Authors: Rachael A. Muscatello (1Vanderbilt University Medical Center, Department of Psychiatry and Behavioral Sciences, Nashville, TN, USA; 2Vanderbilt University Medical Center, Vanderbilt Kennedy Center, Nashville, TN, USA), Mark E. Klemencic (1Vanderbilt University Medical Center, Department of Psychiatry and Behavioral Sciences, Nashville, TN, USA; †Icahn School of Medicine at Mount Sinai, Department of Medicine, New York, NY, USA), Rachel Calvosa (2Vanderbilt University Medical Center, Vanderbilt Kennedy Center, Nashville, TN, USA; ‡Temple University, Department of School Psychology, Philadelphia, PA, USA), Blythe A. Corbett (1Vanderbilt University Medical Center, Department of Psychiatry and Behavioral Sciences, Nashville, TN, USA; 2Vanderbilt University Medical Center, Vanderbilt Kennedy Center, Nashville, TN, USA; 3Vanderbilt University, Department of Psychology, Nashville, TN, USA)
Categories: Article, autism, social behavior, intervention, social adaptation, depression
Source: Journal of autism and developmental disorders
Authors: Rachael A. Muscatello, Mark E. Klemencic, Rachel Calvosa, Blythe A. Corbett
Psychosocial interventions have been designed for autism spectrum disorder (ASD), but most focus on youth. Adaptation to social context was examined before and after a pilot randomized control trial of a theatre-based intervention for autistic adults. Depressive symptom impact on treatment outcomes was also investigated.
Sixty-one adults, ages 18–40 years with ASD, randomized to an experimental (EXP, n = 29) or waitlist control (WLC, n = 32) group, participated in a social interaction paradigm (Contextual Assessment of Social Skills; CASS) pre- and post-intervention. Depressive symptoms (Beck Depression Inventory) were examined as a predictor of behavioral change on the CASS post-intervention.
Before the intervention, adults in the EXP and WLC group had significant behavior adaptation between the Interested and Bored CASS conditions (corrected p < .01). Analysis of covariance of a Social Adaptation Index, controlling for pretest scores, was not significant for group (EXP vs WLC) (F(1,36) = 0.398, p = .532, ηp^2^ = 0.011) or sex (F(1,36) = 0.160, p = .691, ηp^2^ = 0.004). Depression and perception of peer behavior were correlated at pretest (r = −.277, p = .032), but pre-intervention depressive symptoms were not predictive of post-intervention behavior, nor was there a depression-by-group interaction (p > .05).
The intervention did not significantly impact social adaptation as measured by behavioral changes across social interaction contexts; however, autistic adults were already able to differentiate social conditions prior to the intervention. The CASS may not be optimal for broadly measuring treatment effects in adults. Paradigms are needed to more comprehensively measure behaviors that may be enhanced following social skills interventions.
The trial was pre-registered with ClinicalTrials.gov (Identifier: NCT04349644)
Social connection and interaction are inherent aspects of human experience; however, for some, these encounters may be more challenging or stress- and anxiety-provoking. Autism spectrum disorder (ASD)^1^ is a neurodevelopmental condition characterized by difficulties in social communication and interaction (American Psychiatric Association, 2013). Thus, significant research has been devoted to understanding how and why social communication may be more challenging for autistic individuals, as well as focusing on strategies and ways to support and enhance social connectedness across the lifespan. This has culminated in the development of numerous psychosocial programs for autistic individuals (e.g., Ke et al., 2018 for review); however, the majority of these interventions have targeted younger, school-aged children and adolescents. There is a scarcity of programs specifically for autistic adults (e.g., Spain & Blainey, 2015 for review), thereby highlighting the need to understand the unique needs of adults on the spectrum and the need for effective strategies to promote positive, fulfilling social relationships.
As the diagnostic prevalence of autism increases, so too does the number of autistic individuals who reach adulthood each year. Currently, an estimated 700,000 autistic individuals will turn 18 years old over the next 10 years (Shattuck et al., 2020), while the number of young adults receiving first-time autism diagnoses rapidly increases as well (Grosvenor et al., 2024). Therefore, understanding the unique social needs and demands of autistic adults is of critical importance. It is also relevant to acknowledge the significant heterogeneity of symptom profiles on the spectrum, warranting thoughtful consideration in terms of treatment planning.
