MANDS FOR INFORMATION USING “WHO?” AND “WHICH?” IN THEPRESENCE OF ESTABLISHING AND ABOLISHING OPERATIONS

M. ALICE SHILLINGSBURG

MARCUS AUTISM CENTER AND EMORY UNIVERSITY SCHOOL OF MEDICINE

AND

CRYSTAL N. BOWEN, AMBER L. VALENTINO, AND LAURA E. PIERCEMARCUS AUTISM CENTER

Treatments designed to teach mands for information have included prompting and differentialreinforcement, as well as procedures to manipulate the relevant establishing operation (EO).However, previous studies have not included relevant abolishing operation (AO) conditions toensure that the mand is under relevant antecedent control. Data on listener responses (i.e., use of theinformation) are also absent in the literature. The current study shows differential responding underEO and AO conditions and reports listener responses that demonstrate use of the providedinformation. Three participants, diagnosed with an autism spectrum disorder, learned to mand forinformation using “who?” and “which?” questions exclusively under EO conditions. In addition,each participant responded to the information provided to access a preferred item. Generalization ofthe “which?” mand for information was also demonstrated across novel stimuli.

Key words: mands for information, establishing operation, abolishing operation, generalization

A mand is a type of verbal operant that isevoked by an establishing operation (EO) andreinforced by consequences specific to that EO(Skinner, 1957; Sundberg, 2007). Mands, orrequests, can take a variety of forms, and thereinforcer specified by the mand also varies. Forexample, a mand can specify tangible items oractivities but may also specify information as thereinforcer. Mands for information may take theform of “wh–” questions (i.e., who?, what?,when?, where?, why?, and which?). Unlike mandsfor a specific item or activity, mands forinformation do not result in immediate accessto something tangible, but instead provide anindividual with the information necessary toaccess preferred items or activities. For example,the mand for information “Who has my movie?”

might evoke a verbal statement specifying thelocation of the movie (i.e., information), which inturn sets the occasion for listener behavior leadingto access to the item. In this way, it is possible forthe information provided to become a condi-tioned reinforcer. Manding for information isreadily apparent in typically developing children(Brown, 1968); however, this skill often has tobe explicitly taught to children with autism andother language delays (Endicott & Higbee,2007).

Although several early studies have sought toteach children with disabilities to mand forinformation (Bondy & Erickson, 1976; Hung,1977; Knapczyk, 1989; Twardosz & Baer, 1973),these studies failed to manipulate the relevant EOsystematically. For example, Twardosz and Baer(1973) taught the response “what letter?” in thepresence of an unknown letter. However, it is notclear that there was an EO in place to establish thename of the letter as a reinforcer. Further, theexperimenters delivered a tangible reinforcer forresponding. Thus, it appears that in these studiesrequests for information were maintained by

We thank Brittany Lee and Francine White for theirassistance with data collection.

Amber L. Valentino is now at Trumpet Behavioral Health.Laura E. Pierce is now at the University of Kentucky.

Address correspondence to Alice Shillingsburg, MarcusAutism Center, 1920 Briarcliff Road NE, Atlanta, Georgia30329 (e-mail: alice.shillingsburg@choa.org).

doi: 10.1002/jaba.101

JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2014, 47, 136–150 NUMBER 1 (SPRING)

136

tangible reinforcers (e.g., tokens) rather thanaccess to the information.More recently, researchers have begun to focus

on teaching mands for information by manipu-lating antecedent variables to contrive an EO,thereby increasing the value of the information asa reinforcer (Betz, Higbee, & Pollard, 2010;Endicott & Higbee, 2007; Lechago, Carr, Grow,Love, & Almason, 2010; Shillingsburg, Valentino,Bowen, Bradley, & Zavatkay, 2011; Sundberg,Loeb, Hale, & Eigenheer, 2002;Williams, Donley,& Keller, 2000). For example, in a study with two4-year-old children with autism, Williams et al.(2000) began teaching trials by opening a box andmaking a comment about the contents such as“Oh, I like this one!” while shielding the contentsfrom the child. Because the child did not see whatwas in the box, this verbal stimulus served as anantecedent stimulus to evoke the mand “What’sthat?” Subsequently, “Can I see it?” and “Can I haveit?” were also taught. Similarly, Sundberg et al.(2002) conducted two experiments with threechildren who had been diagnosed with autism inwhich the experimenters manipulated the EOsto chain together two mands for information. InExperiment 1, the authors instructed the child to“get —” (an item previously removed from a toybox), which evoked the response “where—?” fromthe child. In Experiment 2, after the child emittedthe mand “where —?,” the authors presentedadditional verbal stimuli, such as “I gave it tosomebody,” which in turn evoked the mand“who?” These procedures were replicated byEndicott andHigbee (2007)with four preschoolerswith autism with similar results.In the above-mentioned studies, researchers

contrived EOs in order to increase the reinforcingvalue of the information by presenting specificantecedent stimuli (e.g., instructing a child toget a missing item that was hidden). Althoughcreative and effective methods to contrive therelevant EO were described, appropriate controlconditions were lacking. Specifically, thesestudies did not arrange conditions in which theabsence of the EO was programmed. Given that

