Teaching social skills to children with autism using point-of-view video modeling.

Teaching social skills to children with autism using point-of-view video modeling. Abstract Video-modeling (VM) is a widely used instructional technique thathas been applied to the education of children with developmentaldisabilities. One form of VM that lacks in-depth analysis ispoint-of-view video modeling (POVM). The current study investigated theuse of POVM to teach three children diagnosed with autism to initiateand maintain a conversation with a conversant. Using a multiple baselineacross scripts design, the participants were taught to engage in botheye contact and vocal behavior without the presentation of a vocaldiscriminative stimulus from the conversant. The treatment packageincluded both the presentation of the target video and reinforcement oftarget behavior. Although this combination proved successful forincreasing the social behavior of two participants, prompts werenecessary to achieve acquisition for a third. These data suggest thatwhile POVM may be a successful technique for teaching some socialskills, limitations exist that should be further investigated. ********** The effective use of video modeling to help remediate thebehavioral excesses and deficits of children with autism is welldocumented (Bellini & Akullian, 2007). This strategy has been shownto help establish a variety of skills, including those related to jointattention (e.g., LeBlanc et al., 2003), play (e.g., D'Ateno,Mangiapanello, & Taylor, 2003), self help (e.g., Shipley-Benamou,Lutzker & Taubman, 2002), academic instruction (e.g., Kinney,Vedora, & Stromer, 2003), communication (Wert & Neisworth,2003), and community survival (e.g., Haring, Kennedy, Adams &Pitts-Conway, 1987). Additionally, video modeling is potentially moreeffective than teaching through in vivo modeling (Charlop-Christy, Le& Freeman, 2000), and can improve the effectiveness of instructionalprompts (Murzynski & Bourret, 2007). The use of videos to teach social skills has been examined in arecently expanding body of literature. The majority of studiesinvestigating social skills instruction via video models, however,focused on relatively simple behaviors. For example, Bidwell andRehfeldt (2004) used video models and contingent praise to teach adultswith severe disabilities to initiate an interaction by bringing a cup ofcoffee to an adult peer. Nikopoulos and Keenan (2004) demonstrated thatvideo models alone were sufficient for teaching three children withautism to initiate an interaction by gesturing or vocally requesting anadult to join the child in play. A few studies investigated video-based training for more complexsocial skills. Using video models alone, Maione and Mirenda (2006)obtained increases in the frequency of social initiations and responsesof a young boy with autism during two different play contexts. Theparticipant watched videos of two adults engaging in appropriateverbalizations and playing with the target activities. With theimplementation of video modeling, the frequency of theparticipant's use of both scripted and unscripted verbalizations(including initiations and responses) increased during these playsessions. However, reinforcement, video feedback, and prompting wereneeded to increase behavior in a third play context. The authorsreported that some of the modeled statements were novel, while othersalready existed in the child's repertoire. Charlop and Milstein(1989) showed that video models and reinforcement increasedconversational responding for three children with autism. Prior to theintervention, each child exhibited utterances of three to four words inlength. The children were taught scripted exchanges consisting ofstatements with up to eight words per utterance. Each scripted exchangeinvolved an appropriate response to the conversant's question,followed by a reciprocal question to the conversant. While this targetrepresents the most complex set of social behavior taught through videomodels to date, all of the exchanges were initiated by the conversant.Thus, further investigation of the utility of video models for teachingcomplex social skills, including those involving initiation ofconversation in the appropriate context, is warranted. Several authors suggest that video modeling is effective because itreduces the amount of irrelevant stimuli in the learning environment,increasing the likelihood that the participant will focus on the mostrelevant cues (Charlop-Christy et al, 2000; Krantz, MacDuff, Wadstrom,& McClannahan, 1991). If so, video formats that further reduceirrelevant stimuli may help promote learning. One format that may serveto reduce additional irrelevant stimuli in the learning environment ispoint-of-view modeling (POVM). In this type of modeling, the cameraangle is presented at the participant's eye level and shows onlywhat the participant might see within the context of the targetedactivity, skill, or context (i.e., from his or her own viewpoint).Depending on the target skill, the participant might view a specificsetting or a pair of hands completing a task. One potential advantage of POVM over the typical, or scene view,video model is that it further restricts the stimuli to those that aredirectly related to the target behavior, eliminating the necessity ofidentifying optimal characteristics of the model (Hine & Wolery,2006). The extent to which POVM has been utilized is unclear, however,because many prior studies did not include detailed descriptions of thevideo modeling procedures. To date, only four studies have explicitlyevaluated the POVM technique (Alberto, Cihak, & Gama, 2005; Hine& Wolery, 