Article review

profileTherealdeal00
McDowell2015.pdf

Behavioral Interventions Behav. Intervent. 30: 333–351 (2015) Published online 4 August 2015 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/bin.1419

ANALYSIS OF LIVE MODELING PLUS PROMPTING AND VIDEO MODELING FOR TEACHING IMITATION

TO CHILDREN WITH AUTISM

Logan S. McDowell1, Anibal Gutierrez1* and Kyle D. Bennett2

1Department of Psychology, Florida International University, Miami, FL 33199, USA 2Department of Teaching and Learning, Florida International University, Miami, FL 33199, USA

Previous researchers have demonstrated that training in imitation can significantly improve the learning capabilities of children diagnosed with autism spectrum disorder (ASD) and that children within this population show a preference for video presentations. Video-based instruction has been used to teach a variety of behaviors to individuals with ASD. However, only a small number of studies have examined the use of video modeling to teach initial imitation. Furthermore, there are limited and conflicting data on the effectiveness of a video modeling procedure that does not incorporate prompting when used to teach imitation to young children with ASD. Thus, the purpose of this study was to evaluate a video-modeling-alone procedure and a live-modeling-with-prompting procedure for teaching imitation to young children with ASD. The results suggest that the live modeling with prompting procedure was more effective, and implications related to this finding are discussed. Copyright © 2015 John Wiley & Sons, Ltd.

Generalized imitation has been repeatedly shown to be a necessary component of a typical developmental trajectory (Dawson & Adams, 1984). Imitation is a prerequi- site for the development of important skills including communication (Cihak, Smith, Cornett, & Coleman, 2012), play skills (D’Ateno, Mangiapanello & Taylor, 2003), and observational learning (Bandura, 1977). Imitation allows children to learn through observation and to reproduce the actions of a model, skills essential for suc- cess in a typical classroom setting (Ledford & Wolery, 2011). Children with autism spectrum disorders (ASD), however, are known to demonstrate a deficit in imitation, not only as compared to typically developing peers but also in a more pronounced manner than children with intellectual or other developmental disabilities (Dawson & Adams, 1984). This pronounced deficit has resulted in decades’ worth of research on the most effective methodologies for teaching children with ASD to imitate (Ledford & Wolery, 2011).

*Correspondence to: Anibal Gutierrez, Department of Psychology, Florida International University, Center for Children and Families, 11200 SW 8th Street, AHC1, Miami, FL 33199, USA. E-mail: [email protected]

Copyright © 2015 John Wiley & Sons, Ltd.

334 L. S. McDowell et al.

One of the most common methods for teaching imitation skills to children with ASD is discrete trial training (DTT; Ledford & Wolery, 2011). In DTT, a therapist presents the child with an instruction, waits for an appropriate response, and provides reinforcement. If the child does not produce the target response independently, the therapist implements a prompting procedure to evoke the target behavior, and this prompting tactic is generally considered a necessary component of DTT when teach- ing imitation and other important skills (MacDuff, Krantz, & McClannahan, 2001).

Video Modeling

Over the past several years, there has been a marked interest in video modeling (VM), another procedure used to teach children with ASD (Bellini & Akullian, 2007). VM is the presentation of previously recorded video footage of a model performing a certain behavior used to evoke new behaviors from participants, and it has been used to train a variety of skills in both children and adults (Bidwell & Rehfeldt, 2004; Taber-Doughty et al., 2011). Behaviors that have been targeted for intervention through VM include self-help (Lasater & Brady, 1995), social commu- nication (Maione & Mirenda, 2006), expressive labeling (Charlop-Christy, Le, & Freeman, 2000), and play skills (Lydon, Healy, & Leader, 2011; D’Ateno, Mangiapanello & Taylor, 2003), to name a few.