the manipulated antecedents are partially respon-sible for the emission of the mand, it is possiblethat after learning to mand in the presence of theEO, the child may then emit the mand insituations with very similar antecedents but whenthe information is not needed (i.e., when the EOis not present). For example, a child who learnsthe mand “where?” when an item is missing andgiven the instruction “Get your —” may beginasking “where?” in response to the sameinstruction even if the item is clearly visible.Similarly, acquisition of the mand “Who has it?”in the presence of the verbal stimulus, “I gave it tosomebody,” might result in the same “Who hasit?” response in the presence of the verbalstimulus, “I gave it to Jerry.” In the latter examplethe information is already given, therefore, anabolishing operation (AO) rather than an EO forthe information is present. In these examples,because the putative mand is emitted regardless ofwhether the EO is in effect or not, the responsemay not actually function as a mand. Alterna-tively, the response might be under stimuluscontrol of common aspects of the learningenvironment.One approach to address this issue is to bring

the learner’s behavior under differential control ofsituations in which information is needed (i.e., anEO is present) and situations in which it is notneeded (i.e., an AO is present). Therefore,acquisition of mands for information could befacilitated by presenting trials that vary only as towhether the information is needed or not, butother antecedents and consequences are heldconstant. At least two previous studies haveapproximated such conditions. In the Sundberget al. (2002) study, a free-access condition waspresented periodically, in which information wasnot needed (i.e., the target item was available),similar to an AO condition. However, no datawere presented that would allow comparison ofresponding between occasions when the itemwas present or not present. Ingvarsson andHollobaugh (2010) also incorporated a similarprocedure by presenting known and unknown

MANDING WHO AND WHICH 137

questions to participants when teaching theresponse “I don’t know, please tell me” (IDKPTM).In their study, children were taught to emitIDKPTM when presented with unknown ques-tions, essentially teaching participants to mand forinformation (i.e., the correct answer). Theparticipants’ IDKPTM response generalized onlyto unknown questions (EO condition), and didnot generalize to known questions (AO condition),suggesting that the EO to obtain the neededinformation was present.When an individual mands for information, a

second person responds by providing information(e.g., verbal directions) that allows the individualto behave as a listener in order to access theterminal reinforcer. The reinforcer for the mand isthe information provided by the second person,but the listener behavior evoked by the informa-tion is an important indicator of the practicalvalue of teaching this repertoire. Therefore, it isimportant to measure not only the putative mandbut also the listener behavior occasioned by thepresentation of the information. Some recentresearch on mands for information has includedan assessment of listener behavior to ensure thatthe information provided can be used to access theterminal reinforcer (Betz et al., 2010; Endicott &Higbee, 2007; Lechago et al., 2010). To that end,researchers have conducted a preassessment todetermine if the participants can follow instruc-tions to go to a particular location or a particularperson. However, we are not aware of studies onmanding for information that have reported dataon relevant listener behavior evoked by thepresentation of verbal information. Informationprovided after a mand likely takes a slightlydifferent form than an instruction, which istypically what is provided during the preassess-ment. For example, an instruction may take theform of “Go to the bookshelf,” whereas informa-tion provided in response to “Where is my toy?”may take the form of “Your toy is on thebookshelf.” It may be useful to observe whetherinformation provided after a mand results in asimilar response of going to the location.

In addition to response acquisition, generali-zation across contexts is an important issue toconsider when teaching mands for information.Generalization should be assessed and pro-grammed, because teaching mands for informa-tion in each relevant setting would becomecumbersome and limit the functionality of themand repertoire. Several previous studies haveexamined generalization of mands for informa-tion with varying results. Betz et al. (2010) taughtmands for information using “where?” to threepreschool children with autism. Betz et al. used averbal cue and subsequently probed for generali-zation of the mand for information with theverbal cue using novel toys and in novel settings.In these situations, generalization occurred.However, when these authors probed forgeneralization in naturally occurring behaviorchains (e.g., giving the participant an empty boxof crayons following the instruction “It’s time tocolor”) without the verbal cue, generalizationwas not observed. Training without the verbalcue was necessary to produce generalization.Treatment procedures to teach mands forinformation should include not only an assess-ment of generalization but also procedures topromote generalization if it does not occur.The primary purpose of the current study

was to expand the literature on procedures toteach mands for information using “who?” and“which?” under AO conditions (i.e., informationnot needed) and EO conditions (i.e., informationneeded) and subsequent listener behavior (i.e., useof the information). In addition, for the “which?”mand, we examined subsequent listener behavior(i.e., use of the information) and generalizationfrom teaching trials to novel situations.

METHOD

Participants, Setting, and MaterialsThree children who attended an intensive

behavioral intervention clinic to address languageimpairments participated in the study. Ian was a12-year 4-month old boy who had beendiagnosed with autism. His repertoire consisted

138 M. ALICE SHILLINGSBURG et al.

of over 100 tacts for items, people, and actionsand use of carrier phrases when tacting. Heemitted intraverbal behavior, including intra-verbal responses to fill-ins and “what?” and“where?” questions. He vocally emitted multi-ple-word mands and consistently responded as alistener to verbal instructions. Jeb was an 8-year-old boy who had been diagnosed with autism. Hisvocal verbal repertoire was similar to that of Ian,and consisted of frequent mands in the form ofshort phrases, consistent listener responding toverbal instructions, approximately 100 tacts, useof carrier phrases, intraverbal fill-ins, and re-sponding to “what?” and “where?” questions. Jenwas a 6-year 1-month-old girl who had beendiagnosed with partial fetal alcohol syndrome andpervasive developmental disorder not otherwisespecified. She responded consistently as a listenerto verbal instructions and had a vocal verbalrepertoire that consisted of approximately 50 to100 tacts of items, people, and actions. Sheresponded to several personal information and“what?” questions. Her mand repertoire consistedof high rates of multiple-word vocal phrases.All trials were conducted in a classroom with at

