Abstract
This article is intended to synthesize the broader literature investigating the effectiveness and salient features of interventions designed to enhance the social competence of youth and adults with autism spectrum disorder. Outcomes for adults with autism spectrum disorder remain poor with only minimal improvement shown for decades. Among 796 articles reviewed, 42 representative social skill intervention studies met the selection criteria and were coded descriptively for design elements and findings. The review synthesizes and classifies the major categories and issues associated with the key features of the intervention (e.g., the intervention method, agent, tools, and measures), nature of the learning tasks, the implementation setting and length, research methodology, and key findings related to social competence.
Introduction
Foundational to autism spectrum disorder (ASD) are consistent difficulties with social communication/interaction, and the presence of restrictive/repetitive patterns of behavior, interests, or activities (Maye, Kiss, & Carter, 2017). Prevalence rates suggest that 1 in 68 children are diagnosed with ASD, and that boys are 4.5 times more likely to have an ASD than girls (Centers for Disease Control and Prevention, 2016). Although there are some common diagnostic features, there is also great heterogeneity associated with ASD. For example, almost half (44%) of the children diagnosed with ASD score in the average or above average range (score > 85) on IQ measures, and approximately 32% score at or below 70 (i.e., the range required for a diagnosis of intellectual disability [ID]; Centers for Disease Control and Prevention, 2016). This variability is also evident in the way social challenges manifest as some individuals with ASD may appear disinterested in peers or to avoid social interaction, while others actively, although generally ineffectively, seek out social interaction (N. O. Davis & Carter, 2014).
ASD is a lifelong disability. As children with ASD enter adolescence, many of the characteristics associated with ASD may improve (e.g., language), but the limitations with social competence tend to persist in adolescence and adulthood (Magiati, Tay, & Howlin, 2014; Tobin, Drager, & Richardson, 2014; Volkmar, Reichow, & McPartland, 2014). Social competence generally refers to the effective use of social behaviors or skills to achieve desired goals (Odom, McConnell, & Brown, 2008). Moreover, social competence is context dependent and therefore requires the successful selection and application of social skills to match the appropriate context. The social difficulties experienced by individuals with ASD are often attributed to the application of discrete social behaviors and social cognition (Bellini, Gardner, & Markoff, 2014). More specifically, the defining features of ASD include limitations in developing the social communication skills necessary for social reciprocity, using and interpreting nonverbal communication, and acquiring and maintaining social relationships (Maye et al., 2017).
For individuals with ASD, difficulties with social competence may be more pronounced in adolescence as expectations related to social competence evolve over time and become increasingly complex (L. K. Koegel, Koegel, Miller, & Detar, 2014; Laugeson & Ellingsen, 2014). Moreover, like their peers without disabilities, individuals with ASD become more interested in forming relationships during the adolescent years. At the same time, youth with ASD are becoming increasingly aware of their social challenges that interfere with their ability to establish lasting relationships (L. K. Koegel et al., 2014; Laugeson & Ellingsen, 2014; Mendelson, Gates, & Lerner, 2016). This desire to connect with others but inability to effectively do so has led to a number of negative consequences for youth with ASD, including victimization, depression, anxiety, lack of friendships, and loneliness (L. K. Koegel et al., 2014; Laugeson & Ellingsen, 2014). These increasingly complicated social demands and challenges can create more stressors that severely limit the ability of youth with ASD to transition into adult social roles, especially in the domains of peer relationships, risk taking and sensation seeking, and cognitive control (Picci & Scherf, 2015). Regrettably, because these challenges persist into adulthood, adults with ASD are more likely to be isolated, which can lead to additional problems, including anxiety and/or depression (Volkmar et al., 2014).
The social competence of children with ASD is linked to postschool outcomes. In fact, social competence in childhood is a powerful predictor of adult outcomes (Howlin, 2014), and outcomes for adults with ASD remain poor with only minimal improvement shown for decades (Magiati et al., 2014). For example, difficulties with social competence are linked to the persistent reports of un- or underemployment and dependence in adulthood (Tobin et al., 2014) as ineffective social skills create a barrier to a postsecondary education (L. K. Koegel et al., 2014) and to obtaining and maintaining employment (Schall, Wehman, & Carr, 2014). More specifically, rates of employment range from less than 50% (Roux et al., 2013; Shattuck, Cooper, Sterzing, Wagner, & Taylor, 2012; Taylor & Seltzer, 2011) to 56% (Chiang, Cheung, Li, & Tsai, 2013). These employment rates are far lower than their typically developing peers (i.e., 98%; Howlin, 2013), and young adults with ASD may be at particular risk for unemployment when compared to youth with other disabilities (Roux et al., 2013; Shattuck et al., 2012). Also disconcerting, Taylor and Seltzer (2011) reported that 25% of individuals with ASD participated in few or no day activities after leaving high school, and this number was 3 times higher for participants with ASD and no comorbid ID.
Rates of individuals with ASD seeking postsecondary schooling are also low but are expanding, with rates between 31.95% and 50% reported (Chiang et al., 2013; Shattuck et al., 2012; Taylor & Seltzer, 2011; Wei, Yu, Shattuck, McCracken, & Blackorby, 2013). The rates for students with ASD are lower than their peers with other disabilities, including learning disabilities, language impairment, sensory impairment, other health impairment, and traumatic brain injury, but higher than youth with ID (Wei et al., 2013). These low rates are disheartening as many individuals with ASD achieve academically at the same or higher levels than their typical peers and could likely succeed at college with appropriate supports (VanBergeijk, Klin, & Volkmar, 2008).
The social challenges associated with ASD that impede the ability to make friends, form romantic relationships, attain and keep employment, and live independently significantly diminish quality of life in adulthood (Laugeson & Ellingsen, 2014; Picci & Scherf, 2015). As a result, transition planning in adolescence must include an emphasis on social competence goals necessary for individuals with ASD to navigate current and future social environments. Such interventions will have to focus on not only social skills necessary for effective social interaction (e.g., turn taking, initiating, responding) but also social cognition to include perspective taking, interpreting nonverbal social cues, emotion recognition and understanding, emotion regulation, and problem solving (Laugeson & Ellingsen, 2014). Additionally, one indicator of a successful transition (e.g., employment, independent living) from high school to adulthood involves self-determined behavior or the ability to act as one’s own “causal agent” (Wehmeyer, Shogren, Zager, Smith, & Simpson, 2010, pp. 477). Engaging in self-determined behavior requires the use of complex social skills such as engaging in and understanding social communication, problem solving, emotion regulation but also self-advocacy or the ability to negotiate with and/or persuade others (Wehmeyer & Shogren, 2017). Current evidence from the National Longitudinal Transition Study indicates that individuals with ASD have limited involvement in the transition process and goals reflect low postschool expectations. That is, the data suggest that individuals with ASD were less likely than individuals with other disabilities to have goals related to college, vocational training, employment, and independent living (Wehmeyer & Shogren, 2017). These data demonstrate a need to teach the skills necessary for social competence to promote self-determined behavior. Yet social competence remains underaddressed in educational settings (Bellini et al., 2014) even though there is an emerging literature base addressing the social needs of youth and adults with ASD (L. K. Koegel et al., 2014; Laugeson & Ellingsen, 2014).