Video modeling may be particularly appropriate for individuals with ASD because of a tendency toward stimulus over-selectivity; the propensity to focus on one stimulus at the expense of other important stimuli in the environment (Reed, 2012). The video presentation allows for the narrowing in of the individual’s focus to the video itself while simultaneously helping to prevent them from over-selectively focusing on unrelated stimuli (Cardon & Azuma, 2012). It has also been shown that children with ASD demonstrate a preference for video presentations above live presentations (Cardon & Azuma, 2012). In combination, the preference for video presentations and the attention-narrowing effects may make VM a beneficial interven- tion modality for children with ASD.

Another potential benefit of VM is that it has been lauded as more effective than traditional DTT procedures in achieving generalization (Patterson & Arco, 2007). Moreover, VM may also be ideal for behavior interventions because of its presumed low cost and potential ease of implementation (Bellini & Akullian, 2007; Charlop-Christy et al., 2000). Unlike other interventions that may require extensive training, VM is purported to be relatively simple to execute. However, while the term ‘video modeling’ may suggest that the video itself is sufficient to evoke the target behavior, and this has been demonstrated in the literature (see Charlop-Christy et al., 2000; Cardon & Wilcox, 2011), in many instances, the presence of an active therapist implementing prompting and fading strategies is incorporated as a necessary

Copyright © 2015 John Wiley & Sons, Ltd. Behav. Intervent. 30: 333–351 (2015)

DOI: 10.1002/bin

Video Modeling 335

component of the training procedure (Banda, Dogoe, & Matuszny, 2011). In these instances, the ease of implementation, which is often considered a benefit of this methodology, might be considered significantly reduced.

Video Modeling for Teaching Imitation

There is some speculation in the literature as to whether VM is appropriate for teaching imitation based on the assumption that imitation skills are a necessary pre- requisite for the use of this methodology (Kleeberger & Mirenda, 2010). Indeed, Rayner (2011a, 2011b) posited that imitation skills are likely needed for VM inter- ventions to be effective. This is a valid issue; however, there is a paucity of empirical literature available evaluating whether this is the case, and more research is needed in this area (Kleeberger & Mirenda, 2010; Rayner, 2011a, 2011b). Never- theless, recent studies have provided promising results teaching young children imitative behaviors using VM alone and VM with additional techniques (Cardon, 2012; Cardon & Wilcox, 2011; Kleeberger & Mirenda, 2010).

In one such study, Cardon and Wilcox (2011) compared VM to reciprocal imita- tion training, another procedure used for teaching imitation, to determine the effectiveness of each technique in developing imitation skills among young children with ASD. In that study, six children between the ages of 20 and 48 months with a diagnosis of ASD, and three children without disabilities between the ages of 20 and 24 months were taught to imitate using one of these two methodologies. Behav- iors targeted in this study included object imitation with a series of age appropriate toys. Cardon and Wilcox reported that both treatments were effective for both acqui- sition and maintenance of the learned behaviors. Of particular importance in this study, however, was that external prompting was not used during the VM condition. This finding is promising as it provides an initial example of children with ASD learning imitation skills using VM alone, and it adds to the argument that VM may be a more efficient procedure than those used traditionally in DTT.

In a related study, Kleeberger and Mirenda (2010) examined the utility of VM on the acquisition of imitating toy play activities and finger play songs with a 4-year-old child with ASD. The intervention procedure incorporated a series of phases including (i) VM alone, (ii) VM with relevant features of the video highlighted by a caregiver, and (iii) VM with external prompting and social praise. Unlike the Cardon and Wilcox (2011) study, however, the imitative behaviors failed to develop until exter- nal prompting and social praise were added, which highlights the importance of exploring the effectiveness of the VM alone technique to teach imitation specifically when targeting young children. Finally, Cardon (2012) evaluated the effects of parent implemented VM to teach

imitation skills to young children with ASD. Four children between the ages of 24

Copyright © 2015 John Wiley & Sons, Ltd. Behav. Intervent. 30: 333–351 (2015)

DOI: 10.1002/bin

336 L. S. McDowell et al.

and 50 months participated in that study along with their caregivers. Results showed that parents could successfully implement VM tactics and that the children’s imita- tion skills increased substantially. A noted limitation of the study, however, was that physical prompts were used as part of the treatment package. Thus, it is difficult to determine the relative effects of VM separate from the response prompts used. Moreover, these results, when combined with that of Kleeberger and Mirenda (2010), suggest that additional prompting may be needed for some young children with ASD when learning imitation skills.

