285

JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2003, 36, 285–296 NUMBER 3 (FALL 2003)

THE ROLE OF SPECIFIC CONSEQUENCES INTHE MAINTENANCE OF THREE TYPES OF QUESTIONS

GLADYS WILLIAMS

APPLIED BEHAVIORAL CONSULTANT SERVICES, NEW YORK

LUIS ANTONIO PEREZ-GONZALEZ

UNIVERSITY OF OVIEDO, SPAIN

AND

KIM VOGT

DAVID GREGORY SCHOOL, PARAMUS, NEW JERSEY

This research replicated and extended a study by Williams, Donley, and Keller (2000).In that study, children with autism received a box with an object inside and learned toask ‘‘What’s that?,’’ ‘‘Can I see it?,’’ and ‘‘Can I have it?’’ to have the name of the object,to see the object, and to get the object, respectively. The purpose of the present researchwas to determine if the three questions (a) were three independent repertoires of behavior,(b) constituted three instances of a single functional response class, or (c) belonged to achain of behavior. The 3 boys with autism who participated responded independently toeach question when the consequences for each question were altered. This indicates thatthe three target responses were three independent repertoires of behavior, each one re-inforced and maintained with its specific consequences. Thus, this procedure serves toteach children with autism to ask questions with flexibility according to a variable context.

DESCRIPTORS: question asking, verbalizations, autism, language acquisition, ver-bal behavior, establishing operations

Asking questions is a complex verbal be-havior that for most children with autismdoes not emerge until it is explicitly taught.Many researchers have been successful atteaching these children to ask questions suchas ‘‘What’s that?,’’ ‘‘What’s inside the box?,’’

Partial support for this research comes from a grantfrom Cajastur to the first two authors. We thank Jo-seph Spradlin and Jennie Keller for reviewing previousversions of the manuscript; all the teachers who par-ticipated in this research; the principals of the SchoolSanta Marıa del Naranco in Oviedo, Spain, and theDavid Gregory School in Paramus, New Jersey; andthe children’s parents. Details of the procedure anddata-collection protocol are available upon request tothe first author.

Address correspondence to Gladys Williams, Ap-plied Behavioral Consultant Services, 66 Re-gency Circle, Englewood, New Jersey 07631 (e-mail:gladyswilliams2003@hotmail.com) or to Luis A.Perez-Gonzalez, Departamento de Psicologıa, Univer-sity of Oviedo, Plaza Feijoo s/n, Oviedo 33003, As-turias, Spain (e-mail: laperez@sci.cpd.uniovi.es).

‘‘Where is it?,’’ and ‘‘Can I see it?’’ (Koegel,Camarata, Valdez-Menchaca, & Koegel,1998; Shabani et al., 2002; Sundberg, 2000;Taylor & Harris, 1995; Williams, Donley, &Keller, 2000). Williams et al. taught 2 chil-dren with autism to ask three questions, ina stepwise fashion, about a closed box witha toy inside. The experimenter first showedthe closed box to the child. When the childasked the first question (‘‘What is that?’’),the experimenter said the name of the hid-den object. When the child asked the secondquestion (‘‘Can I see it?’’), the experimentershowed it to him. When the child asked thethird question (‘‘Can I have it?’’), the exper-imenter gave the object to the child.

The study by Williams et al. (2000) raisedan important question: Did these three typesof questions have the same functions as theyhave in everyday interactions? In a typically

286 GLADYS WILLIAMS et al.

developing person, question asking has con-sequences that vary according to the situa-tion. Therefore, it is important to know ifthis type of teaching leads to establishing arepertoire of three mands, each one main-tained by its proper functions in a variedcontext.

The results of Williams et al. (2000) mayhave been due to three possible factors: First,it is possible that the responses to the threequestions were independent of the other re-sponses, each one maintained by its specificconsequence. That is what presumably oc-curs in everyday life, because typically de-veloping children may ask only one of thesequestions according to the situation. Second,each response could have been establishedbecause initially each one produced access tothe item. In fact, during the initial stages ofteaching, the object followed each questionat the moment of teaching it. According tothis hypothesis, any of the three responseswould produce the final item and, therefore,would be equivalent to each other: Theywould have formed three response forms ofa single operant response class (e.g., Dona-hoe & Palmer, 1994). Third, it is possiblethat this type of teaching led to a sequenceof responses in which each question wasmaintained by the onset of the next questionor the terminal reinforcer (the object). Theobject, thus, would ultimately reinforce theentire sequence. Therefore, the teacher’s re-sponses to the first and second questionswould reinforce each question by presentingthe opportunity to ask the following ques-tion, as in a response chain (Skinner, 1934).If that happens, the interruption of the se-quence by removing the consequence couldresult in the absence of any further behavior.

