Evaluation of a Sibling-Mediated Imitation Intervention for Young Children with Autism

Abstract

Parents and peers have been successful at implementing interventions targeting social interactions in children with autism; however, few interventions have trained siblings as treatment providers. This study used a multiple-baseline design across six sibling dyads (four children with autism) to evaluate the efficacy of sibling-implemented reciprocal imitation training. All six typically developing siblings were able to learn and use contingent imitation, four of the six siblings were able to learn and use linguistic mapping, and all six siblings increased their use of at least one component of the imitation training procedure. Three of the four children with autism showed increases in overall imitation and all four showed evidence of increases in joint engagement. Parents and siblings reported high satisfaction with the intervention and ratings by naïve observers indicated significant changes from pre- to post-treatment. These results suggest that sibling-implemented reciprocal imitation training may be a promising intervention for young children with autism.

Autism is a pervasive developmental disorder that emerges early in childhood and has profound effects on social functioning and communication (American Psychiatric Association [APA], 2000). Because of the significant social difficulties of children with autism, a large number of interventions for these children have focused on the development of social skills (see McConnell, 2002, for review). Interventions that have focused on teaching social interaction skills in natural settings and have involved social interaction partners with whom the child with autism is most likely to interact, such as parents and peers, have shown the best evidence of generalization and maintenance of these skills (Rogers, 2000). Research indicates that parents (e.g., Drew et al., 2002) and peers (e.g., McGee et al., 1992; Pierce & Schreibman, 1995) can be successfully taught to implement intervention strategies, and these interventions lead to improvement in social communication skills for the children with autism. Much less research, however, has been devoted to training siblings as intervention providers.

Involving siblings in social skill interventions for children with autism may be particularly successful as siblings spend a significant amount of time with each other in a variety of situations. Further, siblings’ natural play patterns may provide more opportunities for social interactions and play initiations by the children with autism than those of parents (El-Ghoroury & Romanczyk, 1999). However, typical siblings also make few attempts to interact with their siblings with autism during free play sessions (El-Ghoroury & Romanczyk, 1999), suggesting that they may benefit from instruction in techniques to facilitate social interaction with their brother or sister with autism.

The few studies that have examined sibling-mediated interventions for children with autism indicate that this approach can be successful. For example, siblings have successfully used prompting and reinforcement strategies to increase play behaviors in their siblings with autism (Celiberti & Harris, 1993; Coe et al, 1991). In a more extensive intervention, typical siblings of children with ASD were taught a series of skills in ten lessons, including staying in physical proximity to the child with ASD, gaining the child’s attention, creating opportunities for social interaction, organizing and negotiating about play ideas, offering and asking for help, asking for clear directions, and providing verbal feedback. The typical siblings were able to effectively use these skills with the child with ASD, leading to increases in social interaction (a targeted skill) and joint attention (a non-targeted skill) in the children with ASD (Tsao & Odom, 2006).

Reciprocal imitation, in which partners imitate one another in extended turn-taking sequences, may offer another important treatment target for children with autism and their siblings, since it can facilitate social coordination between peers (Eckerman, Davis, & Didow, 1989). Reciprocal imitation training (RIT) is an intervention designed to increase reciprocal imitation skills in children with autism in a naturalistic social context (Ingersoll, 2008). RIT has been successful at increasing imitation in 2- to 4-year-old children with autism, with generalization to different settings, therapists, and materials, and maintenance over time (Ingersoll, 2010; Ingersoll et al., 2007; Ingersoll & Schreibman, 2006). The intervention also increased joint attention, language, and play skills; these skills are theoretically related to imitation, but were not directly targeted in the intervention (Ingersoll & Schreibman, 2006). Mothers of children with autism have also successfully implemented RIT with their children (Ingersoll & Gergans, 2007). This success as a parent-implemented treatment and the simplicity of the intervention suggest that it could be adapted for use as a sibling-implemented intervention.

In the present study, we trained typically developing children to implement RIT with their younger brothers with autism. We assessed siblings’ ability to correctly implement the strategies involved in RIT and examined the effect of the intervention on imitation and joint engagement in the children with autism. We also examined whether skills learned during the course of the intervention generalized to different materials, settings, and typically developing play partners and maintained over time. Further, we examined whether naïve observers were able to detect differences in the sibling interactions from pre- to post-treatment. Finally, we measured the acceptability of the intervention to parents and typically developing siblings.

Methods

Participants

Four young children with autism and six typically developing siblings participated. Target children were diagnosed by a professional with expertise in autism, and exceeded the cutoff scores for “autism spectrum” on the social, communication, and social+ communication algorithms of the Autism Diagnostic Observation Schedule-Module 1 (ADOS; Lord et al., 2000). At intake, children were administered the Bayley Scales of Infant Development, 3^rd Edition cognitive scale (Bayley, 2006) to estimate nonverbal cognitive skills. Children’s primary caregivers completed the MacArthur-Bates Communicative Development Inventory (CDI; Fenson et al., 1993) as a measure of vocabulary development. Because the children were older than the norming sample ranges on the Bayley and CDI, age equivalents will be reported instead of standard scores..

