jeab-96-03-291

THE IMPACT OF BODY-PART-NAMING TRAINING ON THE ACCURACY OF IMITATIVEPERFORMANCES IN 2- TO 3-YEAR-OLD CHILDREN

VERA CAMOES-COSTA, MIHELA ERJAVEC, AND PAULINE J. HORNE

BANGOR UNIVERSITY , WALES

A series of three experiments explored the relationship between 3-year-old childrens ability to nametarget body parts and their untrained matching of target hand-to-body touches. Nine participants, 3 perexperiment, were presented with repeated generalized imitation tests in a multiple-baseline procedure,interspersed with step-by-step training that enabled them to (i) tact the target locations on their ownand the experimenters bodies or (ii) respond accurately as listeners to the experimenters tacts of thetarget locations. Prompts for on-task naming of target body parts were also provided later in theprocedure. In Experiment 1, only tact training followed by listener probes were conducted; inExperiment 2, tacting was trained first and listener behavior second, whereas in Experiment 3 listenertraining preceded tact training. Both tact and listener training resulted in emergence of namingtogether with significant and large improvements in the childrens matching performances; this wastrue for each child and across most target gestures. The present series of experiments provides evidencethat naming the most basic form of self-instructional behaviormay be one means of establishinguntrained matching as measured in generalized imitation tests. This demonstration has a bearing onour interpretation of imitation reported in the behavior analytic, cognitive developmental, andcomparative literature.

Key words: naming, imitation, listener behavior training, tact training, matching training, youngchildren, manual gestures

_______________________________________________________________________________

Researchers of different theoretical persua-sions agree that imitation is a key driver ofdevelopment in infancy and childhood, andthat its determinants deserve careful experi-mental investigation. Behavior analysts distin-guish between a directly trained repertoire ofmatching relations, established through dis-criminative reinforcement like any other oper-ants (Skinner, 1953, pp. 119120), and gener-alized imitation, a repertoire of emergentmatching relations (Catania, 1998, p. 228). Thislatter kind of imitative repertoire, which couldenable young children to learn new behaviorsquickly and without the need for directtraining, has been the focus of numerousexperiments (e.g., Baer & Deguchi, 1985; Baer& Sherman, 1964; Erjavec, Lovett, & Horne,

2009; Horne & Erjavec, 2007; Kymissis &Poulson, 1994; Poulson & Kymmissis, 1988;Poulson, Kyparissos, Andreatos, Kymissis, &Parnes, 2002; Poulson, Kymissis, Reeve, Andrea-tos, & Reeve, 1991; Steinman, 1970; Waxler &Yarrow, 1970). Traditionally, the methodologyfor examining generalized imitation consists ofthe presentation of discrete trials on each ofwhich the child observes a different modeledaction to which he or she is asked to respond.Following some modeled actions, correctmatching responses result in the delivery ofreinforcers, but matching responses to theremaining models are not reinforced. It is thechildrens responses to the latter models thatare of interest, as matching of unreinforcedprobes is considered to provide evidence ofgeneralized imitation. If unreinforced match-ing is shown to be sensitive to changes in thecontingencies provided for reinforced respons-es, this is considered to be evidence thatgeneralized imitation had been established asa higher-order class of behavior (Catania,1998). Thus, evidence of generalized imitationhas been reported in infants (e.g., Poulsonet al., 1991, 2002), normally developing chil-dren (e.g., Baer & Sherman, 1964; Catania,1998; Sherman, Clark, & Kelly, 1977), andchildren from special populations (e.g., Baer,

We are grateful to the children for their participation inthe study. We also wish to thank the staff at the UniversityNursery and Childcare Centre, Tir Na nOg, Bangor, fortheir support, kindness, and help. The research reportedin this paper had been supported by a Ph.D. Scholarshipawarded to Vera Camoes-Costa by the School of Psychol-ogy, Bangor University.

Address correspondence to Dr. Mihela Erjavec, Schoolof Psychology, Bangor University, Brigantia Building,Penrallt Road, Bangor, Gwynedd, LL57 2AS, UnitedKingdom (e-mail: [email protected]; web: www.psychology.bangor.ac.uk/playlab).

doi: 10.1901/jeab.2011.96-291

JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 2011, 96, 291315 NUMBER 3 (NOVEMBER)

291

Peterson, & Sherman, 1967; Garcia, Baer, &Firestone, 1971; Peterson, 1968).In the recent literature, additional necessary

controls have been identified and incorporatedinto tests for generalized imitation. First, it hasbeen shown that all such tests ought to employpretests for novelty of the target actions present-ed as unreinforced probes. Second, researchersalso need to ensure that parents and othercaregivers remain unaware of the experimentalprocedures, particularly the modeled targetbehaviors, for the duration of the study. Thesesteps safeguard against false positivescaseswhere target responses may be extraexperimen-tally established as trained matching relationseither prior to or during the experiment (seeErjavec & Horne, 2008; Erjavec et al., 2009;Horne & Erjavec, 2007 for a more detaileddiscussion of these points and relevant data).Third, in tasks that model an action directed at aparticular object, it has been shown that theaction component may not be necessary toevoke the target response (see Zentall, 2006).For example, Horne, Erjavec, and Lovett (2009)tested infants imitation of a particular targetbehavior. Depending on assigned condition, theinfants were either shown a mitten falling froma puppets arm (affordance demonstrationcontrol), or an experimenter pointing to amitten on the puppets arm (social enhance-ment control), or an experimenter removingthe mitten from the puppets arm (targetbehavior modelling condition). This study dem-onstrated that infants in the affordance andsocial enhancement control conditions pro-duced as many mitten removals in subsequenttest trials as those who had seen the full targetbehavior of the experimenter removing themitten. Because object affordances and socialenhancement are potential confounding sourc-es of control when target behaviors are actionson objects, the authors concluded that, toprovide a strong test of generalized imitation,future studies should employ novel, empty-handed gestures (i.e., those that do not involvetouching or holding objects) as target behaviors(see Horne & Erjavec, 2007, for further discus-sion of this issue and relevant data).Recent studies examining the determinants

of imitation in infants and young children thathave employed these experimental controlshave failed to replicate the results of earlierstudies. For example, neither extensive expo-sure to modeling nor multiple-exemplar

matching training led to the emergence ofnovel untrained matching relations in infants(Erjavec et al., 2009; Horne & Erjavec, 2007).These results indicate that the imitative abili-ties of infants and young children may havebeen overestimated in the earlier behavioranalytic literature, and for similar reasons, alsoin cognitive-developmental studies (e.g., seeHurley & Chater, 2005). Clearly, more re-search is needed to identify the conditionsunder which young children may show emer-gent matching of novel behaviors.One possibility is that childrens growing

verbal repertoires may alter the way in whichthey respond to behaviors modeled by others.Although experimenter-generated instruc-tions have long been identified as onedeterminant of matching responses in school-and nursery-age children (see Baer & Deguchi,1985, for a review), the effects of self-instruc-tions on young childrens imitative responseshave not yet been examined directly. Horneand Lowe (1996, 1997; also see Lowe & Horne,1996) propose that learning to name objectsand events in their environment fundamental-ly changes the way in which children behave.They define naming as a higher-order bidirec-tional relation, in which the speaker respondsas a conventional listener to his or her ownverbal responses. Via this intraindividualspeakerlistener relation, when a child seesan object (e.g., a shoe) and tacts it (shoe),the child next responds as a listener to thattact by looking once again at the object, and soon. Naming is therefore the earliest form ofself-instruction: This circular speakerlistenerrelation enables the child to maintain his orher attention on a particular object for as longas that particular cycle of speaking andlistening continues. In addition, wheneverthe child names a particular object (e.g., thechilds shoe), the listener response includeslooking at other objects that he or she haslearned to call by the same name. Naming istherefore an important means of establishingcategory relations between objects (e.g., thechilds shoe and the wide variety of adultsshoes). Indeed, common naming (but notcommon listener behavior) has been shown toestablish untrained categories in young chil-dren, even between objects that have nofeatures in common (see Horne, Hughes, &Lowe, 2006; Horne, Lowe, & Harris, 2007;Horne, Lowe, & Randle, 2004; Lowe, Horne, &

292 VERA CAMOES-COSTA et al.

Hughes, 2005; Lowe, Horne, Harris, & Randle,2002). Therefore, it is possible that when a childobserves her mother touching her own foot, thechild may tact what her mother has done bysaying touch foot or simply foot and thenrespond in turn as a listener to that utterance bylooking at her own foot and touching it. In thisway, the self-instructional effects embodied inthe name relation may alter the way in which achild (or adult) responds on generalized imita-tion tests. Some recent findings suggest that thisis indeed the case. For example, many empty-handed gestures used as target behaviors inimitation tests involve touching a specific bodyparta shoulder, an elbow, a palm, a foot, andso on. It is well documented that infants andyoung children accurately match some of thesemodels, but respond incorrectly to the remain-ing models by touching a different part of theirown body, or using an incorrect movement (e.g.,Bekkering & Wohlschlager, 2002; Bekkering,Wohlschlager, & Gattis, 2000; Erjavec et al.,2009; Gleissner, Meltzoff, & Bekkering, 2000;Horne & Erjavec, 2007). Erjavec and Horne(2008) have demonstrated that toddlers re-sponses to hand-to-body target gestures tend tobe more accurate for those actions that fre-quently feature in naming and matching gamesthat the children play with their caregivers (e.g.,the nursery rhyme Heads, shoulders, knees andtoes) than for comparable actions that had nosuch training history (also see Camoes-Costa,Erjavec & Horne, 2010). However, the separatecontributions of matching training, namingtraining, and listener training have not yet beenassessed directly.Let us consider in more detail how learning

to name a target body part may change achilds performance in an imitation test.Figure 1 (left panel) illustrates the namingrelation for foot (adapted from Horne & Lowe,1996, p. 201). A child, who has learned toname this part of the body, sees her foot, andthen says foot. This response automaticallygenerates the auditory stimulus /foot/ towhich the child responds as a listener bylooking once again at her foot. This namerelation can be evoked by seeing a foot orhearing /foot/; therefore, it can be re-evokedeach time the foot is seen, or through self-echoing. As caregivers train this relation, theyare likely to point not only to the childs footbut also to other feettheir own and siblingsfeet, feet on toys, pet animals, birds, and so on.