The findings of these studies represent a discrepancy in the literature and raise ques- tions about whether VM alone can be an effective tool for teaching imitation skills to some young children with ASD. If VM can be effective without prompting, it could potentially be a more efficient methodology than others typically used, such as live modeling (LM) procedures that include prompting and fading typical of DTT. How- ever, Cardon (2012) and Kleeberger and Mirenda (2010) had to implement additional response prompts in addition to VM to help the participants acquire initial imitation skills. This is of particular importance to study because response prompting and fading procedures can be difficult to implement as well as require additional training on the part of professionals and caregivers. Consequently, these added requirements might decrease the perceived ease of implementation of VM-based interventions when used for developing imitation skills among young children with ASD.

Given the conflicting evidence about the need for additional response prompting with VM procedures, it seems prudent to evaluate if VM alone can be effective for teaching imitation skills to young children with ASD. Moreover, given the evidence and historical precedence of using a LM procedure inclusive of prompts, it seems reasonable to compare these approaches. Therefore, the purpose of the current study was to examine whether a VM alone procedure can be effective as an intervention for increasing imitation in young children with ASD while comparing it to a traditional LM intervention package that included prompting. By design, this allows for a comparison of interventions that represent two different treatment approaches, rather than a direct comparison of presentation modalities alone. An additional purpose, given the results of the Kleeberger and Mirenda (2010) study, was to examine VM with prompting should the VM alone intervention not be successful in helping the participants acquire imitation skills. As such, three research questions were posed.

(1) Will children with ASD acquire imitation skills when using VM alone? (2) Will there be a difference between a VM alone procedure and a traditional LM

with prompting intervention package? (3) If participants do not acquire the targeted imitative behaviors under the VM alone

condition, will the addition of a VM with prompting procedure assist the partic- ipants in acquiring the targeted responses?

Copyright © 2015 John Wiley & Sons, Ltd. Behav. Intervent. 30: 333–351 (2015)

DOI: 10.1002/bin

Video Modeling 337

METHOD

Participants

The participants included four male children between the ages of 26 and 42 months, previously diagnosed with ASD. All participants’ diagnoses were confirmed via an Autism Diagnostic Observation Schedule (ADOS; Lord, Rutter, Di Lavore & Risi, 2001) conducted by a research-reliable experimenter. Additionally, an experimenter assessed participants using the Mullen Scales of Early Learning for descriptive pur- poses (henceforth referred to as the Mullen; Mullen, 1995) and the Motor Imitation Scale (MIS; Stone, Ousley & Littleford, 1997). To be included in the study, participants had to have a diagnosis of ASD and exhibit low levels of pretreatment imitation as demonstrated by the results of the Motor Imitation Scale (MIS). Any score at or below 50% (16 out of a potential 32 points) was considered sufficiently low. The MIS was originally designed by Stone, Wolf, and Littleford (1997) to assess

imitation skills present in young children. The assessment is designed to loosely coincide with the steps of imitation originally outlined by Jean Piaget (1962). The assessment involves the presentation of 16 actions. These actions are broken down into meaningful actions, or those that a child would be likely to have seen prior to the assessment (e.g., shaking a maraca), and meaningless actions that he or she is less likely to have encountered previously (e.g., walking a hairbrush across a table). A portion of the assessment is dedicated to motor actions without objects (e.g., clapping or waving), and the rest are motor actions with objects (e.g., pushing a car or shaking a noisemaker). Each of the 16 items is scored on a scale from 0 to 2 (0 meaning no attempt at imitation, 1 meaning burgeoning imitation, and 2 meaning successful point-to-point imitation of the model). Scores on the assessment can range from 0 to 32. As previously stated, a score at or below 16 was considered a sufficient indication of low levels of pretreatment imitation for the purposes of this study. A university Institutional Review Board approved this study. Participants were