These hypotheses have programmatic im-plications: Due to the lack of spontaneityand language production shown by childrenwith autism, it is important to know the de-gree of variability and flexibility these ques-tions have. Moreover, it is important to de-

velop strategies to teach these children toproduce each type of question with enoughvariability, according to the specific states ofdeprivation, and according to the particularconsequences those responses have for typi-cally developing individuals.

The main purpose of the present researchwas to study further the variables that influ-ence question asking. We initially taughtchildren the three questions used by Wil-liams et al. (2000). Once the children beganto ask the three questions, we manipulatedtwo variables: First, we provided no conse-quences for the second question (‘‘Can I seeit?’’). Second, we introduced unpleasant ob-jects. Children’s performance should indicatewhether each question serves a distinct op-erant function, whether the three responsesformed a single response class, or whetherthe three operants formed a response se-quence.

METHOD

ParticipantsThe participants were 3 boys who had

been diagnosed with autism by several in-dependent professionals and officials of theeducational system. Jim, 2 years 9 monthsold, received intensive behavioral in-hometeaching. Erik, 4 years 5 months old, at-tended a special needs behavioral program.Jim and Erik were U.S. residents and thusreceived instruction in English. Dario, 9years 4 months old, attended a special needsbehavioral program in Spain and thus hisinstruction was given in Spanish. All thechildren displayed mands (e.g., ‘‘I want toeat candy’’) and tacts (e.g., ‘‘This is a cat’’;they could tact at least 100 objects); repeatedstatements of at least four words (e.g., ‘‘It iscold today’’); responded to basic social ques-tions such as ‘‘What is your name?’’; andselected objects and pictures, without errors,to the instruction ‘‘Give me the [name ofthe object].’’ All 3 students could say ‘‘No!’’

287ASKING QUESTIONS

and rejected items they disliked by pushingthem away with their hands or feet, or byturning away or covering their faces. Noneof them, however, asked questions prior tothe onset of this study.

Materials

The experimenters provided about 80small boxes varying in shape, color, and tex-ture, with attractive objects hidden inside(e.g., a sparkling spinning wheel, a shiny redrace car). The objects were novel for eachchild but were similar to other objects pre-viously used as reinforcers to teach otherskills. The unpleasant objects were selectedbased on information from the child’s moth-er, who in some instances provided some un-pleasant objects from home (e.g., a used cu-neiform sponge from the bathroom, a slimyjelly-like form toy). We also used live insectsobtained outdoors (a worm, a spider, a snail)in some trials with all children.

Procedure

Overview. The treatment package, a rep-lication with an extension of Williams et al.(2000), was evaluated with a multiple base-line design across the three response forms.When all of the children acquired the ques-tions, we varied the consequences to thequestions in several phases. The child satacross from the experimenter in the class-room or in the child’s teaching room. Onetrial consisted of presenting one box with ahidden object for 20 s. The presentation ofone box and the eventual production of one,two, or three questions constituted one trial.One session consisted of 10 trials, that is,the presentation of 10 boxes, one at a time.Usually, we conducted one session per day.We recorded data on each question in eachtrial. The prompted and self-initiated re-sponses were recorded separately, and onlythe self-initiated responses were consideredcorrect. After the child had met criterion oneach specific question, that question was

never again prompted, even if the child nev-er said it again. If the child asked ‘‘Can I seeit?’’ before asking ‘‘What’s in the box?’’ theexperimenter said ‘‘Sure, you can see it,’’ andshowed the object in the box to the childbut did not give it to him. Or if the childasked ‘‘Can I have it?’’ before asking the firsttwo questions, the experimenter said ‘‘Sure’’and gave him the box with the object inside.

