Video Modeling Training Effects on Types of Attention Delivered by Educational Care-Providers

Abstract

We evaluated the effects of abbreviated (i.e., one-session) video modeling on delivery of student-preferred attention by educational care-providers. The video depicted a novel care-provider interacting with and delivering attention to the student. Within a concurrent multiple baseline design, video modeling increased delivery of the targeted attention for all participants as well as their delivery of another type of attention that was not trained although these effects were variable within and between care-providers. We discuss the clinical and training implications from these findings.

Video modeling (VM) is one of several procedures for training care-providers of people who have autism spectrum disorder (ASD) and related developmental disabilities (DiGennaro Reed, Hirst, & Howard, 2013; Lerman, LeBlanc, & Valentino, 2015). During training, care-providers watch a video model demonstrating skills they are expected to demonstrate under natural conditions. Some variations of VM training include depicting surrogate performers or actual care-providers interacting with a service recipient (Nikopoulos, Luiselli, & Fischer, 2016), inserting text within videos (Rosales, Gongola, & Homlitas, 2015), and adding voiceover scripts (Vladescu, Carroll, Paden, & Kodak, 2012). Research has shown that VM can effectively train care-providers to implement functional analysis (Moore & Fisher, 2007), discrete trial instruction (Catania, Almeida, Liu-Constant, & DiGennaro Reed, 2009; Vladescu et al., 2012), and preference assessment (Rosales et al., 2015).

There are several practical advantages to VM, including the range of skills that can be targeted and standardization of training protocols. As well, it is possible to promote generalization from training by showing videos of skills being displayed in natural as opposed to simulated settings. Furthermore, different video models of the same skill can be constructed in order to accommodate care-providers performing at different competency levels. Nonetheless, VM as a training methodology is still emerging, only a few care-provider skills have been evaluated, and not all variations of training have been fully explored.

This report describes the effects of an abbreviated VM training protocol for teaching care-providers to deliver attention that was hypothesized to be preferred by a student with ASD. Using attention as positive reinforcement is a fundamental intervention procedure within applied behavior analysis. However, the types of attention can vary and have dissimilar effects on behavior based on verbal, physical, and content characteristics (Fisher, Ninness, Piazza, & Owen-DeSchryver, 1996; Piazza et al., 1999). Since VM incorporates visual representation of desired behavior, it may provide a medium to train specific types of attention that a written intervention plan alone may overlook. In the present study, we conducted one-session VM training, evaluated immediate training effects, and reported post-training outcome.

Method

Participants and Setting

The participants were four care-providers at a non-public school for students with ASD and related developmental disabilities. Kevin and Doug had been employed at the school for more than 5 years, while Cathy and Amanda had less than 1-year tenure. All of the participants worked in the same classroom from 9:00 a.m.–3:00 p.m. on weekdays, serving six students who had learning and behavior challenges. The participants were responsible for implementing each student’s individualized educational program and behavior support plan (BSP).

During the study, the participants were observed interacting with and instructing Josie, an 18-year-old female with ASD, communication deficits, and aggressive behavior. Kevin and Doug had previous experience with Josie for approximately 2 years; Cathy and Amanda had less than 1 year of experience with her. Observations included a second student who was paired with Josie and each participant.

Measurement

An observer recorded two dependent measures, reciprocal and differential attention, during a 10-min period when each participant was assigned to work with Josie according to her daily classroom schedule. The dependent measures were selected following several observations that the authors conducted before the study. These observations suggested that Josie enjoyed and preferred certain types of social attention from the participants that featured (a) open-ended praise, (b) brief conversational exchanges, and (c) co-occurring verbalizations such as singing songs and chanting. The positive effects from such attention were inferred by observing Josie smiling, approaching the participants, and responding compliantly to instructions.