recruited through various research programs conducted at a university located in the southeastern USA. Four children with ASD participated in this study. Participants’ caregivers provided written consent allowing their child to participate. Pseudonyms are used throughout this article to protect the participants’ privacy. The first participant, Craig, was a 31-month-old child with ASD. Additionally, he

had an Early Learning Composite score of 49 indicating a descriptive categorization of ‘very low’ as determined by the Mullen. His MIS score was 8 out of 32. The sec- ond participant, Liam, was a 26-month-old child with ASD. He had an Early Learning Composite score of 55 that indicated a descriptive categorization of ‘very low’ on the Mullen. Moreover, he scored 2 out of 32 on the MIS. The third partici- pant was James. He was a 42-month-old child with ASD. His composite score on

Copyright © 2015 John Wiley & Sons, Ltd. Behav. Intervent. 30: 333–351 (2015)

DOI: 10.1002/bin

338 L. S. McDowell et al.

the Mullen was 49, which is in the descriptive category of ‘very low’. Finally, his score on the MIS was 15 out of 32. Michael was the fourth participant. He was a 39-month-old child. His composite score on the Mullen was 49, considered to be in the category ‘very low’. His MIS score was 14 out of 32. (Note that Michael did not complete the study due to time expiring.)

Setting and Materials

The study was conducted at a summer treatment camp for students with ASD (camp ended before Michael could complete the study). Sessions were conducted in one of three rooms. All of the rooms were the same size, approximately 7 ft × 5 ft. One room had a window and the others did not. There were child-sized tables and chairs in each room. Relevant toys were brought to the session rooms and included a small plastic tambourine, toy drum, toy maraca, wooden puzzle piece shaped like a cow, plastic cow figurine, plastic hammer, stacking ring toy, toy piggy bank with plastic coins, plastic screw, and a small plastic cup. Additionally, reinforcing items (e.g., food) were brought to the rooms.

Videos were recorded and played using an iPad®. In each video, a research assis- tant was filmed sitting behind a small table in front of a white wall. At the onset of the video, the research assistant said, ‘Do this’ and performed a simple activity with an object. Each video clip lasted between 3 and 5 s.

Dependent Measures and Data Collection

Target imitative behaviors were selected based on the participants’ MIS scores; however, selected targeted behaviors were not assessed in the MIS. All behaviors included in the study involved the use of an object, produced a sound, and were visible to the participant. These three criteria were included because they have been previously highlighted in the literature as components that decrease the difficulty related to producing an imitative response (Ingersoll, Schreibman & Tran, 2003). The targeted imitative behaviors are presented in Table 1.

All sessions were video-recorded for data collection. We collected data on the num- ber of trials during which the child successfully imitated the model. Point-to-point imitation of the model was recorded as ‘yes’, and any other response was recorded as ‘no’. The dependent variables were percentage correct and trials-to-criterion for the acquisition of the target behaviors. Percentage correct was calculated by dividing the number of correct trials by the total number of trials completed during a session and then multiplying by 100. Trials-to-criterion was calculated by counting the number of trials needed for the participant to successfully imitate the model according to our preset criteria.

Copyright © 2015 John Wiley & Sons, Ltd. Behav. Intervent. 30: 333–351 (2015)

DOI: 10.1002/bin

Video Modeling 339

Table 1. Behaviors targeted for each participant.

Craig Liam James Michael

Dyad Live Video Live Video Live Video Live Video

1

2

Bang hammer

Shake maraca

Stack blocks

Bang drum

Coin bank

Bang hammer

Ring tower

Tambourine

Push cow

Flip cup

Puzzle piece

Turn screw

Push cow

N/A

Walk puzzle piece N/A

Experimental Design and General Procedures

We used an alternating treatments design consisting of baseline and comparison (LM intervention package and VM alone) conditions. Additionally, we added a VM + prompt condition in cases where the target behavior(s) did not develop during the VM alone condition. This added condition was not intended as part of the comparison; rather, it was used for clinical purposes in an attempt to teach each skill to the participants.