Baseline. The experimenter presented onebox at a time, 10 times, for 20 s. The ex-perimenter did not explicitly prompt any be-havior, but opened the box with an objectinside and made a comment about the ob-ject (e.g., ‘‘Oh, this one is great!’’ or ‘‘I reallylike this one.’’). The experimenter did notshow the object to the child. She changedto a new box if 20 s passed with no responsefrom the child.

Teaching the first response form (‘‘What’s inthe box?’’). The experimenter held the boxand made a comment about the object in-side the box. She then prompted the childto repeat the question ‘‘What’s in the box?’’by modeling the question in a firm tone ofvoice. When the child repeated the question,the experimenter told the child what was in-side the box and gave the box with the ob-ject inside to the child. The child could playwith the toy for approximately 20 s; then, anew box was presented. When the child re-peated the question correctly for two con-secutive trials, the experimenter faded theechoic prompt by providing a partial promptfor two more consecutive trials. For example,the experimenter provided the word‘‘What’s’’ instead of providing the wholequestion. The prompt was gradually reduceduntil the child asked ‘‘What’s in the box?’’without any prompt. If the child producedthree consecutive errors (saying the questionincorrectly or saying nothing within 20 s ofpresenting the box), the experimenter pro-vided the full echoic prompt again. Thisprocedure was repeated until transfer of con-trol from the echoic prompt to the box itself

288 GLADYS WILLIAMS et al.

was accomplished. That is, the echoicprompt was eliminated when the child ini-tiated the question without any prompt eachtime a box was presented. Criterion consist-ed of 90% or better correct unprompted re-sponses in two consecutive sessions. At thispoint, the teaching of the second responseform began, and no more prompts were pro-vided for the first question.

Teaching the second response form (‘‘Can Isee it?’’). This procedure started the same wayas the first response form. However, whenthe child self-initiated ‘‘What’s in the box?,’’the experimenter told him the name of theobject inside the box but neither showed itnor gave it to him. The experimenter thenprompted the child to repeat ‘‘Can I see it?’’When the child repeated the question cor-rectly, the experimenter said ‘‘Sure, I canshow it to you,’’ and gave the box to thechild. The procedure to fade and eliminatethe echoic prompt and the criterion to moveon to teach the next response form were thesame as in the teaching of the first responseform. During this phase, the echoic promptwas presented only after the child emittedthe first response (‘‘What’s in the box?’’). Butonce the second question reached criterion,the child’s second question resulted in theexperimenter saying ‘‘Sure, I can show it toyou’’ and giving the box to the child, evenif the child skipped the first question (thishappened only on two occasions with 1child).

Teaching the third response form (‘‘Can Ihave it?’’). The procedure started the sameway as the first and second response forms.When the child self-initiated ‘‘What’s in thebox?,’’ the experimenter told him the nameof the object inside but neither showed itnor gave it to him. When the child self-ini-tiated ‘‘Can I see it?,’’ the experimentershowed him the object inside the box butdid not give it to him. The experimenterthen prompted the child to repeat ‘‘Can Ihave it?’’ When the child repeated the ques-

tion correctly, the experimenter said, ‘‘Ofcourse, I can give this toy to you,’’ and gavethe box to the child. The echoic prompt waspresented only after the child emitted thesecond response (‘‘Can I see it?’’). The pro-cedure to fade and finally eliminate theechoic prompt was the same as for the firstand second response forms. Once this thirdquestion reached criterion, the experimentergave the child the object after the childasked this question, even if the child skippedthe other two questions. The child couldemit any of the three questions during the20-s trials. After criterion had been obtainedfor the third response, all the prompts wereeliminated for the rest of the study. Thisphase was extended several more sessions toensure stability.

‘‘No!’’ to the question ‘‘Can I see it?’’ Thepurpose of this phase was to see if the othertwo operants would be maintained if one ofthe operants (‘‘Can I see it?’’) was eliminat-ed. Immediately after the child asked ‘‘CanI see it?’’ the experimenter said ‘‘No!’’ anddid not show the object to the child. Whenthe child asked ‘‘Can I have it?’’ the experi-menter gave him the box. No prompts wereprovided in this phase, and the contingen-cies of reinforcement were the same as in theprevious phases (except that for the secondquestion the experimenter said ‘‘No!’’) evenif the child skipped one or two questions. Ifthe child did not ask any question, the ex-perimenter changed to a new box after 20 s.If the three questions were a response chain,the two responses emitted before removingthe reinforcer (in this case, the questions‘‘What’s in the box?’’ and ‘‘Can I see it?’’)should be extinguished and the questionemitted after the reinforcer should remain(e.g., Skinner, 1934). If, alternatively, allthree response forms decreased or weremaintained together, this would indicatethat the three responses are instances of asingle response class. If the children respond-

289ASKING QUESTIONS

ed with an alternative pattern, that wouldindicate alternative sources of control.