The purpose of this systematic review is to synthesize the broader literature investigating the effectiveness and salient features of interventions designed to enhance the social competence of youth and adults with ASD. The changing social demands as children enter adolescence will need to be addressed early and well into adulthood as social competence is highly context dependent. It is unlikely that any single intervention will be effective for all adolescents and adults with ASD, because although all individuals with ASD inherently have trouble with social communication, their specific needs will vary. Even though there are recent systematic reviews on the intervention research of social skills development for individuals with ASD, reviews often (a) target a specific social skill intervention such as group-based social skills training (e.g., Cappadocia & Weiss, 2011; Kaat, Lecavalier, & Aman, 2014; Miller, Vernon, Wu, & Russo, 2014; Reichow & Volkmar, 2010) or cognitive-behavioral therapy (e.g., Sukhodolsky, Bloch, Panza, & Reichow, 2013; Weston, Hodgekins, & Langdon, 2016), (b) did not include adults or focused on young children (e.g., Bellini, Peters, Benner, & Hopf, 2007; Gates, Kang, & Lerner, 2017; Reichow, 2012; Reichow & Volkmar, 2010; Wang, Parrila, & Cui, 2013; Whalon, Conroy, Martinez, & Werch, 2015), or (c) included studies focused on adolescents and adults with significant IDs (e.g., Walton & Ingersoll, 2013). Moreover, previous attempts to synthesize social skill interventions for individuals with ASD typically emphasize a single research method (e.g., either single-case research design [SCRD] or group comparison). Therefore, to extend and enrich prior research this systematic review will (a) investigate the broad social skill literature targeting youth and adults with ASD and no comorbid ID, (b) highlight the diversity and complexity of current social skill intervention approaches, and (c) synthesize both single-case and group comparison social skill intervention studies. It is our intent in this systematic review to not only estimate intervention effectiveness but also identify the salient design and implementation features along with theoretical underpinnings that may define and categorize diverse social skills interventions. It is our hope that this synthesis can help guide educators tasked with addressing the social competence of youth and adults with ASD in their selection of effective interventions.
Method
Search Procedures and Inclusion Criteria
Electronic databases Academic Search Complete, Education Full Text, ERIC, MEDLINE PsychArticles, and PsychINFO were searched for peer-reviewed articles from 1995 to 2015 using the following search terms: (1) “autism” or “pervasive developmental disorder” or “Asperger,” (2) “intervention” or “treatment” or “practice” or “strategy” or “therapy” or “program” or “procedure,” and (3) “adolescent” or “adult” or “youth.” This initial search yielded 8,115 articles. Of those, 796 met the criteria for further screening (e.g., title indicated the study included a psychosocial intervention targeting youth or adults with ASD; see Figure 1). Following the initial electronic search, previous reviews of the social competence literature for individuals with ASD and relevant autism journals (e.g., Journal of Autism and Developmental Disorders; Autism: The International Journal of Research & Practice) were searched. Forty-two studies met the following inclusion criteria: (1) adolescents and young adults with ASD age 13 and older (mean age at least 13 years to include participants in and beyond high school), (2) described participants as youth/adults without a moderate or severe ID (e.g., working toward a general diploma; verbal/nonverbal IQ scores of 70 or above), (3) the outcome measure targeted social competence or a behavior that interfered with socialization (i.e., depression, anxiety), and (4) the included intervention was psychosocial (e.g., pharmacological). In addition, study designs that lacked a demonstration of experimental control (e.g., AB designs), and program evaluation studies were excluded. The first and second author screened all studies to determine whether the inclusion criteria was met. A sensitivity analysis, via the one-study removed analysis in Comprehensive Meta-Analysis software 2.2 (Bernard, Borokhovski, Schmid, Tamim, & Abrami, 2014; Borenstein, Hedges, Higgins, & Rothstein, 2008; Schmid et al., 2014), was conducted to examine whether choosing the mean age of 13 (vs. choosing the minimum age of 13) as a selection criterion would skew the results. The sensitivity analysis results indicated that the current mean age selection criterion did not skew overall effect sizes (ESs) and the direction of effect was consistent.
Search and article delimitation process.
Coding Procedures
The 42 articles meeting criteria were initially coded descriptively for design elements and findings (i.e., experimental design, participants, setting, intervention methods, target behaviors, outcome measures, and findings). When conducting the literature search and initial content coding, the researchers developed an analytical framework to guide their review (see Cooper, 2017). Cooper’s (2017) guidelines were selected because they align with professional standards associated with the conduction of systematic reviews and meta-analyses (e.g., Campbell Collaboration Systematic Reviews Guidelines, 2016; Preferred Reporting Items for Systematic Reviews and Meta-Analyses, Moher et al., 2015). The current review is a systematic, descriptive review of both single-case and group comparison research design studies with ES estimates; therefore, we followed the guidelines of conducting a systematic review that are applicable to both single-case and group comparison research design. Using this framework, two coders coded all articles independently based on the research purpose, methods, number and age of participants, interventions, dependent variables, the implementation setting and duration, research methodology, and key findings. These initial coding variables are defined in Table 1. Codes specific to social competence interventions and outcomes were based in the social competence literature for individuals with ASD, and these variables were discussed among coders to establish consistency in terminology. That is, coders initially defined the codes and then met to review codes by practicing coding with a set of single-case and group comparison articles to ensure that the codes fully represented study variables. The agreed-on codes served as the underpinning for a refined analytical framework of the study features (Table 2). Both coders then recoded a randomly selected 60% of the articles, again independently, using the refined coding framework. The average intraclass correlation (interrater reliability) for the codes of key study features was .95, with a range of .91 to .98. After more formal reviews and discussion on differences among the coding sets, coders reached a 100% agreement on the final codes.
Definitions of coding variables
Design features and findings from reviewed studies
Note. In each study, an averaged effect size was reported for the outcomes of the same scale or instrument, and the effect sizes of social competence–irrelevant outcomes were not included in the table. All effect sizes were calculated by comparing the posttest or treatment condition against the pretest or control condition.
Lower score in the scale signifies less impairment, therefore we removed the negative sign of the effect size because it signifies a positive effect. bWith regard to effect size of hierarchical linear modeling in Lerner, Mikami, and Levine (2011), we used the formula suggested by Floman, Hagelskamp, Brackett, and Rivers (2016) and Rosnow, Rosenthal, and Rubin (2000).
Different from the well-established experimental group design (GD), the evaluation standard for SCRD is only emerging. In this review, we applied the evaluative method (Reichow, 2011; Reichow, Volkmar, & Cicchetti, 2008) to assess study quality of SCRD studies. The evaluative method for SCRD studies provides a coding scheme (i.e., high, acceptable, or unacceptable) defining the quality of evidence of six primary (i.e., participant characteristics, independent variable, dependent variable, baseline condition, visual analysis, and experimental control), and six secondary (i.e., interobserver agreement, kappa, blind raters, fidelity, generalization or maintenance, and social validity) indicators. The evaluative method was selected for several reasons: (1) there is evidence of reliability (Cicchetti, 2011), (2) it was designed specifically for the identification of evidence-based practices for individuals with ASD, (3) the criteria align well with quality indicators (Cicchetti, 2011; Wendt & Miller, 2012), and (4) it has been effectively used in several reviews and meta-analyses/reviews related to ASD (e.g., Gevarter et al., 2016; Knight & Sartini, 2014; Spector, 2011; Whalon et al., 2015).
Effect Size Estimates
For this synthesis, ES estimates were calculated, but not aggregated. Several studies defined social competence differently including either continuous behavioral observation (i.e., typical in SCRD studies), standardized informant measures (e.g., Social Skills Improvement System, Social Responsiveness Scale), or direct performance measures (e.g., emotion recognition test). The number of measures and constructs assessed can limit the usefulness of an aggregated ES estimate (Cooper, 2017). In addition, the design features including interventions were wide-ranging, which would fail to lead to any meaningful generalizations. Because of the variability in outcome and design included in this review, ES estimates were provided as only a descriptive indicator of an effect.