This name relation will therefore come toinclude a variety of other stimuli that a childmay name and, in so doing, categorize themas feet (this relation may also include a varietyof conventional listener responseslookingat, pointing to, touching, kicking, putting onand pulling off socks or shoes, and so on,depending on the childs learning history).When, in an imitation context, the child seesan experimenter modeling a foot-touch, shemay name the target body part as foot,overtly or covertly, and respond to her ownutterance by orienting to and touching herown foot (see Figure 1, right panel). Thiscould happen even if the child has never beendirectly trained to produce a matching re-sponse to a model of a hand-to-foot targetaction. Thus an emergent matching responsemay be emitted in a generalized imitation test.However, if the child has not yet learned toname the target body part touched during amodeling demonstration, then naming cannotfacilitate matching of the target behavior.In the present study, we employed multiple

baseline designs in three experiments toexplore the relationship between young chil-drens ability to name target body parts andtheir untrained matching of target hand-to-body touches. Participants were presented withrepeated generalized imitation tests in whichmodels of four trained (baseline) hand-to-bodytouches were interspersed with four unrein-forced (target) body touches; the novelty of alltarget behaviors was established at the outsetfor each individual child. The separate effectsof training the children to (i) tact the targetlocations, and (ii) respond as listeners to theexperimenters tacts of the target locations ontheir subsequent matching of the target bodytouches, were next investigated. We also exam-ined whether prompts for on-task naming oftarget body parts enhanced the childrensgeneralized matching performances.This research complied with British Psycho-

logical Society guidelines for research withchildren and was approved by the School ofPsychology Ethics Committee.

EXPERIMENT 1METHOD

Participants

Three typically developing girls who attendedthe University Nursery and Childcare Centre

BODY-PART NAMING AND IMITATION IN CHILDREN 293

Tir Na nOg in Bangor at least 2 days a weekwere recruited by parental consent to partici-pate in this experiment. Participants are re-ferred to by short alternative names to preserveconfidentiality. At the start of tact training theywere aged 33 months (Emma), 34 months(Anna), and 35 months (Mol). Table 1 showsthe total number of sessions administered toeach child, their ages at the end of theprocedure, and the General Quotient scoreson the Griffiths Mental Developmental Scales(Luiz et al., 2006) obtained for 2 of the children(the remaining child left the nursery before thistest could be administered).

Settings and Apparatus

Sessions were conducted in a speciallydesigned quiet testing room at the nursery.During the sessions, the child and the exper-imenter sat comfortably in an inflatable boat,on beanbags, facing each other. A large teddybear toy (Teddy) was seated on the edge of theboat, facing the child. Age appropriate toysand stickers were used during play breaksbetween test trials and after testing. Theseitems were kept hidden in a closed box and inTeddys backpack between presentations. Twowall-mounted digital video cameras were em-ployed to record the behavior of the child andthe experimenter. Audio and visual inputsfrom the two cameras and a hidden radiomicrophone were fed into a split-screen videorecorder located in a separate audiovisualsuite. JVC SR-VS10 VHS/DV recorders, with

stop- and slow-motion viewing facilities, wereused for recording and coding.The visual stimuli employed were manual

gestures performed live by the experimenter(see Table 2). These gestures consisted oftouches to different parts of the body andwere chosen based on the relative frequencieswith which they appear in the naming (tactand listener) repertoires of 2- to 3-year-oldchildren (see Camoes-Costa et al., 2010). Thus,the touches to body parts that feature mostfrequently in childrens naming repertoireswere designated as baseline gestures; converse-ly, the touches to body parts that seldom evokecorrect tact and listener responses were em-ployed as target gestures.

Procedure

A multiple-baseline procedure was em-ployed; each child participated in all conditionsof the experiment. Training was presented tothe children in a staggered manner to demon-strate experimental control of any resultingchanges in target behavior. The flow of theprocedure is illustrated in Figure 2 (left panel).

Baseline Matching Training and Identification ofNovel Targets

Familiarization. The experimenter estab-lished a good rapport with the children duringunstructured daily play sessions in the nurseryplayroom before inviting each child to partic-ipate in one-to-one play with toys in the testroom. The child was asked to show Teddy what

Fig. 1. (adapted from Horne & Lowe, 1996, p. 201). Left Panel: A child, who has learned to name foot, sees her foot,and then says foot. Upon hearing this self-produced verbal stimulus /foot/, the child shows conventional listenerbehavior; she orients to the foot, moves it, touches it, and so on. Thus naming can be evoked by seeing a foot or hearing /foot/; it can be re-evoked by seeing a foot again or through self-echoing. This illustrates the bidirectional relationbetween the tact, listener behavior, and the object, or a class of objects or events. Right Panel: In an imitation task, seeingthe experimenter touch her foot may evoke the childs naming of this body-part. Thus a child may covertly or overtly sayfoot; hearing her own utterance /foot/ may then direct the childs attention to her own foot, leading her to touch thislocation on her own body (listener behavior). This illustrates how an apparently emergent matching response may comeabout as the result of naming.


the experimenter was doing by repeating theactions shown; the experimenter said, Canyou show Teddy how you do this? beforecommencing the first trial. The experimentermodeled eight different hand-to-body gesturestwice per session and instructed the child toDo this before she presented each gesture.The body parts touched by the experimenteron modeling trials were those that did notfeature frequently in the naming repertoires ofyoung children who attended the nursery (seeCamoes-Costa et al., 2010). At this early stageof the procedure, the experimenter determinedwhich of the corresponding hand-to-body touch-es were not part of the individual childrenstrained matching repertoires and could there-fore be employed as target gestures in theexperiment (see Horne & Erjavec, 2007). If thechild correctly matched a gesture more thanonce during the first two sessions of itspresentation, this indicated that the gesturealready featured in the childs trained matchingrepertoire, and consequently this target body-part touch was replaced with another; thereplacement gestures were likewise tested overtwo sessions. This continued until four noveltarget gestures were identified for each child. Noreinforcers were delivered following any of thechildrens responses (accurate matches or mis-matches), but the children were allowed to playwith Teddys toys at the end of each session. Inthis and subsequent training conditions, chil-dren were tested at least three times a week.Each session lasted approximately 1520 min.

Baseline matching training. This conditionestablished reliable, prompt, and correctmatching responding to the verbal request,

Do this, followed by modeling of a hand-to-body gesture on each trial. The details of thispart of the procedure were presented inprevious publications (see Erjavec et al., 2009;Horne & Erjavec, 2007) and only a summary isprovided here: In each training session therewere three modeling trials of each of the fourbaseline gestures (12 trials per session), with upto three presentations of the modeled gestureper trial (as necessary). The modeled gestureswere presented in a predetermined random-ized order, with the added constraint that nomore than two trials of the same gesture couldoccur in succession. Matching of the fourbaseline gestures was trained under continuousreinforcement, to a criterion of 5 out of 6correct responses per gesture, over two consec-utive sessions. When matching performancemet the 100% reinforcement criterion, rein-forcement rate was reduced to 50% on avariable-ratio 2 (VR 2) schedule. The intermit-tent reinforcement criterion was 11 out of 12correct responses across three trials per gesturewithin a single session.

Matching tests: Target (and baseline) gestures.Next, the children were presented with twomodeling trials of each of four untrainedtarget gestures and each of four trainedbaseline gestures; target and baseline trialswere delivered in a prerandomized sequence(total of 16 trials per session). At the start ofeach session the experimenter asked, Shallwe play the game? and before she modeledeach gesture she prompted the child, Canyou do this? Childrens correct matchingresponses to baseline models were intermit-tently reinforced on a VR 2 schedule, but their

Table 1

Childrens gender, target relations assigned to each participant, ages at start of namingintervention in Experiment 1, 2 or 3, total number of sessions administered in each experiment,and childrens general quotient scores on the Griffiths Mental Developmental Scales (GMDS).