2006; Schreibman, Whalen, & Stahmer, 2000;Shipley-Bena-mou et al., 2002) and these studies included a few specificprocedural features in common while other features varied. For example,participants received prompts and praise for watching the video model ineach case and some degree of generalization and maintenance ofbehavioral gains was observed in each study. However, many aspects ofPOVM differed across studies. For example, only Alberto et al. (2005)prompted rehearsal during video viewing and incorporated least-to-mostprompts during post-viewing practice sessions. The delay between videoviewing and practice opportunities was not specified in this study andin Schreibman et al. (2000), whereas practice occurred immediately afterthe presentation of the video in Shipley-Benamou et al. (2002) and Hineand Wolery (2006). Finally, participants received reinforcement forcorrect responding during practice sessions in some studies (e.g.,Shipley-Benamou et al., 2002; Hine & Wolery, 2006) with no explicitmention in others. The above investigations demonstrated the effectiveness of POVM forteaching self-help skills, play skills, and compliance with transitions.However, additional research is needed to determine the ease with whichsocial and communication skills - two primary core deficit areas forchildren diagnosed with autism--may be acquired through this teachingapproach. The purpose of the current study was to investigate theefficacy of POVM for teaching children with autism to initiate andmaintain social interactions with others. The extent to which theseskills generalized across materials and maintained over time also wasevaluated. Method Participants Participants were three children diagnosed with mild-moderate tosevere autism by an independent psychologist. During the course of thestudy, all participants received behavior analytic services at a privatecenter for 6 hours each day, 5 days per week. They lived at home withtheir parents and received various other therapies outside of theprivate center (e.g., occupational therapy, auditory integrationtraining, dietary supplementation). Each child's language abilitiesand autism severity were assessed prior to the study using the PreschoolLanguage Scale, Fourth Edition (PLS-4; Zimmerman, Steiner, & Pond,2002) and the Childhood Autism Rating Scale (CARS; Schopler, Reichler,& Renner, 1988), respectively. Zhane was 5 years, 5 months at thetime of the study and had attended the center for 2.5 years. Hisreceptive language abilities were assessed to be the age-equivalent of 2years, 3 months, and his expressive language abilities were 2 years, 9months. Zhane's autism severity score was 39, which falls in thesevere range of symptomatology. Randall was 8 years, 2 months and hadattended the center for 2 years, 3 months. His receptive languageabilities were assessed to be the age-equivalent of 3 years, 4 months,and his expressive language abilities were 3 years, 1 month.Randall's autism severity score was 35.5, which falls in themild-moderate range of symptomatology. Janet was 4 years, 4 months andhad attended the center for 10 months. Her receptive and expressivelanguage abilities both were assessed to be the age-equivalent of 3years, 10 months. Janet's autism severity score was 32.5, whichfalls in the mild-moderate range of symptomatology. The children wereselected for the study because they did not engage in spontaneous socialinitiations but could imitate three- to four-word sentences. None of theparticipants had exposure to video models as an instructional strategyprior to the study. Setting and Stimulus Materials All sessions were conducted in a small (2.4 m by 4.6 m) room at theday treatment center. The room contained a child-sized table and chairs,a filing cabinet, a bookcase, and a tripod-mounted video camera, as wellas any materials relevant to the session (described below). Duringtraining sessions, a portable DVD player played the video of thedesignated script. Three scripted sequences of social initiations were prepared (seesample script in Table 1). Each script focused on a different situationthat would set the occasion for a social initiation by the participant.These activities were selected in consultation with the day treatmentcenter supervisor and incorporated free-play items available duringbreaks from instructional time. Each script modeled on the video wasassociated with specific materials. The "Get Attention" scriptinvolved getting a conversant's attention for the purpose ofdisplaying a creation made with a marker and a dry erase board. The"Request Assistance" script was designed to teach a requestfor a conversant's assistance in attaining and opening a claspedplastic box that contained a bottle of bubble solution. The "Sharea Toy" script involved offering a Mr. Potatohead[R] doll to aconversant and then requesting it back again. In addition, two sets ofgeneralization materials were selected for each script. The scriptedstatements were designed to allow for different materials to bepresented in the situation. For example, the conversation created for"Share a Toy" could be used to offer and request any toy, notjust the Mr. Potatohead[R] doll used in the video and practice sessions(see Table 1). The materials shown in each video clip and the materialsused to evaluate generalization are listed in Table 2. Relevantmaterials were present in the baseline and post-viewing practicesessions for each target script.Table 1An Example Script: "Share a Toy" Conversant ParticipantExchange Actions Statement Actions Statement1 enters the looks up from the "Hey there!" room, looks toy, looks at the at the conversant participant2 maintains "Hi!" looks