For SCRD studies, ES estimates were calculated using Tau-
For GD studies, the ES estimates were measured by calculating Cohen’s d statistic (Cohen, 1977) using Comprehensive Meta-Analysis software (https://www.meta-analysis.com). For several studies that involved the nonparametric Wilcoxon signed-rank test, we calculated ES estimates by converting the Z value produced by SPSS 19.0 to Cohen’s d using a Web-based ES calculator (http://www.polyu.edu.hk/mm/effectsizefaqs/calculator/calculator.html). For a single study using hierarchical linear modeling (Lerner, Mikami, & Levine, 2011), we followed the formula suggested by Floman, Hagelskamp, Brackett, and Rivers (2016) and Rosnow, Rosenthal, and Rubin (2000), by calculating Hedge’s g from the t value generated by the multilevel modeling and then converting the g to Cohen’s d.
To calculate Hedge’s g from t values generated by multilevel analysis, we use the following:
To calculate Cohen’s d from Hedge’s g, we use the following:
Although the majority of dependent variables measured an increase in positive prosocial behaviors, some studies also included measures with lower scores indicating less impairment or a reduction in inappropriate or interfering behaviors (i.e., anxiety, inappropriate interactions). For these measures, we removed the negative sign because it signified a positive effect.
Findings
SCRDs were used in 14 studies (34%) and GD in 28 studies. The majority of SCRD studies applied a multiple baseline design (n = 9; 69%) of which seven (78%) were across participants, and two (22%) across behaviors. Two (15%) SCRD studies employed a reversal design, and two (15%) used a multielement design (e.g., ABCDCD). The GD studies adopted randomized controlled trials (RCTs, n = 14, 50%), within-subjects pre–post test designs (n = 12; 43%), and quasi-experimental comparison GDs (n = 2, 7%). Table 2 provides design components, ES estimates, and a summary of findings. Descriptive syntheses of these SCRD and GD studies, organized by the salient themes or patterns that highlight the types of interventions, major intervention goals and strategies, and research methods with quality indicators, are provided below.
Findings From SCRD Studies
Forty-two adolescents/adults participated in the 14 SCRD studies, and studies included a range of 1 to 7 youth/adult participants with a median of 3. Age was reported for 36 participants (86%), and the mean age was 16 (ages 13–26). The majority of participants were male (n = 34; 81%). Nineteen (45%) participants were identified as having Asperger syndrome, 10 (24%) autism, 3 (7%) pervasive developmental disorder, and 10 (24%) an ASD. Target behaviors varied across studies, and the majority of studies specifically addressed more than one social behavior. Table 3 categorizes the outcome measures across studies, and provides a frequency count of each target behavior. With the exception of one study, skills taught and measured focused on social interactions with others (e.g., initiations, responses, social interactions, social engagement). One study (Lovett & Rehfeldt, 2014) measured performance on a computerized perspective taking probe (i.e., emotion recognition).
Intervention types of SCRD studies
Note. SCRD = single-case research design.
The 14 reviewed SCRD studies are categorized by intervention type, including direct instruction, technology-based instruction, or naturalistic interventions. Direct instruction interventions included a combination of adult directed instructional techniques (e.g., modeling, role-play/rehearsal, and feedback) to teach participants with ASD target social skills (e.g., social skills groups). Technology-based interventions relied on a form of technology to deliver or enhance instruction to youth/adults with ASD (e.g., virtual reality, computer emotion recognition software). Naturalistic interventions involved the facilitation of natural learning opportunities that required the use of a target social skill (e.g., peer-mediated interventions, incorporating preferences in activities).
Direct Instruction Interventions
Seven SCRD studies (50%) included direct instruction teaching methods (K. M. Davis, Boon, Cihak, & Fore, 2010; Dotson, Leaf, Sheldon, & Sherman, 2010; Gutman, Raphael-Greenfield, & Salvant, 2012; Hagopian, Kuhn, & Strother, 2009; K. Mitchell, Regeher, Reaume, & Feldman, 2010; Newman, Buffington, & Hemmes, 1996; Nuernberger, Ringdahl, Vargo, Crumpecker, & Gunnarsson, 2013). A total of 20 (48%) youth/adults participated in these studies, and ranged in age from 13 to 23, and the remaining 3 were described as teenagers. A specific age was provided for 14 participants, and the mean age of these students was 17. Seven (35%) participants were described as having average or above average intelligence or performing academically in the average or above average range, and three (15%) participants had a mild ID. The cognitive ability of 10 (50%) participants was not provided. The social skills of 10 (50%) participants were described. Two (10%) participants were reported to initiate conversations with others about preferred interests, and 8 (40%) were noted to have trouble engaging with others socially in general. Studies occurred in a pullout setting in the participant’s school (K. M. Davis et al., 2010; Gutman et al., 2012; Newman et al., 1996), a clinic (Dotson et al., 2010), participant living facility (Nuernberger et al., 2013), or inpatient unit for problem behavior (Hagopian et al., 2009).
Four of these studies used a combination of direct instruction procedures including (1) defining the target behavior/skill, (2) providing a description of and/or rationale for the target behavior/skill, (3) modeling the steps necessary to execute the target behavior/skill, (4) sharing examples/nonexamples of the target behavior/skill, (5) role-playing or rehearsing the target behavior/skill, (6) reinforcement/praise following use of the target behavior/skill, and/or (7) giving corrective feedback. In one study, direct instruction was provided prior to baseline and the Power Card strategy was added during intervention (K. M. Davis et al., 2010). The Power Card served as a cue for participants to maintain a conversation that focused on the interests of others. The card included an example of the participant’s self-identified “hero” (e.g., baseball player, cartoon artist) engaged in a conversation.
The remaining two studies paired positive reinforcement with prompting (Newman et al., 1996) or corrective feedback (Hagopian et al., 2009). In one study, the participant earned two small candy bars after meeting a predetermined criterion (85% reduction of inappropriate comments) at the end of a conversation. If the participant engaged in inappropriate social behavior (e.g., interrupting others, touching others without permission, turning away), then the researcher immediately provided corrective feedback by labeling and defining the inappropriate behavior (Hagopian et al., 2009). In the second study, the researcher taught the participant to self-reinforce appropriate responses during conversation (e.g., relevant comment/question/response). Initially, the researcher modeled providing a token for an appropriate response, and in subsequent phases, external reinforcement was faded to a verbal prompt, then to self-reinforcement without any prompting (Newman et al., 1996).
The majority of direct instruction interventions were short lasting from 7 to 30 minutes and occurred over a 4- to 12-week period or 5 to 26 sessions. The longest sessions ranged from 90 to 120 minutes, and occurred for approximately 34 sessions (Hagopian et al., 2009). Three studies (43%) incorporated instructional activities outside of the initial teaching context to encourage generalization. In one study, role-play scenarios initially occurred in an individualized setting. Later sessions took place in other parts of the school and involved additional school staff and peers, and final sessions were conducted in the community (Gutman et al., 2012). In the second study, participants were taught skills individually, and once they reached 100% on a task analysis during a rehearsal, instruction moved to the living center with peers present. Participants were prompted to initiate a conversation with a peer, and immediately after the conversation, participants received feedback in a one-to-one setting (Nuernberger et al., 2013). In the third study, homework was assigned and caregivers received training to provide prompting, reinforcement, and error correction to encourage effective use of the target social skills at home.
Following participation in interventions incorporating direct instruction, participants increased appropriate verbal (e.g., gaining the attention of the communication partner, initiating responding, continuing and ending an interaction) and/or nonverbal (e.g., maintaining eye contact, standing at an appropriate distance from the conversation partner) conversational skills with ES estimates ranging from .81 to 1.0 (K. M. Davis et al., 2010; Dotson et al., 2010; Gutman et al., 2012; Newman et al., 1996; Nuernberger et al., 2013), and in one study, participants decreased inappropriate social interactions while gradually increasing appropriate conversational skills (ES = 0.94; Hagopian et al., 2009). Findings from one study were mixed (K. Mitchell et al., 2010), with all participants showing improvement in one target skill (i.e., greetings) but only one participant showing an increase across the three target behaviors (ES = 0.42). Although increases are noted across studies, the extent to which participants transferred newly acquired social skills to their daily interactions with others is unclear as the majority of studies measured skill use in contrived or highly structured (e.g., role-plays) settings rather than naturally occurring activities/routines (n = 4; 57%).