Research demonstrates that outcomes can vary widely across individuals in terms of close friendships, romantic relationships, employment, and education (Mason et al., 2021; Steinhausen et al., 2016; van Heijst & Geurts, 2015; Yew et al., 2023). Social competence challenges can persist into adulthood (Clarke & Lord, 2024; Seltzer et al., 2004) and may negatively impact desirable outcomes (Clarke & Lord, 2024). Yet the amount of supports available to individuals drastically decreases when one turns 18 years, especially after secondary school (Shattuck et al., 2011). Without adequate support, many adults with ASD can become increasingly isolated from their communities and be at greater risk of experiencing significant loneliness and poor mental health outcomes, such as anxiety or depression (Grace et al., 2022; Myers et al., 2015; Taylor et al., 2017).
Throughout the lifespan, loneliness and isolation are commonly reported amongst autistic individuals (e.g., Grace et al., 2022; Schiltz et al., 2024), with frequently reported occurrences of peer rejection (Feldman, 2022), bullying (Maiano et al., 2016; Sterzing et al., 2012; van Roekel et al., 2010), and more restricted peer support networks (van Asselt-Goverts et al., 2015). There is a well-established relationship between loneliness and increased risk for depression in both nationally-representative (Cacioppo et al., 2006) and autistic (Grace et al., 2022), populations. In turn, social withdrawal can be both symptomatic of and/or contribute to depressive symptoms (American Psychiatric Association, 2013). A recent meta-analysis found a 12.3% pooled prevalence rate of depression in ASD, with some estimates as high as 50% lifetime prevalence (Hudson et al., 2019). Others report a 2.9 times greater risk for an autistic adult to be diagnosed with depression compared to nonautistic peers (Croen et al., 2015). Heightened depression can further lead to additional social withdrawal, and fewer interactions with one’s peers may exacerbate loneliness - contributing to a cycle of social isolation, lack of peer support, and depression. Further, in research of autistic youth, adaptive, prosocial skills (e.g., initiating and maintaining friendships, showing empathy, following social norms) were directly correlated with lower depression scores (Pascoe et al., 2023). Importantly, this was distinct from social communication difficulties core to autism (e.g., poor eye contact, limited reciprocal interaction, etc.), which were not associated with depression. As such, elucidating the relationship between depressive symptoms and social behavior in autistic adults is warranted, as is investigation of the extent to which outcomes of peer-based social competence interventions may be impacted by one’s psychiatric profile.
Objective and ecologically-valid measures are necessary to observe and quantify differences in behavior within a naturalistic environment, allowing researchers to determine how social functioning differs in autistic adults and whether various interventions effectively modify observed behavior. The Contextual Assessment of Social Skills (CASS; Ratto et al., 2011) was developed by Dr. Ratto and colleagues to serve as a reliable indicator of peer interaction behavior. Specifically, the CASS consists of two role-play scenarios with trained research confederates. In one role play, the trained peer demonstrates social interest and positive feedback (Interested condition), while in the other, the confederate acts bored and disinterested (Bored condition). The CASS allows investigators to observe an important aspect of conversational skills- the ability to recognize and respond to nonverbal social cues, such as body language, facial expression, or tone of voice. For example, in a conversation with someone who expresses boredom through lack of eye contact or short, disinterested responses, appropriate social adaptation would typically drive one to modify their behavior by increasing the number of questions asked or topic changes in an effort to keep the conversation going (Burgoon, 1995). In this way, recognizing the emotion of another and adapting one’s behavior accordingly can enhance conversational quality and rapport. In the CASS, comparisons of behavior when interacting with an interested versus bored peer thereby represent social adaptability or the extent to which an individual adapts behavior to changing social context.
In the initial study of twenty adolescents and young adults with ASD and twenty without (Ratto et al., 2011), three out of the four primary outcomes (Asking Questions, Topic Changes, Quality of Rapport) significantly differed based on context (Bored vs. Interested) and diagnosis (ASD vs. Nonautistic Controls), while the fourth primary outcome (Overall Involvement) was higher in controls but did not differ by context. Overall, the study showed that autistic adolescents and adults do show differences in behavior based on context; however, these changes or adaptation to context are on a reduced scale relative to nonautistic peers (Ratto et al., 2011). Additional research in school-aged children and adolescents with ASD similarly found differences in behavior between Interested and Bored conditions, though significant findings were limited to asking more questions in the Bored condition and higher quality of rapport in the Interested (Simmons et al., 2021). Nevertheless, both youth and adults can identify differences in the confederate’s behavior, correctly rating confederates as more bored and less engaged in the Bored condition compared to the Interested, although the level of change between contexts is less than that seen in nonautistic adolescents and adults (Ratto et al., 2011; Simmons et al., 2021). Finally, there is some early evidence of biological sex-based differences in behavior on the CASS, with one recent study of autistic children and early adolescents citing behavioral differences in male and female participants, especially in terms of greater Vocal Expressiveness in female participants (Corbett et al., 2021).