Participant Gender Target relations

Age at startof training

(months/days)

Total numberof sessions

administered

Age at endof testing

(months/days) GQ scores

Experiment 1 Emma Female T2, T5, T6, T8 33/00 70 sessions 35/11 117Anna Female T1, T4, T7, T8 34/14 57 sessions 36/08 ____Mol Female T3, T5, T8, T9 35/29 113 sessions 38/16 120

Experiment 2 Jack Male T2, T4, T9, T14 31/17 79 sessions 35/16 132Gina Female T2, T7, T8, T10 31/24 85 sessions 36/01 120Mila Female T1, T4, T11, T14 33/24 80 sessions 37/24 120

Experiment 3 Fin Male T3, T7, T8, T9 28/03 44 sessions 30/10 110Carl Male T2, T3, T8, T9 32/04 51 sessions 35/14 116Fex Male T2, T8, T12, T13 34/07 83 sessions 38/22 113


responses to the target models were neverreinforced. In the first two sessions only, anytarget gestures to which more than one correctresponse was emitted were replaced, to ensurethat none of the targets featured in the childrens(extraexperimentally) trained matching reper-toires. The criterion for performance of thebaseline gestures was 13 out of 16 (81%) correctover two consecutive sessions. If the criterion wasnot met, the experimenter was required toretrain baseline responding before conductingthe next test sessionhowever, this was notnecessary in this or any subsequent conditions ofthe study.The number of test sessions was staggered

across the participants; they received eitherthree, or six, or nine sessions before moving

on to the next training condition. This allowedthe effects of the subsequent interventions tobe compared with those of repeated unrein-forced presentations of the target gestures,and to control for nonexperimental eventsthat may occur over time.

Tact Training: Target Locations on the Child

The aim of this training condition was todetermine whether teaching the children to tacteach of the four target locations on their own bodywould facilitate their matching of the modeledtouches to those same locations on the experiment-ers body on subsequent matching test trials. Thistested the hypothesis that naming of target bodylocations can be instrumental in childrensproduction of matching responses.

Table 2

Description of movement modeled and accepted responses variations for each baseline gesture(B1B4) and target gesture (T1T15) used during training and testing. The experimenter alwaysused her left hand for modeling hand-to-body touches but children could respond witheither hand.

Baseline/Target gesturesBehavior modeled bythe experimenter Accepted response variations

B1 Nose Tips of fingers touching nose Touching noseB2 Ear Tips of fingers touching right ear Touching either ipsilateral or contralateral earB3 Neck Tips of fingers touching neck

(Adams apple)Touching anywhere on the neck

B4 Lips Tips of fingers touching lips Touching any area of the lipsT1 Temple Tips of fingers touching

right templeTouching either ipsilateral or countralateral temple

(excluding head/hair, forehead and ear)T2 Bridge of foot Tips of fingers touching bridge

(arch) of right footTouching contralateral bridge of the foot (excluding

top of foot, sole, toes and heel)T3 Armpit Tips of fingers touching right

armpitTouching contralateral armpit (excluding upper arm)

T4 Thigh Tips of fingers touchingmiddle of left thigh

Touching ipsilateral thigh (excluding knee and hip)

T5 Crook of arm Tips of fingers touching crookof right arm

Touching crook of contralateral arm (excludinglower and upper arm)

T6 Crown Tips of fingers touching middleof crown

Touching the top of the head (excluding forehead,back of the head, or temple)

T7 Ankle Tips of fingers touching leftankle bone

Touching the area of either ankle (excluding shin,calf or any part of the foot)

T8 Wrist Tips of fingers touchingright wrist

Touching contralateral wrist (excluding lower armand back of the hand)

T9 Upper arm Tips of fingers touchingmiddle ofright upper arm

Touching contralateral upper arm (excluding crockof arm and shoulder)

T10 Lower arm Tips of fingers touching middleof right lower arm

Touching contralateral lower arm (excluding crockof arm, elbow, wrist and pulse)

T11 Shin Tips of fingers touching middleof left shin

Touching shin area of either leg (excluding ankle,top, side or back of the knee and calf)

T12 Calf Tips of fingers touching middleof right calf

Touching calf area of either leg (excluding ankle,shin, and top, side or back of the knee)

T13 Thumb Tip of index finger touchingmiddle of right thumb

Touching or grabbing contralateral thumb(excluding other fingers)

T14 Hip Tips of fingers touching middleof right hip bone

Touching hip area (excluding waist line or top of thigh)


Tact training was administered in two stagesfor each child. First, the child was trained toaccurately tact two of the four target body parts.The experimenter pointed to a target locationon the childs body (e.g., thumb) and said,Look! Tell Teddy what this is called? If thechild responded correctly, the experimenterexclaimed enthusiastically, Yes, it is the thumb!Clever girl! Occasionally (on 33% of trials) shealso presented a toy for the child to play with asadditional reinforcement. If the child producedan incorrect tact or no tact, the experimenterstated, This is the thumb, before asking thechild, Can you say thumb? If the child failedto respond, the experimenter repeated presen-tation of the latter training sequence up to twomore times on each trial beforemoving onto thenext. Tact trials for each target location werealternated in the training sessions in a preran-domized sequence. As tact responding becamemore proficient, the experimenters prompt wasabbreviated to, What is this? and the rein-forcement rate was reduced to 25%. Finally, thechilds tact responses were tested in extinction.

The criterion was seven out of eight correct tactresponses per target body location.Following this, three test sessions for match-

ing of target (and baseline) gestures wereconducted, as described earlier (see Matchingtests: Target (and baseline) gestures). In this andall subsequent test blocks, if the child matchedeach target gesture on at least four out of sixtrials over three test sessionsshowing consis-tent matching of all target gesturestheremaining training and test conditions wouldnot have been administered; instead, the childwould have progressed to the final tests(Listener Test and Follow Up; see below).The second stage of the tact training

condition was administered next; the childwas trained to tact the remaining two targetbody locations, exactly as described for thefirst two, followed by another three-sessionblock of matching tests.

Tact Training: Target Locations on the Experimenter

The aim of this condition was to determinewhether teaching the children to tact each of

Fig. 2. Flowcharts showing the stages of Experiment 1, Experiment 2, and Experiment 3.


the four target locations on the experimentersbody would facilitate their matching of themodeled touches to those same locations onsubsequent matching test trials.Procedurally, this tact training condition was

identical to the previous one, except that theexperimenter pointed to the target locationson her own body, rather than on the childsbody, while training the child to produce thetarget tacts. As in the previous trainingcondition, tact training was first administeredto criterion for two of the target bodylocations, followed by a three-session block ofmatching tests. This was followed by tacttraining of the remaining two target locations,and by three more sessions of matching tests.

Mixed Tact Training: All Target Gestures

The aim of this training condition was toensure that the tact responses trained previ-ously were maintained over time for all fourtarget body locations, regardless of whetherthese were on the experimenters or thechilds body.

Target locations on the child. This training wasalso administered in two stages. First, one tacttest trial was conducted for each of the fourtarget locations. If the child responded cor-rectly on all four trials, matching tests werereadministered. If any errors occurred on thetact test trials, mixed tact training was con-ducted at a progressively thinner reinforce-ment schedule until criterion was reached inextinctionthe child produced at least 11correct responses over 12 consecutive tacttrials, across the four targets, with 3 trials pertarget body location. The matching tests werethen administered, exactly as before, exceptthat from this point onwards in the procedurethe four tact responses to the childs bodyparts were probed once each before each testsession to establish that tact responding to alltarget locations was maintained. If tact perfor-mance fell below criterion, mixed tact trainingwas readministered and criterion performancereestablished before proceeding; if the childresponded correctly to the four tact probes,one matching test was conducted. Probing fortact responses and matching tests was repeateduntil three test sessions were completed.

Target locations on the child and the experimenter.One tact test trial was conducted for each targetlocation (four per child and corresponding fourper experimenter). If tact performance was

errorless a matching test was conducted. If anyerrors were recorded, mixed tact training wasadministered to criterion (as described above).Thus at this point, prior to each matching testsession, the child was required to show errorlesstacting of all target locations that the experi-menter pointed at on the childs body and onher own body. In order to evaluate the effectsof repeated matching tests over time on theaccuracy of childrens matching of target ges-tures, and to compare these with the effects ofthe next intervention, the number of matchingtest sessions was staggered across children; theycompleted three, or six, or nine sessions beforeproceeding to the next training condition.

Matching Tests with On-Task Tact Prompts forTarget Locations

The aim of this training condition was todetermine whether childrens untrained match-ing would become more accurate if, on eachmatching trial, the experimenter touched atarget location on her own body andmaintainedthis gesture while she asked the child to tact thattarget location then prompted the child totouch the same location on her own body.The children were once again invited to teach

Teddy about body parts. Tact performance wasreviewed and if necessary retrained (as in MixedTact Training: All Target Gestures). After the childdemonstrated errorless tact responding, amatching test session was conducted in the sameway as for previous matching tests except thatthe experimenter provided tact prompts duringmodeling of the corresponding target gestures.There were three trials per target gesture (12trials per session). On each trial, the experi-menter first said, Whats this? while pointingto a target location on herself. If the childdid not respond within 3s, the experimenterprompted the child to do so by saying Tellme! up to two more times. If the childresponded incorrectly, the experimenter provid-ed corrective feedback as in previous tact training.When the child produced a correct tact, themodeling instruction was presented, up to threetimes, as needed; if the child still did not respond,the next trial was presented. As in previousmatching tests, there were no scheduled conse-quences for any of the childs responses to themodeled target gestures. Each child completedthree sessions.Next, the children were presented with

matching tests, without tact probes, as in the


previous conditions; the number of sessions wasonce again staggered across the 3 children(three, six, or nine sessions were administered).