Quality ratings: The majority of primary indicators were rated acceptable or high with a few exceptions, including the description of participants in one study (Newman et al., 1996) and the dependent variable in a second study (Gutman et al., 2012). In the second study, the authors indicated they recorded frequency counts of target behaviors, but some definitions were inconsistent with frequency data (e.g., is able to greet or say goodbye; Gutman et al., 2012). In another study, fewer than three data points and high levels of the target behavior demonstrated at baseline for one or more behaviors of two participants limit interpretation of intervention effectiveness (K. Mitchell et al., 2010). In addition, all studies included evidence of interobserver agreement at acceptable levels (>80%). Only one study (K. M. Davis et al., 2010) included treatment integrity data, which suggested the intervention was implemented as intended (>80%). One study indicated that raters were blind to the treatment condition (K. Mitchell et al., 2010), and three studies provided mixed evidence of generalization (K. M. Davis et al., 2010; Dotson et al., 2010; K. Mitchell et al., 2010). Data in follow-up exceeded baseline levels in two studies (Gutman et al., 2012; Nuernberger et al., 2013), and maintenance data were mixed in two studies (Dotson et al., 2010; K. Mitchell et al., 2010). Evidence of social validity was provided in five studies (K. M. Davis et al., 2010; Dotson et al., 2010; Gutman et al., 2012; K. Mitchell et al., 2010; Nuernberger et al., 2013).
Technology-Based Interventions
Three SCRD studies (7%) incorporated technology-based interventions to address the social skills of youth and adults with ASD (Lovett & Rehfeldt, 2014; Mason, Rispoli, Ganz, Boles, & Orr, 2012; State & Kern, 2012). These studies included computer-based multiple exemplar instruction (Lovett & Rehfeldt, 2014), video modeling (VM; Mason et al., 2012), or a comparison of video feedback with in vivo self-monitoring (State & Kern, 2012). A total of six adolescents/adults participated in these studies. Participants ranged in age from 14 to 26. An age range was reported for three participants (17–18), and the mean age of the remaining 3 participants was 20 (range 14–26). Three participants were described as reading on grade level (Lovett & Rehfeldt, 2014), one as performing above grade level (State & Kern, 2012), and no information about cognitive ability was provided for two participants (Mason et al., 2012). The social skills of three participants were described (Mason et al., 2012; State & Kern, 2012). Two participants were described as engaging in inappropriate interactions with others (e.g., making noises, interrupting others, making inappropriate comments), and one participant had difficulty engaging in interactions with others and reading the emotions of others. Interventions occurred in a one-to-one setting in the participant’s school (boarding school, university) or school and home.
In one study, three adolescents with Asperger syndrome were taught perspective taking through a series of scenarios depicting interactions in the context of a social situation presented on PowerPoint slides. Depending on the scenario, the participant was required to change perspectives between characters, locations, or times. Scenarios varied in level of complexity and consisted of deictic relations (i.e., here-there, now-then, or I-you) dependent on the perspective of the speaker. Each trial consisted of two to four sentences describing the perspective of the character with a corresponding picture, questions about the scenario, and four response options provided on the bottom of the slide. In the intervention condition, participants received multiple exemplar instruction. That is, 36 scenarios were presented in increasing levels of difficulty. The word “correct” appeared on the slide following a correct response, and “try again” after an incorrect response. The trial was repeated until the participant responded correctly. Participants were offered a 3- to 5-minute break every 15 minutes. Sessions occurred three times a week for 30 to 45 minutes for 4 to 6 weeks.
In the VM study (Mason et al., 2012), four 1- to 2-minute videos were developed to teach four target social skills: eye contact, facial expressions, turn taking, and sharing emotions. The videos included instructions for using each skill written and narrated in the video immediately preceding a direct, unscripted model of two peers engaged in the target behavior. Participants watched the video of a target skill on a laptop and then engaged in a conversation with a facilitator for 5 minutes. After 5 minutes of conversation, the researcher directed the participant to stop conversing and showed the video again. This continued for a 50-minute session providing four to six opportunities to watch the video and practice the modeled skill. The number of sessions varied by target skill ranging from 4 to 11.
The third study compared the effectiveness of video feedback and in vivo self-monitoring on the inappropriate social behaviors of an adolescent with ASD (State & Kern, 2012). Across conditions, the participant engaged in 15 minutes of game play at school with his teacher and peer, and at home with his mother. In addition, the interventionist modeled five appropriate behaviors and five inappropriate and labeled each. In the video feedback condition, the student viewed a video of a randomly selected 5 minutes taken from a 15-minute activity session during which he was engaged with a peer, his teacher, or his mother in game play. The video was paused the video every 15 seconds, and the interventionist and target student recorded whether the student was engaged in appropriate and inappropriate behavior. In the in vivo self-monitoring condition, during the 15-minute game play activity, the student wore a watch that vibrated at the end of a 1-minute interval cueing him to document whether he was engaged in appropriate or inappropriate interactions on his self-monitoring sheet. In both conditions, the student earned points to receive computer time. The study lasted between 25 and 30 sessions per condition.
Findings from these studies were variable, with ES estimates ranging from .42 to .81, but all participants experienced some benefit from technology-based interventions. Following computer-based multiple exemplar training, participants made gains on perspective taking measures within three or four sessions (ES = 0.73). It is noteworthy that all participants demonstrated some degree of perspective taking ability at pretest including mastery on a measure that directly stated the character’s perspective in the provided scenario. Findings from the VM study were mixed. Visual analysis suggested some overlap with baseline levels, and the ES estimate was 0.42. In the comparison study, in vivo self-monitoring produced a greater reduction in inappropriate social interactions than the video feedback condition with an ES of 0.81. Appropriate interactions were variable across conditions, but the student continued to demonstrate appropriate interactions while decreasing inappropriate interactions/noises. The authors suggest that for some learners with ASD, decreasing inappropriate social interactions may be insufficient to show an increase in appropriate social skills without targeted instruction.
Overall, similar to the direct instruction interventions, the extent to which participants were able to apply their newly learned skills in real-world situations following technology-based interventions is a bit unclear. First, few studies incorporated technology (n = 3; 23%). In addition, one study produced mixed findings following VM (Mason et al., 2012), and another found video corrective feedback less effective than providing a low-tech (i.e., vibrating watch) cue to monitor performance in real time (State & Kern, 2012). In addition, it is difficult to determine if multiple exemplar training or VM interventions would have an impact during authentic social situations as skills/behaviors were measured in controlled contexts (e.g., computer scenarios, interactions with research assistants).
Quality indicators: All primary quality indicators were in the acceptable to high range in one study (State & Kern, 2012), suggesting the intervention was explained with replicable precision, and the intervention had a notable impact on social behavior. In one study (Lovett & Rehfeldt, 2014), fewer than three data points were provided for 2 of the 3 participants, and baseline data indicated that participants met mastery criteria (>80%) for at least one target skill prior to intervention; however, participants did show improvement on measures not mastered in baseline, and demonstrated some evidence of generalization to video-based scenarios with greater improvement following the addition of a prompt indicating the emotion felt by the character. Also, 2 of 3 participants showed improvement on one of two standardized, distant measures. In the third study, visual analysis indicated overlap with baseline and a minimal shift in level or trend. In addition, the authors included an ES estimate that suggested variability ranging from small to large effects (Mason et al., 2012). All studies demonstrated reliability by calculating percentage agreement or including a kappa coefficient (Mason et al., 2012). In one study (State & Kern, 2012), interobserver agreement was above 80% for all primary measures (e.g., inappropriate social interactions, noises) but below 80% (i.e., 72%) for a secondary dependent measure (i.e., appropriate social interactions). In no studies was intervention fidelity monitored, or were coders blind to the phase or condition. Maintenance data were collected in one study and was mixed (Mason et al., 2012), and evidence of social validity was provided in one study (State & Kern, 2012).