Few studies have examined whether the CASS is sensitive to change following a social skills intervention. An earlier pilot study examined change in CASS ratings in five young adults with autism without intellectual disability following a social skills intervention (White et al., 2015). The study focused specifically on the Interested CASS condition, finding that four of the five participants demonstrated more conversational involvement after the intervention, while other behaviors such as asking questions, topic changes, and rapport changed in some, but not all, individuals. Another study in Latinx adolescents reported improvements in observed behavior, though the changes did not reach the level of statistical significance (Veytsman et al., 2023). Most recently, a large multisite randomized control trial in autistic youth examined behavioral changes in the CASS Interested following a peer-mediated intervention, finding no differences between the experimental and control conditions on three of the behavioral domains (Vocal Expressiveness, Rapport, and Social Anxiety). However, there was an observed indirect effect where improvements in face memory for the experimental group led to changes in behavior on the CASS at a two month follow-up (Corbett et al., 2023). Though findings are somewhat variable, the literature to this point cites some differences in behavior on the CASS in autistic versus non-autistic individuals and in response to interventions, supporting its utility as an objective measure of social interaction behavior.
The number of psychosocial interventions specifically designed for autistic adults continues to be lacking compared to those for children and adolescents (see Pallathra et al., 2019; Schweizer et al., 2023 for recent literature reviews). To address the need for additional randomized control trials (RCTs) in autistic adults, a recent RCT by our team examined the efficacy of a peer-mediated, theatre-based intervention, finding that adults in the theatre program had significant post-test gains on several social measures including an event-related potential (ERP) measure of incidental face memory and self-report social communication and motivation (Corbett et al., 2025). While the study examined social behavior, the primary outcomes for the RCT were limited to two subscales (vocal expressiveness, quality of rapport) of the CASS-Interested. For these domains, no significant changes were observed for those participating in the SENSE Theatre^®^ intervention compared to the control group. However, several studies of the intervention in youth have reported notable behavioral changes, such as increased group or cooperative play (Corbett et al., 2010; Corbett et al., 2016; Corbett et al., 2014) or improved vocal expressiveness and rapport on the CASS (Corbett et al., 2023) with novel peers. Further, the initial analysis of SENSE Theatre^®^ for adults did not examine mental health outcomes and their potential influence on treatment efficacy.
The goal of the current study was thus two-fold. First, the study aimed to expand upon previous investigation of social behavior changes following the SENSE Theatre^®^ intervention for adults (Corbett et al., 2025) to more comprehensively examine whether the intervention was efficacious in modifying social behavior and awareness. Specifically, the study compared observable behavior and self-reported perceptions of interaction quality across two social conditions (Interested and Bored) before and after the theatre-based intervention (Aim 1). At pre-intervention testing, it was hypothesized that participants would demonstrate observable behavior change on the nine CASS domains between the Interested versus Bored contexts (Hyp 1.1a), which is consistent with results of the original study of the CASS (Ratto et al., 2011). We further predicted that participants would accurately detect a difference in confederate behavior between the Interested and Bored contexts (Hyp 1.1b). At post-testing following the theatre intervention, we predicted the experimental (EXP) group would show greater behavioral adaptation from the Interested to Bored context (Hyp 1.2), demonstrating an enhanced ability to adapt to differences in one’s behavior in social settings. As Ratto and colleagues (2011) found that autistic adults were less perceptive of changes in peer behavior in the Interested vs. Bored conditions relative to nonautistic controls, we expected the EXP group would show an increased awareness of change in peer behavior (Hyp 1.3) following the intervention, compared to adults in the Waitlist Control (WLC) group.