Direct Matching Training of All Target Gestures

The aim of this training condition was toestablish whether childrens responses to alltarget models in the matching tests could befurther improved by matching training.The matching training of target gestures was

conducted in much the same way as forbaseline gestures at the outset of the study(see Baseline Matching Training). Shaping andputting-through procedures were used whennecessary. As each childs performance becamemore accurate, the experimenter graduallyreduced the reinforcement rate. Matchingtraining was complete when a child produced11 out of 12 correct responses across the fourtarget gestures, tested in extinction. This wasfollowed by three matching test sessions.

Listener Behavior Test

A listener behavior test was conducted todetermine whether the childrens success onthe matching task (i.e., before direct matchingtraining) correlated with their listener re-sponses to the named target locations on thechilds body. One session was administeredwith 4 trials per target body location (16 trialsin all) presented in a predetermined random-ized order. On each trial, the experimenterasked the child to touch a named bodylocation (e.g., Can you touch your wrist?).There were up to three prompts per trial, ifnecessary, and the childrens listener respons-es were not reinforced or corrected.

Follow Up

One matching test session was administeredevery 2 to 3 weeks, until five sessions wereconducted. This completed the experimentalprocedure.

Coding of Childrens Responses

Each session was recorded on videotape andcoded using predetermined response criteriato identify, on each training and test trial,whether a child produced a correct response,an incorrect response, or failed to emit aresponse. The movement sequences consid-ered as correct responses to each modeledgesture and listener prompts are listed in

Table 2. For each target, the boundary withother targets was pre-set in a manner thatenabled the coders to determine whether achilds response on a matching trial approxi-mated the antecedent model as opposed toany other target or baseline model, or none ofthese. The response criteria excluded behav-iors that children naturally produce at this age,such as clapping, kicking, touching clothesor near objects, extending hands, mouthingfingers, rubbing eyes, and scratching any partof the body. If a child performed a correctresponse immediately after an incorrect re-sponse, it was coded as a self-correction andcounted as correct in the final analyses.Conversely, an incorrect response emittedimmediately after a correct response wascoded as a correct-to-incorrect and countedas incorrect (see Erjavec & Horne, 2008;Horne & Erjavec, 2007). Such multiple re-sponses were very infrequent. In matching testsessions, they occurred on 3% of baseline trials(range: 26%) and on 3% of target trials(range: 25%). Overall, self-corrections werescored eight times more frequently thancorrect-to-incorrect responses. Childrens tactresponses also needed to be entirely correct.Also recorded was whether a reinforcer was

delivered on a particular baseline gesturematching trial, the number of models (one,two, or three) per trial required to evoke aresponse, and the form of each incorrectresponse.

Interobserver Agreement

A second scorer, familiar with infant re-search but blind to the aims of the presentexperiment, independently coded 25% ofsessions selected on a random basis. Interob-server agreement was calculated for eachtraining and test phase by dividing the numberof agreements by the total number of codedresponses then multiplying the result by 100.An agreement was defined as two independentobservers assigning the same response code ona given trial. Agreement per phase rangedfrom 98% to 100%.

RESULTS AND DISCUSSIONChildrens Matching of Baseline Gestures

The four trained matching relations for allchildren are listed in Table 2 (top panel). Eachchild completed training in three sessions


(the minimum required by the mastery criteria,see Procedure). The children continued torespond correctly to baseline models through-out the procedure in all conditions (seeFigures 3, 4, and 5).

Childrens Performance in Tact Training Sessions

Across 3 children and 12 target behaviors,the children learned to tact the relevantlocations on their own bodies in an averageof 26 trials (range: 578 trials). Many fewertrials were needed to subsequently train tactresponses for the corresponding target loca-tions on the experimenters body (8 trials onaverage; range: 019 trials). This pattern ofresults was observed in each childs data (seeFigures 3, 4, and 5). Indeed, after tact trainingon the childs body, several of these tacts (3 forAnna and 3 for Emma) were in place at theoutset of tact training on the experimentersbody; for these tacts, Anna and Emma did notrequire mixed tact training. Mols tact rela-tions were more fragile and all requiredextensive mixed training (see Figure 4).Overall, the results show that tact training

often generalized from target locations on thechilds body to the corresponding locations onthe experimenters body; this effect rangedfrom moderate savings in the number oftraining trials required to establish each tactto full emergence of tacts without training.

Childrens Matching of Target Gestures

The sets of four gestures that were identifiedas targets for each child are listed in Table 1 anddescribed in Table 2 (bottom panel). Childrensmatching of each target gesture in the test trialswas classified as either (i) consistent, if correct onat least two-thirds (66100%) of the trials; (ii)intermittent (3366% correct trials); (iii) infre-quent (132% correct trials); and (iv) unmatched(no matches). Note that for each child, thenumber of matching tests following each train-ing condition may differ between targets, inaccordance with the staggered and pair-wiseintroduction of the independent variables (e.g.,tact training on the childs body) throughoutthe procedure. Therefore, in order to calculatepercent matching per phase for each target,matching responses were summed over all thematching tests that were conducted immediatelyafter one training phase and before the begin-ning of the next.

In the matching tests that followed baselinematching training and prior to the first round oftact training (i.e., across Test 1 sessions for thefirst pair of targets, and Test 1 and Test 2 forthe second pair of targets), out of a total of 12untrained target gestures across all 3 children,6 were infrequently matched, and the remain-ing 6 were not matched at all. Our data showno evidence that repeated modeling of targetgestures resulted in increased matching, whichis consistent with our previous findings (Erja-vec & Horne, 2008; Erjavec et al., 2009; Horne& Erjavec, 2007).After staggered target tact training on the childs

body, the childrens matching performancesimproved for some of the gestures, but no targetwas matched consistently. Looking at Test 2 andTest 3 scores for the pairs of target gesturestrained first, and at Test 3 and Test 4 scores forthe pairs of gestures that were trained second foreach child, 5 out of 12 targets were nowmatchedintermittently, 6 infrequently, and 1 target wasnot matched (see Figures 3, 4, and 5). In thetests administered after staggered target tacttraining on the experimenters body, but before thenext round of training (i.e., across Test 4 andTest 5 for the first pair of targets, and in Test 5for the second pair), the children matched 4 oftheir targets consistently, 1 intermittently, 5infrequently, whereas 2 remaining targets werenot matched. Next, following mixed tact trainingof all target gestures on the child (Test 6), thechildren matched 6 of the targets consistently, 2intermittently, and 4 infrequently. Finally, fol-lowing mixed tact training of all target gestures on theexperimenter (Test 7), the children matched 5 ofthe targets consistently, 2 intermittently, and 5infrequently. Clearly, tact training administeredin the absence of modeling was effective inestablishing and/or increasing matching ofuntrained target gestures in subsequent tests.However, after 20 (Emma), 25 (Anna), and 35(Mia) tact training sessions, consistent match-ingdefined a priori as correct responding onat least two thirds of trials (see Procedure)didnot develop across all targets for any of theparticipants. The data show no evidence thattraining the children to tact the target locationson the experimenters body resulted in largerimprovements in their matching of the targetgestures than when they were trained to tactthese same locations on their own bodies.The children were next taught to tact the

location touched by the modeler before they


responded to the Do this matching prompton test trials. This intervention aimed toincrease the incidence of (covert) naming ofthe target body locations during subsequent(unprompted) matching tests. In matchingtests following this intervention (Test 8), thechildren matched 6 out of 12 target gesturesconsistently; 2 intermittently, 3 infrequently,and 1 target was not matched. However, thistact-prompt intervention was sufficient toestablish consistent matching for 1 child,Anna, across all of her target gestures (seeFigure 3). The remaining 2 participants re-quired direct matching training, administerednext. As expected, direct matching trainingresulted in consistent matching of all targetbehaviors for these 2 children (see Test 9 forMol, Figure 4, and Emma, Figure 5); thisoutcome replicated our previous findings(Erjavec et al., 2009; Horne & Erjavec, 2007).Two children participated in the listener-

behavior test. On 16 trials per child, respond-ing was 100% for Mol and 94% for Emma,

showing that listener responses, which werenever directly trained, emerged as a result oftact training. This is consistent with theexisting literature (e.g., Horne et al., 2007;Lowe et al., 2002, 2005) and predictions of thenaming account of Horne and Lowe (1996).Due to her leaving the nursery, Anna did nottake part in either Listener or Follow Up tests.In the Follow Up, Emma and Mol were

presented with one matching test session every2 to 3 weeks until five sessions had beenconducted. On average they matched two oftheir eight targets at 100%, 3 consistently, andthe remaining 3 intermittently. This shows thatthe majority of the target matches, whendirectly trained, were well maintained overthe following 3 months in the absence ofreinforcement or corrections. Childrens in-correct responses on all matching test trialswere noted and examined. These responsesare reported in Appendix A. The childrenoften touched, in response to the targetmodel, the body parts adjacent to the target

Fig. 3. Correct matching responses emitted by Anna in the eight baseline trials (plotted together at the top of eachfigure and represented by open circles) and the eight target trials (plotted separately for the four target behaviors, withtwo trials per gesture, and represented by filled circles) presented during each session of the multiple baseline procedurein Experiment 1. Training and testing was administered in a staggered manner (see text) over five experimental trainingconditions. Each training phase is labeled and denoted by a shaded area; the bold lines indicate where training wasapplied in a staggered manner to pairs of target actions. Underneath, the number of training trials that needed to beadministered before criterion performance was attained (T) and the proportion of trials on which a child respondedcorrectly in probing sessions (T) are presented for each target gesture. Bold figures indicate the relations that were inplace at the outset (i.e., only 05 trials were required to reach criterion performance).