least one therapist present. The room containedtables, chairs, shelves, and other teachingmaterials typically found in a classroom setting.Additional children and instructors were presentbut did not interact with the participants duringthe study. Study materials included small plasticcups, small cardboard boxes, small paper bags,and classroom cabinets for each participant.Containers were differentiated with stickers thatrepresented different characteristics, includingcolors, numbers, letters, or pictures, and werechosen individually for each participant based onpreviously mastered skill sets in his or her clinicaltreatment programming. Preferred edible andtangible items were also included, and variedaccording to each participant’s preferences.Preferred items were identified based on theparticipant’s verbal requests and informal assess-ments including observation and presentingchoices of items to the participant.

MeasurementTrained observers collected trial-by-trial data

using paper and pen. Observers scored a correctindependent response, correct prompted re-sponse, incorrect response, or a nonresponse formands for information emitted by the participantsand correct, incorrect, or nonresponse for theparticipants’ approach to the person or container.Mands for information using “who?” were scoredas correct if the participants said either “who?” or“who has it?” Mands for information using“which?” were scored as correct if the participantemitted the response “which—?” and named thespecific container in which the item was hidden.For example, if the therapist indicated that apreferred item was in one of the boxes, a correctresponse consisted of “which box?” Alternatively,if the therapist indicated that the item wasin one of the cabinets, a correct responseconsisted of “which cabinet?” Correct indepen-dent mands were those emitted before the prompt.Correct prompted mands were those emittedwithin 5 s of the presentation of the prompt.An approach response consisted of approachingthe correct person (“who?”) or the correctcontainer (“which?”) within 5 s of the locationbeing provided. All other responses were scoredas incorrect.The primary dependent variable was the

cumulative number of correct independentmands for information during EO and AOconditions. The mastery criterion in the EOcondition was 9 of 10 consecutive trials withindependent target responses.

Interobserver Agreement and Treatment IntegrityA second observer collected data on the target

mands “who?” and “which?” simultaneouslywith, but independently of, the experimenterduring preassessment, baseline, mand training,posttraining probes, and generalization probetrials. Trial-by-trial interobserver agreement wascalculated by dividing the number of agreementsby the number of agreements and disagreementsand converting the ratio to a percentage. An

MANDING WHO AND WHICH 139

agreement was defined as both the primary andsecondary observers recording a response ascorrect, incorrect, or a nonresponse within atrial. Interobserver agreement for “which?” forIan, Jeb, and Jen was 97%, 97%, and 92%,respectively, and was calculated for 41%, 53%,and 54% of trials, respectively. Agreement for“who?” was 96%, 100%, and 87% for Ian, Jeb,and Jen, respectively, and was calculated for 52%,43%, and 54% of trials. During the alternatingcondition, agreement was 93%, 89%, and 98%for Ian, Jeb, and Jen, respectively, and wascalculated for 40%, 37%, and 33% of trials.Treatment integrity was measured via a six-

item checklist of therapist behaviors required tocomplete a trial. Behaviors included providingthe participant’s preferred item to a therapist(“who?”) or placing the item under a designatedcup (“which?”), presenting the appropriatediscriminative stimulus, and implementing theappropriate delays and consequences. Treatmentintegrity was assessed for Ian during “which?,”“who?,” and alternating trials during 43%, 37%,and 22% of trials, respectively, and averaged99%, 100%, and 100%. Integrity was assessedfor Jeb during 34%, 21%, and 33% of trials for“which?,” “who?,” and alternating trials, respec-tively, and averaged 98%, 98%, and 100% forthese measures. For Jen, treatment integrity wasassessed for “which?,” “who?,” and alternatingtrials during 19%, 18%, and 22% of trials,respectively, and averaged 100% for all measures.

DesignAn adapted alternating treatments design

(Sindelar, Rosenberg, & Wilson, 1985) wasused to compare the effects of mand training inthe presence of an EO or an AO withinparticipants. A nonconcurrent multiple baselinedesign across participants was also used todemonstrate replication of the treatment effectsof mand training.Mands for information “which?” and “who?”

were targeted concurrently, but in separate trialblocks. A block of “who?” trials was conducted a

minimum of 30min after completing a block of“which?” trials. The number of trials conductedin each block for each mand varied each day(range, 2 to 13 for Ian, 2 to 19 for Jeb, and 5 to 15for Jen).

ProcedurePreassessment. The purpose of the preassess-

ment was to evaluate each participant’s ability tofollow vocal instructions, to ensure that theparticipant could use the information provided inresponse to the mand. During the preassessmentfor the “which?” scenario, the participant waspresented with an array of nine opaque contain-ers, each with a unique symbol or characteristic.The containers included cups, bags, boxes, andcabinets. All participants were able to tact eachtype of container. While the participant was notlooking, the therapist placed a preferred itemunder a designated container such that the itemwas no longer visible to the participant. Thetherapist then presented the demand “Give [handme or pick up] the — cup” while specifying aparticular color, number, letter, or picture. Theparticipant was given 5 s to emit a response. Ifthe participant selected the correct container, thepreferred item could be consumed. If he or sheselected the incorrect container or did not select acontainer, reinforcement was not available andthe therapist proceeded to the next trial. Eachcontainer with a specific characteristic wasprobed three times. If the participant correctlyfollowed a particular instruction three times, itwas considered to be part of the participant’slistener repertoire. The preassessment continueduntil nine characteristics for each container wereidentified. The selected containers with thedifferent characteristics were subsequently usedwhen teaching the mand “which?”One containerwas selected for teaching trials, and the otherswere used in subsequent generalization probes(Table 1).During the preassessment for “who?” at least

three additional therapists were present inaddition to the primary therapist and participant.