Aim 2 of the study sought to examine the impact of mental health outcomes, namely, depression, on treatment outcomes in this population. We expected elevated depressive symptoms to be associated with less social engagement with an interested peer before (Hyp 2.1) and after (Hyp 2.2) SENSE Theatre^®^. Because the recruited sample from the source RCT was strengthened by a nearly 1 ratio of male to female participants (Corbett et al., 2025), we also explored possible sex-based differences in behavior before and after the intervention.
The research was carried out in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki). The Vanderbilt University Medical Center Institutional Review Board approved the study. Informed written consent was obtained from adult participants and a Surrogate Rider consent was obtained for six legally conserved participants prior to inclusion in the study.
Adult participants, aged 18–40 years, were recruited from a broad community sample via targeted outreach to medical and health-related services, clinics, research registries, regional autism/disability organizations, schools, and social media platforms. All participants were required to have a confirmed diagnosis of ASD and full-scale intelligence quotient (IQ) above the cut-off for intellectual disability (IQ > 70). Those with a significant report of current uncontrolled aggression, based on phone screening and responses on the Adult Behavior Checklist (Achenbach and Rescorla, 2003), were excluded to ensure the safety of all participants, peers, and staff.
Detailed descriptions of sample recruitment, enrollment and attrition, including the study CONSORT diagram, are available in the initial manuscript describing the primary outcomes of the RCT (Corbett et al., 2025). In short, adults were randomly assigned to either an experimental (EXP) or waitlist control group (WLC) following informed consent. Of the total eligible sample of 64 adults, pre-intervention analyses included 61 adults (EXP = 29, WLC = 32), while post-intervention analyses were conducted on the 47 adults (EXP = 23, WLC = 24) who completed at least 7 of the 10 intervention sessions (see Corbett et al., 2025 for CONSORT). As shown in Table 1, the mean age was 24.51 years across the groups (WLC = 24.69 years, EXP = 24.3 years). The sex distribution was nearly evenly distributed with 30 (49.2%) female participants and 31 (50.8%) male participants. The racial and ethnic characterization of the sample was comprised of 9.8% Black, 75.4% White, 1.6% Asian, and 13.1% multiracial. There were 4.9% Hispanic and 95.1% non-Hispanic participants.
Participants were required to have a confirmed diagnosis of ASD based on DSM-5 criteria (American Psychiatric Association, 2013) and corroborated by the Autism Diagnostic Observation Schedule (ADOS-2; Lord et al., 2012). The ADOS-2 is a semi-structured interview and series of activities to assess behaviors indicative of autism, administered by research-reliable personnel. A Module 4 score consistent with ASD based upon the revised algorithm (Hus & Lord, 2014) was required for inclusion. If participants did not meet the cut-off criteria (score of 8), clinical judgement determined inclusion/exclusion. One female participant in the WLC group scored below the cut-off with a score of “6”; however, the participant was included based upon clinical judgement from a licensed clinical psychologist.
Inclusion criteria also included an IQ ≥ 70 as measured by the Wechsler Abbreviated Scale of Intelligence (WASI-II; Wechsler, 2011, a measure of cognitive ability used to obtain a quick, reasonable estimate of an adult’s intellectual functioning. The decision to exclude those with intellectual disability was based on previous research examining the efficacy of the treatment with youth that was also inclusive of individuals with IQ ≥ 70 (Corbett et al., 2016; Corbett et al., 2023).
The study also utilized an informant-based screening tool for ASD, the Social Communication Questionnaire-Lifetime Version (SCQ) (Rutter et al., 2003), completed by an adult with first-hand knowledge of the participant’s functioning during early childhood and/or at present. Scores of 15 or higher are highly suggestive of ASD; however, participants were not excluded based on SCQ score.
The Contextual Assessment of Social Skills (CASS) (Ratto et al., 2011) is a direct observation measure of social behavior and adaptation, explicitly developed for use with individuals with ASD. The CASS consists of two brief, 3-minute role-play conversations with two similar-age, opposite-sex, unfamiliar confederates. In the Interested condition (CASS-I), the research confederate is instructed to show interest and engagement through verbal and non-verbal behavior, while in the Bored condition (CASS-B), a different confederate acts bored and provides little direct eye contact or verbal responses. All confederates were adults trained by the research staff and compensated for their participation. Confederate training included observation of protocols, demonstration of protocols with a research team member to ensure competence, and review of a standardized document with suggested behaviors and questions. Many confederates were college students or early graduate students, making them slightly younger than the average age of study participants. Consistent with the original CASS validation study (Ratto et al., 2011), in all cases the CASS-I was completed prior to the CASS-B.