Naturalistic Interventions
Four SCRD studies (29%) incorporated naturalistic interventions to enhance the socialization of youth and adults with ASD (L. K. Koegel, Ashbaugh, Koegel, Detar, & Regester, 2013; R. L. Koegel et al., 2012; R. Koegel, Kim, Koegel, & Schwartzman, 2013; Schmidt & Stichter, 2012). A total of 16 (41%) adolescents/adults with ASD participated in these studies. Participants ranged in age from 13 to 23 (M = 16). Participants attended middle/junior high school (R. L. Koegel et al., 2012; Schmidt & Stichter, 2012), high school (R. Koegel et al., 2013), or college (L. K. Koegel et al., 2013), and all intervention activities occurred in their typical educational/community settings. Participants were reported to perform in the average to below average range academically. Social skills were described for 13 (81%) participants, and 3 (19%) were referred to the intervention because of concerns related to social skills. Descriptions indicated that participants rarely engaged in social interaction/were socially isolated (n = 10; 63%), or were socially rejected by peers (n = 3; 19%). All studies sought to improve the social interactions between adolescents/adults with ASD and their peers (e.g., initiations, responses, social engagement). Interventions ranged in duration from 5 to 33 sessions.
Three studies emphasized participation in social activities/events with same age typically developing peers. In two studies, learners with ASD in Grades 6 through 11 engaged in club activities that incorporated their preferred interests (R. L. Koegel et al., 2012, R. Koegel et al., 2013). If an existing club was not available, a club was created (e.g., movie trivia; comic books). All students were welcome to attend the club, and teachers advertised the clubs by posting flyers around the school to recruit members. In a third study (L. K. Koegel et al., 2013), university students with ASD participated in a structured social planning intervention led by a clinician at the University autism center. Initially, the clinician assessed student interests and dislikes and created a menu of community and university social activities consistent with the participant’s interests (e.g., math club, fitness classes). Participants were asked to select at least one activity from the menu to attend weekly. During weekly sessions, participants brought their planners and cell phones to schedule weekly events as well as program directions and contact information in their phones. In these meetings, participants discussed social skills appropriate for engaging in adult social/group activities (e.g., inviting peers to attend events, asking questions about others interests, ending conversations). Time was also dedicated to any topics that participants brought to the meeting. Support was faded as the participants began to search for social activities independently.
The fourth study (Schmidt & Stichter, 2012) was designed to promote the generalization of skills taught in a social skill curriculum by incorporating a peer-mediated intervention. Initially participating students with ASD received social skills instruction on sharing ideas, turn taking, problem solving, determining emotion/feelings of self and others, and identifying and interpreting facial expressions. Following direct social skills instruction, three typically developing peers willing to participate, who attended school regularly, and demonstrated positive social skills and behavior were selected for peer-mediated instruction. Selected peers participated in six 40-minute sessions of direct instruction. Peer-mediated instruction occurred in two phases: (1) initiation and (2) proximity. Peers were taught using direct instruction methods (i.e., adult modeling, rehearsal, feedback, and positive reinforcement) to initiate and maintain an interaction. Specifically, peers were taught to sit next to or across from their peer buddy, gain their peer buddy’s attention, initiate a conversation, and respond to their peer buddy during lunch. In the peer-mediated proximity phase, peers were instructed to sit next to or across from their buddy, not to initiate, and if their buddy initiated to respond with only one conversational turn. Throughout the peer-mediated intervention, the peers participated in six 40-minute check-up sessions for feedback and to ask questions. The focal student and peers were monitored during 10-minutes of their lunch period.
Naturalistic studies led to increases in social engagement or interaction with typically developing peers (ES = 0.75 to 0.98; R. L. Koegel et al., 2012; R. Koegel et al., 2013; Schmidt & Stichter, 2012), frequency/rate of social initiations (ES = 0.90 to 0.98; R. L. Koegel et al., 2012; R. Koegel et al., 2013), number of social activities (ES = 0.97; L. K. Koegel et al., 2013), and higher ratings of social satisfaction (L. K. Koegel et al., 2013). In addition, participants in one study noted an impact on quality of life indicators such as hosting a party, moving out of parent’s home and into an apartment, dating, initiating study sessions, and increased GPA (R. Koegel et al., 2013). Also, peer-mediated instruction evidenced greater generalization of target social skills to lunch and math settings than direct social skill instruction alone (Schmidt & Stichter, 2012). In this study, all three participants experienced greater gains in responding (ES = 0.75) than in maintaining an interaction (ES = 0.35) and initiating an interaction (-.05), and the authors speculated that this may reflect the emphasis on peers initiations, which may create greater opportunities for learners with ASD to respond.
Quality indicators: All primary indicators were rated in the acceptable or high range. Two studies provided evidence of maintenance (L. K. Koegel et al., 2013; R. Koegel et al., 2013), and in one study two participants maintained increases from the peer-mediated condition to a peer proximity–only condition (Schmidt & Stichter, 2012). Evidence of generalization was demonstrated in two studies. In one study, 2 of 3 participants generalized their performance to a math setting that provided opportunities for group work (Schmidt & Stichter, 2012), and in one study participants had difficulty generalizing to new clubs that did not incorporate a preference or opportunities to engage with peers (e.g., computer; R. Koegel et al., 2013). Interobserver agreement was demonstrated through percentage agreement (>80%) across all studies, and with a kappa coefficient in one study (Schmidt & Stichter, 2012). Treatment integrity data were collected in one study, and showed a high rate of fidelity (Schmidt & Stichter, 2012). In no studies were raters blind to the phase or condition. In addition, all studies met the secondary indicator criteria for social validity, and some provided additional evidence of social validity (e.g., quality of life, socialization satisfaction).
Findings From Group Design Studies
In the 28 GD studies, 800 youth/adults with a documented ASD participated. Studies included as few as 6 and as many as 73 participants. Participants ranged in age from 10 to 65. Table 2 provides design components of each study reviewed along with ES estimates and a summary of findings. A series of salient themes and meaningful patterns have emerged from the data of group comparison studies. These themes and patterns outline the predominant goals and strategies included in intervention programs, frequently used and recently emerging trends in social skills training programs, and the potential issues in the current social skill intervention research.
Types of Intervention Goals
GD studies investigated a wide range of interventions. Similar to SCRD, many GD studies incorporated direct instruction techniques to promote knowledge learning and behavioral social skills practice. Yet mindfulness-based therapy, cognitive behavioral therapy (CBT), and the arrangement of supportive environments for social activity participation were also prevalent in the GD studies. Similar to SCRD naturalistic interventions, many GD intervention programs focused on employing environmental and natural reinforcement, (semi) structured social interaction experiences, and interactive agents (e.g., peers) or tools (technology-based learning systems) as natural stimuli and prompts.
In GD studies, the targeted social competencies varied. They can be categorized as follows: social problem and stress management (n = 7, 25%, ES = 0.22 to 3.46), socialization participation (n = 6, 21%, ES = −0.01 to 1.27), peer interaction and relationships (n = 6, 21%, ES = −0.01 to 3.23), emotion recognition (n = 5, 18%, ES = −0.95 to 1.14), and general social skills development (n = 5, 18%, ES = −0.75 to 1.25).
Social problem and stress management
More than reinforcing the behavioral practice of social interaction skills, interventions in seven GD studies focused on training cognitive strategies and social-cognitive skills for surfacing, analyzing, and managing daily life social problems and stresses. For example, Cashin, Browne, Bradbury, and Mulder (2013) examined the use of narrative therapy to help young people with high-functioning autism (HFA; n = 10) process social problems and reduce their psychological distress. The 5-hour narrative therapy requested participants to surface narratives of their perceived problems and then perform narrative-based problem analysis, interpretation, and solution identification. The study indicated a statistically significant reduction in the psychological distress measure from pre- to posttest (ES = 1.52), though the difference in life-management difficulty score and the reduction in cortisol (a distress indicator) did not reach statistical significance.