140 M. ALICE SHILLINGSBURG et al.

The primary therapist provided a preferred itemto one of the additional therapists when theparticipant’s attention was diverted and in-structed the participant to “go to [show me orpoint to] [individual’s name].” Following theinstruction, the participant was given 5 s torespond. If the participant approached the correctindividual, reinforcement was provided. If he orshe approached the incorrect individual or didnot respond, no reinforcement was provided andthe therapist moved to the next trial. Participantswere required to approach at least three additionaltherapists correctly in order to proceed with theintervention. Only therapists whom the partici-pant had correctly identified and approached inthe preassessment were included when teaching“who?”“Which?” scenario conditions. During the

“which?” scenario trials, nine opaque cups wereplaced upside down and side by side in front ofthe participant. Colored cups were used duringteaching for Ian and Jeb, and animal cups wereused for Jen. A highly preferred item was placedunder one of the cups without the participantseeing its location. The location of the preferreditem was different on each trial and wasrandomized among the nine cups. An emptycandy wrapper (e.g., an empty Skittles bag) orsome reference to the item (e.g., related items)was placed in sight of the participant to serve as astimulus that signaled the availability of thepreferred item. A trial began when the participantemitted a mand for the preferred item that wasunder the cup (e.g., “May I have a Skittle?”). Inthe EO condition, the therapist presented theverbal stimulus associated with the EO (i.e., “Youcan have a Skittle; it’s under one of these cups.”).

Any further information regarding the location(e.g., which cup the Skittle was under) was notprovided, potentially contriving a scenario inwhich emitting a mand for information using“which?” would be appropriate. In the AOcondition, the therapist presented the verbalstimulus associated with the AO, which includedinformation regarding the location of the itembeing provided (e.g., “Your Skittle is under theorange cup.”). This created a scenario in whichemitting the mand for information using“which?” would not be appropriate (the informa-tion of the location of the item had already beenprovided).“Who?” scenario conditions. During “who?”

trials, at least three additional therapists werestanding or sitting near the participant’s table. Ahighly preferred itemwas given to one of the extratherapists when the participant was not looking,and the therapist hid the item out of his or herview. The target therapist varied randomly fromtrial to trial. The primary therapist then placed anempty candy wrapper (e.g., an empty Skittlesbag) or something to signal the availability of theitem (e.g., a related item) in view of theparticipant. A trial began when the participantemitted a mand for the preferred item. Aprocedure similar to the “which?” scenario wasused to assess EO control. During the EO trials,the therapist provided the verbal stimulusassociated with the EO (e.g., “One of yourtherapists has your chip.”) to contrive a scenarioin which the mand for information using “who?”was appropriate. During the AO trials, thetherapist provided the verbal stimulus associatedwith the AO (e.g., “Brittany has your chip.”) tocontrive a scenario in which emitting the mand

Table 1“Which?” Scenario Teaching and Generalization Targets

Teaching Generalization

Ian Colored cups Colored boxes, numbered bags, numbered cups, numbered cabinetsJeb Colored cups Colored boxes, numbered bags, numbered cups, numbered cabinetsJen Animal picture cups Animal picture bags, animal picture cabinets, animal picture boxes

MANDING WHO AND WHICH 141

for information using “who?” would not beappropriate (the needed information had alreadybeen provided).During both “who?” and “which?” trials,

the participants were allowed to guess wherethe preferred item was located (i.e., guess thecontainer or the person), regardless of whetherthey manded for information. In other words,access to the containers or therapists was notblocked. This was done so that blocking access tothe cups during an EO trial and lack of blockingaccess during AO trials would not serve as a cue toindicate that a vocal response was required,thereby introducing undesirable stimulus con-trol. In addition, attempts to guess incorrectlyrather than manding for information mayfacilitate acquisition of the mand becauseincorrect guesses would not lead to access tothe preferred item, whereas manding for infor-mation would likely result in accessing the correctcontainer or person on each trial. Only oneguessed response was permitted before the trialwas terminated.Baseline and posttraining probes. Baseline

probes were conducted for the “which?” scenariofollowed by baseline probes for the “who?”scenario separated by a minimum of 30min. EOand AO condition trials were interspersed.During EO trials, if the participant emitted thecorrect mand for information (e.g., “Which cup?”or “who?”) the therapist immediately providedthe location or information (e.g., “The Skittle isunder the purple cup.”). If the participantresponded incorrectly or did not respond, noinformation or reinforcement was provided andthe therapist moved to the next trial. If theparticipant did not emit a response but attemptedto select a container or therapist and selected thecorrect one by chance, he or she was allowedaccess to the reinforcer.During AO trials, if the participant selected the

correct location or therapist, he or she wasallowed access to the reinforcer. If the participantdid not select a location or therapist or selectedthe incorrect location or therapist, no reinforce-

ment was provided and the therapist moved tothe next trial. If the participant emitted the mand“which?” or “who?” the trial was terminated.Treatment. During treatment, EO and AO