‘‘No!’’ to the question ‘‘Can I see it?’’ andhiding the box contingently. Hiding the boxwas introduced when the response ‘‘No!’’ tothe question ‘‘Can I see it?’’ did not decreasethat question. After the child asked ‘‘Can Isee it?’’ the experimenter said ‘‘No!’’ and hidthe box behind her back. No prompts weregiven in this phase, and the contingencies ofreinforcement remained the same for‘‘What’s in the box?’’ and ‘‘Can I have it?’’

Return to original contingencies of reinforce-ment. Conditions returned to the same con-tingencies of reinforcement as in the phaseof teaching the third question, but noprompts were provided. The child’s respons-es were reinforced accordingly, even if heskipped questions. For example if the childsaid ‘‘Can I have it?’’ before asking ‘‘What’sin the box?’’ the experimenter told him‘‘Sure, you can have it’’ and gave him thebox. If the child did not emit any questionduring the 20-s period, the experimenterchanged the box and started a new trial.This phase was repeated intermittentlythroughout this investigation because wewanted to know if the questions would re-cover under the initial conditions of teach-ing.

Some unpleasant objects. Conditions re-mained the same as in the previous phase,with the exception that the experimenterpresented, in a random fashion, five trialswith unpleasant objects in one session (10trials). This phase was introduced to explorefurther the conditions that maintained re-sponding to the three questions.

All unpleasant objects. The experimenterpresented all 10 trials with unpleasant ob-jects to show the child. The purpose of thisphase was to see clearly the effects intendedin the previous phase. The contingencies ofreinforcement for each question were thesame as in the phase of teaching the third

response, even if the child skipped any ques-tion; no prompts were provided.

‘‘No!’’ to the question ‘‘Can I have it?’’ Thisphase was introduced to study further therole of this consequence on the recovery ofthe question ‘‘Can I see it?’’ When the childasked ‘‘Can I have it?,’’ the instructor said‘‘No!’’ and waited to see if the response ‘‘CanI see it?’’ reemerged within the remainingtime. When the child asked ‘‘Can I see it?’’the experimenter said ‘‘Sure!’’ and gave thebox with the object inside to the child.These contingencies of reinforcement,slightly different from previous phases, wereapplied even if the child skipped the firstquestion (‘‘What’s in the box?’’). Noprompts of any kind were given in thisphase, and no additional time was given.

Measurement and Interobserver Agreement

The first self-initiated question of each re-sponse form constituted the behavior forthat particular trial. The child, however, hadthe opportunity to ask three questions pertrial. The experimenter did not count thesubsequent questions that sometimes fol-lowed each response form (repetitions suchas ‘‘Can I see it? Can I see it?’’). The dataon each of the questions were recorded sep-arately by the experimenter and an observer.An agreement was scored when both record-ed the same response or responses in all con-ditions within the 20-s trial. A disagreementwas scored when the experimenter and theobserver recorded one or more different re-sponses. The observer was present on 31%of the sessions for Jim, 31% of the sessionsfor Erik, and 45% of the sessions for Dario.Interobserver agreement was 100%.

RESULTS

The 3 children learned to ask the threequestions. Although the various conditionsdid not always produce the same effects ineach child, all of them recovered and main-

290 GLADYS WILLIAMS et al.

tained the three questions throughout thestudy. This finding replicates that of Wil-liams et al. (2000).

Figure 1 shows the results for Jim. Duringbaseline, Jim asked no questions in any ofthe three forms. During the intervention toteach the first question, he asked five to 10questions in three sessions and maintainedthe 10 questions (the maximum he couldask in a given session) for seven more ses-sions. He met acquisition criterion for thesecond question rather fast, in that he wentfrom one to 10 in two sessions. He acquiredthe third question in two sessions. Thereaf-ter, he continued asking these three ques-tions at very high levels (nine or 10 ques-tions per session). During the phase of‘‘No!’’ to the question ‘‘Can I see it?,’’ thequestion decreased gradually to zero in thefourth session, while the other two questionsremained at high levels (near 10). A returnto the original contingencies brought backthe second question to previous levels (near10 questions per session) and maintained theother two questions at similar levels. In thenext phase, with some unpleasant objects,Jim decreased the questions ‘‘Can I see it?’’and ‘‘Can I have it?,’’ but he asked the firstquestion (‘‘What’s in the box?’’) on all trials.In the next phase, the experimenter present-ed only unpleasant objects. Jim almoststopped asking the second and third ques-tions at the same time that he maintainedthe first one in a consistent manner. A returnto normal conditions brought back all threequestions to 10.

Figure 2 shows results for Erik. Erik didnot ask any questions during baseline, buthe learned to ask the three questions to reachlevels of 10. He met acquisition criteria forthe first question in eight sessions, the sec-ond question in three sessions, and the thirdquestion in four sessions. The three questionforms were maintained for two more sessionsat levels of 10 (or near 10). In the phase of‘‘No!’’ to the question ‘‘Can I see it?’’ Erik

stopped asking the question (‘‘Can I see it?’’)in two sessions, whereas he continued to askthe other two questions. A return to theoriginal contingencies did not recover thesecond question (‘‘Can I see it?’’). Then, theexperimenters introduced a new phase tostudy whether the child would ask this ques-tion again. When the child asked ‘‘Can Ihave it?,’’ the experimenter said ‘‘No!’’ anddid not give the object to the child. If heasked ‘‘Can I see it?’’ the experimenter toldhim ‘‘Sure, you can see it,’’ and gave himthe object. In this phase, Erik again askedthe second question (immediately reachedthe maximum level of 10 questions per ses-sion); moreover, he continued asking thethird question (‘‘Can I have it?’’). In thephase with some unpleasant objects, Erikasked fewer questions for ‘‘Can I have it?’’(six per session) and also fewer questions for‘‘Can I see it?’’ (from zero to four). He con-tinued to ask the first question (‘‘What’s inthe box?’’) in every trial. The next phaseconsisted of showing only unpleasant ob-jects. Erik continued to ask the second ques-tion (‘‘Can I see it?’’) at a low frequency, andhe almost stopped asking the third question(‘‘Can I have it?’’). In the return to the phasewith the original contingencies, Erik did notask the second question, but he asked thethird question again, as he did in the pre-vious phase with the original contingencies.For this reason, we again introduced thesame phase of ‘‘No!’’ to the question ‘‘Can Ihave it?’’ In only one session in this phase,Erik acquired the second question again. Heasked the first question (‘‘What’s in thebox?’’) consistently throughout all the phas-es.

Figure 3 shows Dario’s results. Duringbaseline, Dario asked no questions in any ofthe three forms. Some difficulty was notedwith Dario’s intelligibility; for this reason,we taught him to say ‘‘What’s that?’’ insteadof ‘‘What’s in the box?’’ During the inter-vention, he met acquisition criterion for

291ASKING QUESTIONS

Figure 1. Frequency of questions asked by Jim during the presentation of 10 boxes with hidden objectsinside during baseline, teaching of the three response forms, maintenance, removal of consequences for asking‘‘Can I see it?,’’ and the introduction of unpleasant objects (see text for details).

292 GLADYS WILLIAMS et al.

Figure 2. Frequency of questions asked by Erik during the presentation of 10 boxes with hidden objectsinside during baseline, teaching of the three response forms, maintenance, removal of consequences for asking‘‘Can I see it?,’’ and the introduction of unpleasant objects (see text for details).

293ASKING QUESTIONS

Figure 3. Frequency of questions asked by Dario during the presentation of 10 boxes with hidden objectsinside during baseline, teaching of the three response forms, maintenance, removal of consequences for asking‘‘Can I see it?,’’ and the introduction of unpleasant objects (see text for details).