Each dyad was composed of two behaviors. One behavior was randomly placed into the VM alone treatment condition, while the other was placed in the LM + prompting treatment condition. One dyad of behaviors was taught at a time, and the second dyad of behaviors was targeted following the successful acquisition of both behaviors in the first dyad. Each dyad was considered complete when both of the behaviors assigned to it reached a mastery criterion of 80% correct across two consecutive sessions. We conducted sessions at least twice a week. Sessions consisted of the presentation of five trials. A least three sessions were conducted per day for each treatment type (VM alone and the LM + prompting package) during the comparison condition. A minimum of 1 h was required between treatment condi- tions in order to avoid multiple treatment interference (Wolery, Gast, & Hammond, 2010). Data were collected on the percent of correct independent responses produced during each session. Sessions lasted approximately 2–10 min. This time variation was due to no prompting being provided in the VM alone condition, which could decrease the amount of time required during these sessions. This contrasts with the LM + prompting and VM + prompting sessions where prompting could have been pro- vided and thus had the potential to increase session times. The first session of the comparison condition for each participant started with VM. Subsequent days may have started with VM alone or LM + prompting depending on how many sessions were conducted the previous day. Equivalency of the behaviors was determined based on visual and auditory feedback as well as a logical analysis conducted by

Copyright © 2015 John Wiley & Sons, Ltd. Behav. Intervent. 30: 333–351 (2015)

DOI: 10.1002/bin

340 L. S. McDowell et al.

the researchers (Wolery et al., 2010). Finally, different verbal discriminative stimuli (SD) were used in the different treatments to assist participants in discriminating between the different conditions (Wolery et al., 2010).

Procedures

Pre-Baseline Procedures. A researcher administered the Autism Diagnostic Observation Schedule (ADOS), Mullen, and the MIS to each participant before the study. (The MIS was also administered at the end of the study to measure any gains made by participants.) Additionally, a researcher conducted a preference assessment by asking parents to list preferred items. At the beginning, and throughout each session, a researcher placed 1–5 preferred items in front of the child, and he or she was given the opportunity to make a selection. Preferred items were varied throughout sessions to avoid satiation with any particular item.

Baseline. During baseline, children were presented with a live researcher who said, ‘Do this’ and performed the target actions. Upon completion of the action, the researcher handed the object used to the child and gave them 2–3 s to respond. Data were collected on whether the child successfully imitated the target behavior. When necessary, the therapists prompted the children to look at the relevant model prior to presenting the SD, and they were successful in reorienting their gaze in instances where attention may have wandered. This occurred throughout all conditions of this study. A video alone baseline was not conducted because (i) we were interested in determining the behaviors the participants could not imitate via a live model; (ii) Cardon and Wilcox (2011) and Charlop-Christy et al. (2000) showed that video alone was a successful treatment to teach imitation skills to children with ASD and, thus, may not have been representative of the pre-intervention behavior; and (iii) the alternating treatments design does not require a pre-intervention baseline. However, because no prompting was incorporated during the initial VM alone treatment condition, the first several data points could be interpreted to represent the child’s initial ability to reproduce the actions of the video model in a baseline-like situation.

Comparison Condition: Live Modeling + Prompting. The comparison condition began following baseline. The LM treatment incorporated prompting, which took place over a number of phases. During the first phase, the researcher provided the child with a discriminative stimulus (SD) to imitate the behavior by saying, ‘Do what I do’ and then performed the behavior. The researcher then waited 2–3 s to see if the child responded independently. If the child did not produce the response, the re- searcher moved to phase two. During this phase, the researcher provided the SD

and modeled the behavior, and then immediately provided hand-over-hand guidance

Copyright © 2015 John Wiley & Sons, Ltd. Behav. Intervent. 30: 333–351 (2015)