Reciprocal attention (RA), the dependent measure targeted for training, was defined as participants verbalizing to Josie by (a) stating open-ended, labeled praise, or behavior description that encouraged reciprocity (e.g., “You’re doing a good job staying in your seat, aren’t you?”), (b) responding within a conversational exchange (e.g., Josie: “I want to say hi to mommy and daddy.” Participant: “What do you want to say to mommy and daddy?”), and (c) singing, whistling, or chanting in unison. The second dependent measure, differential attention (DA), was not targeted for training and was defined as participants simply acknowledging Josie’s behavior without promoting a reciprocal exchange (e.g., “Nice job staying in your seat,” or, “I like how you are keeping safe hands.”). Using partial interval recording, the observer scored occurrences of RA and DA during consecutive 10 s intervals throughout the 10-min period.

As a secondary outcome measure, we reviewed behavior incident reports that the participants completed every time they implemented a physical escort with Josie as a consequence for her aggressive behavior (i.e., slap, bite, push, hair pull) toward them and peers. The steps for physically escorting Josie were described in her BSP. We quantified the incident report data as the frequency of physical escorts each month.

Interobserver Agreement

A second observer conducted partial interval recording simultaneously and independently with the primary observer during 31% of observations. Interobserver agreement (IOA) was computed for RA and DA by dividing the number of intervals both observers agreed that the behaviors occurred by the total intervals and multiplying by 100. Average IOA across participants and experimental phases was 92.4% (range 89–98%) for RA and 91.1% (range 84–97%) for DA. Due to limited staff availability, IOA could not be routinely assessed for physical escorts.

Procedures, Experimental Design, and Evaluation

This study evaluated the effects of abbreviated VM training within a multiple baseline across participants design. In addition to visual analysis of RA and DA data, we also evaluated results statistically through Non-Overlap of All Pairs (NAP; Parker & Vannest, 2009). The NAP is a non-parametric method for determining overlap between each baseline datum and each intervention datum; NAP scores are strongly correlated with the R ² effect size.

Baseline

The participants were observed interacting with Josie under conditions that were in effect before the study. As noted previously, the observations were conducted according to a daily classroom schedule that assigned each participant to work with Josie and another student. The participants were not given instructions and did not receive feedback before or following observations, respectively.

Video Modeling Training

The VM training protocol was implemented individually with each participant during a single session conducted by the same trainer (senior author). The one session per participant lasted approximately 15 min and included four basic components. First, the trainer explained the purpose and general methodology of the VM training, reviewed the definition of RA, and verbally described examples and non-examples of the behavior. Second, the participants watched a 7-min video of a novel care-provider interacting with Josie. The video was recorded on an iPad device, then downloaded and viewed on a desktop computer. The care-provider shown in the video delivered RA to Josie while she received instruction and transitioned between activities. All of the qualities of attention defined by RA were included in the video model. The care-provider delivered 32 instances of RA during the 7-min video. Third, the trainer discussed the video model’s depiction with them, encouraged questions, and answered any inquiries. This training interaction focused on the examples of RA illustrated in the video. Fourth, the trainer concluded the session by instructing the participants to deliver RA to Josie as seen in the video and during their classroom activities with her. No further VM training occurred following the single session. Identical to the baseline phase, the participants did not receive instructions or feedback preceding or following observations.

Follow-Up

While under baseline conditions, the participants were observed interacting with Josie 1 month from conclusion of the post-VM training phases.

Procedural Integrity

A second person observed the trainer during 100% of the VM training sessions with the participants. Using a behavior-specific checklist, the observer recorded whether the trainer accurately implemented the steps of the training protocol: (a) explain the purpose of training, (b) review RA definition, (c) discuss examples and non-examples of RA, (d) have participants watch the video, (e) answer questions, and (f) direct the participants to deliver RA with Josie in the classroom. Procedural integrity (accurately implemented steps/total steps × 100) was 100% for all participants and training sessions.

Results

Figure 1 shows the percentage of intervals in which the participants delivered RA and DA during baseline, post-VM training, and follow-up phases. In baseline, the mean percentage of recording intervals that Kevin delivered RA was 30% (range, 19–43%). Data depicting RA during this phase were moderately variable and showed a slightly increasing trend. Following VM training, the mean percentage of intervals with RA increased to 55% (range, 40–67%), with data indicating low variability and a slightly decreasing trend. Kevin’s mean percentage of intervals with DA also increased from 16% (range, 10–25%) in baseline, where data were stable and showed an increasing trend, to 34% (range, 17–52%) post-VM training. During this phase, data for DA were variable with a slightly increasing trend. At follow-up, Kevin was no longer working in the classroom.

Fig. 1.

Percentage of intervals with reciprocal and differential attention

Cathy’s mean percentage of intervals with RA during baseline was 7% (range, 0–14%), with low variability and no trend. Following VM training, she delivered RA during a mean of 44% (range: 8–68%) of intervals. Data for RA during this phase were variable and showed a decreasing trend. Follow-up revealed a mean percentage of intervals with RA of 28% (range, 7–42%), with data again being variable but indicating an increasing trend. Her mean percentage of intervals with DA increased slightly across phases, from 30% (range, 22–42%) in baseline, to 33% (range, 13–53%) in post-VM training, to 39% (range, 20–48%) at follow-up. Visual analysis of these data suggested a decreasing trend in baseline, a slightly increasing trend in the post-VM phase, and no trend in follow-up. Data for DA were somewhat variable across all three phases.

The mean percentage of intervals with RA for Doug during baseline was 19% (range, 7–25%), with variable data and no evident trend. Following the VM training, his delivery of RA increased to a mean percentage of 33 (range, 9–62%). Data during this phase were variable with a slightly decreasing trend. His mean percentage of intervals with RA at follow-up was 28% (range, 14–47%), with variable data and an increasing trend. For DA, the mean percentage of intervals was 20% (range, 11–35%) in baseline, 3% (range, 18–58%) in post-VM training, and 32% (range, 17–50%) at follow-up. Doug’s delivery of DA showed an increasing trend in baseline, no trend in the post-VM phase, and a slightly increasing trend in follow-up. Across all phases, data for DA were variable.

Finally, Amanda’s mean percentage of intervals with RA during baseline was 15% (range, 7–25%); these data were stable with a slightly increasing trend. Following VM training, the mean percentage of intervals that Amanda delivered RA increased to 42% (range, 33–67%). Data for RA during this phase were variable with a very slight decreasing trend. During follow-up, her mean percentage of intervals with RA was 27% (range, 25–28%); Data were stable with a very slight decreasing trend. The mean percentage of intervals with DA was 43% (range, 35–57%) in baseline, 39% (range, 17–58%) in post-VM training, and 50% (range, 40–67%) at follow-up. Amanda’s delivery of DA showed a decreasing trend in baseline, an increasing trend in the post-VM phase, and an increasing trend in follow-up. Across all phases, data were variable.

Figure 2 indicates the frequency of physical escorts the participants applied with Josie each month. During a 3-month baseline phase, the average frequency per month was 472 (range 274–574). After VM training, average frequency decreased to 120 (range 85–146).

Fig. 2.

Frequency of physical escorts per month

The NAP effect size scores for RA are presented in Table 1. Parker & Vannest (2009) indicated that scores between 0 and 0.65 are considered weak effects, scores between 0.66 and 0.91 are considered moderate effects, and scores between 0.92 and 1.0 are considered strong effects. Table 1 results found moderate to strong effects for post-VM training and follow-up phases for the three participants with complete data, although there was an evident decrease in NAP scores from post-VM to follow-up.

Table 1.

NAP effect size by participant and phase

Post-VM training			Follow-up
Participants	NAP	Effect size	NAP	Effect size
Kevin	0.92	Strong	--	--
Cathy	0.96	Strong	0.90	Strong
Doug	0.82	Moderate	0.68	Moderate
Amanda	1.00	Strong	0.96	Strong

Discussion

Results of this study suggest that VM may be an effective method for training direct-service care-providers (Catania et al., 2009; Moore & Fisher, 2007; Rosales et al., 2015; Vladescu et al., 2012). Specifically, the participants increased their delivery of reciprocal attention following abbreviated VM training, notwithstanding the variable response trends that were documented. The type of attention was selected because it appeared to be preferred by the target student. Incidental data also suggested that physical intervention with the student decreased after all of the participants completed training. Another effect was the apparent generalization of VM training to non-trained but desirable attention. This outcome may have stemmed from the relative similarity between the two types of attention (RA and DA) that were measured during the study.

Response trends for Kevin, Cathy, and Doug revealed an immediate increase in the delivery of RA after the one-session VM training, followed by variable and gradually decreasing delivery. Our abbreviated training model was dictated by limited availability of dedicated trainers at the school. Nonetheless, the results of post-VM training suggest that brief VM sessions may have to be re-implemented periodically to counter drops in performance by care-providers.

Similarly, all of the participants delivered more reciprocal attention following VM training but improvement was modest and less evident at the 1-month follow-up. Once again, allocation of further training resources at the school were restricted and prevented the evaluation of potentially helpful performance enhancing tactics. For example, delivery of reciprocal attention by the participants might have increased to higher levels if they received additional VM sessions, training was extended until they achieved specified learning criteria (Catania et al., 2009), or observational feedback from the trainer was made available.

It is also the case that the VM training protocol included review of the response definition for RA, trainer instructions, and discussion about the video model with the participants. Some or all of these procedures may have contributed to the results. However, it is customary to combine response clarification, instructions, and similar procedures when conducting VM training with care-providers (DiGennaro Reed et al., 2013; Nikopoulos et al., 2016).

Recall that the participants in the study were instructors in the same classroom. This arrangement posed a concern for possible observational learning. Indeed, the baseline data for Doug and Amanda revealed a gradually accelerating trend in percentage of intervals with reciprocal attention. As a control for observational learning, the trainer did not discuss the VM protocol or features of reciprocal attention with the participants until each of them started training. We also did not detect any evidence that the participants shared their training or focus on reciprocal attention with each other at any time during the study. An alternative experimental approach, although not available in this study, would have been evaluating training with participants in separate classrooms.

One study limitation was that the attention preferences selected for Josie were based on direct observation and not formalized preference assessment. Of relevance, Natof & Romanczyk (2009) advised that types of attention should be empirically documented and are likely to vary considerably among children with developmental disabilities. Other characteristics of attention, not evaluated in the current study, would be an instructor’s tone of voice, physical mannerisms, and enthusiasm when interacting with a child.

The frequency of physical escort data served as a proxy measure of the effects from VM training. That is, the participants had less physical intervention with Josie after being trained and subsequently delivering higher percentages of reciprocal attention. Of course, these results are correlational and represented cumulative and aggregated incident report data among all of the participants. Additionally, the frequency measure did not take into account the number of days or hours each day that Josie was in the classroom. Note too, that use of physical escort in Josie’s BSP was based on functional behavioral assessment (FBA) and not a confirmatory functional analysis. Thus, whether the VM training effects on delivering reciprocal attention actually improved Josie’s behavior remains unclear.

One implication from this study is that using VM as a training modality may be indicated when instructing care-providers to implement methods that are detailed and intricate, involve multiple steps, and must be tailored to an individual service-recipient. Abbreviated VM training may also have an immediate positive effect with care-providers, notwithstanding need for booster training should their performance decrease over time. Future research should continue to evaluate alternative training formats, procedures that can be combined with conventional VM, and post-training maintenance of acquired skills. Social validity assessment of care-provider acceptability and satisfaction will also enhance implementation of VM training (Nikopoulos et al., 2016).

Compliance with Ethical Standards

Funding

No funding was received to support this study.

Ethical Approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed Consent

Informed consent was obtained from all individual participants included in the study.

Conflict of Interest

The authors declare that they have no conflict of interest.