Ryan was 46 months old at intake. He attained scores of 41 months on the Bayley and 30 months on the CDI. Ryan participated with his 9-year-old brother, Mark. Daniel was 53 months old at intake and scored 27 months on the Bayley and greater than 16 months on the CDI. He participated with his 13-year-old sister, Lily. Patrick was 57 months old at intake. He attained scores of 25 months on the Bayley and greater than 16 months on the CDI. He participated with his 8-year-old brother, Eric, and his 9-year-old sister, Andrea. Chris was 45 months old at intake and scored 27 months on the Bayley and greater than 16 months on the CDI. He participated with his 10-year-old sister, Melissa, and his 8-year-old sister, Kristen.

In families with more than one child who fell in the target age range, all siblings were invited to participate. In two families, two siblings were interested in participation. For these families, sessions were conducted separately for each sibling and baseline lengths for the two siblings were staggered by two weeks. All siblings were reported to be developing typically, although Mark’s mother reported that he may have some undiagnosed sensory concerns.

Setting and Materials

All sessions were conducted in the participants’ homes. One room in the home was chosen based on convenience for the family and for filming, and all sessions were conducted in this room. Seven to eight pairs of duplicate toys were provided by the sibling trainer at the beginning of baseline. Toys were chosen based on parent input, developmental appropriateness for the child with autism, and interest to the typically developing siblings. The children were encouraged to use these toys during the baseline and training sessions; however, they often used other toys that were accessible in the filming space in addition to these toys.

Procedure

A multiple-baseline design was conducted across dyads. The trainer (first author) visited the participants’ homes twice a week throughout baseline and treatment. She was a graduate-level student with approximately two years of experience working with young children with ASD. She had been trained in direct implementation of RIT by the second author and had been using RIT for approximately one year. Sessions were videotaped for later scoring.

Baseline

During baseline sessions, the sibling was instructed to “Play with your brother like you usually would.” Most baseline sessions lasted for 10 minutes (the goal length). However, occasionally sessions were shorter (5–9 minutes) due to lack of target child cooperation.

Treatment

Sibling Training

The trainer taught the siblings to use the RIT techniques during two, 15- to 30-minute treatment sessions per week for ten weeks. During training, the trainer made use of a manual written in child-friendly language and augmented with pictures depicting the intervention techniques. This manual is available from the first author by request. When introducing a new intervention technique, the trainer explained the technique and read through the relevant portion of the manual with the sibling, role-played the technique with the sibling while giving instruction and feedback, and demonstrated the technique with the child with autism while explaining her actions. In addition, after each technique was taught, a poster depicting the technique was placed in the playroom to remind the sibling to use the technique. Finally, the sibling used the technique while playing with the child with autism for 10 minutes. During half of this period, the trainer provided live feedback for the sibling. During the other half, the trainer did not provide any suggestions or feedback. The half of this 10-minute session during which feedback was provided alternated from session to session. To encourage the siblings to use the intervention strategies, parents were provided with a sticker chart to track siblings’ playtime outside of training sessions as well as how often the sibling practiced RIT techniques with the child with autism. Parents were asked to place one sticker on the chart for each 30 minutes of playtime between siblings, and an additional sticker on the chart for each session in which siblings practiced RIT techniques. Siblings were reinforced for overall playtime and practicing RIT techniques with their choice of a small prize provided by the trainer each time they accumulated 10 stickers. Siblings were not initially reinforced for time spent during training sessions. However, after the second week of training, Mark was reinforced with up to two stickers for making an effort to use the techniques correctly, as he found it difficult to maintain engagement and effort during the sessions as time went on.

Treatment Phases

To facilitate sibling learning, the intervention techniques were introduced in four phases. During Phase I (2 weeks), the sibling was taught to imitate the actions of his or her brother with autism, including actions with toys, body gestures and movements, and vocalizations (contingent imitation). During Phase II (2 weeks), the sibling was taught to use simplified language (e.g., “Boy is driving”) to describe items and activities that were the attentional focus of the child with autism (linguistic mapping). In Phase III (3 weeks), the sibling was taught to alternate between imitating his brother and providing opportunities for his brother to imitate him, as well as to provide social praise for imitation attempts by the child with autism. Finally, in Phase IV (3 weeks), the sibling was taught to use gentle physical guidance to help his brother with autism imitate if he did not do so spontaneously. See Ingersoll (2008) for a more detailed description of the intervention techniques.

