurred in the twoyounger groups but not in the older-adultgroup.

Individual Participants’ Data

For each participant, Appendix 2 displaysthe percentages of selection of the elementtypes in the two experiments. A negativedifference indicates a larger percentage ofselection in Experiment 1. There was a cleartendency in the younger two groups to preferthe originally less-often-selected element.

To further demonstrate the reversal thatoccurred in Experiment 2, a criterion of 20%

or greater in the selection of the previouslymore-often-selected element and the previous-ly less-often-selected element was imposed. Ofthe 8 participants in each group, 7 of theyoungest group of participants and 6 of themiddle-aged group of participants met thecriterion, but only 3 of the elderly group ofparticipants met it. A chi-squared analysisrevealed a significant difference between thenumber of participants who met criterionbetween the youngest and oldest groups ofparticipants [x2(1) 5 4.27; p , .04, W 5 .516].The difference between the middle-aged andoldest group was not significant [x2(1) 5 2.29;p , .13], nor was there a significant differencebetween the youngest and middle group ofparticipants.[x2(1)5.41; p ..05]. These resultssuggest that the use of the verbal punisher hadthe least impact on the oldest group ofparticipants.

DISCUSSION

The results from Experiment 2 demonstrat-ed that the enhancement of stimulus controlby underselected elements was facilitated inthe two younger groups of participants. How-ever, this effect was not observed in the oldestgroup of participants. These findings suggestthat different processes may be involved in theoverselectivity observed in the oldest group.Specifically, the emergence of the enhance-ment effect in the two younger groups

Fig. 3. Mean percentage and standard error of selection of the originally more-often-selected stimulus elements andthe originally less-often-selected elements in the test phase for the three age groups in Experiment 2. The originalpercentages of selection also appear for comparison.

376 LOUISE MCHUGH and PHIL REED

supported the overshadowing rather than theattentional-deficit account of stimulus over-selectivity, as the less-often-selected stimuluselements in Experiment 1 were, in fact,learned. That facilitation did not emerge inthe oldest group of participants may lendsupport to the attentional-deficit account ofstimulus overselectivity for that age group.

GENERAL DISCUSSION

The results from Experiment 1 demonstrat-ed stimulus overselectivity in all three agegroups of adult participants in a simplediscrimination task. The effect was more pro-nounced when participants also were requiredto complete a distractor task. The effect wassimilar to that reported by Reed and Gibson(2005). In addition, a developmental trend inoverselectivity emerged: Overselectivity wasmore likely to occur in later adulthood. InExperiment 2, participants in the two youngerage groups demonstrated differentially greaterstimulus control by the previously less-often-selected stimulus elements. This effect was notevident in the oldest group.

Two hypotheses were considered. The first isthat overselectivity of stimulus elements repre-sents the failure to attend to them in the firstplace (attentional deficit). The second is thatone element overshadows the other (over-shadowing). The results of Experiment 2provide some support for the second hypoth-esis. The results for the two younger agegroups of participants showed that the less-often-selected stimulus elements in Experi-ment 1 were learned.

Previously, Dube et al. (1999) used a mea-sure of eye movement to assess overselectivityin individuals with mental retardation. Theirfindings suggested that participants who over-select failed to attend to all the relevantstimuli. If stimuli are not attended to, thenthey cannot control behavior. This finding isin line with learning theories of differentialstimulus control (e.g., Mackintosh, 1975;Rescorla & Wagner, 1972; Sutherland &Mackintosh, 1971). However, the present studyfound that, for the two younger groups ofparticipants, previously less-often-selectedstimulus elements gained control of behaviorwhen the previously more-often-selected ele-ments were punished, a finding that stronglysuggests the originally underselected elements

did not go unattended. Still, the same effect wasabsent in the older-adult group of participants.Either overshadowing is less likely to occur inthe elderly, or they may experience an atten-tional deficit. Deciding between these hypoth-eses is a matter for further research (see Reed,2006). For example, future research mightexamine selectivity in participants with diag-nosed attentional deficits (e.g., ADHD). Fur-thermore, the performance of such participantson the acquisition task involving compoundstimuli could be compared with that of partic-ipants who ordinarily manifest stimulus over-selectivity (e.g., those with ASD), thus possiblyproviding insights into the behavioral processesby which these two categories of disorder can bedifferentiated further.