trials were conducted as in the baseline probeswith the exception that a constant prompt delay(Schuster, Gast, Wolery, & Guiltinan, 1988)was implemented to teach the mands “who?”and “which?” during EO trials. The procedureconsisted of an initial 0-s delay that was presentedfor a minimum of the first eight EO trials. DuringEO trials, when the therapist presented the verbalstimulus associated with the EO, the therapistimmediately provided a vocal prompt relevant tothe mand being taught (i.e., “Which cup?” or“Who has it?”) and allowed 5 s for the participantto imitate the response. Correct promptedresponses resulted in the therapist providing theinformation needed to retrieve the preferred item.When at least eight trials with a 0-s delay hadbeen conducted and the participant had re-sponded correctly to the prompt for threeconsecutive trials, a 2-s prompt delay wasincorporated to allow an opportunity for anindependent response. Correct prompted andindependent responses resulted in informationregarding the location of the preferred item. Aprompt for the correct response was given after 2 swithout a response or contingent on an error.Mastery criteria consisted of emitting correctindependent mands for information for 9 of10 EO trials and the absence of manding in atleast the last three AO trials. During AO trials,following the mand for the preferred itememitted by the participant, the therapist providedthe verbal stimulus associated with the AO,which included the information to access thepreferred item. No consequences were provided ifthe mand for information was emitted during anAO trial (this never happened for Jeb or Jen).Alternating scenarios. Subsequent to mastery

of both mands for information, EO and AO trialsfor “which?” and “who?” were alternated in aquasirandom fashion. These trials were identicalto those conducted in the baseline probe phase.

142 M. ALICE SHILLINGSBURG et al.

Generalization probes. Generalization probeswere conducted for the mand “which?” withpreferred items located in containers other thanthose used during teaching trials. During thisphase, the therapist conducted trials similar to thebaseline probes, but placed the objects in a varietyof containers such as colored or numbered bags,boxes, and classroom cabinets (see Table 1). Ifgeneralized use of the mand “which —?” did notoccur, additional strategies (described below)were used to promote generalization. Generaliza-tion probes for the mand “who?” were notconducted.Tact training. Booster tact training was im-

plemented if the “which —?” response did notgeneralize to the new containers. Although allparticipants emitted tacts for all containers in thepreassessment, one did not exhibit generalizedresponding to untaught containers during gener-alization probes (described in results) and insteadcontinued to emit “which —?” specifying thecontainer name used during teaching rather thanthe new container. Booster tact training wasconducted to assess whether recent practice ofeach container’s name would facilitate accuratemanding. Tact training consisted of presentingthe container (i.e., bag, box) and asking “What isit?” If the participant did not respond, a vocalprompt to label the item was provided, followedby a partial vocal prompt, and another indepen-dent opportunity. After the participant hadcorrectly tacted the container on five consecutiveindependent opportunities, a generalizationprobe for the mand “which —?” using thatcontainer was conducted.Multiple exemplar training. If booster tact

training did not produce generalization of themand “which—?” additional teaching trials withthe novel containers were conducted. Teachingtrials and mastery criteria were identical to theteaching trials with the first taught container.One additional container was taught to masteryfollowed by a probe of the remaining untaughtcontainers to assess generalization. If generaliza-tion did not occur, an additional container was

taught followed by a probe of untaught contain-ers, and this continued until all containers weretaught or generalization occurred.

RESULTS

Figures 1 through 6 show the results of mandtraining and generalization probes for all threeparticipants. Figure 1 (left) shows baseline andposttraining of mands for information using“which?” for Ian (top), Jeb (middle), and Jen(bottom). Figure 1 (right) shows baseline andposttraining approach responses (i.e., listenerresponses) to select the correct container. Inbaseline none of the participants emitted themand “which?” in EO or AO conditions, andthey reliably selected the correct container only inthe AO condition when the information hadalready been provided. After mastery of the“which?” mand (see Figure 2), posttrainingprobes were conducted. During posttraining(Figure 1) each participant emitted the mandfor information using “which?” only under EOconditions and did not emit the mand under AOconditions (i.e., information already provided).In addition, when the participants manded andinformation was provided, they began to respondas listeners by approaching and selecting thecorrect container. In other words, they used theinformation provided to access the reinforcers.Ian and Jen used the information 100% of thetime during both EO and AO conditions, and Jebused the information during 100% of EO trialsand 80% of AO trials.Figure 2 shows the teaching trials for “which?”

for all three participants. Ian met criteria to fadeto a 2-s delay after 24 trials of alternating EO andAO trials and met mastery criteria after 53 trials.Jeb met criteria to fade to a 2-s delay after 18 trialsof alternating EO and AO trials and met masterycriteria after 61 trials. Jen met criteria to fade to a2-s delay after 16 trials of alternating EO and AOtrials and met mastery criteria after 33 trials.Approach responses were seen throughout allteaching trials when information was given and

MANDING WHO AND WHICH 143

occurred during 86% of EO trials and 100% ofAO trials for Ian, 91% of EO trials and 96% ofAO trials for Jeb, and 94% of EO trials and 75%of AO trials for Jen.