Fig.4.

DataforMol;otherwiseas

forFigure

3,exceptthattheshaded

triangles

representgesturesthatalreadyfeaturedin

theMolstrained

matchingrepertoirewhich

werereplacedbynovelgestures.

Fig.5.

DataforEmma;otherwiseas

forFigure

4.


location, showing lack of discrimination be-tween specific body locations. For example, inresponse to modeled touches to the armpit,Mol touched her tummy on 32%, her chest on23% and her upper arm on 26% of trials.Theoretically at least, listener behavior train-ing, which includes manual correction follow-ing errors, might be expected to improvediscrimination between body parts. This wasinvestigated in the next two experiments.

EXPERIMENT 2METHOD

Participants

Two typically developing girls, Gina (31months) and Mila (33 months) and one boy,Jack (31 months) took part.

Settings, Apparatus, and Procedure

The setting and apparatus were as describedin Experiment 1. The baseline and targetbehaviors allocated to each child are shown inTable 1. The flow of the experimental proce-dure is illustrated in Figure 2 (middle panel).The sequence of interventions was: (i) baselinematching training; (ii) tact training (locationson the child and experimenter); (iii) tactprompts during matching tests; (iv) listenertraining; and (v) tact prompts with namingfeedback. As an addition to the proceduredescribed for Experiment 1, it was planned toreview previously trained tact and listenerbehavior responses at the start of eachmatching test, and if necessary retrain them.However, this prolonged the procedure, re-ducing the childrens willingness to partici-pate; tact/listener reviews were thereforeadministered only prior to the first, fourth,and seventh matching test sessions presentedafter each intervention. These reviews werealso administered prior to any new training/intervention, which made the mixed tacttraining sessions administered in Experiment1 unnecessary.

Listener Behavior Training

Listener behavior training was administeredin two stages for each child. First, the child wastrained to respond accurately to the names fortwo of the four target body locations. Theexperimenter asked Can you show Teddywhere is your (target body-part name)? then

waited for the child to respond. If the childtouched the correct location, the experiment-er exclaimed enthusiastically, Yes, thatsright. Clever girl (or boy)! and occasionallyalso gave the child a toy to play with.Otherwise, the experimenter said, Here itis! and moved the childs hand to touch theappropriate location, before asking the childagain, So, can you now touch your (repeattarget body-part name)? A correct touch wasreinforced; if the child failed to respond, thelatter training sequence was repeated up totwo more times on each trial. Listener trials foreach target location were alternated in eachtraining session in a prerandomized sequence.As the child became more proficient atcorrectly locating the named body parts, theexperimenter gradually faded the prompts,and reduced the reinforcement rate to 25%.Finally, the childs listener behavior was testedin extinction. The criterion was seven out ofeight correct trials per body part.After listener training for the first two target

locations was completed, three matching testswere conducted as described in Matching tests:Target (and baseline) gestures. Next, listenertraining was conducted for the remainingtwo target locations, followed by matchingtests, staggered across children.

Matching Tests with On-Task Tact Prompts forTarget Locations and Naming Feedback

The aim of this training condition was todetermine whether the childrens untrainedmatching would become more accurate if oneach matching trial (i) the child first tactedaccurately the target location, and (ii) followinga correct matching response the experimenterprovided feedback by saying, Well done, youretouching your (target body-part). Nine suchtrials were conducted for each target location,followed by another block of matching tests.Finally, direct matching training and follow

up tests were implemented as in Experiment 1.

Coding and Interobserver Agreement

Childrens responses were coded according tothe criteria described for Experiment 1. In thematching test sessions, multiple responses oc-curred on 4% of baseline trials (range: 35%)and on 4% of target trials (range: 35%). Overall,self-corrections were scored four times morefrequently than correct-to-incorrect responses.


A second scorer independently coded 32%of sessions selected on a random basis.Interobserver agreement in each phase rangedfrom 95% to 100%.


The four gestures that were trained asbaseline matching relations to all childrenwere the same as in Experiment 1 (seeTable 2, top panel). Each child completedmatching training in the minimum threesessions. The children continued to respondcorrectly to baseline models in all tests (seeFigures 6, 7, and 8).

Childrens Performance in Tact and ListenerBehavior Training Sessions

Across 3 children and 12 target responses,the children learned to tact the target loca-tions on their own bodies in an average of 27trials (range: 661 trials). Replicating theresults of Experiment 1, tacts trained on thechilds body generalized to target locations onthe experimenters body. For Mila and Jack, alltacts were at criterion without training (seeFigure 6 and 7, respectively), and the remain-ing child (Gina) required 519 trials to meetthe criterion, considerably fewer than in thepreceding target tact training on her own body(see Figure 8).It is well documented in the literature that

tact training, administered to children of thisage, establishes the whole naming relation;this is evident in the emergence of listenerbehavior in tests that use well established,simple responses such as pointing to a wholeobject (e.g., Lowe et al., 2002; 2005). Bycontrast, correct listener responses in thepresent tests required children to makeaccurate discriminations between the adjacentbody parts. Nevertheless, the data show thatMila and Jack emitted correct listener respons-es to half of their target locations on the firsttrial; the remaining listener responses re-quired only 13 trials, on average, to reachcriterion performance (range: 322 trials; seeFigures 6, 7, and 8).


In the tests following baseline matchingtraining, out of a total of 12 target gestures(across all children), 1 target was matched

Fig.6.

DataforMila.Otherwiseas

forpreviousfigu

res.


Fig.7.

DataforJack;otherwiseas

forpreviousfigu

res.

Fig.8.

DataforGina;otherwiseas

forpreviousfigu

res.


intermittently, 7 infrequently, and 4 remainingtargets were not matched at all. As inExperiment 1, there was no evidence thatrepeated presentation of modeling and re-sponse opportunities led to improvements inchildrens matching of target gestures.Following tact training of target locations on

the childs body, one target gesture wasmatched consistently, three intermittently, sixinfrequently, and two remaining targets werenot matched. After tact training of targetlocations on the experimenters body, thechildren matched one target gesture consis-tently, six intermittently, four infrequently,and the remaining gesture was not matched.These data replicate the results of Experiment1 and show that training the children to tacttarget locations resulted in some correctmatching of target gestures in subsequenttests. However, as in Experiment 1, none ofthe children showed consistent matchingacross all of their target body parts.In the tests administered after the on-

task tact prompts intervention, the childrenmatched five targets consistently, three inter-mittently, one infrequently, and three werenot matched. Although this intervention in-creased childrens correct matching of sometargets, it was not sufficient to establishconsistent matching of all gestures for any ofthe children (see Figures 6, 7, and 8). Theseresults were similar to Experiment 1, in whichonly 1 participant consistently matched alltarget gestures at this point in the procedure.

Listener Behavior Training was administerednext. In the matching tests that followed, 6 outof 12 targets were matched consistently, 4intermittently, and 2 infrequently. Althoughchildrens matching continued to improve, noneof the children consistently matched all of theirtarget gestures. They proceeded to the next stageof training where on each trial the experimenterfirst evoked the childs tacting of target bodylocation, then presented the matching promptDo this, and finally provided naming feedbackfor any correct matching responses. In thesubsequent matching tests, the children consis-tently matched nine of their target gestures, butthe remaining three were matched only infre-quently. This intervention improved the chil-drens performances considerably, but for eachchild one target gesture remained poorlymatched (see Figures 6, 7, and 8). Therefore,direct matching training was conducted next.

Direct matching training resulted in consis-tent matching of all target behaviors for allchildren (12 targets in total) in repeated testsadministered in extinction, as in Experiment 1.In Follow Up, all children were given onematching test session every 2 to 3 weeks. Acrossall children, 11 target gestures continued to bematched consistently in the first three followups, but by the fourth follow up there wasevidence of deterioration in matching of two ofthe eight target gestures tested. At the fifthfollow up, only two out of four targets testedwere matched. This pattern of maintenance inextinction for some target matches, but declinein others, is similar to that in Experiment 1.Childrens incorrect responses on all match-

ing test trials are reported in Appendix A. Thesedata show that, as in Experiment 1, childrenserrors mostly consisted of touches to body partsadjacent to the target locations. Listener behav-ior training, which was intended to improvechildrens discriminations between adjacenttarget body parts, was administered late in theprocedure. By the end of listener training Jackmatched three of his four targets at close to100%, but showed no improvement for hisremaining target. Gina showed transient im-provement of matching for one target, but adecline in another; and there was no discernibleeffect of listener training on Minas targetmatching. The next experiment was designedto investigate whether listener behavior trainingmay be more effective in improving childrensmatching of target gestures if administeredearlier in the procedure. We also wanted toestablish whether the effect of tact training,which was demonstrated in Experiment 1 andExperiment 2, would be greater if the childrenunderwent tact training after they had learnedlistener responses for each target body part.