CBTs were employed in three other studies (i.e., McGillivray & Evert, 2014; Reaven, Blakeley-Smith, Leuthe, Moody, & Hepburn, 2012; White et al., 2013) to teach stress-coping strategies and social problem-solving skills to 86 participants with ASD ranging in age from 12 to 25 years. The CBT interventions were manualized, integrated multiple intervention components (e.g., direct instruction, discussion, modeling, feedback, and homework), and were delivered through group or dyad (adult-led group, peer-integrated group, or parent-child dyad) therapy sessions, and all lasted more than 9 weeks (with 1.5- to 2-hour weekly sessions). Surfacing common stressful situations, identifying common emotion and cognitive distortions, and then learning coping strategies, resources, and alternate ways of thinking were salient elements of those CBT programs. Reaven et al. (2012) also included adaptations to support learners with ASD, including token reinforcement and a visual structure of the intervention routine. The three studies reported the efficacy of CBT in reducing self- or parent-reported depression, social anxiety, and social responsiveness deficits (ES = 0.22 to 1.03), but the effect of CBT on anxiety lacked statistical evidence.
Three recent studies (i.e., Kiep, Spek, & Hoeben, 2015; Pahnke, Lundgren, Hursti, & Hirvikoski, 2014; Spek, van Ham, & Nyklíček, 2013) examined the efficacy of a mindfulness-based intervention on the stress management of people with HFA. One hundred and twenty youth and adults between the ages of 13 and 65 participated in these studies. Interventions sessions occurred for 40 to 60 minutes over 6 to 12 weeks in a clinic or school setting. The intervention typically integrated meditation, sensory- and movement-based exercises, discussion, and homework. All studies included some adaptations to mindfulness exercises specifically for individuals with ASD. For example, in two studies, exercises related to examining one’s own thoughts and the use of metaphors were omitted because of the information-processing difficulties associated with ASD (Kiep et al., 2015; Spek et al., 2013). Participants were expected to engage in daily mindfulness exercises for 12 to 60 minutes. Unanimously, these studies reported significant, positive effects of mindfulness treatment on decreasing symptoms of depression, anxiety, or perseveration (ES = 0.87 to 3.46) and promoting prosocial behavior (ES = 0.31) or positive affect (ES = 2.79). The variation in time spent engaged in mindfulness activities across settings suggests a need for more research related to intensity.
Socialization participation
Six social skill intervention studies aimed to improve quality of life by encouraging active socialization and participation in recreational programs or leisure activities. These studies examined semistructured recreational programs or self-chosen leisure activities that occur in natural, community-based settings and included 168 adolescents and adults with ASD. In two studies, programs were specifically designed to encourage participation of individuals with ASD in theatre or music activities (Corbett et al., 2014; Hillier, Greher, Poto, & Dougherty, 2012). Corbett et al. (2014) created a 2-week peer-mediated theatre summer camp in which peers were trained to support their peers with ASD in the participation of theatre games and preparation of a theatre performance. Hillier et al. (2012) developed a 6-week teacher-facilitated music-making program to reinforce active participation in and positive attitudes toward socialization. Both studies reported positive effects of the interventions: improved socialization, duration of interaction with peers, performance, and attitudes toward peers; improved self-esteem; and reduced anxiety (ES = 0.64 to 1.11).
In three studies, individuals with ASD participated in a variety of leisure or recreational activities that matched participants’ personal interests and choices. García-Villamisar and Dattilo (2010, 2011) examined a yearlong leisure program comprised of the activities of games, exercise, crafts, events, and club activities. Participants chose their preferred leisure activities. The studies reported significant and positive effects of the leisure program on quality of life, stress, and adaptive behavior (ES = 0.81, 1.20 and 1.27, respectively). Similarly, Palmen, Didden, and Korzilius (2011) studied the efficacy of a 6-month community leisure program developed to teach individuals with ASD to plan and arrange leisure activities of interest. Participants increased their leisure satisfaction and engagement and decreased their need for leisure supports after the program (ES = 0.77). Hesselmark, Plenty, and Bejerot (2014) compared the effectiveness of participation in a group-based recreational program against group-based CBT for 36 weeks. In the recreational activity intervention, an activity list (including activities pre-selected by participants and therapists) was created and each participant chose a weekly activity from the list. The authors reported that both recreational program and CBT interventions improved self-reported quality of life (ES = 0.17 and 0.42 respectively) but did not have a significant impact on measures of psychiatric symptoms or autism symptomology. Overall, there was no statistical difference between the two interventions.
Peer interaction and relationship
Another salient goal of the social skill interventions was to promote the peer interactions and peer relationships of youth and adults with ASD. In particular, a group of researchers at UCLA (Gantman, Kapp, Orenski, & Laugeson, 2012; Laugeson, Ellingsen, Sanderson, Tucci, & Bates, 2014; Laugeson, Frankel, Gantman, Dillon, & Mogil, 2012; Laugeson, Frankel, Mogil, & Dillon, 2009; Schohl et al., 2014) have conducted a series of studies examining a peer interaction program for adolescents and young adults called Program for the Education and Enrichment of Relational Skills (PEERS). Studies investigating PEERS included 209 individuals with ASD ranging in age from 12 to 23. Four studies occurred in a clinic setting and one in a self-contained school. The manualized, parent-assisted PEERS program lasted 12 to 14 weeks, and provided participants with direct instruction and rehearsal of social skills and strategies to build peer relationships. The program assigned peer interaction homework and encouraged the use of electronic social media for peer relationship building. All five studies reported the treatment group outperformed the comparison groups on measures of desirable social behavior (ES = 0.83 to 1.32), quality of peer play (ES = 0.71 to 1.15), social skills knowledge (ES = 1.88 to 3.23), and friendship quality (ES = −0.01 to .14), while also reducing feelings of loneliness (ES = 1.06), social anxiety (ES = 0.16 to 0.95), and other autism-related symptoms (ES = 0.72 to 1.40).
One study examined the effect of a group-based manualized social skills intervention, called SDARI (or Socio-Dramatic Affective-Relational Intervention), on the social skills of 17 adolescents (11–17) with ASD (Lerner et al., 2011). SDARI was implemented in a 6-week summer program and consisted of 29 5-hour sessions. SDARI focused on child–child relationship building, and included multiple cooperative play and gaming activities for reinforcing social interactions with peers. The authors reported that the intervention program increased social assertion (ES = 1.31) and reduced errors in nonverbal emotion recognition (ES = 1.14).
Emotion recognition
Emotion recognition is a key social skill targeted by five of the reviewed GD studies. Eighty-eight adolescents and adults participated in these studies, and reported ages ranged from 12 to 52. These intervention programs used direct instruction (Barnhill, Cook, Tebbenkamp, & Myles 2002) or computer-based direct instruction incorporating videos or pictures (Faja, Aylward, Bernier, & Dawson, 2008; Faja et al., 2012; Golan & Baron-Cohen, 2006) to teach emotion recognition. Direct instruction included role-play, modeling, discussion, and feedback. Most studies focused on facial expression recognition, and two studies (Barnhill et al., 2002; Golan & Baron-Cohen, 2006) also included the training of emotion recognition from voices. Among the five studies examining interventions for emotion recognition, three (Faja et al., 2012; Golan & Baron-Cohen, 2006; Lerner et al., 2011) reported a significant effect on the participants’ emotion recognition ability, including facial recognition, face memory task performance, emotion recognition scale result, and nonverbal emotion recognition in voice (ES = 0.37 to 1.14). The other two studies did not find significant results. In general, the mixed findings of the five studies made the effect of the treatment on emotion recognition unclear.