at the toy, "I'm playing eye contact, looks at the with Mr. sits at the conversant Potatohead[R]. table3 looks at the "That looks at the toy, "Would you like toy, looks looks like looks at the to play?" at the fun!" conversant participant4 looks at the "Yes! looks at the toy "May I play toy, looks Thank while the some more?" at the you!" conversant plays, participant looks at the conversant5 looks at the "Sure. looks at the toy, "Thank you!" toy, looks Here you plays with the at the go." toy, looks at participant conversantTable 2Materials for Video-Model and Generalization Scripts "Get Attention" "Request "Share a Toy" Assistance"Video Model Dry erase board Clasped plastic Mr. box Potatohead[R] Dry erase marker Bottle of blowing bubbles Screw-top plastic containerGeneralization Playdoh[R] Key-locked Plastic toy busSet A shape sorter Bottle of blowing bubblesGeneralization Interlocking Bottle of Plastic toySet B building blocks blowing dragon bubblesNote. For "Share a Toy", generalization to a third and fourth toy weretested for Randall. Generalization Set C included a plastic toydinosaur and Set D included a twirling toy. Each video model began with a brief visual introduction (separateslides which showed "1", "2", "3","GO!") followed by three repetitions of the target script.Subsequent to the first presentation, a brief visual transition("READY!", "GO!") preceded the two consecutivepresentations. The total durations for the "Get Attention","Request Assistance", and "Share a Toy" videos were2:21, 2:33, and 2:31, respectively. All camera angles on the video wereshot from the first person perspective (i.e., POVMs) as illustrated inthe Figure 1 screen shots that correspond with the script in Table 1.During filming, the camera was swiveled on the tripod to mimic naturalhead movements and brief (e.g., 2 s to 3 s) eye contact with theconversant, who was an unfamiliar graduate student. A female adult whowas not in view (the first author) spoke the target verbalizations.Because of this person's proximity to the camera, theparticipant's lines were spoken more loudly than theconversant's lines in the final videos. All videos were recorded ina location unfamiliar to the children. Response Measurement and Reliability All post-viewing practice sessions were videotaped for datacollection purposes. During these sessions, the tripod and camera wereplaced in an unobtrusive position over the left shoulder of theconversant (i.e., the adult with whom the participant practiced thetarget skills) to adequately capture eye contact. Data were collected onthe behavior in the target script, as well as on all of thechildren's novel vocal behavior. All scripts were composed of fivespecific exchanges (see example in Table 1). For the purpose of thisstudy, an exchange was defined as eye contact and vocal behavior fromthe child that occurred prior to the vocal behavior of the conversant.Each script began with a social initiation from the child in the form ofa greeting ("Get Attention": "Hi there!";"Request Assistance": "I'm glad to see you!";"Share a Toy": "Hey there!"). Correct and incorrectvocal behavior and eye contact were scored for each exchange. A correctvocal behavior was scored if the child said the exact sentence from thevideo or a sentence that differed by no more than two words (added ordeleted) from the target script (e.g., "Circle" instead of"It's a circle."). For the initial social exchange, anyappropriate greeting (e.g., "Hi", "Hello", and thescript examples above) said by the child was scored as correctregardless of the modeled greeting for that script. Correct eye contactwas scored if the child looked at the conversant for any amount of timeimmediately prior to, during, or following (within 2 s) the target vocalbehavior. These data were collected using pen-and-paper data sheets thatlisted the target vocal behavior for each exchange. Each sheet alsoincluded space to transcribe novel vocal behavior; however, no increasesin appropriate novel language occurred for any of the children duringtreatment and therefore no data are presented for this measure. Thenumber of exchanges consisting of both correct eye contact and vocalbehavior was totaled for each post-viewing practice session. Data werecollected during sessions by the experimenter and were verified byvideotape at the end of each day. A secondary observer collected data from video independently during38%, 42%, and 43% of post-viewing practice sessions for Zhane, Randall,and Janet, respectively. These data were compared for each instance ofeye contact and vocal behavior during a session. An agreement was scoredif both the primary and secondary observer mutually recorded theoccurrence or nonoccurrence of a specific vocal or play target behavior.Interobserver agreement was calculated by dividing agreements byagreements plus disagreements and multiplying by 100. Across allsubjects, scripts, and conditions, interobserver agreement averaged 93%(range, 70% to 100%). Observers also collected data on theconversant's presentation of scripted statements to ensureintegrity. Across all subjects, scripts, and conditions, accuracy ofscripted conversant behavior averaged 99% (range, 80% to 100%). Datawere not collected on other forms of conversant or experimenter behavior(e.g., reinforcer and prompt delivery). [FIGURE 1 OMITTED] Experimental Design and Procedure The study employed a multiple baseline across behaviors (scripts)design. Initial script assignment was counterbalanced acrossparticipants. Each participant began treatment on one of the threescripts while baseline data were collected for the remaining twoscripts. Probes for generalization to novel