DOI: 10.1002/bin

Video Modeling 341

to ensure that the child engaged in the response. The child was then provided with reinforcement. If the child successfully engaged in imitating any part of the re- sponse, the researcher continued to phase three, which involved partial physical guidance. This transition could occur either within a session or between sessions depending on the participant’s accuracy. In phase three, the researcher provided the same SD and model, and then touched the child’s hand to prompt the response. One trial consisted of the presentation of the SD and the performance of the behav- ior, in cases in which no prompting was incorporated. In cases in which prompting was incorporated, one trial consisted of the presentation of the SD, the delay, the prompt, and the eventual performance of the behavior. The next trial began with the next presentation of the relevant SD. The child received reinforcement after each instance in which he or she produced the target response, both prompted and unprompted.

Comparison Condition: Video Modeling Alone. Video modeling alone sessions were alternated with LM + prompting sessions. The therapist sat next to the child at a table and placed an iPad® across from the child. The appropriate video was displayed on the iPad® screen. The therapist then directed the child’s attention to the screen, presented the instruction, ‘Let’s watch the video’, and pressed play on the iPad®. The video was composed of a model sitting at a similar table and provid- ing the SD, ‘Do this’, prior to modeling the appropriate action. Each video lasted approximately 3–5s. Once the video was completed the therapist said, ‘Your turn’, and provided the child with the same object used in the video. The therapist then waited 2–3s for the child to respond. If the child responded correctly, the therapist provided an appropriate reinforcer. If the child did not respond or responded incor- rectly (i.e., he or she performed an action which did not share point-to-point similarity with the modeled response), the therapist moved to the next trial and said, ‘Let’s watch the video again’, and replayed the video. Prompting procedures were not used during the VM alone condition (Cardon & Wilcox, 2011). This continued until all five trials of the session were completed. Video Modeling + Prompting. If the behavior targeted in the VM alone condi-

tion was not acquired by the time the behavior in the LM condition was acquired, and did not demonstrate an increasing trend, a VM + prompting condition was added. This condition involved the researcher playing the video and giving the participant 2–3 s to independently respond. If the participant did not respond, or was incorrect, the researcher followed the same prompting procedure that was used in the LM + prompting condition (i.e., full physical prompt, then partial physical prompt). Reinforcement was provided for prompted and independent correct responses. The purpose of the VM + prompting condition was to avoid a situation where a child was repeatedly exposed to a video with no indication of learning occurring as this could potentially lead to boredom and a refusal to participate in the

Copyright © 2015 John Wiley & Sons, Ltd. Behav. Intervent. 30: 333–351 (2015)

DOI: 10.1002/bin

342 L. S. McDowell et al.

VM sessions (Kleeberger & Mirenda, 2010). This additional phase was not intended to determine whether VM + prompting was a more effective technique, as the previous exposure to the videos during the VM alone phase may have influenced the relevant behavior acquisition but rather in an attempt to help the children acquire the skills.

Interobserver Agreement

Independent observers were trained on the data collection procedures prior to coding the videos. Interobserver agreement (IOA) data were collected for 100% of baseline sessions and 52% of sessions across the comparison and VM + prompting conditions. Trial-by-trial IOA data were collected on whether point-to-point imitation of the model occurred. Data were recorded as a ‘yes’ or ‘no’. Agreement was scored when both observers coded either a ‘yes’ or ‘no’ for the same trial. Trial-to-trial IOA was calculated by dividing the total number of agreements by the total number of agreements and disagreements and multiplying by 100. The overall IOA was 98.9% (98.4–100%).

RESULTS

Figure 1 displays the percentage of correct imitative responses emitted by the participants. Percent correct is presented on the y-axis, and sessions are pre- sented on the x-axis. The participants’ baseline levels were zero (with the exception of one data point for Liam, which was 20% correct). These data were stable with a flat trend. During the comparison condition, the LM + prompting package was either more effective or more efficient for all four participants. VM alone was effective for two of the four participants. Where VM alone was effective, the LM + prompting intervention package was more efficient in that participants reached the criterion faster. VM + prompting was effective for three of the four participants requiring it. Finally, results of the post-treatment MIS …