Behavior during the CASS is scored across two summed domains (Number of Questions Asked, Number of Topic Changes) and seven domains scored on a 7-point Likert scale (Vocal Expressiveness, Gestures, Positive Affect, Social Anxiety, Kinesic Arousal, Overall Quality of Rapport, and Overall Involvement). Higher scores indicate more socially adaptive behaviors (e.g., more expressiveness, engagement, fewer anxiety behaviors, etc.). A total score from Asking Questions, Topic Changes, Overall Involvement, and Quality of Rapport was calculated based on the original validation study (Ratto et al., 2011). A Social Adaptation Index (SAI; Ratto et al., 2011; White et al., 2015) was then calculated by subtracting the Interested total from the Bored total, where higher change scores were more indicative of social behavior adaptation in response to changing social context across the Interested and Bored conditions. Because normative social adaptation would incur a reduction in quality of rapport from Interested to Bored, the Quality of Rapport domain was reverse scored prior to calculation of the SAI so that an increase on all four items was indicative of normative social behavior adaptation. Behavioral coding raters were trained to reliability (Cohen’s κ ≥ 0.80) and were naïve to group assignment or assessment period when coding video recordings. Two of the study authors (MK and RC) oversaw behavioral coding and conducted routine fidelity checks to maintain coding calibration.
Participant perception of the confederate’s behavior was indexed using the Conversational Rating Score (CRS; Ratto et al., 2011). The CRS was completed after both the Interested and Bored conditions, and participants were asked to rate the research confederate’s interest based on five items using a 7-point Likert scale. Higher total scores indicate greater perceived social interest.
The Beck Depression Inventory (BDI-2; Beck et al., 1996) is a 21-question self-reported measure of depressive symptoms over a two-week time frame. Each item is rated on a scale of 0–3 and summed to create a total score, where higher scores indicate more elevated symptoms of depression. The widely-used measure demonstrates strong reliability (⍺ = 0.92 – 0.93; Beck et al., 1996) and construct validity (r = .93; Beck et al., 1996), including strong psychometric properties for use with autistic adults (Williams et al., 2021).
The SENSE Theatre^®^ intervention included ten 2-hour sessions, once per week, as well as an additional two days of dress rehearsals and a final performance. The intervention relies on three core tenants – highly trained peers, theatrical play, and a performance approach. SENSE Theatre^®^ has demonstrated efficacy in both children/adolescents (Corbett et al., 2011; Corbett et al., 2016; Corbett et al., 2014) and in adults (Corbett et al., 2025). A detailed description of the intervention, including modifications of the program to be suitable for adult participants, can be found in the published results of the RCTs primary outcomes (Corbett et al., 2025).
Following initial diagnostic and eligibility testing, enrolled participants were randomly assigned to either the Experimental (EXP) or Waitlist Control (WLC) group. Participants in both groups completed pre-intervention testing, including the CASS interaction paradigm and the BDI-2 survey described above. After pre-testing, the EXP group received the 10-week intervention, after which time both the EXP and WLC groups completed post-testing with repeated administration of the CASS and BDI-2. The WLC group received the SENSE Theatre^®^ intervention after the final follow-up visit.
It was first hypothesized that participants would demonstrate behavior differences between the interested versus bored contexts and would accurately detect a difference in confederate behavior between the two conditions (Hyp 1.1). To examine this hypothesis, differences in social behavior across conditions (Interested vs. Bored) were quantified using Wilcoxon Signed-Rank Test as all of the CASS outcome variables except for Bored Condition Asks Questions and Bored Condition Topic Changes were non-normally distributed. The nine outcome variables of the CASS, as well as participant rating of confederate behavior on the CRS, were examined. P-values were corrected for multiple comparisons using the Benjamini-Hochberg procedure (Benjamini & Hochberg, 1995), and effect size r was calculated according to (Rosenthal, 1994).
As described above, total scores based on standardized (Z-score) values of the four primary outcomes (Asking Questions, Topic Changes, Overall Involvement and Quality of Rapport) were calculated. The difference in total score between the Bored and Interested conditions provided a Social Adaptation Index (SAI) to quantify normative social behavior adaptation across conditions.