sets of materials wereconducted throughout all baseline and treatment phases. Once aparticipant attained mastery on the first intervened-upon script (seebelow), treatment began on a subsequent script. Therapists from thetreatment center acted as conversants and were randomly rotatedthroughout all conditions and across all participants. The adultpresented in the video model was not employed by the day treatmentcenter and was never present. Baseline. During baseline sessions, the child sat at a table withthe relevant stimulus materials for the target script. One adultconversant was present. The child was instructed that the conversantwould return shortly and that he or she should play nicely at the tableuntil then. No video was presented. Within 20 s of exiting the treatmentroom, the conversant re-entered the room by knocking lightly on thedoor, stepping into the room, and closing the door behind her. Theconversant performed each action and stated each assigned line within 10s of the beginning of an exchange in the script (see Table 1),regardless of participant behavior. For example, if the participant didnot respond within 10 s of the conversant entering the room, theconversant stated the scripted line of Exchange 1 and waited another 10s for the participant to initiate the next exchange. This procedureensured that each participant behavior could occur with equalopportunity in every session. No contingencies were programmed for eyecontact or vocal behavior. Video plus food. This treatment phase was conducted by one adultacting as the conversant and a second acting as the trainer (the firstauthor) during video viewing and practice sessions. The trainer wasresponsible for setting up session materials, delivering reinforcers,and cueing the conversant (as described below). The trainer was constantacross all treatment conditions for all participants. The participant was seated at the table with the DVD player and thevideo model for the target script. The trainer sat behind theparticipant for the duration of the video viewing and the post-viewingpractice session. Prior to beginning the video, the trainer stated,"Let's watch a movie!" During viewing, reinforcement wasperiodically provided contingent on attention to the video model (i.e.,every 10 to 15 s for eye contact with the television screen) using fooditems identified via a multiple-stimulus-without-replacement preferenceassessment (DeLeon & Iwata, 1996) conducted immediately before eachtraining session. Immediately following the video viewing, the trainerinitiated a post-training practice session by placing the relevantmaterials on the table and stating, "Let's practice."These sessions were identical to baseline with the following exceptions.Food items were delivered by the trainer contingent on scriptedexchanges with the conversant (see Table 1). Immediately following thechild's scripted behavior, the conversant engaged in the scriptedvocal response. If the child did not engage in any part of an exchange,the trainer cued the conversant when 10 s had elapsed by holding up thenext written statement on an index card behind the participant and outof the participant's view. A participant attained mastery if any 8(out of 10) scripted eye contact and vocal behaviors occurred persession across three consecutive sessions. Under this criterion, a giventrial might have included the vocal behavior without the eye contact,for example, and have been counted as one correct behavior (i.e.,mastery was not based on the occurrence of both correct behaviors onevery exchange). Video only. During this condition (Janet only), the video model wasshown as described above and the adult roles were the same as describedabove, but no additional components of the treatment package were ineffect. That is, no food reinforcers were delivered during video viewingor during the post-viewing practice session. During the video-plus-foodcondition, Janet began to orient towards the trainer instead of theconversant each time the conversant spoke suggesting that conversantstatements became discriminative for the delivery of food reinforcersalthough they were only delivered contingent on correct participantbehavior. Previous research using video models indicated that it waspossible for some participants to acquire skills through video modelingwithout the inclusion of programmed reinforcer delivery (e.g.,Charlop-Christy et al., 2000, see Krantz et al, 1991, for a review).Therefore, this condition was implemented to test for acquisition in theabsence of arbitrary reinforcers, as well as to eliminate thestrengthening of inappropriate stimulus control (i.e., looking for afood item each time the conversant spoke). Sessions were otherwiseidentical to those in the video-plus-food condition. Least-to-most prompts. This phase was introduced to facilitationacquisition of the target behaviors for Janet after she did not meet themastery criteria during prior conditions. These sessions were identicalto video-plus-food sessions with the following alteration. If a correctvocal response did not occur within 10 s of an opportunity duringpost-viewing practice sessions, the trainer instituted a three-stepleast-to-most prompting procedure (Horner & Keilitz, 1975). First,the trainer provided a gesture prompt by pointing to the conversant. Ifthe child did not engage in the vocal response within 2 s to 3 s, thetrainer continued to provide a gesture prompt along with a partialverbal model of the beginning of the child's scripted line (e.g.,"I'm glad ..."). If the child did not engage in thetarget behavior within 2 s to 3 s, the trainer combined the gestureprompt with a full vocal model of