294 GLADYS WILLIAMS et al.

‘‘What’s that?’’ in eight sessions, ‘‘Can I seeit?’’ in four sessions, and ‘‘Can I have it?’’ insix sessions. Subsequently, he asked thesethree questions in almost every trial. Duringthe subsequent condition (‘‘No!’’ to thequestion ‘‘Can I see it?’’), the response ‘‘CanI see it?’’ was not affected initially, as he con-tinued asking this question. We introduceda new phase in which the experimenter said‘‘No!’’ and hid the box immediately after heasked ‘‘Can I see it?’’ Asking this questiondecreased, and it did not occur in the lastthree sessions. He continued to ask the othertwo questions. In the subsequent phase(some unpleasant objects), the first question(‘‘What’s that?’’) was maintained (at the fre-quency of 10 questions in 10 trials). Mean-while, he stopped asking the second (‘‘CanI see it?’’) and the third (‘‘Can I have it?’’)questions. The experiment was then discon-tinued for external reasons.

DISCUSSION

The 3 children learned to ask the threequestions. These results replicate those ofWilliams et al. (2000). Once all the childrenacquired the three questions, we applied dif-ferent consequences to two questions to seehow the other questions were affected.When we responded ‘‘No!’’ to ‘‘Can I seeit?,’’ 2 children gradually decreased askingthis question, whereas they continued to askthe other two questions. The other child de-creased this question after an additional pro-cedure was implemented in a subsequentphase (the experimenter said ‘‘No!’’ and hidthe box). He also continued to ask the othertwo questions.

The response pattern displayed when theexperimenter said ‘‘No!’’ to the question‘‘Can I see it?’’ weakens the hypothesis thatthe three question types constitute a re-sponse class. Had they constituted a re-sponse class, removing the consequence (theputative reinforcer) of a response should re-

sult in identical effects in the three classes,either maintenance or a reduction of thethree responses at the same time. This effectdid not occur. On the contrary, the secondresponse form (‘‘Can I see it?’’) decreased(extinguished by the lack of the former con-sequence, punished by ‘‘No!,’’ or some com-bination), while the other two responseforms were maintained. Thus, it indicatesthat this question was not in the same re-sponse class as the other two questions. Thismanipulation, however, did not tell us any-thing about whether the first and the thirdresponse forms (‘‘What’s in the box?’’ and‘‘Can I have it?’’) are members of the sameresponse class.

This performance also weakens the hy-pothesis that the three responses were achain of responses. Had they constituted aresponse chain, the first response (‘‘What’sin the box?’’) would have decreased as a re-sult of removing the reinforcer that main-tained the subsequent responses in the chain.Thus, maintenance of that question suggeststhat the three questions were not membersof a response chain.

The second manipulation consisted of in-troducing unpleasant objects on some or alltrials. In those conditions, the children asked‘‘Can I see it?’’ and ‘‘Can I have it?’’ less andless frequently. Still, they asked ‘‘What’s inthe box?’’ when new boxes were presented.This response pattern rules out the possibil-ity that the first and third questions formeda single response class, because one of themremained at a high frequency while the otherdropped (for the same reasons as in the firstmanipulation). This response pattern alsorules out the possibility that the three re-sponses were members of a chain. Had theybeen members of a chain, the three respons-es would have been maintained by the finalreinforcer (the object). Then, the removal ofthe object should have resulted in the ex-tinction of the three responses. This did notoccur. Thus, this response pattern in one

295ASKING QUESTIONS

sense rules out the possibility that the threeresponses were members of a response chain.

Another way to interpret the results is thatquestion asking served the function ascribedby the contingencies of reinforcement in anygiven condition. For example, it may be thatsaying ‘‘Can I see it?’’ originally was rein-forced by gaining access to the item, buteventually it was reinforced by merely seeingthe item. These differential contingencieswere programmed and, hence, the operantfunction of the questions necessarily shiftedin response to newly arranged contingenciesof reinforcement.

The results observed after the teachingphases suggest that each question was main-tained by its own consequence. The firstquestion (‘‘What’s in the box?’’) was main-tained by the answer the child received tothat question. The experimenter’s responsesmaintained the question independently ofthe fact that the experimenter said the nameof a pleasant object (a potential reinforcer)or an unpleasant object. We may supposethat ‘‘What’s in the box?’’ would have beenextinguished if the experimenter discontin-ued responding to that question by not giv-ing the name of an object (either pleasant orunpleasant). We did not manipulate this var-iable in the current study.