General social skills development
Five GD studies examined the interventions focused on teaching basic social skills and/or social cognition skills necessary for social competence development. For example, two studies by Schmidt, Stichter, Lierheimer, McGhee, and O’Connor (2011) and, Stichter et al. (2010) examined the effect of a comprehensive, 10-week social competence intervention program delivered twice a week on the emotion recognition, theory of mind, and executive functioning of 33 participating youth (ages 11–14) with ASD. The Social Competence Intervention specifically targets greeting/acknowledging others, recognizing facial expressions, sharing ideas, turn taking, recognizing feelings, and problem solving. The intervention embedded opportunities to build on and practice previously taught skills. Instructional procedures included metacognitive strategies, self-monitoring and self-regulation. Following participation in this comprehensive intervention, participants increased their scores on executive functioning (ES = 0.72 to 0.88) and emotion recognition (ES = 0.60 to 1.25) measures. The findings on the treatment effect on theory of mind (ES = 0.29 to 0.36) and social deficits reduction (ES = 0.83 to 1.25) were mixed, with one study showing nonsignificant results.
Similarly, McMahon, Vismara, and Solomon (2013) evaluated a 22-week social skills training program with 14 children and adolescents (ages 10–16; M = 13) that spent 1.5 hours a week practicing vocalizations and interacting with others. The program comprised direct instruction, game or playground play, joke telling, and homework. The study results were mixed, indicating participants increased their peer interactions (ES = 0.5), yet reduced adult interactions. Participant initiations (ES = −0.75) also decreased, but their responding (ES = 0.85) increased following the program. Tse, Strulovitch, Tagalakis, Meng, and Fombonne (2007) examined the effectiveness of a 12-week, group-based social skills training program addressing social communication skills (e.g., initiating and maintaining conversation, listening, negotiating, and responding to teasing) via group discussion, role-play, and games. The study reported the program decreased participants’ (n = 32, ages 13–18) social responsiveness deficit (ES = 0.39) and improved their social competence development (ES = 0.42).
P. Mitchell, Parsons, and Leonard (2007) studied the potential of using a virtual reality–supported simulation to teach social reasoning (e.g., deciding where to sit in a coffee shop or on a bus) to six teenagers with HFA. Participants significantly improved their ability to make a social judgment (e.g., where to sit) following the 6 weeks of the virtual simulation (ES = 1.1).
Intervention Strategies and Principal Learning Processes
Among the social skill interventions examined by the GD studies, different desirable learning processes underlie the selection and implementation of component strategies or activities. A reasonable attempt to classify the interventions reported is to delineate the nature of the learning processes performed by participants. The principal learning processes were (1) knowledge and procedure learning (n = 16, 57%), activated via direct instruction (e.g., modeling, feedback, role-play, discussion, or homework); (2) experiencing and observing (n = 8, 29%), via participating in social activities (e.g., a recreational or leisure program) or simulated social tasks (e.g., reasoning in a virtual reality–based café scenario); (3) strategy learning (n = 5, 18%), through mindfulness-based or cognitive behavioral exercises that aim to reinforce desirable patterns of thinking or behavior; and/or (4) interacting with a purposefully arranged environment (n = 3, 11%), by interacting with technology-integrated programmed prompts (e.g., video and audio recordings for emotion recognition) or assistive tools (e.g., a mobile application or software).
Of the 28 group comparison studies, more than 50% focused on teaching social knowledge and skills in a highly structured (or manualized), multicomponent, researcher-led instructional program. Differently, eight studies involved participants in experiential learning in a naturalistic, semistructured recreational program (Corbett et al., 2014; García-Villamisar & Dattilo, 2010, 2011; Hesselmark et al., 2014; Hillier et al., 2012; Lerner et al., 2011; Palmen et al., 2011) or a simulation-based role-play scenario (P. Mitchell et al., 2007). Five other studies taught cognitive strategies through mindfulness-based exercises or CBT (e.g., Cashin et al., 2013; Kiep et al., 2015; McGillivray & Evert, 2014; Pahnke et al., 2014; Spek et al., 2013). The three remaining studies presented programmed prompts, such as computerized visuals or a multimedia courseware, to activate the practice of emotion recognition (Faja et al., 2008; Faja et al., 2012; Golan & Baron-Cohen, 2006).
Almost all GD studies employed adult-led interventions that were usually facilitated by researchers or school teachers. Only one study (Corbett et al., 2014) examined a peer-mediated intervention program, and four (García-Villamisar & Dattilo, 2010, 2011; P. Mitchell et al., 2007; Palmen et al., 2011) used interactive, environmental arrangements (e.g., an educational software or a computer simulation) to facilitate the practice of desirable social skills.
Technology was integrated in 12 studies. The technology implemented included computer-assisted multimedia and video (Corbett et al., 2014; Faja et al., 2008; Faja et al., 2012; Golan & Baron-Cohen, 2006; Pahnke et al., 2014), social media for electronic communication (Gantman et al., 2012; Laugeson et al., 2009; Laugeson et al., 2012; Laugeson et al., 2014; Schohl et al., 2014), a digital authoring tool (Hillier et al., 2012), and virtual reality–based simulation (P. Mitchell et al., 2007).
Research Methodology and Quality
The GD studies adopted RCT experiments (n = 14), within-subjects pre–post designs (n = 12), and quasi-experimental comparison GDs (n = 2). Among the RCT studies, seven (50%) had a total sample size smaller than 30 and hence may have lacked statistical power. Among the within-subject design studies, six (around 67%) had a sample size smaller than 15. Most studies evaluated a broad variety of intervention outcomes, including outcomes that were specifically targeted or aligned with the intervention or social competence/skills. Only six studies collected follow-up measures or transferred skill performance (Hesselmark et al., 2014; Kiep et al., 2015; Laugeson et al., 2012; McGillivray & Evert, 2014; Pahnke et al., 2014; Schmidt et al., 2011). Notably, almost all interventions in the GD studies lasted 4 or more weeks, with one to two sessions per week. Hpwever, very few (e.g., Gantman et al., 2012; Lerner et al., 2011) included and reported the treatment integrity data.
Discussion
Overall Findings
The current review of social skills interventions designed for youth and adults with ASD indicated a sturdy and rising trend of empirical research addressing social skill/knowledge, peer interaction, and participation in social activities. Overall, studies employing SCRD were effective in promoting social behavior of youth and adults with ASD, with all studies reporting at least some gains. Similarly, group comparison studies generally provided statistical evidence supporting the effectiveness of the interventions in promoting diverse social learning outcomes except for the four studies that examined emotion recognition.
Although SCRD studies addressed social interaction or conversational skills, many measured target behaviors in contrived contexts such as role-play scenarios, or a one-to-one interaction in a controlled setting (n = 6; 43%), making it difficult to determine the extent to which participants were able to effectively interact with others in their everyday activities. The remaining 8 (57%) SCRD studies measured social interactions or conversations in natural, authentic contexts (e.g., game play, lunch, group work, living room in a group home). The aforementioned observation was not obvious with GD studies, which usually used more global, standardized measures of social competence, quality of life, stress/anxiety, and autism symptomology. Yet many of these studies failed to provide solid theoretical underpinnings or design rationales explaining the association between the intervention and the change of social skills or behaviors that were not directly related to the intervention.
Difficulty in Classifying and Comparing Interventions and Social Skill Outcomes
There are several problems associated with classifying, synthesizing, or replicating social-skill interventions for youth/adults with ASD: (1) lack of a rich or comprehensive description of specific instructional procedures and learning activities in the intervention, (2) inconsistent operationalized definitions and terms for the intervention strategies (e.g., modeling, feedback) employed, and (3) lack of theoretical underpinnings for the intervention activities in comparison with other social skill programs. Most interventions claimed to involve multiple instructional strategies, such as modeling, reinforcement, role-play, discussion, feedback, and homework; yet few studies fully explained methods used to implement each component or the theoretical rationale for their use.