the child's scripted line (e.g.,"I'm glad to see you"). In this phase, food items weredelivered if behavior occurred independently or with only a partialmodel; edibles were not delivered if a full model was used. In a laterrepetition of this phase, the trainer was eliminated from thepost-viewing practice session and the conversant delivered prompts andfood items. This alteration was made because, as mentioned above, Janetbegan to attend to and engage with the trainer instead of theconversant. The mastery criteria were identical to those in thevideo-plus-food condition. Generalization. Probes with the generalization materials wereconducted throughout all conditions of the study using the proceduresdescribed in the baseline condition. Maintenance. The procedures were identical to those in the baselinecondition with a modification introduced for Randall and Janet after adecrement in responding was observed. Contingent food items werereintroduced to determine if this decrease in behavior was due toextinction effects (labeled "Food Only" on graphs). Sessionswere identical to those in the video-plus-food condition except that thechildren did not watch the video prior to the practice session, and thetrainer was not present during these sessions. Instead, the conversantdelivered food items contingent on correct behavior. Follow-Up. Procedures were identical to those in baseline.Follow-up data were collected 10 days after the last maintenance sessionfor Zhane only, as his rapid performance during treatment allowed forfollow-up assessment during the study timeline. Results For each participant, two figures are presented with the firstillustrating the number of correct (i.e., both eye contact and vocalbehavior) exchanges in each session while the second depicts data on theoccurrence of eye contact and scripted vocal behavior separately.Because either behavior could occur in the absence of the other on eachexchange, the information presented in the second figure provides a moresensitive analysis of behavior over the course of intervention.Furthermore, the mastery criterion was based on the independentoccurrence of eye contact and vocal behavior, regardless of whether theyoccurred together during the same exchange (i.e., any 8 of 10 behaviorsacross three consecutive sessions). However, data in the first figurealso are important to evaluate because reinforcement was deliveredcontingent upon a fully correct exchange (i.e., both aspects correct fora given exchange). Zhane's performance is depicted in Figures 2 and 3. During the"Request Assistance" script (top graph of each figure), Zhanedid not engage in any correct exchanges (Figure 2) during baseline witha gradual increase during the video-plus-food condition and mastery in14 sessions. A return to the baseline condition during maintenanceproduced a brief decrease in exchanges with subsequent maintenance at orabove the mastery criterion and skill maintenance at follow-up. Nocorrect exchanges were observed during the generalization probes untilthe maintenance condition, and generalization sessions for both sets ofstimuli met mastery levels during the follow-up condition. Figure 3shows that eye contact occurred more frequently than vocal behavior inbaseline. With the implementation of the video-plus-food condition, eyecontact increased more rapidly than vocal behavior, and eye contactcontinued to occur at a higher frequency than vocal responses duringmaintenance. For the "Share a Toy" script (middle graphs), nocorrect exchanges (Figure 2) occurred during baseline but increasesoccurred once treatment was implemented with mastery in 9 sessions.Correct exchanges continued during the maintenance and follow-upsessions. In the generalization probes, no correct responding occurreduntil treatment and the increases were not maintained during themaintenance and follow-up phases. Figure 3 illustrates that no correctvocal responses occurred during baseline, though eye contact didincrease during baseline. The video-plus-food condition resulted inincreases in both eye contact and vocal responses that maintained atapproximately the same frequency during maintenance and follow-up.Similar results were obtained for the third script, "GetAttention" (bottom graphs), with the mastery criterion forexchanges (Figure 2) obtained in 6 sessions of intervention andmaintained during maintenance and in follow-up. Eye contact and vocalbehavior (Figure 3) increased simultaneously during intervention;however, little generalization to the novel materials was observed. Randall's performance is depicted in Figures 4 and 5. Duringthe "Share a Toy" script (top graphs), Randall did not engagein any correct exchanges (Figure 4) during baseline, and there was noincrease during the video-plus-food condition. Figure 5 shows that botheye contact and vocal responses occurred at baseline levels. However,Randall's therapists at the day treatment center reported that hewas using the scripted vocal behavior appropriately during hisextra-experimental teaching sessions. It was hypothesized that thepresentation of the video immediately before practice sessions may havecreated an abolishing operation for responding. Therefore, a baselineprobe was conducted, during which Randall responded with all 10 of themodeled behaviors. Subsequent baseline sessions were conducted (labeled"Maintenance" on the figures), but there was a drasticdecrease in correct exchanges after 4 sessions. Figure 5 shows that adecrease occurred for both eye contact and vocal behavior. Because achange from treatment to baseline conditions included the removal ofboth the video viewing component and the delivery of