Following the theatre intervention, we predicted the experimental (EXP) group would show a greater change in SAI (Hyp 1.2), as well as increased awareness of changes in peer behavior (Hyp 1.3), compared to adults in the Waitlist Control (WLC) group. A series of analysis of covariance (ANCOVA) models examined group-based differences in post-intervention scores while controlling for pre-intervention values.
For Aim 2 examining depressive symptoms and behavior during the CASS-I condition, we expected depressive symptoms to be associated with less social engagement with an interested peer before (Hyp 2.1) and after (Hyp 2.2) SENSE Theatre^®^. Pearson bivariate correlations, or Spearman rho correlations in cases of non-normally distributed variables, assessed the relationship between BDI-2 scores and CASS Total score, CASS SAI score, CASS-Interested individual domains, and CRS confederate rating at pre-intervention. Further, we ran separate linear regression models with post-intervention CASS total, SAI, or CRS rating as outcome variables. Predictive variables were entered into the model in a stepwise fashion, starting with group (EXP vs. WLC), sex (Male vs. Female), and pre-intervention score for the outcome, followed by post-intervention BDI-2 total score, and finally a BDI-2 total score by group interaction.
Biological sex was included as a between-group factor in all analyses to explore whether there were significant differences in male versus female participants on the primary outcomes of interest.
Prior to analysis, means and standard deviations for primary behavioral and depression outcome variables were calculated by group (EXP and WLC) and condition (Interested and Bored). The descriptives are presented in Table 2.
Results from the paired-samples t-tests are displayed in Table 3. Following correction for multiple comparisons, it was found that autistic adults used more gestures, had more positive affect, less social anxiety, and more conversational rapport in the interested condition compared to the bored. In the bored condition, participants asked more questions and changed the topic more frequently (all corrected p < .001).
Additionally, participants were able to accurately identify a change in the behavior of the research confederate, with significantly higher ratings of conversational quality in the Interested compared to the Bored condition (see Table 3, corrected p < .001).
An ANCOVA examined the group- and sex-based differences CASS SAI at post-test, controlling for pre-test scores. There was no significant effect for group (F(1,36) = 0.398, p = .532, ηp^2^ = 0.011), nor was there a significant difference in CASS SAI post-test scores in male versus female participants (F(1,36) = 0.160, p = .691, ηp^2^ = 0.004).^2^ Similarly, the interaction effect for group and sex was not statistically significant (F(1,36) = 0.016, p = .900, ηp^2^ = 0.000).
An additional ANCOVA was conducted to investigate group- and sex-differences in participant rating of change in peer behavior from Interested to Bored conditions. There were no significant effects for group (F(1,42) = 0.106, p = .747, ηp^2^ = 0.003), sex (F(1,42) = 0.690, p = .411, ηp^2^ = 0.016) or group*sex interaction (F(1,42) = 0.300, p = .587, ηp^2^ = 0.007).
Correlation analyses revealed that the reported level of depressive symptoms on the BDI was not significantly associated with CASS Total Score in the Interested condition (r = .042, p = .752). Additionally, the BDI was not significantly correlated with any of the individual domains of the CASS or with social adaptation between Interested and Bored (SAI) (all p > .05). There was a significant correlation with depression and perceived peer behavior, such that those with greater depressive symptoms reported peer behavior as less engaged during the CASS Interested condition (r = −.277, p = .032). However, depression scores were not significantly associated with awareness of change in peer affect from Interested to Bored conditions (r = .189, p = .149).
Additional regression models assessed the extent to which depression symptoms prior to the intervention predicted changes in social behavior and awareness after the intervention. In all models, pre-intervention BDI total was not a significant predictor of post-intervention behavior, nor was there a significant depression by group interaction (all p > .05). Moreover, there was no statistically significant evidence to suggest that depressive symptoms prior to the intervention were associated with social behavior/awareness after the intervention, nor did the relationship vary by group (EXP vs. WLC) or sex (male vs. female). All regression model results are shown in Table 4.
There were concerns regarding ceiling effects, as the domain scores associated with the SAI, especially overall involvement and rapport, were left/negatively skewed at pre-testing (skewed towards higher/more normative scores). Therefore, subsequent exploratory analyses were conducted in participants with a pre-intervention SAI that scored below one standard deviation from the mean. In a sample of 10 participants (WLC = 6, EXP = 4), there was a large effect size for between-group improvement in SAI score, evidenced by improved social behavior adaptation in the EXP compared to WLC group (F(1, 9) = 2.43, p = .163, ηp^2^ = 0.258) following the peer-mediated intervention.
The current study examined changes in social behavior in response to different social interaction contexts as well as following the peer-mediated SENSE Theatre^®^ intervention for adults with ASD. The primary results showed that autistic adults demonstrated behavioral adaptation to context evidenced by differences in behavior based on a research confederate’s display of social interest or boredom. During the interested condition, the autistic adults engaged in more prosocial behavior including gestures, positive affect, and conversational rapport. They also evidenced fewer signs of social anxiety. These responses contrasted with the bored condition, in which autistic adults asked more questions, changed the topic more frequently and accurately observed less conversational interest by the peer. These behavioral responses were largely consistent with Ratto et al., (2011), who also reported some level of change across contexts (Interested vs. Bored) for asking questions, topic changes, and overall rapport in autistic adults. However, the overall sample in the current study demonstrated a trend-level difference in Overall Involvement from Interested to Bored conditions, whereas the autistic sample in Ratto and colleagues’ (2011) study did not show significant change over contexts.
Regarding hypothesized treatment effects, it was predicted that those receiving the intervention would demonstrate greater social awareness and adaptation to changes in the behavior of a novel research confederate. However, there was no change post-treatment in the social adaptation index, a measure of the extent to which someone alters their social behavior across changing social conditions (Interested vs. Bored). It is possible that since the participants already evidenced robust adaptation to the contexts prior to the intervention, there was limited opportunity for growth or change on this index. It is noteworthy to add that many of the participants anticipated the contrasting conditions on re-exposure to the paradigm such that at post-intervention statements revealed they expected to meet someone who was “interested” and another who was “bored.”
Future studies will likely benefit from a naturalistic, behavioral paradigm that allows for engagement with peers without anticipation of the precise elements of the encounter. For example, previous studies of the SENSE Theatre^®^ intervention in youth found robust change in cooperative play in a naturalistic playground paradigm (Corbett et al., 2019; Corbett et al., 2016; Corbett et al., 2014). A similar study utilizing the CASS in youth also found treatment effects for some behaviors on the CASS; however, these effects were mediated by brain-based changes as measured by event related potential (Corbett et al., 2023). It was further noted this effect was significant at the two-month follow up testing (Corbett et al., 2023). In contrast, the current study examined behavioral changes only immediately after the intervention. Thus, as with in youth (Corbett et al., 2023), it may be that brain-based changes to the intervention (Corbett et al., 2025) temporally precede observable behavioral changes, where notable difference in behavior would not emerge until later. It is also plausible that the treatment does not directly enhance engagement with a novel peer during brief encounters. Notably, exploratory analysis of those who scored at least one standard deviation below the mean at pre-intervention found a large effect size for improvement in social behavior adaptation following the intervention. In contrast to earlier studies (Ratto et al., 2011; White et al., 2015), the current study’s sample already showed a pre-treatment change in Overall Involvement, one of the primary outcome domains of the CASS, across contexts that approached statistical significance. Therefore, it is plausible that ceiling effects prevented notable change in social behavior within the larger group, and future studies will need to expand upon this hypothesis in a much larger sample with a broader range of social behavior strengths and challenges. Overall, future studies of intervention effects for adults may benefit from more naturalistic and unpredictable behavioral paradigms, testing across multiple time points post-intervention to better understand the time course of intervention response, and ensuring the study sample is inclusive of individuals with a wide range of autistic traits.
For the second aim, we sought to determine whether self-reported, pre-intervention depressive symptoms were associated with social behavior both before and after the intervention, which were largely not found. The only significant correlation was for participants endorsing elevated depressive symptoms; they rated peers as “less interested” during the Interested Condition of the CASS. This finding appears consistent with the profile of depressed individuals who often experience more negative interpretation of the behavior of others and higher rejection sensitivity (e.g., Ayduk, 2001; Normansell & Wisco, 2017). Nevertheless, there were no significant predictive effects of depression at pre-intervention with behavior at post-intervention, and it does not appear that depressive symptoms prior to the intervention influenced social behavioral change (or lack thereof) in response to the intervention. It is relevant to note that the peers were reliably trained to behave in a certain way (i.e., interested or bored during conversation) regardless of the behavior or level of engagement of the participant.
There has been some evidence for reduced depressive symptoms following social skills programs in adolescents (Parenteau et al., 2024; Schiltz et al., 2018; Yoo et al., 2014). For example, an RCT of the PEERS^®^ social skills intervention (Laugeson et al., 2012) reported significant reduction in self-reported depression symptoms as well as an association between fewer mood symptoms and increased social contact (Schiltz et al., 2018). These studies highlight the interconnection between social interaction and depressive functioning, emphasizing that improved peer interactions may also lead to benefits in mood. Our study sought to examine the reverse effect by determining whether affective symptoms prior to the intervention predicted post-intervention outcomes in social awareness and behavior adaptation. The lack of significant findings in our study may be informative, suggesting that one’s affective profile might not necessarily be predictive of their potential response to the treatment; however, future research in a larger sample of adults measuring a range of social outcomes (e.g., behavior, awareness, cognition, communication) is needed to potentially replicate our findings. It will be beneficial to compare treatment outcomes across a large range of affective profiles, including those with clinically confirmed diagnoses of depression, to more clearly elucidate the extent to which depressive symptoms may influence or predict one’s likelihood to benefit from the SENSE Theatre^®^ intervention.
Finally, due to the often-reported sex-based distinction in diagnostic and depression profiles, comparisons between male and female participants were explored. Nonetheless, there were no sex-based differences for depression, behavioral responses or associations with other variables of interest in this pilot RCT, suggesting no difference in treatment response on the included variables for male and female participants.
There are various potential explanations to the null findings in the study. First, it is very possible that the CASS does not reflect natural, real-world conversations suggested by the comments that some participants were aware of the approaching bored condition at post-testing. Indeed, other studies have solely utilized the Interested condition to examine change in social behavior in an effort to more closely represent typical, naturalistic conversations (Corbett et al., 2023; Dolan et al., 2016; van Pelt et al., 2020). Secondly, the SAI may not be optimal to be used as a reliable treatment outcome measure. Finally, it could be limitations of the intervention in modifying specific behavior. Since significant social behavior changes (e.g., social communication, motivation) have been reported, it does not appear to simply be a lack of efficacy of the treatment. With that said, it may suggest that the treatment does not directly enhance engagement with a novel peer during brief encounters.
Strengths of the study included a nearly even ratio between male and female autistic participants despite the common 1 male-to-female ratio (Maenner et al., 2023). The sample was also robustly characterized using strict diagnostic criteria, clinical judgement and gold standard measures. The study employed a naturalistic paradigm and considered the possible influence of depression symptomology in treatment outcome.
Despite these strengths, the sample was not fully representative based on cognitive, racial or ethnic profile with non-Hispanic, white participants with average intellect being over-represented. Therefore, the extent to which these findings generalize more broadly to others is undetermined. The study relied on self-reported depression symptoms, and there was no clinical determination of whether an individual met criteria for a diagnosis of major depressive disorder or similar condition, nor was there inclusion of information regarding other co-occurring diagnoses. Therefore, future studies would be strengthened by including clinical assessment of co-morbid psychiatric symptoms, while also considering different categories of symptoms within a diagnosis (e.g., emotional, physical, or cognitive symptoms of depression). Finally, confederate behavior was not independently coded during the task, and future studies could be enhanced by assessing observable change in confederate behavior, such as to examine whether participants’ depressive symptoms are associated with a peer’s social interaction behavior.
In summary, the intervention did not appear to significantly impact social adaptation as measured by perceived and behavioral changes across social interaction contexts. Importantly, the autistic adults were already able to differentiate social conditions prior to the intervention indicating that more naturalistic, complex, and unpredictable social paradigms may be needed to illustrate behavioral change following social skills interventions. It was also shown that self-endorsed depressive symptoms did not predict changes in behavior following the intervention. Despite these findings, prior research has found SENSE Theatre^®^ to be a promising intervention to enhance social competence (e.g., social motivation, incidental face memory) in adults with autism (Corbett et al., 2025). Although prior and current research did not find significant treatment effects in social behavior adaptation on the CASS paradigm in autistic adults with average to above average intelligence, future endeavors may be helpful in exploring a range of social skills and behaviors across several naturalistic contexts that can be enhanced following social skills interventions.