In fact, generally, few of the reviewed articles included an explicit discussion of the epistemological perspectives or cognitive theories of ASD that guided the development of their interventions. A number of the interventions reviewed may potentially leverage benefits related to social cognition, such as theory of mind (Premack & Woodruff, 1978) and executive functioning (Denckla, 1996; Rajendran & Mitchell, 2007). Yet, because the reviewed studies focused on reporting their research process and results without providing a clear theoretical basis or rationale, such a hypothesis lacks evidence.
In addition, the lack of a common definition for social skills made it difficult to make comparisons across studies (Laugeson & Ellingsen, 2014; Rao, Beidel, & Murray, 2008). Studies included direct observational measures as well as rating scales completed by caregivers, teachers, and/or self. To limit subjectivity, it has been suggested to pair standardized norm-referenced measures with direct observational measures conducted by raters who are blind to the study conditions (Laugeson & Ellingsen, 2014). In particular, SCRD studies incorporated direct observation measures, but raters were not blind to the study conditions, and these measures often only assessed discrete social skills without measuring broad social outcomes (e.g., friendship, employment, social problem solving) associated with adult social roles. The intervention programs in GD studies were frequently evaluated using multiple scales that may or may not be aligned with the designated or primary intervention objectives, and the authors failed to provide an explicit rationale for the inclusion of each and varied outcome measure. Among a large battery of measured direct and indirect outcomes, positive findings were discovered on certain scales and not others, with no solid explanation explaining the discrepancies. Such an observation makes us concerned about the purposefulness in the study design and external validity of the findings, especially with the studies reporting significant findings on only secondary or superlative outcomes.
In general, the intervention studies are mainly conducted following a product evaluation approach. Studies often focused on evaluating the effectiveness of a local application or a single curriculum. This approach makes it difficult to compare and synthesize interventions targeting similar learning goals or learner groups, and to extract findings that may contribute to the design heuristics and theoretical insights that will guide future intervention development. In other words, few studies contributed to the knowledge base that is required for the development and understanding of a “genre” of social intervention programs or tools. Because of the broad array of interventions and target behaviors, it is difficult to determine which interventions are most appropriate for teaching social skills to youth and adults with ASD, and under what circumstances or conditions these interventions are most effectivw.
Problems With Generalization
Single-case studies measuring generalization reported mixed findings. Difficulty generalizing skills to new, novel, or even natural contexts was noted in direct instruction, technology-based, and naturalistic studies, suggesting a need to intentionally plan for generalization by incorporating practice of newly developed social skills in novel or authentic contexts. That is, many adolescents and adults with ASD may require instruction and practice in multiple, natural environments or contexts with family members, coworkers, peers, and others they are likely to interact with in those contexts. For example, PEERS incorporates several opportunities to practice target skills in multiple contexts through homework activities, and has also included caregivers and peers as intervention agents (Laugeson & Ellingsen, 2014). Other studies have incorporated peer-mediated instruction to help adolescents with ASD transfer skills learned in social skill group lessons to lunch and academic settings (e.g., Schmidt & Stichter, 2012).
Few group comparison studies collected follow-up data to measure the learning outcome maintenance or transfer. The promising findings frequently reported by the reviewed studies should be interpreted and generalized with caution. Mixed results with multiple, discrete measures from a single study made it difficult to judge the overall learning effectiveness of the intervention or determine the association between intervention program components and specific learning processes or outcomes to be reinforced. Social skill intervention programs were typically treated as a package lacking sufficient explanation of or rationale for the research design and/or salient program features (e.g., the selection, assignment, and dosage of specific component activities). For example, the description of the manualized instruction typically involved only a list of the activity elements, such as modeling, feedback, role-play, discussion, and/or homework, without explaining how these strategies were delivered or sequenced. The functional or operationalized definition and procedures of each component activity across interventions or studies may exist in the cited curriculum but were not well explained in the published papers.
Fidelity data were rarely collected. Thus, it is difficult to judge the implications of the study findings if it is unclear whether the intervention was implemented the way it was intended, which is especially salient when individuals beyond the intervention creators are the implementers including peers, caregivers, and educators.
Agency and Intensity
Some studies incorporated choice in the intervention including participant preferences in social events, activities, and communication partners. As individuals with ASD enter adolescence, they report lower rates of self-determination, satisfaction, and autonomy (Wehman et al., 2014). Incorporating opportunities for choice in interventions is an important element for not only motivation and engagement but also personal agency.
Although duration of the intervention was frequently reported, the intensity of the intervention was often less clear. Intensity can be defined in terms of dose to include the number of quality instructional episodes, the frequency of instructional episodes, as well as the total duration of instruction (Warren, Fey, & Yoder, 2007). It is unclear how many opportunities were often provided to practice target social skills, as well as the quality of each learning opportunity.
In spite of the diversity in the objectives, targeted forms of learning, intervention strategies, and tools used in GD studies, the intervention programs that produced promising results generally included one or more of the following components: structure (or manualized), longer duration (e.g., 8+ weeks), respected learner agency (i.e., self-chosen recreational activities), natural social communication partners (i.e., parents, peers), an emphasis on authentic and contextualized learning, and purposefully planning for generalization and transfer of newly learned skills.
Future Research
There is an emerging trend in integrating technology into social skills interventions. The computer-assisted intervention materials included video and interactive multimedia, and more recently social media, a collaborative authoring tool (e.g., music making application), and a virtual learning system (e.g., virtual reality). Yet technology-supported social skill interventions are still limited, and the research on whether and how technological environments enhance social participation and learning is still lacking.
No studies addressed social skill instruction in employment settings. This is unfortunate as challenges related to social skills place individuals with ASD at a particular disadvantage in work settings (Schall et al., 2014). In addition, few studies targeted interactions that are specific to postsecondary educational settings such as universities. There is a need for more studies that address the social skills of youth and adults with ASD in postsecondary settings as the number of young adults with ASD attending universities or other postsecondary programs is expanding (Volkmar et al., 2014). These settings require a different set of social skills to navigate courses taught by faculty likely unfamiliar with ASD, interactions with roommates or housemates, and so on.
Deficits in social communication, cognition and social skills extend to adulthood and present a new set of challenges as relationships become increasingly complex (Laugeson & Ellingsen, 2014). Individuals with ASD described as high-functioning often desire relationships with others and are aware of their difficulties interacting with others, which can place this population at greater risk for depression and/or anxiety (Laugeson & Ellingsen, 2014). Yet studies providing evidence of changes in the quality of friendships/relationships are still limited. There is also a lack of studies that feature peers and parents as the leading facilitators of the social interventions for youth or adults with ASD.
Particularly rare is empirical research that examines the salient design features of an intervention program, by extracting the functional relationship between program features, implementation contexts, and specific learning outcomes. It may be meaningless to borrow a static program and use it in a novel and potentially mismatched context to achieve program replication. The field needs research to identify salient program features in relation to learner characteristics and contextual demands, which will help both educators and practitioners make informed decisions in selecting, transferring, and customizing an existing intervention program or application for varied contexts and learner groups.
Footnotes
Authors
FENGFENG KE is an associate professor of the Educational Psychology and Learning Systems Department at Florida State University’s College of Education. Her main research interests include inclusive design of e-learning, computer-supported collaborative learning, and digital game-based learning for diverse learners.
KELLY WHALON is an associate professor at Florida State University’s College of Education. Her research interests focus on intervention strategies designed to improve the academic achievement and communication skills of children with autism spectrum disorder (ASD).
JOONMO YUN is a doctoral candidate of the Program in Special Education at Florida State University’s College of Education. His research interests focus on special education, literacy education, quantitative methods in social science, and educational assessment.