Generalization Probes

During baseline, at the end of the 10-week treatment, and at a 1-month follow-up, generalization sessions were conducted to determine whether skill gains generalized to untrained situations and maintained over time. To assess generalization to new materials, dyads were taped for 10 minutes playing with a set of novel toys provided by the experimenter. To assess generalization to a new setting, dyads were filmed interacting in a different setting in the home. To assess generalization of skills acquired by the child with autism, the target child was filmed interacting with another play partner not involved with the treatment. This child was a similar-aged female family friend for Patrick, Ryan, and Daniel and another sibling in the family for Chris (his 5-year-old sister). Finally, to assess maintenance of gains over time, the siblings were filmed approximately 1 month after completion of the intervention, interacting in the original setting and with the original materials.

Dependent Measures

The 5-minute, un-coached portion of each session was scored in order to examine the siblings’ ability to use the intervention strategies independently.

Sibling Behaviors

To determine if siblings learned to implement RIT techniques correctly, siblings’ fidelity of implementation of each intervention technique was coded from the videotaped play sessions. Contingent imitation and linguistic mapping were scored using 10-sec interval scoring. Contingent imitation was defined as the sibling performing an action with or without a toy either simultaneously to the child performing the action or immediately following the child performing the action. Linguistic mapping was defined as the sibling commenting on or describing the child’s current focus on interest in a non-directive way. Asking questions or giving directions was not considered linguistic mapping. Use of the imitation training procedure was scored by recording all instances in which the sibling 1) provided an appropriate model for imitation (the child saw the model, the sibling used an attention-getting or descriptive phrase, and the sibling used a toy the child was already interested in), 2) provided a physical or verbal prompt for imitation, and 3) provided praise for correct spontaneous or prompted imitation. Use of each strategy was recorded separately and converted to rate per minute of correct use by dividing total instances of correct implementation by the length of the session.

Target Child Behaviors

Imitation was scored as a total percentage of models imitated. Joint engagement was scored as the percentage of 10-sec intervals in which the behavior occurred for at least half the interval. Joint engagement was defined as the target child and the sibling interacting with the same object, and the target child showing evidence of awareness that the sibling was involved in the interaction (e.g., exchanging materials, eye contact, watching the sibling; Kasari et al., 2010). Intervals in which the child spent 50% or more of the interval off camera or interacting with an adult were excluded from analyses of joint engagement.

Inter-rater Reliability

All videotapes were coded by either the first author or undergraduate raters who were trained until reaching an acceptable reliability criterion (Kappa ≥ .60, Pearson’s r ≥ .70) on at least three consecutive videos. Twenty-five percent of the behavioral observations were scored by a second rater who was blind to the treatment phase. For measures scored using interval scoring, Cohen’s Kappa was calculated to determine inter-rater reliability. Cohen’s Kappa was .63 for contingent imitation, .58 for linguistic mapping, and .72 for joint engagement. For frequency measures, Pearson’s r was used to calculate inter-rater reliability, with follow-up t-tests to determine whether raters showed mean differences (Hartmann, 1977). The correlations between raters were .64 for modeling, .82 for prompting, .83 for praise, and .81 for imitation. All correlations were significant at p < .001 and follow-up t-tests indicated no significant differences between raters.

Social Validity

Ratings by Naïve Observers

To investigate whether changes in the children’s behaviors and interactions were socially meaningful (Schwartz & Baer, 1991), 65 undergraduate students viewed 2-minute videotaped samples of each sibling dyad, taken from the third and fourth minute of the generalization to new materials session. These sessions were chosen because the pre-treatment session occurred during baseline and the post-treatment session occurred at the end of treatment and the materials were standardized across clips. The first 35 observers saw Eric, Lily, and Melissa during baseline and Andrea, Mark, and Kristen during post-treatment. The other 31 observers saw Andrea, Mark, and Kristen during baseline and Eric, Lily, and Melissa during post-treatment,, so that each observer saw each dyad only one time. The observers were told that they would be viewing videos of siblings playing together and were kept blind to the circumstances of the videos, including the siblings’ point in treatment and the fact that one child had autism. After viewing each segment, the observers completed a brief questionnaire about how the interaction using a 7-point likert-type scale, with lower scores indicating less positive or less typical behavior and higher scores indicating more positive or more typical behavior.

Treatment Acceptability

To investigate the acceptability of the intervention for parents and siblings, both groups were administered a short questionnaire regarding their satisfaction with the intervention after training was completed (Ingersoll and Gergans, 2007).