EXPERIMENT 3METHOD

Participants

Three typically developing boys aged 28months (Fin), 32 months (Carl), and 34 months(Fex) at the start of listener behavior trainingparticipated in this experiment (see Table 1).

Settings, Apparatus, and Procedure

The setting and apparatus was as described inExperiment 1. The baseline and target gesturesallocated to each child are shown in Table 1.


The experiment consisted of the same trainingconditions described in Experiment 2, but inthe following order: (i) listener behaviortraining; (ii) target tact training of the childsbody and then on the experimenters body; (iii)matching tests with on-task tact prompts fortarget locations; (iv) matching tests with on-tasktact prompts for target locations and namingfeedback; (v) direct matching training of alltarget gestures; and (vi) follow up sessions. Theflow of the procedure is illustrated in Figure 2(right panel). Testing and retraining (asneeded) of tact and listener responses wasconducted at the start of the matching testsessions as in Experiment 2.

Coding and Interobserver Agreement

Childrens responses were coded as in Experi-ment 1. In the matching test sessions, multipleresponses occurred on 6% of baseline trials(range: 38%) and on 4% of target trials (range:36%). Overall, self-corrections were scored twotimes more frequently than correct-to-incorrectresponses.A second scorer independently coded 33%

of sessions selected on a random basis.Interobserver agreement in each phase rangedfrom 99% to 100%.


Fex and Fin completed baseline matchingtraining in three sessions (the minimumrequired to meet the criterion) and Carlneeded four sessions to do so. Figures 9 and10 show that the children continued torespond correctly to the baseline models inall tests.

Childrens Performance in Listener Behavior andTact Training Sessions

In this experiment, listener behavior train-ing was administered first. Across 3 childrenand 12 target responses, training children totouch the correct body locations after hearingthe experimenters naming of these locationstook 44 trials per target behavior, on average(range: 1186 trials). Compared to tact re-sponses trained first in Experiments 1 and 2(accomplished in M 5 26/27 trials), listenerbehavior was clearly more difficult to establish.A comparison between the present results andthose of Experiment 2, where listener behavior

training was administered late in the proce-dure, confirms that tact training had beeneffective in establishing and/or improvingchildrens corresponding listener responses4 out of 12 were in place at the outset ofExperiment 2 listener training, 2 required onlythree or four training trials, and the remaindertook many fewer trials to train (M 5 13) thanwas the case in the present experiment (M 544).Likewise, the present data show that, once

the childrens listener responses were at crite-rion, in most cases the corresponding tactresponses were either already in place orrequired very few trials to meet the set criterion(see Figures 9 and 10). At the start of tacttraining for target locations on the childs body,all four tacts were in place for Fex, three for Fin,and two for Carl; the remaining three tacts tookbetween 8 and 10 trials to reach criterion. Forthe 2 children who underwent tact training fortarget locations on the experimenters body, alleight tacts were in place at the outset. Overall,these data show that listener behavior trainingsufficed to establish naming of most target bodylocations, and that the resulting name relationsincluded corresponding locations on thechilds and the experimenters bodies. Thisoutcome differs from the findings of previousnaming studies (Horne et al., 2004; 2006) inwhich listener training did not invariably resultin the emergence of the corresponding tactrelations.


The sets of four gestures that were identifiedas targets for each child are listed in Table 1and described in Table 2 (bottom panel). Inthe generalized imitation tests conducted afterbaseline matching training, out of a total of 12target gestures (across all children), 1 targetwas matched intermittently, 6 infrequently,and 5 were not matched at all. In the testsessions that followed listener behavior train-ing, the children matched 4 of their targetsconsistently, 3 intermittently, and 4 infre-quently; the remaining target was notmatched. These data show that listener train-ing, conducted in the absence of modeling,was effective in establishing or increasingmatching of target gestures in the subsequentmatching tests.In the matching tests administered after tact

training for target locations on the childs


Fig.9.

DataforFex;otherwiseas

forpreviousfigu

res.

Fig.10.

DataforFin,an

dCarl;otherwiseas

forpreviousfigu

res.


body, 5 out of 12 targets were matchedconsistently, 3 intermittently, 3 infrequently,and 1 target was not matched. Fin participatedin only one test session after tact trainingbefore dropping out of the study. Carldropped out of the procedure after tacttraining of target locations on the experiment-ers body was completed but before thesubsequent matching test could be adminis-tered (see Figure 10). After tact training fortarget locations on the experimenters bodywas completed, out of a total of six targetgestures, three were matched consistently, oneintermittently, and two infrequently. Thispattern is similar to the results of the previousexperimentsat the end of the tact trainingintervention, all children matched some oftheir target gestures consistently, but no childdid so across all their targets.Only one child, Fex, participated in the

matching tests with on-task tact prompts fortarget locations intervention. In the generalizedmatching tests that followed, he matched threeof his four target gestures consistently, but theremaining target infrequently. His perfor-mance was similar to those of 5 out of 6participants in Experiment 1 and Experiment2, who by the end of this intervention consis-tently matched mostbut not alltarget ges-tures. After the next interventionmatchingtests with on-task tact prompts for targetlocations with naming feedbackwas adminis-tered, Fexs matching was errorless for three ofhis targets, but the remaining gesture was stillmatched only infrequently. As in our previousstudies, the direct matching training thatfollowed resulted in consistent matching of alltargets for this child. Over his four Follow-Upsessions, Fex matched three target gesturesconsistently, but the remaining gesture wasonce again matched only infrequently.Childrens incorrect responses on all match-

ing test trials are reported in Appendix A; theirpattern of errors was similar to those found inthe previous two experiments.

Analysis of Childrens Matching Responses AcrossAll Experiments

Considering the results of Experiment 1,Experiment 2 and Experiment 3 together,there were 9 participants and 36 target behav-iors. The percentages of childrens correctmatching responses to the experimentersmodeling of target gestures in repeated match-

ing tests were analyzed statistically to explorethe effects of various types of training admin-istered in these experiments. In all cases wheredata were available for at least 3 participants, wecalculated the effect sizes as indices of changeacross conditions; wherever the sample size waslarge enough (4 or more participants) we alsoemployed repeated-measures t tests.

Effects of the Tact Training Interventions

Tact training administered before listener-behavior training. In Experiment 1 and Exper-iment 2, target tact training was administeredon the childs body first, and then for the samelocations on the experimenters body. InExperiment 1, this was followed by a mixedtact-training intervention.Across the 6 children and 24 target gestures

in Experiment 1 and Experiment 2, meantarget matching in all generalized imitationtest sessions administered before target tact-training on the childs body was 8% (range: 613%). After this training, but before the nexttraining commenced, matching was recordedon a mean of 27% of trials (range: 1543%).Statistically, this difference was significant, t(5)5 24.98, p 5 .004, with a very large effect size(Cohens d 5 2.94; see Brace, Kemp, &Snelgar, 2006). Indeed, all children were morelikely to match their target responses after thisintervention. After target tact training on theexperimenters body, but before the nexttraining commenced, target matching wasrecorded on a mean of 38% trials (range:1758%). The difference between the chil-drens performances on the tests administeredbefore and after this intervention was notstatistically significant, t(5) 5 21.68, p 5 .152,but the effect size was large (Cohens d 5 .90).All children were more likely to match theirtarget responses after this intervention.Across 3 children and 12 target gestures in

Experiment 1, in the test sessions prior tomixed tact training, mean target matching was44% (range: 3658%). After mixed tact train-ing, but before the next training commenced,correct target matches were recorded on amean of 51% trials (range: 3858%). Theeffect size was medium (d 5 .55); only 1 childwas more likely to emit correct target respons-es after this intervention.Overall, these tests confirm that the tact

training interventions significantly increasedchildrens matching of target gestures; tact


training of target locations on the childs body,administered first, resulted in the biggestincreases in target matching; the subsequenttarget tact training for the same locations onthe experimenters body, and mixed training,had smaller effects.

Tact training administered after listenerbehavior training. In Experiment 3, the childrenwere given tact training later in the procedure,after listener behavior training. In their case, andacross 12 target gestures, before target tacttraining on the childs body, mean targetmatching was 48% (range: 2265%). After thistact training, correct target matches were record-ed on amean of 48% trials (range: 2765%). Theeffect size was small (d 5 0.01); only 1 child wasmore likely to match the target responses afterthis intervention, which clearly did not add to theeffects of the earlier listener-behavior interven-tion (see below). Only 1 participant, Fex, alsounderwent target tact training on the experi-menter; in his case, this intervention improvedthe percentage of correct matching across thefour target gestures from 27% to 67%.