response-contingentfood items, a food-only condition was initiated and correct exchangesmet mastery in 4 sessions and responding continued in maintenance (only2 sessions were conducted due to study timeline). Across all phases,Randall's exchanges during generalization probes did not increaseabove baseline levels. With the implementation of treatment for the "RequestAssistance" script (middle graphs), correct exchanges graduallyincreased to mastery in 15 sessions and continued in maintenance, thoughgeneralization was limited (Figure 4). Eye contact increased duringbaseline for this script when intervention began with the first script(Figure 5) and maintained with the introduction of video-plus-foodcondition, although both eye contact and vocal behavior continued atapproximately the same frequency during maintenance. During baseline forthe "Get Attention" script, correct exchanges never exceededone (the initial greeting; Figure 4, bottom graph). With initiation ofthe video-plus-food condition, Randall's responding reached masteryin 14 sessions and continued in maintenance, though littlegeneralization occurred. Although eye contact increased during baselinewith this script as well, the behavior decreased prior to theintervention. Similar increases in eye contact and vocal behavioroccurred during the video-plus-food condition (Figure 5). Data for Janet are displayed in Figure 6 and Figure 7. Janet'sbaseline responding during the "Get Attention" script (topgraphs) was at zero levels with little increase after 12 sessions in thevideo-plus-food condition. Both eye contact and scripted vocal behaviorremained infrequent (Figure 7). Anecdotally, it was noted that Janet wasattending more to the experimenter seated behind her (who provided thefood reinforcers) than to the target conversant, although no eye contactor other forms of attention were delivered by the experimenter. Tocontrol for this behavior, the video-only phase was initiated after areturn to baseline; however, correct exchanges did not increase (Figure6) and both eye contact and vocal behavior decreased (Figure 7). Theleast-to-most-prompts condition was then implemented correct exchangesquickly increased with mastery in 10 sessions. Figure 7 shows that therewas a more rapid increase in vocal responses than eye contact. However,an immediate decrease in exchanges occurred during a return to baseline.Because the video, prompts, and response-contingent food items had beenremoved, a food-only condition was introduced based on the assumptionthat the removal of reinforcement had extinguished correct responding.Nonetheless, correct exchanges did not increase under this condition. Itwas hypothesized that the decrease in exchanges during maintenance couldinstead have been due to the absence of the experimenter (who hadpreviously delivered prompts). To establish stimulus control in thepresence of the conversant alone, the conversant began to provideprompts in the next phase (labeled "Conversant Prompts" on thegraph) and correct exchanges met mastery in 2 sessions with maintenanceacross 3 additional sessions. Figure 7 shows that both eye contact andscripted vocal responses increased concurrently. In the next phase, theprompts were removed while the delivery of response-contingent fooditems continued and Janet's performance maintained. Across allphases, generalization to new stimuli did not occur. Similar results were obtained for the "RequestAssistance" script (middle graphs). Treatment began with theleast-to-most-prompts condition, and correct exchanges reached masteryin 5 sessions. As with the first script, a food-only condition wasintroduced briefly, but behavior decreased to only one correct exchangeimmediately. Correct exchanges returned to mastery in 2 sessions andmaintained across 3 additional sessions during the conversant-promptscondition, and maintained when prompts were removed in the followingphase. Figure 7 shows that frequency of eye contact increased duringbaseline but decreased during the video-plus-food condition. As with thefirst script, vocal behavior increased more rapidly than eye contactwhen least-to-most prompts were introduced, but both behaviors occurredat approximately the same level during the conversant-prompts andfood-only conditions. There was limited generalization across allconditions. With the implementation of the least-to-most-promptscondition for the "Share a Toy" script (bottom graphs),mastery was met in 6 sessions and correct exchanges continued in theconversant-prompts and food-only conditions, though no generalizationoccurred. It can be seen in Figure 7 that both eye contact and vocalbehavior increased simultaneously with the implementation of treatmentfor this script. Discussion Results of this study are inconclusive regarding the overalleffectiveness of POVM to teach social exchanges to children with autism.Responding on all three scripts came under the control of the video andreinforcement contingencies for 1 of the 3 participants (Zhane). For asecond participant (Randall), two scripts were readily taught using thevideo modeling package intervention whereas an additional scriptrequired modification. For a third participant (Janet), response promptswere necessary to increase the frequency of eye contact and socialinitiations. For all participants, eye contact appeared to generalize acrossbaselines to some extent and was acquired and maintained somewhat moreoften than scripted vocal behavior (see Figures 3, 5 and 7). There aretwo likely explanations for these findings. First, the eye contactmodeled in each video involved the same topography of shifting gaze fromthe materials to the person in all three