Figure 3 (left) shows baseline and posttrainingmands for information using “who?” Figure 3(right) shows approach responses (i.e., use of theinformation). In baseline, none of the participants

0

5

10

0 2 4 6 8 10 12 14 16 18 20 22 240

5

10

0 2 4 6 8 10 12 14 16 18 20 22 24

Trials Trials

0

5

10

0 2 4 6 8 10 12 14 16 18 20 22 24

0

5

10

0 2 4 6 8 10 12 14 16 18 20 22 24Cu

mul

a�ve

App

roac

h -W

hich

0

5

10

0 2 4 6 8 10 12 14 16 18 20 22 24

0

5

10

0 2 4 6 8 10 12 14 16 18 20 22 24

Cum

ula�

ve M

and

-Whi

ch

Ian

Jeb

Jen

Ian

Jeb

Jen

EO

AOEO

AO Baseline Pos�raining Baseline Pos�raining

Figure 1. Cumulative record of independent responses for manding “which?” (left) and cumulative record for approachbehavior (right) during baseline and posttraining probes for Ian, Jeb, and Jen.

Figure 2. Cumulative record of independent and prompted responses for manding “which?” (left) and cumulative recordfor approach behavior (right) during mand training for Ian, Jeb, and Jen.

144 M. ALICE SHILLINGSBURG et al.

emitted the mand “who?” in the EO or AOconditions, and they reliably approached thecorrect person only when the information hadalready been provided in the AO condition. Aftermastery during “who?” teaching trials (seeFigure 4), posttraining probes were conducted.During posttraining each participant emitted themand for information using “who?” under EOconditions and did not emit the mand under AOconditions. In addition, when the participantmanded and information was provided, he or sheused the information to approach the correctperson. Ian used the information 100% of thetime during both conditions. Jeb used theinformation during 100% of EO trials and81% of AO trials, and Jen used the informationduring 100% of EO trials and 92% of AOtrials.Figure 4 shows teaching trials for “who?” for all

three participants. During treatment, Ian metcriteria to fade to a 2-s delay after 34 trials ofalternating EO and AO trials and met masterycriteria after 94 trials. Jeb met criteria to fade to a2-s delay after 18 trials of alternating EO and AO

trials and met mastery criteria after 36 trials. Jenmet criteria to fade to a 2-s delay after 16 trials ofalternating EO and AO trials and met masterycriteria after 57 trials. Ian, Jeb, and Jen used theinformation during EO trials for 84%, 94%, and93% of trials and 98%, 83%, and 89% of AOtrials, respectively.Figure 5 shows the results of the assessment of

alternating “which?” and “who?” trials with eachparticipant. All three participants differentiallyemitted “which?” and “who?” correctly duringEO trials. Jeb and Jen also did not emit mands forinformation in the AO condition for bothscenarios. Ian, however, emitted the mand“who?” in the AO condition on four of thefirst five trials. During these trials, it appearedthat Ian emitted “who?” before the therapistcould tell him which person had the preferreditem. Beginning with Trial 6, the therapistdiscontinued saying “Yes, you can have yourSkittle” and instead simply stated the name of theperson immediately. Following this modification,Ian never asked “who?” in the AO condition. Theparticipants used the information during the

0

10

20

0 2 4 6 8 10 12 14 16 18 20 22 240

5

10

0 10 20

Trials Trials

0

5

10

0 2 4 6 8 10 12 14 16 18 20 22 24

0

5

10

0 2 4 6 8 10 12 14 16 18 20 22 24

Cum

ula�

ve A

ppro

ach

-Who

0

5

10

0 2 4 6 8 10 12 14 16 18 20 22 24

0

5

10

0 2 4 6 8 10 12 14 16 18 20 22 24

Cum

ula�

ve M

and

-Who

Ian

Jeb

Jen

Ian

Jeb

Jen

EOAO

AO

EO

Baseline BaselinePos�raining Pos�raining

Figure 3. Cumulative record of independent responses for manding “who?” (left) and cumulative record for approachbehavior (right) during baseline and posttraining probes for Ian, Jeb, and Jen.

MANDING WHO AND WHICH 145

“which?” alternating assessment for 100% of alltrials during both EO and AO trials. During the“who?” alternating assessment, Ian, Jeb, and Jenused the information during 100%, 83%, and

100% of EO trials and 60%, 100%, and 100% ofAO trials, respectively.Figure 6 shows the results of the generalization

probes for the mand for information “which?”

Figure 4. Cumulative record of independent responses for manding “who?” (left) and cumulative record for approachbehavior (right) during mand training for Ian, Jeb, and Jen.

Figure 5. Cumulative record of independent responses for manding “which?” and “who?” (left) and cumulative recordfor approach behavior (right) during alternating conditions for Ian, Jeb, and Jen.

146 M. ALICE SHILLINGSBURG et al.

Before teaching the mand “which?” with num-bered cups, all containers (see Table 1) wereprobed in baseline. No manding for informationwas observed with any containers. Followingmastery of the mand “which?” using thenumbered cups, probes to assess manding usingnovel containers were conducted. After learningto mand “which cup?” both Ian and Jeb emittedmands for “which box?,” “which bag?,” and“which cabinet?,” demonstrating generalized useof the mand “which?” Use of information duringthe EO trials occurred during 91%, 78%, and100% of EO trials and 94%, 100%, and 92% ofAO trials for Ian, Jeb, and Jen, respectively.

Generalized responding was not observedwith Jen following mastery of the mand “whichcup?” Instead she emitted “which cup?” ratherthan the name of the specified container.Therefore, tact training to label each containerwas conducted. Tacts of the containers wereconsidered mastered following five trials ofindependent labeling (data available from thefirst author). Tact training was not sufficient toproduce the mand “which?” using the correctcontainer name. Therefore, we directly taught asecond exemplar (“which box?”). Mastery of thismand did not result in generalization; therefore,trials of “which cup?” and “which box?” were

Figure 6. Cumulative record of independent responses for manding “which?” during generalization probes for Ian, Jeb,and Jen. Solid lines indicate whenmand training occurred. Dotted lines indicate that a separate generalization probe occurredor tact training of the containers used in the generalization probes. PT¼ posttraining; GP¼ generalization probe.