REFERENCES

Ardle, J. J., Ferrer-Caja, E., Hamagami, F., & Woodcock, R.(2002). Comparative longitudinal structural analysesof the growth and decline of multiple intellectualabilities over the life span. Developmental Psychology, 38,115–142.

Broomfield, L., McHugh, L., & Reed, P. (in press). Theeffect of observing response procedures on thereduction of over-selectivity in a match to samplestask: Immediate but not long term benefits. Journal ofIntellectual Disabilities.

Cumming, W. W., & Berryman, R. (1965). The complexdiscriminated operant: Studies of matching-to-sampleand related problems. In D. I. Mostofsky (Ed.), Stimulusgeneralization (pp. 284–330). Stanford, CA: StanfordUniversity.

Dube, W. V., Lombard, K. M., Farren, K. M., Flusser, D.,Balsamo, L. M., & Fowler, T. R. (1999). Eye trackingassessment of stimulus overselectivity in individualswith mental retardation. Experimental Analysis ofHuman Behavior Bulletin, 13, 267–271.

Dube, W. V., & McIlvane, W. J. (1999). Reduction ofstimulus overselectivity with nonverbal differentialobserving responses. Journal of Applied Behavior Anal-ysis, 32, 25–33.

Gibson, E., & Reed, P. (2005). Stimulus overselectivity inrats. Journal of Autism and Developmental Disorders, 35,851–859.

Kaufman, M. A., & Bolles, R. C. (1981). A nonassociativeaspect of overshadowing. Bulletin of the PsychonomicSociety, 18, 318–320.

Koegel, R. L., & Schriebman, L. (1977). Teaching autisticchildren to respond to simultaneous multiple cues.Journal of Experimental Child Psychology, 24, 299–311.

Lovaas, O. I. (1980). Behavioral teaching with youngautistic children. In B. Wilcox, & A. Thompson (Eds.),Critical issues in educating autistic children and youth (pp.220–233). Washington, DC: Department of Educa-tion, Office of Special Education.

Lovaas, O. I., & Schriebman, L. (1971). Stimulus over-selectivity of autistic children in a two stimulussituation. Behavior Research and Therapy, 9, 305–310.

STIMULUS OVERSELECTIVITY AND AGE TRENDS 377

Lovaas, O. I., Schreibman, L., Koegel, R., & Rehm, R.(1971). Selective responding by autistic children tomultiple sensory input. Journal of Abnormal Psychology,77, 211–222.

Mackintosh, N. J. (1975). A theory of attention: Variationsin the associability of stimuli with reinforcement.Psychological Review, 82, 276–298.

Matzel, L. D., Schachtman, T. R., & Miller, R. R. (1985).Recovery of an overshadowed association achieved byextinction of the overshadowing stimulus. Learningand Motivation, 16, 398–412.

Reed, P. (2006). The effect of retention interval onstimulus overselectivity using a match-to-sample para-digm. Journal of Developmental Disorders, 36, 1115–1121.

Reed, P., & Gibson, E. (2005). The effects of concurrenttask load on stimulus overselectivity. Journal of Autismand Developmental Disorders, 35, 601–614.

Reed, P., & Reilly, S. (1990). Context extinction followingconditioning with delayed reward enhances subse-quent instrumental responding. Journal of ExperimentalPsychology: Animal Behavior Processes, 16, 48–70.

Remington, R. W. (1980). Attention and saccadic eyemovement. Journal of Experimental Human Perceptionand Performance, 6, 726–744.