The second response form (‘‘Can I seeit?’’) could be maintained by the view of theobject and the possibility of asking the thirdquestion. The view of the object played arole. In fact, when the consequence in thatphase was not to have the opportunity toview the object, the question decreased.However, the children still had the oppor-tunity to ask the third question (‘‘Can I haveit?’’). The 3 children continued to ask thisquestion, even though they did not ask thesecond one. This fact suggests that the con-sequence that maintained the second ques-tion was the view of the object. Also, it sug-gests that the opportunity to ask the thirdquestion was not the consequence that

maintained the second question. The thirdresponse form (‘‘Can I have it?’’) was prob-ably maintained by the access to the object.

The antecedent stimuli of the three ques-tions also changed. The antecedent stimulusfor the first question was the presentation ofthe box. The antecedent stimulus for thesecond question could have been the exper-imenter’s response to the first question,which established the occasion to reinforcethat response with the view of the object.The antecedent stimulus for the third ques-tion could have been the response to the sec-ond question. However, when the secondquestion was no longer reinforced, the thirdresponse was emitted just after the experi-menter responded to the first question. Thisfact indicates that the antecedent stimulusfor the third response was not precise. In-stead, it seems that the response to the firstquestion established the conditions for theonset of the second and third questions.

In summary, these data show that each ofthe three questions ultimately was main-tained by different consequences thatchanged over conditions according to chang-es in consequences. This fact indicates thatthe questions did not belong to a responseclass, because not all the questions weremaintained by the same reinforcer—in thiscase, access to the object. The three ques-tions were not a response chain becausewhen one question decreased, regardless ofwhich one, the other two were maintainedat high levels. Given the flexibility of re-sponding according to a variable environ-ment, the teaching methods used here ap-pear to be sufficiently versatile to teach chil-dren with autism question asking thatmatches the functions of question asking ineveryday life.

REFERENCES

Donahoe, J., & Palmer, D. (1994). Learning andcomplex behavior. Boston: Allyn and Bacon.

296 GLADYS WILLIAMS et al.

Koegel, L. K., Camarata, S. M., Valdez-Menchaca, M.,& Koegel, R. L. (1998). Setting generalization ofquestion asking by children with autism. AmericanJournal on Mental Retardation, 102, 346–357.

Shabani, D. B., Katz, R. C., Wilder, D. A., Beau-champ, K., Taylor, C. R., & Fischer, K. J. (2002).Increasing social initiations in children with au-tism: Effects of a tactile prompt. Journal of AppliedBehavior Analysis, 35, 79–83.

Skinner, B. F. (1934). The extinction of chained re-flexes. Proceedings of the National Academy of Sci-ences, 20, 532–536.

Sundberg, M. (2000, May). Teaching mands for in-formation by manipulating the establishing opera-tion. Paper presented at the 26th annual confer-

ence of the Association for Behavior Analysis,Washington, DC.

Taylor, B. A., & Harris, S. L. (1995). Teaching chil-dren with autism to seek information: Acquisitionof novel information and generalization of re-sponding. Journal of Applied Behavior Analysis, 28,3–14.

Williams, G., Donley, C., & Keller, J. (2000). Teach-ing children with autism to ask questions abouthidden objects. Journal of Applied Behavior Anal-ysis, 33, 627–630.

Received June 18, 2002Final acceptance March 26, 2003Action Editor, Timothy R. Vollmer

STUDY QUESTIONS

1. What explanations did the authors propose to account for the findings of the Williams etal. (2000) study, in which children asked a series of related questions about something?

2. Describe the general procedure used to train participants to use the initial questions.

3. What was the rationale for the condition in which the experimenter responded ‘‘No!’’ tothe question ‘‘Can I see it?’’

4. What was the rationale for the condition in which unpleasant items were placed in theboxes?

5. Describe the general effects observed on question asking during (a) the condition in whichthe therapist responded ‘‘No!’’ to the question ‘‘Can I see it?’’ and (b) the condition in whichsome unpleasant items were placed in the boxes.

6. Erik was the only participant exposed to the condition in which the experimenter said ‘‘No!’’to ‘‘Can I have it?,’’ and he continued to ask this question. How would you account forthese results?

7. What conclusions did the authors draw regarding maintaining variables for the three ques-tions?

8. What implications do the test conditions and results of the curent study have for clinicaluse?

Questions prepared by Carrie Dempsey and Stephen North, The University of Florida