MANDING WHO AND WHICH 147

alternated rapidly. Generalization still failed tooccur; therefore, another exemplar was taught(“which bag?”). Following mastery of “whichbag?” generalization of the mand “which cabi-net?” was observed in the classroom.

DISCUSSION

The current study demonstrates successfulacquisition of the mands for information “who?”and “which?” under EO conditions and subse-quent listener responses (i.e., use of the informa-tion) with three participants. All participantsshowed differential responding between con-ditions in which information was needed (EOcondition) versus when it had already beenprovided (AO condition), demonstrating controlof the response by the relevant EO. Allparticipants also emitted the correct mand forinformation when alternating between situationsin which the mand “who?” or “which?” wasappropriate. Generalization of the “which?”mand to novel scenarios was observed for twoof the three participants without additionalteaching. One participant, Jen, required addi-tional strategies to promote generalization of the“which?” mand. Ultimately, multiple exemplarinstruction resulted in generalized use of themand when the preferred item was hidden in oneof several cabinets in the classroom.The current study highlights some potentially

important components to consider when teach-ing mands for information. First, programmingboth EO and AO conditions ensures functionaluse of the mand for information by reducing theprobability of alternative sources of control (e.g.,teaching materials or stimuli associated withtherapist presence). In addition, in the currentstudy the experimenter waited for the child toemit a mand for the hidden preferred itemspontaneously. When the child manded for thehidden item, information regarding the locationwas provided (i.e., it’s under one of the cups, orit’s under the yellow cup). By contrast, in previousstudies, the adult typically initiated the trial

by giving an instruction (e.g., “Let’s play, getyour toy”; Ostryn & Wolfe, 2011; Sundberget al., 2002). We suggest that the currentapproach served to ensure control by relevantEOs rather than potential discriminative stimuli.Another novel component of the present study

was measuring use of the information (i.e.,listener responses) during baseline, treatment,and posttraining, rather than relying solely on apreassessment of listener responses. Failure to usethe information by failing to approach theappropriate container or person could indicateeither absence of an EO for the hidden preferreditem or an inadequate listener repertoire. Severalprevious studies included preassessments toensure appropriate listener responses beforeintervention (Betz et al., 2010; Lechago et al.,2010; Ostryn & Wolfe, 2011); however, noprevious studies have reported data on listenerresponses that constituted use of informationduring the study. Although a preassessment of thelistener repertoire is a good indicator of futureuse of provided information, it is also necessaryto measure listener responses during and aftertreatment to evaluate the functionality of themands being taught.Another component to consider is teaching

more than one mand. Previous researchers havediscussed the potential benefit of teaching morethan one “wh–?” question when targeting mandsfor information to promote differentiated useof the mand under the appropriate conditions(Ostryn & Wolfe, 2011). These researchersindicated a temporary decline in correct use ofthe first mastered mand for information when anadditional mand was targeted for instruction.Thus, for a period of time the participants tendedto emit the newmand during situations when thepreviously mastered mand would have beenappropriate. In the current study, two differentmands for information were taught simulta-neously to promote differentiated responding.Targeting “which?” and “who?” simultaneouslymay have increased the likelihood that theappropriate response would be used under the

148 M. ALICE SHILLINGSBURG et al.

appropriate conditions. Future researchers couldcompare whether teaching two mands simulta-neously rather than sequentially results in fasteror slower overall acquisition.Measurement of generalized responding is

another important component when teachingmands for information. Generalization wasassessed with all three participants, and specificteaching procedures were tested to promotegeneralization of responding when it did notoccur for the “which?” mand. Although generali-zation was observed for two of the threeparticipants without additional programming,one participant (Jen) required multiple exemplartraining with additional scenarios to achievegeneralized use of the mand “which?” It ispossible that Jen’s overall verbal repertoire andlearning history explain the lack of generalization.Both Ian and Jeb had a history of emitting carrierphrases when manding and tacting. Jen hadonly recently learned to emit mand frames anddid not emit carrier phrases when tacting.In addition, she often required more extensiveteaching to acquire new skills. Future researchshould evaluate the probability of generalizedresponding across participants with diverserepertoires.One limitation of the current study was the

lack of a generalization assessment for the mand“who?”Due to staffing limitations, generalizationprobes with alternative people were not feasiblein the clinic setting. Also, we did not assessgeneralization outside the clinic setting, althoughnovel materials were used. Future research mayassess generalization with a variety of people andacross settings.Future research should also address a limita-

tion that was present during the AO conditionsfor both of the mands taught. When a participantemitted “who?” or “which?” mands (during theirrespective trials) under AO conditions, noconsequences were provided and the trial wasterminated, potentially resulting in extinction.This occurred through all phases of the study.Information provided during AO conditions

should not have value as a reinforcer, becausethe child already has the information necessary toaccess the preferred item. However, the abilityto evaluate the differential reinforcing valueof the information was compromised by notdelivering it in the AO condition. Further, thisapproach can potentially lead to differentialreinforcement of the putative mand for informa-tion, which limits the ability to isolate therelevant EO (i.e., lack of information) as thecontrolling variable.Although this was a limitation of the current