Results

Sibling Behaviors

Contingent Imitation

During baseline, all six siblings used low to moderate amounts of contingent imitation. After the onset of Phase I, all six siblings increased their use of contingent imitation strategies. Four of the six siblings continued to use higher-than-baseline levels of contingent imitation through the final phase of training. Eric and Lily maintained their increased use of contingent imitation in a different setting, and Kristen and Lily maintained their use of this strategy with different materials. Mark, Lily, and Kristen all continued to use increased levels of contingent imitation at a 1-month follow-up (See Figure 1). The percentage of data points exceeding the median (PEM) was calculated for contingent imitation. PEM was 95% for Mark, 96% for Lily, 70% for Eric, 65% for Andrea, 83% for Melissa, and 100% for Kristen. PEM for all siblings pooled was 85%. See Table 1 for summary of results of all behaviors.

Figure 1.

Use of contingent imitation and linguistic mapping by siblings. Note: GM generalization materials, GS generalization setting, FU 1-month follow-up.

Table 1.

Summary of Sibling Fidelity of Implementation and Target Child Behaviors Based on PEM and Visual Analysis

Target Child		Ryan	Patrick		Daniel	Chris		Average Across Siblings
Sibling		Mark	Eric	Andrea	Lily	Kristen	Melissa
Contingent Imitation	PEM	95%*	70%*	65%	96%*	100%*	83%*	85%*
	Visual Analysis	Effective	Effective	Slightly effective	Effective	Effective	Effective	----
Linguistic Mapping	PEM	67%	75%*	89%*	100%*	76%*	47%	76%*
	Visual Analysis	Effective	Effective	Effective	Effective	Delayed effect	Ineffective	----
Modeling	PEM	93%*	75%*	53%	67%	21%	80%*	65%
	Visual Analysis	Effective	Effective	Ineffective	Effective	Ineffective	Effective	----
Prompting	PEM	44%	63%	56%	78%*	67%	44%	58%
	Visual Analysis	Effective	Delayed effect	Effective	Effective	Ineffective	Ineffective	----
Praise	PEM	64%	42%	53%	73%*	71%*	27%	55%
	Visual Analysis	Effective	Delayed effect	Effective	Effective	Effective	Ineffective	----
All Treatment	PEM	77%*	66%	65%	85%*	71%*	59%	71%*
Strategies Combined
Imitation (Phase III)	PEM	33%	25%	14%	67%	80%*	33%	42%
	Visual Analysis	Ineffective	Ineffective	Ineffective	Ineffective	Effective	Ineffective	----
Imitation (Phase IV)	PEM	40%	63%	88%*	100%*	89%*	67%	73%*
	Visual Analysis	Ineffective	Effective	Effective	Effective	Effective	Effective	----
Joint Engagement	PEM	82%*	60%	57%	65%	33%	17%	52%
	Visual Analysis	Effective	Delayed effect	Effective	Effective	Decrease, then increase	Ineffective	----

^*Above 70%, the criterion for an effective treatment (Scruggs & Mastropieri, 1998)

Linguistic Mapping

Eric and Lily used linguistic mapping relatively rarely during baseline, while Mark, Andrea, Melissa and Kristen used moderate amounts of linguistic mapping. Following the onset of Phase II, Lily, Eric, Andrea, and Melissa, increased their use of linguistic mapping. Kristen gradually increased use of this strategy to above-baseline levels by the end of treatment. Mark did not increase his use of linguistic mapping during treatment. Lily and Andrea continued to use increased levels of linguistic mapping in a different setting and Eric used a level of linguistic mapping between baseline and treatment levels in a different setting. Lily and Eric continued their increased use of linguistic mapping with different play materials. Lily and Kristen continued to use high levels of linguistic mapping at follow-up, Melissa and Eric’s use of linguistic mapping at follow-up fell between baseline and treatment levels, and Andrea’s use of linguistic mapping dropped back to baseline levels at follow-up (See Figure 1). PEM was calculated for linguistic mapping, using only the baseline data points and treatment data points after the onset of Phase II, as linguistic mapping was not taught until Phase II. PEM was 67% for Mark, 100% for Lily, 75% for Eric, 89% for Andrea, 47% for Melissa, and 76% for Kristen. PEM for all siblings pooled was 76%.

Use of the Imitation Training Procedure

Modeling

Siblings’ rates of correct modeling during baseline varied across siblings. Eric and Andrea presented relatively few correct models during baseline. Mark, Lily, and Melissa presented a moderate number of correct models during baseline, and Kristen presented a high number of correct models during baseline. Following the onset of teaching modeling in Phase III, Mark, Eric, and Melissa increased their use of correct models. Lily did not immediately increase her use of correct models, but gradually increased use to higher-than-baseline levels by Phase IV. Andrea and Kristen did not increase their use of correct models. Mark and Lily continued to use increased rates of modeling with different materials and Eric continued to use increased rates in a different setting. None of the siblings continued using increased rates of modeling at follow-up. PEM was calculated for modeling, using only baseline points and treatment points after the onset of Phase III, as modeling was taught during Phase III. PEM was 93% for Mark, 67% for Lily, 75% for Eric, 53% for Andrea, 80% for Melissa, and 21% for Kristen. PEM for all siblings pooled was 65%. See Figure 2.

Figure 2.

Rate per minute correct use of modeling, prompting, and praise by siblings. Note. GM = generalization materials; GS = generalization setting; FU = 1-month follow-up

Prompting

All six siblings showed relatively low rates of prompting during baseline, with Kristen and Melissa showing somewhat higher rates than the other siblings. Following the onset of teaching prompting in Phase IV, Mark, Lily, Eric, and Andrea, increased their rates of prompting. Kristen and Melissa did not increase their rates of prompting. Mark, Lily, and Eric continued to use prompting with different materials, and Mark and Eric continued to use prompting in a different setting. Eric and Andrea both maintained their use of prompting at follow-up. PEM was calculated for prompting, using only baseline points and treatment points after the onset of Phase IV, as modeling was taught during Phase IV. PEM was 44% for Mark, 78% for Lily, 63% for Eric, 56% for Andrea, 44% for Melissa, and 67% for Kristen. PEM for all siblings pooled was 58%.

Praise

All six siblings showed relatively low rates of praise during baseline. Following the onset of teaching praise in Phase III, Mark, Lily, and Kristen increased their use of praise. After the onset of Phase IV, Mark and Lily further increased their use of praise, and Eric and Andrea increased their use of praise. Melissa did not increase her use of praise. Ryan, Eric, and Kristen continued to use elevated rates of praise both in a different setting and with different materials, Lily continued to use praise with different materials, and Lily, Eric, Andrea, and Kristen continued to use praise at follow-up. PEM was calculated for praise, using only baseline points and treatment points after the onset of Phase III, as modeling was taught during Phase III. PEM was 64% for Mark, 73% for Lily, 42% for Eric, 53% for Andrea, 27% for Melissa, and 71% for Kristen. PEM for all siblings pooled was 55%.

Combined Strategies

PEM was calculated for the use of all strategies for each sibling separately, and for all siblings pooled. PEM for use of RIT strategies was 77% for Mark, 85% for Lily, 66% for Eric, 65% for Andrea, 59% for Melissa, and 71% for Kristen. PEM for all siblings pooled was 71%. See Figure 2.

Target Child Behaviors

Imitation

All of the children with autism showed extremely variable levels of imitation during baseline sessions. Three of the four children with autism showed evidence of increases in imitation performance during the final two phases of treatment, when imitation training was introduced. Daniel, Patrick (with both siblings), and Chris (with Kristen only) showed increases in overall levels of imitation between baseline and the final two phases of treatment, and all three of these children maintained or increased these gains at follow-up. In addition, Chris showed higher levels of imitation at follow-up during his session with Melissa than he showed during either baseline or treatment phases. Ryan decreased his imitation level from baseline to treatment and follow-up. Daniel’s increases in imitation generalized to a different setting and different play materials, and Patrick’s increases in imitation generalized to a different setting (with Andrea), and to different play materials (with Eric; See Figure 3). For imitation, PEM was calculated separately for Phases III and IV, given that different parts of the imitation training procedure were introduced in each phase, and the elements introduced in Phase IV appeared to make a difference in behavior for many of the children. In Phase III, PEM was 40% for Mark, 67% for Lily, 25% for Eric, 14% for Andrea, 33% for Melissa, and 80% for Kristen. PEM in Phase III for all siblings pooled was 42%. In Phase IV, PEM was 33% for Mark, 100% for Lily, 63% for Eric, 88% for Andrea, 67% for Melissa, and 89% for Kristen. PEM in Phase IV for all siblings pooled was 73%.

Figure 3.

Percentage intervals joint engagement and percentage correct imitation by children with ASD. Note. GM = generalization materials; GS = generalization setting; FU = 1-month follow-up.

Joint Engagement

All four target children displayed moderate, but variable, rates of joint engagement during baseline. Ryan, Daniel, and Patrick (with both siblings) all showed increased rates of joint engagement during treatment compared to baseline. At follow-up, these increases were maintained for Ryan and, but dropped back to baseline levels for Patrick. Chris showed decreases in joint engagement during the treatment phase with both siblings. However, with Kristen, he showed large increases in joint engagement at follow-up (and also during the last few treatment sessions), while joint engagement levels remained low with Melissa through follow-up. Patrick’s increases in joint engagement maintained with different materials (with both siblings) and in a different setting (with Eric). Daniel and Chris (with both siblings) also showed increased joint engagement with different materials, but not in a different setting (See Figure 3). PEM for joint engagement was 82% for Mark, 65 % for Lily, 60% for Eric, 57% for Andrea, 17% for Melissa, and 33% for Kristen. PEM for all siblings pooled was 52%. See Table 1.

Social Validity

Ratings by Naïve Observers

Blind observers rated the interactions significantly more positively at post-treatment than at pre-treatment. A MANOVA confirmed significant differences in all behavioral categories examined: Sibling attempts to engage child (Baseline: M=4.9, SD=2.2; Treatment: M=6.0, SD=1.1), F(1) = 138.56, p < .001; Child responds appropriately to sibling (Baseline: M=2.8, SD=1.8; Treatment: M=4.0, SD=1.4), F(1) = 49.39, p < .001; Children’s play is characterized by imitation (Baseline: M=2.7, SD=1.9; Treatment: M=4.1, SD=1.7), F(1) = 74.86, p < .001; Children appear to be having fun (Baseline: M=2.8, SD=1.8; Treatment: M=4.0, SD=1.4), F(1) = 67.31, p < .001; Interaction appears “typical” compared to other children (Baseline: M=4.4, SD=1.4; Treatment: M=5.0, SD=1.4), F(1) = 6.65, p = .01). These patterns were observed across the majority of dyads for all questions except “interaction appears typical,” for which this pattern was observed for three of the six dyads.

Treatment Acceptability

All parents and siblings reported a high level of satisfaction with the intervention. Siblings rated treatment acceptability on a 3-point likert-type scale, with higher scores indicating greater satisfaction. Siblings reported that they enjoyed the intervention (M=2.8, SD=0.4) and that they learned a lot (M=3.0, SD=0.0). They also indicated that, as a result of the intervention, they changed the way that they interact with their brother (M=3.0, SD=0.0) and that playing with their brother was more fun now than before learning the intervention (M=3.0, SD=0.0). Siblings also reported that, on average, the skills were “a little hard” to use (M=2.2, SD=0.4). Parents rated treatment acceptability on a 5-point likert-type scale, with higher scores indicating greater satisfaction. Parents indicated that their typically developing children were able to learn the intervention (M=5.0, SD=0.0), that the children enjoyed the intervention (M=4.5, SD=0.6), and that the quality of play between siblings improved (M=4.6, SD=0.5). They also reported skill improvements for their children with autism in the areas of imitation (M=5.0, SD=0.0), language/communication (M=4.5, SD=1.0), play (M=4.5, SD=0.6), and social engagement (M=4.5, SD=0.6). Parents did not report improvements in fine motor skills, which were not targeted in this intervention (M=3.3, SD=1.3), suggesting that high ratings by parents were not solely due to positive response bias. Finally, parents reported that they would recommend the intervention to others (M=4.8, SD=0.5).

Discussion

Discussion of Results for Siblings

The typical siblings had varying success learning to use the three intervention strategies independently. All six siblings learned to use contingent imitation with their siblings with autism in Phase I. However, only four of the six continued to use contingent imitation after other skills were added. This suggests that some children may have had difficulty remembering to use initial reciprocity strategies once the teaching strategies were introduced.

Four of the six siblings learned to use linguistic mapping appropriately in Phase II, while one sibling (Kristen) slowly increased her use of this skill during treatment, and one sibling (Mark) did not increase his use of linguistic mapping. While several previous studies have incorporated verbal behavior successfully into sibling-mediated interventions (Trent et al., 2005; Tsao and Odom, 2006), the verbal behaviors measured were much less specific (e.g., sibling simply had to make any verbal response, or use any verbal or nonverbal social behavior to be scored as correct) than those required to meet fidelity for linguistic mapping in this intervention. In our study, the sibling had to use descriptive language (not phrased as a question) around the child’s focus of interest. Mark in particular found this quite difficult and frequently continued to phrase his language in question form (e.g., “Are we playing with the binoculars?” rather than “We’re playing with the binoculars”). Therefore, it is possible that Mark and Kristen increased their rate of language, but did not use language that qualified as linguistic mapping. It is also possible that they had difficulty remembering to use linguistic mapping while carrying out the other skills, as the verbal component of the intervention may have been less salient or more difficult to remember than the physical components. Thus, the linguistic mapping taught in this intervention may need to be adapted into a simpler form for siblings, or stressed more during training in order for all siblings to successfully learn this more complex verbal technique.

Four of the six siblings increased their use of modeling during treatment. While Kristen did not increase her use of modeling, this is likely due to her high rates of modeling during baseline sessions. Given the focus of this treatment on reciprocal interactions, increasing her rate of modeling further from baseline might actually have interfered with her use of the other skills, such as contingent imitation. Five of the six siblings increased their use of prompting during treatment. Again, Kristen showed relatively high rates of prompting during baseline, which is likely why she did not increase her prompting rates. Finally, five of six siblings (including Kristen) increased their use of praise for correct imitation. Given the number of strategies the siblings were asked to combine (modeling an action with a toy the child was interested in, ensuring the child saw the action, using a vocalization to describe or draw attention to the action, ensuring the child imitated, and providing praise for correct imitation), it is impressive that the siblings increased their use of such a wide variety of skills with a relatively short duration of training. However, most siblings still used these skills at a lower rate than used by adult therapists trained in RIT (once every 1–2 minutes). The low rate of use of these procedures by the siblings may have limited the target children’s acquisition of imitation skills. Therefore, additional steps should be taken in future studies to increase the rate at which siblings use these strategies. Simply extending the duration of treatment may be helpful, as the siblings had only six sessions to practice the use of modeling, prompting, and praise.

Siblings’ generalization of the use of the skills to a different setting and different play materials was inconsistent. While five of the six siblings were able to show generalization of at least some intervention strategies to untrained situations, no siblings consistently generalized all of their skills to all settings. The siblings were more successful at generalizing their use of the imitation training procedure than contingent imitation and linguistic mapping. The imitation training strategies are more structured and rely less on child behavior (e.g., if the child is not engaged with any activity or object, imitating the child and talking about his interests may be difficult, but introducing a new model and prompting around this may be easier). The siblings’ use of these less child-dependent strategies during generalization probes suggests that changes in child behavior may have been responsible for the inconsistency in generalization of contingent imitation and linguistic mapping. However, it is also possible that the siblings were distracted by the change in setting or materials and had a more difficult time using their skills under these circumstances.

Five of the six siblings showed at least some maintenance of skills at follow-up, and Lily and Kristen maintained all three skills at follow-up. However, the inconsistent use of skills by the other children suggests that more needs to be done to ensure that siblings are able to maintain the skills they have learned. During the treatment period, siblings were reinforced for playtime with their sibling with autism and use of RIT strategies; however, this was discontinued after the treatment period ended. Encouraging parents to continue this reinforcement after training ended might promote continued practice of these skills by siblings and lead to better skill maintenance. Alternatively, involving parents more closely in the sibling training procedure might also help siblings maintain skill gains (Tsao & Odom, 2006).

Discussion of Results for Target Children

Imitation was the primary skill for the child with autism targeted in this intervention. During Phases III and IV (after the sibling was taught to provide models for the child to imitate), Daniel, Patrick, and Chris (with Kristen) all showed increases in overall imitation. While Chris showed slight decreases in imitation during treatment with Melissa, he did show increased imitation at follow-up. In contrast, Ryan showed decreasing levels of imitation from baseline to treatment and follow-up. There are several possible reasons why the children did not show more consistent gains in imitation skills. First, the intervention was delivered at a lower intensity than in previous studies of RIT. If siblings were not implementing the intervention outside of training sessions, the target children would only have received 30 minutes per week of the intervention. It is possible that decreased intensity of intervention (30 minutes/week as compared to 60–180 minutes/week when implemented by therapists or parents) led to fewer gains in imitation for the children with autism. Second, previous studies investigating RIT have used all components of the intervention for 9 (when implemented by parents) or 10 (when implemented by therapists) weeks. In this study, siblings were not taught the final component of the intervention (prompting) until the final 3 weeks of treatment. If prompting for imitation is a key “active” component leading to increases in imitation, it is possible that the decreased duration of intervention led to fewer gains in imitation than were seen in previous studies of RIT. It is also possible that the siblings’ lower fidelity of using the full imitation training procedure compared to trained therapists (only a few sessions fell into the range of one model every 1–2 minutes for the full procedure used by adult therapists) led to fewer gains in imitation. To determine whether sibling-mediated RIT can lead to increases in imitation in children with autism, it would be necessary to provide the full intervention at an increased intensity and/or for an increased duration.

All target children showed patterns that indicated possible effects of the intervention on joint engagement. Ryan, Daniel, and Patrick (with both siblings) showed increases in joint engagement during treatment, and these increases maintained to follow-up for Ryan and Daniel. This finding is consistent with other studies indicating increases in joint attention and social responsiveness when children with autism are imitated by therapists (Field et al., 2001; Ingersoll & Schreibman, 2006) or parents (Lewy & Dawson, 1992). In contrast, Chris’ joint engagement dropped sharply during the initial phases of treatment, in which his siblings began to imitate him. Towards the end of treatment with Kristen, Chris increased his joint engagement, eventually to above-baseline levels. This early drop in joint engagement was likely due to the fact that Chris had a difficult time tolerating contingent imitation. His sisters were encouraged to get close to him so that he would notice being imitated, insist that he share toys (i.e., not allow him to have both pairs of a toy), and follow him when he tried to walk away from the interaction. Initially, this resulted in considerable frustration for Chris; he frequently became unengaged, sought to leave the interaction, or focused intently on the toys in an apparent attempt to “tune out” his sisters. As Kristen persisted with this technique, and began interspersing it with appropriate models and prompts for imitation, Chris began to adjust to this new style of interaction. By the final few sessions, he appeared to be enjoying his sister’s proximity and was participating fully in these joint interactions. In contrast, Melissa had more difficulty persisting with contingent imitation as Chris disengaged, and decreased her use of this technique by the end of treatment. Based on these observations, we would have expected to see a continuing elevation of joint engagement with Kristen had the intervention period continued past the final session. Chris’s above-baseline level of joint engagement at follow-up with Kristen supports this interpretation.

Generalization was inconsistent across children and contexts for imitation and joint engagement, and none of the children maintained their skill gains when interacting with a different child. However, many of the children’s gains were maintained (or even continued to increase) at 1-month follow-up, suggesting that the changes seen during this intervention may persist over time, at least in the original training context. Other studies investigating sibling-mediated social interaction interventions have indicated that siblings with autism have failed to show generalization of skills in different settings (Tsao & Odom, 2006) and with untrained peers (James & Egel, 1986). Given that the skills taught in this intervention were reciprocal in nature, and that siblings spend time together in a wide variety of settings, we hoped that a sibling-mediated RIT approach would promote generalization of skills for the child with autism. The fact that some of the children with autism did show generalization of skills to a different setting and different materials is encouraging. However, it appears that additional steps need to be taken to help children with autism generalize these skills. The two target children who had two siblings participate in training both showed some evidence of generalization and maintenance of skills. It is possible that this aided generalization. However, it is unclear whether they showed better generalization than children who only had one sibling participate. It would be interesting to examine whether involving multiple siblings in intervention aids in generalization of maintenance of skills. It is possible that carrying out some sessions in different settings and with different materials during treatment would help the children better generalize these skills to new situations. It is likely a more intensive and long-term intervention would be necessary to see gains with an untrained play partner. Additional investigation of techniques that promote generalization of skills in response to sibling-mediated interventions is needed.

Social Validity

Ratings of the children’s videotapes by blind observers indicated that changes in both the siblings’ and the target children’s behavior, as well as in the quality of the interaction, were detectable to untrained observers. These data suggest that qualitative changes in the children’s interactions were present that may not have been fully captured by the imitation and joint engagement variables. While it is difficult to determine what factors may have contributed to more positive ratings by blind observers at post-treatment, it is possible that factors such as positive affect or perceived reciprocity may have contributed, given the positive changes for the items “the children appear to be having fun together” and “the children’s play was characterized by imitation.” These findings are an important element of the overall evaluation of this intervention, as they suggest that this intervention produced large enough changes in behavior to be noticeable during a brief interaction by naïve observers.

Both parents and siblings indicated a positive opinion of this treatment overall. Parents felt that their typically developing children were able to learn the intervention, and that it led to increases in the quality of play between siblings and to increases in the child with autism’s imitation skills. Parents also believed that the treatment helped their child with autism to increase language/communication, social engagement, and play skills. Finally, parents felt that the children enjoyed the intervention and that they would recommend it to others. Siblings also reported a positive opinion of the treatment overall. They reported that they enjoyed the intervention, learned a lot, changed their play with their brother, and that the intervention had made playing with their brother more fun. Siblings did report that they found the skills “a little hard” to learn. However, siblings’ use of the majority of the intervention components indicates that they were able to learn the skills, even though they did find it a bit difficult. While ratings from parents and siblings are subjective, they provide important information. Parents and siblings may be able to report on skill increases seen outside of treatment sessions. In addition, it is important for treatments to be perceived and effective and enjoyable by those participating in them in order to promote greater investment in the treatment.

Limitations

This study had a number of limitations. First, while all of the children with autism showed some improvements during the intervention, different children showed different patterns of skill gains. Further research needs to be done to determine which intervention components lead to skill gains for individual children, and what elements of the intervention can be adjusted in order to promote more consistent skill increases in the children with autism. In addition, different patterns of skill gains across children somewhat limits the interpretations of these patterns, as it is possible that changes in child behavior were due to some other variable not measured in this study. In addition, the behavioral coding schemes may not have been sensitive enough to capture all the meaningful changes in sibling and target child behavior. The “modeling” code used for the siblings had lower reliability than is ideal, limiting the interpretation of these data. In addition, ratings by naïve observers appeared to indicate changes in behavior that may not have been captured by the behavioral codes. It is possible that codes that captured changes in dyadic behavior or generalized quality of interactions may have better represented the changes seen as a result of this intervention. Finally, the sibling trainer’s fidelity of implementing the training procedure with the siblings was not systematically measured. Future studies should track the trainer’s fidelity to ensure that the training steps are being implemented as designed.

Conclusions

School-aged siblings were able to learn and use the intervention strategies during a short-term, low-intensity intervention, resulting in positive changes in the behavior of both the sibling and the child with autism. In addition, both siblings and parents found this intervention to be enjoyable and helpful to both the sibling and the child with autism. These findings contribute to the existing evidence that sibling-mediated intervention is a promising approach that may lead to gains both for the child with autism and for the typically developing sibling.