scripts, while target vocalbehavior was different in each case. Second, the action of eye contact(e.g., the motion of the camera) was clearly visible in the video model,whereas the scripted vocal responses were stated by a person not seen onthe video which is a potential drawback of the POVM compared to a scenemodel. However, these findings were not robust, so further analysis ofPOVM for teaching various forms of social behavior should beinvestigated. Although Zhane's frequency of correct exchanges clearlyincreased with the introduction of treatment across all scripts, hisbehavior did not generalize to the materials used during probes for twoof the three scripts. Randall's response to treatment wasperplexing. Anecdotal reports indicated that he had acquired thebehavior shown in the video (i.e., saying lines from the script whileengaging in eye contact), but it is not clear why he did not engage inthese behaviors during post-viewing practice sessions for the firstscript. His mother reported that he frequently engaged in delayedechoing of lines from his favorite movies at home. However, it appearedthat he did not generalize from the video model to the in-vivo practicesession. The intervention was nonetheless effective with the other twoscripts. For Janet, the video model and reinforcement alone wereinsufficient to increase correct exchanges and response prompts werenecessary to increase her eye contact and vocal behavior. The obtainedresults suggested that Janet's responding was at least partlycontrolled by the behavior of the experimenter, who during treatmentdelivered prompts and reinforcement for Janet's exchanges with theconversant. However, the treatment components responsible for theincrease in correct exchanges are unclear. Additional analyses comparingthe efficacy of video modeling alone to the prompting procedure alonewould provide more information about Janet's acquisition of socialbehavior. Although the general treatment effects were replicated across threescripts for each participant, few correct social exchanges occurred inthe presence of materials that did not appear in the videos. Forexample, Zhane could talk about and share Mr. Potatohead[R], but he didnot do so with a toy bus or with a toy dragon. These results suggestthat generalization is unlikely to occur if training is restricted to asingle set of materials. However, the scripts were designed to bemaximally different while including components of social referencing andverbal initiations. The "Get Attention" script involvedshowing off an item that the child had created (a drawing, a model builtfrom Playdoh[R], or a structure built from blocks), the "RequestAssistance" script focused on requesting an out-of-reach item andassistance to open it (a clasped, screw-top, or locked container withbubbles visible inside), and the "Share a Toy" script wasabout sharing a toy and then requesting it back again (Mr.Potatohead[R], a bus, or a dragon). It is unclear why the current study failed to replicate the resultsof previous research on POVM. One possibility is our use of anoff-screen modeled response (e.g., the scripted vocal statements), asmentioned above. When the intended model is not clearly visible oncamera, as in the case of hands manipulating materials, the stimuli thatshould signal behavior may be more ambiguous. Another possibility is thecomplexity of the social exchanges examined in this study. The currentanalysis selected target behaviors that have not been previously studiedusing POVM and that have rarely been studied with traditional videomodels. Participants not only were required to make brief eye contactwith each social exchange, but they also were required to make astatement that was not dependent on the previous statement of anotherperson (i.e., initiation of the interchange). It is possible that theseskills would not have been acquired through traditional scene videomodeling either. Further analysis of the usefulness of the POVMtechnique to teach social skills to children with autism is needed. Anintermediary step between simple social skills (e.g., greetings) andmore complex skills like those assessed here is warranted. It would also be beneficial to determine which components of thepresent procedure were necessary to produce the desired results. Forexample, while a model and reinforcement were sufficient to change thebehavior of one participant, these components were not entirelysufficient for another, and prompts were necessary for behavior changein a third. Also, it may have been unnecessary to include both a trainerand a conversant. From Janet's results, one person may havesufficed to implement the intervention. A component analysis may help toidentify the necessary ingredients for an effective video modelingtreatment package. It is unclear to what extent the addition of anarbitrary reinforcer aided in the acquisition of target behavior. It ispossible that the inclusion of highly preferred activities, assessed foreach participant, may have made the social interaction itself morereinforcing and reduced the need for arbitrary reinforcers. Although notincluded in many other studies of video modeling, reinforcement wasincluded here because of the unlikelihood that parity of behavior alone(i.e., similarity to a model) would have acted as a reinforcer (e.g.,Home & Erjavec, 2007). This assumption should be tested further withvideo modeling techniques. Anecdotally, the intervention was associated with collateraldecreases in self-stimulatory vocalizations in the post-viewing practicesessions for all children. During baseline sessions prior to treatment,both Zhane and Randall mumbled statements to themselves