Effects of the Matching Tests with On-Task TactPrompts for Target Locations

After tact training in all experiments, thechildren were presented with three sessions ofmatching tests with on-task tact prompts fortarget locations. All participants in Experiment1 and Experiment 2, and 1 child in Experi-ment 3, took part. Across these 7 children and28 target gestures, a mean of 45% (range: 1767%) of target trials were matched in the testsessions conducted prior to the on-task nam-ing prompts for matching intervention. Afterthis intervention, but before the next, targetmatches were recorded on a mean of 55%trials (range: 3188%). Statistically, this differ-ence was not significant, t(6) 5 21.59, p 5.164. Only a medium effect size was obtained(d 5 .57) because only 4 children were morelikely to emit correct target responses after thisintervention.

Effects of the Listener BehaviorTraining Interventions

Listener-behavior training administered beforetact training. In Experiment 3, listener behaviortraining was administered to 3 children, across12 target behaviors. Prior to this intervention,the children responded correctly on 9% of the

generalized imitation test trials (range: 317%). Following the intervention, but beforethe next training commenced, the childrenemitted correct matching responses on 48% ofthe trials (range: 2265%). The effect size wasvery large (d5 2.60). Indeed, all children weremore likely to match targets after the listenerintervention.

Listener behavior training administered aftertact training. In Experiment 2, across 3 childrenand 12 target gestures, mean target matchingresponses in the test sessions conducted priorto listener behavior training was 52% (range:3164%). Following listener training, but be-fore the next training commenced, mean targetmatching increased to 59% (range: 4771%).The effect size was small (d 5 .47) and only 1child was more likely to emit correct targetresponses after the listener intervention.Overall, listener behavior training, when

administered first (Experiment 3), was veryeffective in increasing childrens matching oftarget gestures in the subsequent tests; howev-er, this training had no effect when adminis-tered later in the procedure (Experiment 2).

Effects of the On-Task Tact Prompts for TargetLocations and Naming Feedback Intervention

In Experiment 2 and Experiment 3, afterboth listener behavior training and tact trainingwere completed, the children were presentedwith three sessions of matching tests with on-task tact prompts for target locations followedby naming feedback. All participants in Exper-iment 2 and 1 child in Experiment 3 took part.Across these 4 children and 16 target gestures,mean target matching in the test sessionsconducted prior to this intervention was 59%(range: 4771%). After this intervention wasadministered, but before the next trainingcommenced, target matches were recorded ona mean of 72% trials (range: 6481%). Statis-tically, this difference was significant, t(3) 523.82, p 5 .032, with a large effect size (d 51.48). Indeed, all children were more likely toemit target responses after this intervention.

Joint Effects of All Naming Interventions

Considering all naming training that wasadministered to the participants in the threeexperimentstact training and prompts inExperiment 1 and tact and listener trainingand prompts in Experiment 2 and Experiment


3there were 7 participants who completedthe experimental procedures. The 2 childrenwho left the nursery before all scheduledinterventions could be administered alsocompleted a part of this training. Therefore,it was possible to compare the accuracy of 9childrens performances, across 36 targetgestures, to estimate the overall effectivenessof name training.Mean target matching in all generalized

imitation test sessions prior to any interventionwas 8% (range: 317%). After all naminginterventions, but prior to the start of directmatching training, target matches were re-corded on a mean of 65% test trials (range:3388%). Statistically, this difference wassignificant, t(8) 5 29.18, p , .001, and theeffect size was extremely large (d 5 5.40); theeffect was present for each child and acrossmost of the target gestures.

The Effects of Direct Matching Training

Although all childrens matching of targetgestures showed large and significant increasesfollowing the naming interventions, only 1child in Experiment 1 (Anna) developedconsistent matching of all her target gesturesas the result of tact training. To achieve thiscriterion, the remaining children needed toundergo direct matching training. Two chil-dren in Experiment 1, 3 children in Experi-ment 2, and 1 child in Experiment 3 partici-pated in this intervention. Across these 6children and 24 target gestures, mean targetmatching in all matching test sessions admin-istered prior to direct matching training was62% (range: 3381%). After matching train-ing, matching occurred on a mean of 93% testtrials (range: 8896%). Statistically, this differ-ence was significant, t(5) 5 23.86, p 5 .012,and the effect size was very large (d 5 2.96);the effect was present for each child.

GENERAL DISCUSSION

In the present study we aimed to investigatewhether training young children to name fourtarget body parts would be sufficient for thechildren to match, for the first time, an adultstouches to those locations during a general-ized imitation test. The series of three exper-iments we report were designed to test thisnaming hypothesis by comparing imitation ofnovel target gestures by children who either (i)

had learned to name them or (ii) onlyresponded as listeners to those names. Ifuntrained imitation emerged only in childrenwho had learned to name the target stimulithis would provide evidence that target nam-ing is sufficient whereas the correspondinglistener behavior is not. However, our datashow that all 9 children acquired both the tactand listener components regardless of whetherthey were ostensibly trained the tacts (Exper-iment 1 and Experiment 2) or correspondinglistener relations (Experiment 3) at the outset.Consequently, we assume that both the tactand listener training established naming andour analysis of whether naming can in turnestablish matching of novel targets when theseare presented in the context of a generalizedimitation task must rest on whether there was asignificant change in the level of targetmatching after target naming was establishedduring the experiments.

The Effects of All Naming Interventions on NovelTarget Matching

The effects of the naming interventions sodefined were analyzed statistically for thecombined data from all 9 children whoparticipated in the three experiments, therebyencompassing potential untrained matchingof 36 novel target behaviors. The analysesshowed that before the naming interventionswere introduced, mean target behavior match-ing was only 8% in the generalized imitationtests. This initial low level of target matching,here over as many as 12 successive generalizedimitation tests, replicates the findings reportedin our earlier studies on generalized imitation(Erjavec et al., 2009; Horne & Erjavec, 2007).However, in the generalized imitation testsconducted after all the naming interventionshad been implemented, target-matching re-sponses had increased to a mean of 65%. Thiswas a significant change with an extremely largeeffect sizeand the increase in target matchingoccurred in all children. Behavior change ofthis order is impressive given that targetmatching was always tested under extinction.

Limitations on the Effects of Naming Interventions

Although the above data suggest that targetnaming had a large and significant effect onthe development of childrens untrainedmatching of the novel target behaviors, with


1 child, Anna (Experiment 1) showing nearperfect matching following the naming inter-ventions, the untrained matching performanc-es of the remaining children were morevariable. This variability, however, is to beexpected. The naming account predicts thatself-instructional effects will occur if the childproduces the relevant name on-task. ThisSkinnerian approach considers outcomes inprobabilistic rather than mechanistic termsand recognizes that task performance is alwaysa function of multiple sources of control;particularly in young children, other influenc-es on task performance are likely if theprobability of the main independent variableunder study, here on-task naming, is not fullycontrolled. For example, the studies by Horneet al. (2004, 2006, 2007) and Lowe et al. (2002,2005) show that emergent name-based catego-ry sorting of arbitrary stimuli occurred reliablyonly when the experimenter evoked thechilds on-task naming by first pointing tothe sample stimulus, and asking Whatsthis? before indicating the mixed array ofstimuli and saying to the child Can you giveme the others? When the children were onlyasked to look at the sample before sorting thestimuli, many failed to sort the arbitrarycomparisons along common name lines, eventhough they had all previously learned toname the stimuli appropriately, and hadpassed tact review trials immediately beforethe sorting test was conducted. In the presentstudy, although the children took part in anintervention that comprised a tact-promptversion of the generalized matching task (withno consequences for correct target matches)they were never prompted to tact the targetstimulus in any of the generalized imitationtests employed throughout each experiment:The main independent variable, target nam-ing, was therefore left uncontrolled in therepeated tests that measured emergence ofuntrained target matching. Although a strongtest of the self-instructional effects of namingwas not performed in the present experiments,future studies can address this shortcoming.

The Effects of Instructional Context

Another explanation of performance vari-ability is that the self-instructional effects ofnaming pertain to looking rather than touch-ing behaviors. Naming may therefore generatelooking at a target location on the childs and

modelers body, but additional on-task instruc-tions may be required to determine whetherthe child also touches the body part that he orshe names. The instruction Do this given oneach generalized imitation test trial may nothave been sufficient in this regard. Thisinstruction is essentially ambiguous: Shouldthe child touch the named body part on theexperimenters body or on her own body? Incontrast, the sorting instruction employed inthe naming and categorization studies has nosuch ambiguity: Zag (sample name pro-duced by child) plus Where are the others?(experimenter instruction) directs the childsbehavior quite clearly from the sample to thecomparisons. The possibility that on-task in-structions might facilitate the effects of targetnaming in the imitation test context should beinvestigated further.

Discrimination Between Target Body Locations

Although the naming interventions resultedin untrained matching of target behaviors, thechildren did not learn to match all their targetsreliably until they took part in direct matchingtraining. Prior to this, their most frequent errorsconsisted of touching body parts adjacent to thetargets. This is surprising given that the noveltarget responses all terminated on body partsthat are visible (see Table 2) and, throughouttact training, the experimenter touched eachtarget body location while she asked the childWhats this? Likewise, during listener train-ing, the experimenter corrected errors bymoving the childs arm and hand to enablehim or her to touch the location specified by agiven listener stimulus, gradually fading thisputting through procedure as the childlearned to respond correctly. In both tact andlistener training the child was therefore provid-ed tactual and visual stimulation to help him orher discriminate one target location fromanother. It appears that this bimodal stimulationper se was not sufficient to constrain variability inthe childrens otherwise untrained matching ofthe target behaviors. In contrast, the four trainedbaseline matches required the children todifferentiate between their nose, ear, lips orneckvery fine discriminations indeed, andones that rely exclusively on touch (see Table 2).Nevertheless, these matches were reliable andveridical under intermittent reinforcementthroughout the procedures employed in thethree experiments and, once established at the


start, never required correction. Children of thisage as well as infants are easily able to learn finediscrimination of body part locations, andmatchthem, when the relevant matching relations areestablished under contingent reinforcement.

Parity and Cross-Body Mapping

One hypothetical constraint on childrensmatching of target locations modeled on anexperimenters body is that body structure andsize is very different when comparing a 3-year-oldchild with an adult. This might be expected tolimit childrens discrimination of varying de-grees of parity between the modeled behaviorand their own responses. However, this studyprovides good evidence that childrens tacts fortarget locations on their own bodies oftengeneralized without training to the correspond-ing locations of the adults body. This outcomesuggests that cross-body mapping may not be aserious limitation on young childrens targetmatching (see Baer & Deguchi, 1985).

Direct Matching Training and Maintenance

The impact of direct matching training forall matching relations is demonstrated clearlyby the large and significant increases to amean of 93% in target matching whenputting through and contingent reinforce-ment were finally employed. This second largechange in level clearly supports the findings ofour earlier studies on the determinants ofmatching in infants and young children andreaffirms that repeated presentation of genu-inely novel target behaviors in the context oftrained and intermittently reinforced match-ing relations does not of itself result ingeneralized imitation in infants and youngchildren. That said, direct matching trainingdid not guarantee reliable matching of all thenovel targets in the long-term follow-up tests.Target matching averaged 80% over thefollow-up data in all three experiments, butfor some targets fell to baseline levels underthe extinction conditions of these long-termtests. This suggests that intermittent reinforce-ment, as was provided for the baselinematching relations throughout the proceduresin the present and previous studies, may benecessary to maintain even directly trainedmatching relations. This accords fully with aSkinnerian account of how matching is estab-lished and maintained.

Conclusion

Our failure to identify children who learnedonly listener behavior in the course of thelistener training conducted at the outset inExperiment 3 limits our consideration of thefindings to the naming account, which is themain theoretical driver of the present study.However, replication of Experiment 3 with alarger number of participants should increaseour chances of identifying children who onlylearn listener behavior (see Horne et al., 2004;2006) and would enable us to investigatewhether this relation on its own results in anincrease in novel target matching in youngchildrens performances on generalized imita-tion tasks, in which case we could reinterpretour findings in terms of simpler behavioralrelations. Nevertheless, this first study toinvestigate the relationship between namingand imitation in very young children foundlarge and significant increases in matchingresponses following the introduction of thenaming interventions. It appears that targetnaming raises the probability of target match-ing in most cases to levels that would enablecaregivers to fine-tune the relevant matchingrelations in the day-to-day social environmentby providing occasional social reinforcementcontingent on a good match to target. Thisprovides support for our hypothesis thatnaming, and indeed other kinds of verbalself-instruction, may be an important means ofaccelerating the development of imitation, arepertoire that plays an important role inhuman learning throughout the lifespan.

REFERENCES

Baer, D. M., & Deguchi, H. (1985). Generalized imitationfrom a radical-behavioral view-point. In S. Reiss & R.Bootzin (Eds.), Theoretical issues in behavior therapy (pp.179217). New York: Academic Press.

Baer, D. M., & Sherman, J. A. (1964). Reinforcementcontrol of generalized imitation in young children.Journal of Experimental Child Psychology, 1, 3749,doi:10.1016/0022-0965(64)90005-0

Baer, D. M., Peterson, R. F., & Sherman, J. A. (1967). Thedevelopment of imitation by reinforcing behavioralsimilarity to a model. Journal of the Experimental Analysisof Behavior, 10, 405416, doi: 10.1901/jeab.1967.10-405.

Bekkering, H., & Wohlschlager, A. (2002). Action percep-tion and imitation: A tutorial. Attention and Performance,19, 294314, http://hdl.handle.net/2066/62802

Bekkering, H., Wohlschlager, A., & Gattis, M. (2000).Imitation of gestures in children is goal directed. TheQuarterly Journal of Experimental Psychology, 53, 153164,doi: 10.1080/713755872


Brace, N., Kemp, R., & Snelgar, R. (2006). SPSS forpsychologists: A guide to data analysis using SPSS forwindows (Versions 12 and 13) (3rd Ed.). London:Palgrave Macmillan.

Camoes-Costa, V., Erjavec, M., & Horne, P. (2010). Compre-hension and production of body-part labels in 2- to 3-year-old children. British Journal of Developmental Psychology.In Press. doi:10.1348/026151010X523040

Catania, A. C. (1998). Learning (4th ed.). Upper SaddleRiver, NJ: Prentice Hall.

Erjavec, M., & Horne, P. J. (2008). Determinants ofimitation of hand-to-body gestures in 2- and 3-yearold children. Journal of the Experimental Analysis ofBehavior, 89, 183207, doi:10.1901/jeab.2008.89-183

Erjavec, M., Lovett, V. E., & Horne, P. J. (2009). Do infantsshow generalized imitation of gestures? II: The effectsof skills training and multiple exemplar matchingtraining. Journal of the Experimental Analysis of Behavior,91, 355376, doi:10.1901/jeab.2009.91-355

Garcia, E., Baer, D. M., & Firestone, I. (1971). Thedevelopment of generalized imitation within topo-graphically determined boundaries. Journal of AppliedBehavior Analysis, 4, 101112, doi:10.1901/jaba.1971.4-101

Gleissner, B., Meltzoff, A. N., & Bekkering, H. (2000).Childrens coding of human action: Cognitive factorsinfluencing imitation in 3-year-olds. DevelopmentalScience, 3, 405414, doi: 10.1111/1467-7687.00135

Horne, P. J., & Erjavec, M. (2007). Do infants showgeneralized imitation of gestures? Journal of the Exper-imental Analysis of Behavior, 87, 6387, doi:10.1901/jeab.2007.11-06

Horne, P. J., Erjavec, M., & Lovett, V. E. (2009). The effectsof modeling, local stimulus enhancement, and affor-dance demonstration on the production of object-directed actions in 6-month-old infants. British Journalof Developmental Psychology, 27, 269281, doi:10.1348/026151008X297126

Horne, P. J., & Lowe, C. F. (1996). On the origins of namingand other symbolic behavior. Journal of the ExperimentalAnalysis of Behavior, 65, 185241, doi:10.1901/jeab.1996.65-185

Horne, P. J., & Lowe, C. F. (1997). Toward a theory ofverbal behavior. Journal of the Experimental Analysis ofBehavior, 68, 271296, doi:10.1901/jeab.1997.68-271

Horne, P. J., Hughes, J. C., & Lowe, C. F. (2006). Namingand categorization in young children: IV. Listenerbehavior training and transfer of function. Journal ofthe Experimental Analysis of Behavior, 85, 247273,doi:10.1901/jeab.2006.125-04.

Horne, P. J., Lowe, C. F., & Harris, F. D. (2007). Namingand categorization in young children: V. Manual signtraining. Journal of the Experimental Analysis of Behavior,87, 367381, doi:10.1901/jeab.2007.52-06

Horne, P. J., Lowe, C. F., & Randle, V. R. (2004). Namingand categorization in young children: II. Listenerbehavior training. Journal of the Experimental Analysis ofBehavior, 81, 267288, doi: 10.1901/jeab.2004.81-267

Hurley, S., & Chater, N. (Eds.). (2005), Perspectives onImitation: From neuroscience to social science (Vol. 1).London: MIT Press.

Kymissis, E., & Poulson, C. L. (1994). Generalized imitationin preschool boys. Journal of Experimental Child Psychol-ogy, 58, 399404, doi:10.1006/jecp.1994.1041

Lowe, C. F., Horne, P. J., & Hughes, J. C. (2005). Naming andcategorization in young children: III. Vocal tact trainingand transfer of function. Journal of the Experimental Analysisof Behavior, 83, 4765, doi: 10.1901/jeab.2005.31-04

Lowe, C. F., & Horne, P. J. (1996). Reflections on namingand other symbolic behavior. Journal of the ExperimentalAnalysis of Behavior, 65, 315340, doi:10.1901/jeab.1996.65-315

Lowe, C. F., Horne, P. J., Harris, F. D., & Randle, V. R.(2002). Naming and categorization in young children:Vocal tact training. Journal of the Experimental Analysis ofBehavior, 78, 527549, doi:10.1901/jeab.2002.78-527

Luiz, D., Barnard, A., Knosen, N., Kotras, N., Horrocks, S.,McAlinden, P., Challis, D., & OConnell, R. (2006).MDS-ER 28 Griffiths Mental Developmental Scales Extended Revised: 2 to 8 years. Oxford: Hogrefe.

Peterson, R. F. (1968). Some experiments on theorganization of a class of imitative behaviors. Journalof Applied Behavior Analysis, 1, 225235, doi:10.1901/jaba.1968.1-225

Documents

jeab-96-03-291