procedure, only one participant emitted mandsduring AO trials, and this occurred infrequently(only six times throughout the study). Therefore,this methodological shortcoming cannot beconsidered a major flaw of the current study.Nevertheless, future research should programidentical consequences for all responses duringboth AO and EO conditions to isolate the effectsof the relevant motivating operation (MO;Smith, Iwata, Goh, & Shore, 1995).This study extends the literature by highlight-

ing and expanding the role of the relevant MOwhen teaching mands for information to childrenwith autism. Two methodological features helpedascertain that mands were being taught under theappropriate antecedent (MO) conditions: (a) theinclusion of the AO condition and (b) definingtrials based on child initiation. Inclusion of anevaluation of listener behavior also extendsprevious research by evaluating the extent towhich participants were able to use the informa-tion presented to them. Future research shouldfurther evaluate the effectiveness and efficiencyof these procedures, as well as the potential forthe current procedures to produce lasting andgeneralized behavior change.

REFERENCES

Betz, A. M., Higbee, T. S., & Pollard, J. S. (2010).Promoting generalization of mands for informationused by young children with autism. Research in AutismSpectrum Disorders, 4, 501–508. doi: 10.1016/j.rasd.2009.11.007

MANDING WHO AND WHICH 149

Bondy, A. S., & Erickson, M. T. (1976). Comparison ofmodelling and reinforcement procedures in increasingquestion-asking of mildly retarded children. Journalof Applied Behavior Analysis, 9, 108. doi: 10.1901/jaba.1976.9-108

Brown, R. (1968). The development of wh questions inchild speech. Journal of Verbal Learning and VerbalBehavior, 7, 279–290. doi: 10.1016/S0022-5371(68)80002-7

Endicott, K., & Higbee, T. (2007). Contriving motivatingoperations to evoke mands for information in pre-schoolers with autism. Research in Autism SpectrumDisorders, 1, 210–217. doi: 10.1016/j.rasd.2006.10.003

Hung, D. W. (1977). Generalization of “curiosity”questioning behavior in autistic children. Journal ofBehavior Therapy and Experimental Psychiatry, 8, 237–245. doi: 10.1016/0005-7916(77)90061-1

Ingvarsson, E. T., & Hollobaugh, T. (2010). Acquisition ofintraverbal behavior: Teaching children with autism tomand for answers to questions. Journal of Applied BehaviorAnalysis, 43, 1–17. doi: 10.1901/jaba.2010.43-1

Knapczyk, D. R. (1989). Generalization of student questionasking from special class to regular class settings. Journalof Applied Behavior Analysis, 22, 77–83. doi: 10.1901/jaba.1989.22-77

Lechago, S. A., Carr, J. E., Grow, L. L., Love, J. R., &Almason, S. M. (2010). Mands for informationgeneralize across establishing operations. Journal ofApplied Behavior Analysis, 43, 381–395. doi: 10.1901/jaba.2010.43-381

Ostryn, C., & Wolfe, P. S. (2011). Teaching preschoolchildren with autism spectrum disorders to expressivelydiscriminate between “what’s that?” and “where is it?”Focus on Autism and Other Developmental Disabilities,26, 195–205. doi: 10.1177/1088357611421504

Schuster, J. W., Gast, D. L., Wolery, M., & Guiltinan, S.(1988). The effectiveness of a constant time-delayprocedure to teach chained responses to adolescents

with mental retardation. Journal of Applied BehaviorAnalysis, 21, 169–178. doi: 10.1901/jaba.1988.21-169

Shillingsburg, M. A., Valentino, A. L., Bowen, C. N.,Bradley, D., & Zavatkay, D. (2011). Teaching childrenwith autism to request information. Research in AutismSpectrum Disorders, 5, 670–679. doi: 10.1016/j.rasd.2010.08.004

Sindelar, P. T., Rosenberg, M. S., & Wilson, R. J. (1985).An adapted alternating treatments design for instruc-tional research. Education & Treatment of Children, 8,67–76.

Skinner, B. F. (1957). Verbal behavior. Englewood Cliffs, NJ:Prentice Hall.

Smith, R. G., Iwata, B. A., Goh, H., & Shore, B. A. (1995).Analysis of establishing operations for self-injurymaintained by escape. Journal of Applied BehaviorAnalysis, 28, 515–535. doi: 10.1901/jaba.1995.28-515

Sundberg, M. L. (2007). Verbal behavior. In J. O. Cooper,T. E. Heron, & W. L. Heward (Eds.), Applied behavioranalysis (2nd ed., pp. 526–547). Upper Saddle River,NJ: Merrill/Prentice Hall.

Sundberg, M. L., Loeb, M., Hale, L., & Eigenheer, P.(2002). Contriving establishing operations to teachmands for information. The Analysis of Verbal Behavior,18, 15–29.

Twardosz, S., & Baer, D. M. (1973). Training two severelyretarded adolescents to ask questions. Journal of AppliedBehavior Analysis, 6, 655–661. doi: 10.1901/jaba.1973.6-655

Williams, G., Donley, C. R., & Keller, J. W. (2000).Teaching children with autism to ask questions abouthidden objects. Journal of Applied Behavior Analysis, 33,627–630. doi: 10.1901/jaba.2000.33-627

Received July 2, 2012Final acceptance August 22, 2013Action Editor, Einar Ingvarsson

150 M. ALICE SHILLINGSBURG et al.