that weredifficult for others to hear. Once treatment began, these responses didnot occur in practice sessions, although they were noted in probes forgeneralization. Future studies should further analyze the potentialrelationship between video-viewing and verbal behavior that appears tobe maintained by automatic reinforcement. More research needs to be conducted to determine if POVM is limitedin its capacity to teach these or other behaviors (e.g., academic, otherexpressive skills, other social behavior). Future investigations shoulddetermine the characteristics of children who would be considered bestresponders to the point-of-view procedure or skills best suited topoint-of-view perspective in models. Additionally, a comparativeanalysis should test the relative efficacy of traditional video modelingand POVM. Note This research was completed in partial fulfillment of therequirements for the Master's degree for the first author. Theauthors would like to thank Claire St. Peter Pipkin and David P.Jarmolowicz for reviews of earlier drafts of this paper, as well asKathleen Betley and Renee Hogmire for assistance with data collection.The authors would also like to thank the Texas Young Autism Project fortheir assistance with this research. Finally, thanks go out to MaggieStrobel and Alyson Hovanetz for help with creating the video models forthis study. References Alberto, P. A., Cihak, D. F., & Gama, R. I. (2005). Use ofstatic picture prompts versus video modeling during simulation. Researchin Developmental Disabilities, 26, 327-339. Bellini, S., & Akullian, J. (2007). A meta-analysis of videomodeling and video self-modeling interventions for children andadolescents with autism spectrum disorders. Exceptional Children, 73,264-287. Bidwell, M. A., & Rehfeldt, R. A. (2004). Using video modelingto teach a domestic skill with an embedded social skill to adults withsevere mental retardation. Behavioral Interventions, 19, 263-274. Charlop, M. H. & Milstein, J. P. (1989). Teaching autisticchildren conversational speech using video modeling. Journal of AppliedBehavior Analysis, 22, 275-285. Charlop-Christy, M. H., Le, L. & Freeman, K. A. (2000). Acomparison of video modeling with in vivo modeling for teaching childrenwith autism. Journal of Autism and Developmental Disorders, 30, 537-552. D'Ateno, P., Mangiapanello, K. & Taylor, B. A. (2003).Using video modeling to teach complex play sequences to a preschoolerwith autism, Journal of Positive Behavior Interventions, 5, 5-11. DeLeon, I. G. & Iwata, B.A. (1996). Evaluation of amultiple-stimulus presentation format for assessing reinforcerpreferences. Journal of Applied Behavior Analysis, 29, 519-532. Haring, T. G., Kennedy, C. H., Adams, M. J. & Pitts-Conway, V.(1987). Teaching generalization of purchasing skills across communitysettings to autistic youth using videotape modeling. Journal of AppliedBehavior Analysis, 20, 89-96. Hine, J. F. & Wolery, M. (2006). Using point-of-view modelingto teach play to preschoolers with autism. Topics in Early ChildhoodSpecial Education, 26, 83-93. Horne, P. J. & Erjavec, M. (2007). Do infants show generalizedimitation of gestures? Journal of the Experimental Analysis of Behavior,87, 63-87. Horner, R.D. & Keilitz, I. (1975). Training mentally retardedadolescents to brush their teeth. Journal of Applied Behavior Analysis,8, 301-309. Kinney, E. M., Vedora, J. & Stromer, R. (2003).Computer-generated video models to teach generative spelling to a childwith an autism spectrum disorder. Journal of Positive BehaviorInterventions, 5, 22-29. Krantz, P. J., MacDuff, G. S., Wadstrom, O., & McClannahan, L.E. (1991). Using video with developmentally disabled learners. In P.W.Dowrick (Ed.) Practical guide to using video in the behavioral sciences(pp. 256-266). Oxford, England: John Wiley & Sons. LeBlanc, L. A., Coates, A. M, Daneshvar, S., Charlop-Christy, M.H., Morris, C. & Lancaster, B. M. (2003). Using video modeling andreinforcement to teach perspective-taking skills to children withautism. Journal of Applied Behavior Analysis, 36, 253-257. Maione, L. & Mirenda, P. (2006). Effects of video modeling andvideo feedback on peer-directed social language skills of a child withautism. Journal of Positive Behavior Interventions, 8, 106-118. Murzynski, N. T., & Bourret, J. C. (2007). Combining videomodeling and least-to-most prompting for establishing response chains.Behavioral Interventions, 22, 147-152. Nikopoulos, C. K. & Keenan, M. (2004). Effects of videomodeling on social initiations by children with autism. Journal ofApplied Behavior Analysis, 37, 93-96. Schopler, E., Reichler, R. J., & Renner, B. R. (1988). TheChildhood Autism Rating Scale (CARS). Los Angeles: Western PsychologicalServices. Schreibman, L., Whalen, C., & Stahmer, A. C. (2000). The use ofvideo priming to reduce disruptive transition behavior in children withautism. Journal of Positive Behavior Interventions, 2, 3-11. Shipley-Benamou, R., Lutzker, J. R., & Taubman, M. (2002).Teaching daily living skills to children with autism throughinstructional video modeling. Journal of Positive BehaviorInterventions, 4, 165-175. Wert, B. Y. & Neisworth, J. T. (2003). Effects of videoself-modeling on spontaneous requesting in children with autism. Journalof Positive Behavior Interventions, 5, 30-34. Zimmerman, I. L., Steiner, V. G., & Pond, R. E. (2002)., FourthEdition (PLS-4). San Antonio, TX: Harcourt Assessment, Inc. Allison Serra Tetreault and Dorothea C. Lerman University ofHouston - Clear Lake Correspondence to Allison S. Tetreault, Dept of Psychology, WestVirginia University, P.O. Box 6040, Morgantown, West Virginia 26505;e-mail: Allison. Tetreault@mail.wvu.