Rescorla, R. A., & Wagner, A. R. (1972). A theory ofPavlovian conditioning: Variations in the effectivenessof reinforcement and nonreinforcement. In A. H.Black, & W. F. Prokasy (Eds.), Classical conditioning II:Current research and theory (pp. 64–99). New York:Appleton-Century-Crofts.

Schreibman, L., & Lovaas, O. I. (1973). Overselectiveresponse to social stimuli by autistic children. Journalof Abnormal Child Psychology, 1, 152–168.

Sutherland, N. S., & Macintosh, N. J. (1971). Mechanisms ofanimal discrimination learning. New York: AcademicPress.

Trabasso, T., & Bower, G. (1968). Attention in learning. NewYork: Wiley.

White, K. G., & Ruske, A. C. (2002). Memory deficits inAlzheimer’s disease: The encoding hypothesis andcholinergic function. Psychonomic Bulletin and Review,9, 426–437.

Wilkie, D. M., & Masson, M. E. (1976). Attention in thepigeon: A re-evaluation. Journal of the ExperimentalAnalysis of Behavior, 26, 207–212.

Received: September 27, 2006Final acceptance: July 11, 2007

378 LOUISE MCHUGH and PHIL REED

APPENDIX 1

Individual data for each group in Experiment 1. The data are the percentages of selection of the more-often-selected andless-often-selected stimulus elements in the distractor and no-distractor conditions in Experiment 1. Age group (in years)is denoted by number: 1 5 18–22; 2 5 47–55; 3 5 70–80.

Age GroupMore-often-selectedwith no distractor

Less-often-selectedwith no distractor

More-often-selectedwith distractor

Less-often-selectedwith distractor

1 100 90 100 701 100 90 90 801 100 100 90 801 90 80 90 501 100 100 90 901 100 80 90 801 100 100 100 701 100 100 90 801 100 100 90 901 100 90 100 901 100 100 90 901 100 100 100 901 100 100 100 1001 100 80 80 801 100 90 80 601 100 100 90 602 100 80 90 702 100 100 90 902 100 90 90 602 90 80 90 602 100 100 100 902 90 80 90 802 100 100 90 702 90 90 90 902 100 100 100 902 100 90 80 702 100 90 90 802 100 100 90 702 100 100 90 802 100 90 90 702 100 90 80 702 90 90 100 803 70 70 90 803 70 60 70 503 90 80 80 703 90 70 90 503 90 70 100 303 90 70 80 503 80 80 80 603 100 100 80 503 90 80 80 603 100 80 90 403 90 80 70 503 100 70 80 403 90 80 80 403 90 70 80 503 80 60 70 503 90 90 100 30

STIMULUS OVERSELECTIVITY AND AGE TRENDS 379

APPENDIX 2

Individual data for the percentages of selection of the punished stimulus elements (i.e., the originally more-often-selected; 1st column) and the originally less-often-selected (2nd column) stimulus elements in the test phase ofExperiment 2. The 3rd and 4th columns show the differences in the percentages of selection of the two types of elementsbetween Experiments 1 and 2 (specifically, % in Experiment 2 - % in Experiment 1). Age group (in years) is denoted bynumber: 1 5 18–22; 2 5 47–55; 3 5 70–80.

Age Group

% selection of punishedstimulus elements

(more)

% selection ofunpunished stimulus

elements (less)

Difference in % selectionof the more-often-selected element

Difference in % selectionof the less-often-selected

element

1 60 90 240 201 70 100 220 201 50 80 240 01 60 90 230 401 50 70 240 2201 90 90 0 101 60 90 240 201 50 100 240 202 50 90 240 202 70 80 220 2102 50 90 240 302 70 100 220 402 60 90 240 02 60 70 230 2102 50 70 240 02 70 90 220 03 50 70 240 2103 50 50 220 03 70 30 210 2403 50 50 240 03 70 60 230 303 80 30 0 2203 30 70 250 103 40 50 240 0

380 LOUISE MCHUGH and PHIL REED

All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately.