Abstract
Students with autism spectrum disorder (ASD) have marked deficits in social communication skills, which can challenge their ability to participate in academic and social activities. Music therapy is a popular intervention for children with ASD, and although research evidence supporting its effectiveness is growing, methodological flaws limit interpretation. The present study proposes a new measure, the DUACS, to assess outcomes associated with a group music therapy program, Voices Together. Participants were 64 elementary students from nine self-contained autism classrooms who attended 16 weekly intervention sessions. Teacher-reported behavior questionnaires and direct student behavioral measures were administered at baseline, during, and after treatment. Increases in communication and social skills suggested program effectiveness, particularly among students with higher baseline skills. Further investigation of this program approach will help to better understand its mechanisms and for whom it works best.
The prevalence of autism spectrum disorder (ASD) has increased significantly over the past three decades and is currently estimated to affect 1 in 59 school-aged children (Baio et al., 2018). For children with ASD, deficits in communication and social interactions present an ongoing challenge (Wan & Schlaug, 2010). Approximately 30% of individuals with ASD use minimal language or fail to ever acquire spoken language (Tager-Flusberg et al., 2009). Of those who do use speech, many display deficits in back-and-forth conversation and have difficulty using and understanding non-literal language (Tager-Flusberg et al., 2009). Improving these skills is important for building relationships, meaningfully participating in school and community, and ultimately achieving independence (LaGasse, 2014).
One way to help children with ASD improve social communication skills may be through music. Kanner’s (1943) earliest descriptions of ASD included observations of children with noticeable deficits in language and communication, alongside superior musical skills. In more recent years, scientists have garnered evidence supporting Kanner’s observations of enhanced memory for pitch (Heaton et al., 2001) and a stronger ability than typically developing peers to detect changes in melodies (Mottron et al., 2000). Neuroscientific studies have also documented differences in brain activation during music processing between people with ASD and typically developing individuals. For example, Sharda and colleagues (2014) found that the areas of the brain that process both speech and song were more effectively engaged during song than speech for individuals with ASD. No difference in fronto-temporal connectivity between individuals with ASD and typically developing controls was found when listening to sung language. However, when listening to spoken language, fronto-temporal connectivity was disrupted in individuals with ASD (Sharda et al., 2014).
These findings suggest that there are different fronto-temporal mechanisms for speech versus music processing among people with ASD. Therefore, music-based therapies focused on the development of speech may be effectively used to support the developmental of speech processing. The Sharda et al. (2014) findings indicate an impairment to the typical connection between speech and music could be leveraged to support the growth of language through music among individuals with ASD. As Sharda and colleagues (2014) state, While most of the research on non-savant ASD so far has focused on deficits, and what might not be functional in the brain of an individual with autism, the findings from this study are one of the first in a long line of research that might actually capitalize on the potential of the autistic brain to compensate for its losses. (p. 184)
Music therapy has been described as the evidence-based use of music to accomplish individual goals within a therapeutic relationship by a credentialed professional (American Music Therapy Association, 2017). In music therapy, practitioners use both musical experiences (e.g., singing, playing instruments, discussing music) and the relationships they develop with clients to facilitate communication and other social goals (Geretsegger et al., 2014; Peters, 2000). Music therapists utilize predictable session structure with defined goals and outcomes while also providing flexibility that helps participants learn how to respond to less predictable situations. For example, the combination of structured rhythm and dynamic interactions creates an environment of both predictability and anticipation that may help stimulate verbal skills in individuals with ASD (Wigram & Gold, 2006). As discussed in two systematic reviews, intervention studies have shown that music therapy consistently produced greater changes as compared with standard care (i.e., participants’ usual therapies and consultations) among children with ASD with respect to social interaction, verbal and non-verbal communication skills, social-emotional reciprocity, and challenging behavior (Geretsegger et al., 2014; James et al., 2015). Altogether, basic science and systematic reviews of multiple intervention studies lend support to the claim that music therapy is a promising approach for children with ASD (Sharda et al., 2014; Geretsegger et al., 2014).
However, questions about the effectiveness of music therapy for children with ASD remain due to both programmatic and research design decisions. First, the majority of research studies we reviewed had fewer than 25 participants, making it difficult to detect any statistically significant findings (Geretsegger et al., 2014; James et al., 2015). Second, studies did not often use a consistent and standardized measure to examine stability and change in communication and other outcomes of interests (Whipple, 2004). As a field, it is necessary to come up with brief, consistent ways to support the key components of the intervention generalized into everyday living.
Third, many of the interventions did not include any comparison groups or randomize treatment. Randomized control trials (RCTs) are used to test whether an intervention works under optimal situations and tries to determine which mechanisms within an intervention improves well-being and to what extent those improvements generalize beyond the intervention (Patsopoulos, 2011). However, and although they are often referred to as the gold standard research design, a critique of RCTs in this domain is that participation and eligibility can be so constrained that benefits may be overestimated (Broder-Fingert et al., 2017). Furthermore, RCTs are not always feasible or ethical and the credibility of the results might be undermined by the homogeneity in the samples (Broder-Fingert et al., 2017; Geretsegger et al., 2014).
In addition to research design, there have been problems with drawing conclusions due to programmatic design. Music therapy interventions are not monolithic; they differ widely in quality, clinical applicability to the populations served, and links to generalized practice. Each of these need to be addressed and clearly stated in intervention research (Geretsegger et al., 2014). Most of the intervention research we reviewed evaluated the effects of one-on-one music therapy interventions rather than group-based music therapy interventions. In addition, most of the therapy programs did not have a specific structure and curriculum. However, direct interactions with peers in a group setting have the potential to enhance children’s social skills more than individual music therapy and some research on group-based models has shown promising outcomes (Eren, 2015). For example, LaGasse (2014) found that, compared with children who participated in group sessions on social skills, children in group-based music therapy sessions made greater improvements in joint attention with peers, eye gaze toward others, and parental perception of social skills. Similarly, in research on an early pilot of the group-based music therapy program Voices Together (VT), which is the subject of the present investigation, Mendelson and colleagues (2016) found that children increased their communication skills across a range of developmental disabilities. At the same time, the need for group-based music therapy has grown as the field of special education looks to integrate different modalities of therapy into students’ daily routines, raising the importance of research to establish evidence-based approaches.
We developed the present study’s research design and methodology to address some of the limitations of these earlier studies. First, students served as their own control group during a baseline assessment phase before intervention began; the repeated measures approach allowed the researchers to create a quasi-control and treatment group. Second, we sought to mitigate some of the variability in previous findings (e.g., Mendelson et al., 2016) by including only students in self-contained ASD classrooms. By focusing on children with ASD who have similiar deficits and concerns around communication and socialization, we hope to better understand the specific needs of one group of students rather than trying to assess all the unique needs within all special education classrooms. Third, rather than conducting live, in-the-moment coding during the intervention, we developed a systematic measurement tool (DUACS, Duke University Autism Communication and Socialization) that we administered outside the therapy session to assess communication and socialization. By video-recording student responses, we were able to establish a more robust assessment to examine treatment effects and allowed for inter-rater reliability of the responses through multiple reviews of the prompts and responses.
The current study evaluated the effects of the group-based music therapy method (VOICSS™, Vocal Interactive Communication and Social Strategies) using a repeated measures student and teacher approach. VOICSS™ is the core method used by VT, a music therapy intervention developed and delivered by the eponymous non-profit organization which has served hundreds of special education classrooms and students across seven school districts since 2008. VT music therapists use evidence-based interactive songs to teach social-emotional skills, speech, and communication skills. The approach is structured but non-directive, providing students with clear expectations in a comfortable environment (White, 2012).
We derived four interconnected research questions to determine whether and how VT helps to improve children’s communication and socialization: (a) Are there differences between students in communication and socialization at baseline? (b) Do students improve at the same or different rates based on baseline performance? (c) Are there significant differences in baseline or changes from baseline when considering the demographic variables? and (d) Are there significant differences in baseline or changes from baseline when examining the teacher-reported behavior variables? We hypothesized that participation in the VT structured group music therapy model would be associated with significant improvements in communication and social-emotional adjustment skills among elementary school-aged children with ASD, measured through both a student behavioral prompt and teacher report.
Method
Participants
Participants were recruited from nine elementary schools (kindergarten to fifth grade) in one southeastern U.S. school district. To participate in the study, students met the school district’s requirements for placement in a special education self-contained classroom for children with ASD. Students who were not scheduled to be in their classroom during the weekly VT sessions due to other school-related commitments or whose parents did not provide consent were not included in the study.
The final sample consisted of 64 children between the ages of 5 and 11 (M = 8.04; SD = 1.62) years across nine classrooms. Of the 64 participants, 80% were male. Although the schools did not allow us to record the race/ethnicity of the students, across the school district about 30% of the population was non-White, primarily Hispanic and African American (National Center for Education Statistics, 2017). Most of the children had individualized education plans (IEPs) denoting an ASD diagnosis (94%), although clinical diagnosis was not determined by the school, program, or study. Based on the spoken language benchmarks described in Tager-Flusberg et al. (2009), students ranged from preverbal to fully verbal (i.e., using complex language skills; see Table 1), with the highest percentage of students falling in the “sentences” category, with enough language to serve their day-to-day communicative needs, and corresponding to typically developing children between the ages of 30 and 48 months (Tager-Flusberg et al., 2009).
Pre-Intervention Language Level of Students in Music Therapy Study.
Setting, Method, and Design
All intervention elements (see section “Treatment”) and assessment activities occurred in each student’s classroom. The therapy intervention was conducted by VT music therapists each week. Assessment prompts were administered outside the therapy sessions by VT music therapists together with members of the research team three times during a 6-week pre-intervention baseline period and three times during the 16-week intervention. Other research team members coded the video-recorded prompt responses, and a different team member analyzed the results. In addition, teacher-reported measures were collected at the beginning and end of the program.
Treatment
Trained VT music therapists offered the VOICSS™ method for 45 min each week for 16 weeks (White, 2012). Sessions occurred in participants’ classrooms with children sitting in chairs in a semi-circle surrounding the music therapist, who was board-certified, trained in the VOICSS™ methods, and had previous experience working with children with ASD. The VOICSS™ method is designed to improve speech and communication and to build social-emotional skills by incorporating music to engage participants. The method uses specialized music, counseling skills, and an interactive structure to promote language acquisition, communication, and social-emotional skills. The format alternates between therapist-to-peer and peer-to-peer interactions. The VT program helps students develop self-awareness, understand their strengths and challenges in everyday life, identify emotional states within themselves and others, and develop decision-making and problem-solving skills in a peer-group setting. Basic social rules, such as greeting, listening, turn-taking, and gaining someone’s attention, are reinforced, discussed, and practiced in each session.
The VOICSS™ method consists of three specialized elements: music, group process, and routinized curriculum. The music incorporates techniques to make communication more likely, including leaving requests for a response on an unresolved musical note, internal structure for turn-taking that alternates between group and individual response, and directly asking each individual questions that alternate between singing and speaking. Therapists also regularly employ responsive prompting, a strategy consisting of communication prompts that vary in intensity based on each participant’s ability level, to encourage verbal responses within the structure of program songs and during spontaneous group interactions. Consistent with the VOICSS approach, music therapist always assume competency to allow each child to initiate independent language. Prompting is offered as needed and in a sequential manner by asking, waiting, filling in the sentence or word, and then offering the full sentence as needed.
Each VT session begins with a group process where class chooses one student to be the “speaker.” The speaker announces and begins the songs, and the therapist uses the speaker and the songs themselves to encourage social communication in the group. The therapist focuses on individual students through rhythmic and responsive prompting and on the group as a whole through communal decision-making and group singing. Therapists use a non-directive approach to create a positive social group environment. They intentionally comment on positive social communication and interactions, prompting group members to notice others’ prosocial behaviors. One song in each session provides each student with a few minutes to share something personal (e.g., things they are good at, or why making friends at school can be difficult; White, 2012). Each session has a predictable schedule, which is written on a board at the front of the group. The consistent routine keeps the sessions predictable while also allowing the curriculum to evolve to meet the needs of students and the goals of the program.
Measures
The research team collected student behavioral outcomes and teacher-reported student outcomes during the baseline and treatment phases. The behavior outcome measure, DUACS, assessed social and communication outcomes and to see whether VT participation led to broader, more generalized student gains in social communication (Schmid et al., 2019). Teachers reported on student’s spoken language level (Tager-Flusberg et al., 2009) and completed the Pervasive Developmental Disorder Behavior Inventory (PDDBI; I. L. Cohen et al., 2003).
DUACS assessment
Although we hoped to find and use a pre-existing measure, a number of issues made it necessary to create a new one. The majority of existing instruments are either parent- or teacher report, and we wanted to assess the child’s behaviors directly. In addition, many of the measures examining social communication are designed to assess and diagnose deficits, not show changes over time (Gould et al., 2011). Furthermore, many of the standardized behavioral prompts are lengthy and are typically administered in a one-on-one format in a laboratory setting (Gould et al., 2011). We wanted something relatively quick and simple that could be conducted within a school setting. Finally, we wanted to generalize the behavioral components found in VT to more authentic, real-world situations.
The research team developed a set of behavioral prompts that tapped into the core elements of the program, focusing on communication and social-emotional adjustment. The main goal of the prompts was to create a stable set of behavioral activities that reflected changes we expected to see from the intervention and mapped onto known deficits for children with ASD. The resulting measure, DUACS, consists of prompts designed to elicit both gestural/non-verbal responses and verbal responses. It comprises 20 verbal and non-verbal prompts in increasing order of difficulty from “What is your name?” to “What do you do when you are having a hard time in school?” (see the appendix for a copy of the measure and coding scheme). The DUACS measurement was administered and video-recorded by trained VT program and research team members one-on-one with each child outside the therapy session. Assessments were by their nature interactive, but research and program staff were trained to remain neutral and not to respond or encourage the children beyond the scripted prompts.
Five trained research assistants who had not administered the assessments coded the recorded DUACS assessments in the laboratory. Across most items, responses were coded from 0 (not asked) to 3 (a complete and appropriate response). Research assistants were blinded to others’ coding determinations, but videos were not blinded by time and were coded as they were collected. The DUACS total score ranged from a minimum possible score of 5 to a maximum of 61. Inter-rater reliability was calculated using Cohen’s kappa coefficient and ranged from .89 to .97 (J. Cohen, 1960). Using Cronbach’s alphas, we calculated the measure’s internal consistency (.94–.96) and test–retest reliability (.85–.95). Further assessment of the psychometric properties of the DUACS is reported elsewhere (Schmid et al., 2019).
Pervasive Developmental Disorder Behavior Inventory
Teachers completed the PDDBI for each child in their classroom pre- and post intervention (I. L. Cohen et al., 2003). This reliable rating scale was designed to measure responsiveness to intervention in children diagnosed with pervasive developmental disorders such as ASD (I. L. Cohen et al., 2003). The inventory assesses changes across multiple domains including communication, attention, anxiety, and aggression. The PDDBI consists of 180 items rated on a 4-point Likert-type scale (0 = never, 3 = often) that make up 10 subscales. The subscales include maladaptive behaviors (sensory/perceptual approach problems, specific fears, arousal problems, aggressiveness, social pragmatic problems, and semantic/pragmatic problems) and adaptive behaviors (social approach behaviors, learning, memory, receptive language, phonological skills, and semantic/pragmatic ability). Subscale alphas for the PDDBI for the VT sample ranged from .71 to .97.
Spoken language level
Teachers rated their students’ spoken language level on a scale of 1 (preverbal communication) to 5 (complex language) using categories and descriptions from Tager-Flusberg and colleagues (2009). Teachers rated each student on a scale of 1 to 5 with each value indicating a different level of expressed language (e.g., 1 = preverbal communication, 5 = complex language).
Data Analysis
Since repeated DUACS measures were nested within students, we used multilevel modeling (MLM) to analyze study data. The use of MLM has several advantages over repeated measures analysis of variance (ANOVA) in that it allows for within-subject dependence, missing data, varying numbers of assessment by students, and unequal spacing between assessments. Specifically, MLM allows for the unique estimation of variance and covariance, does not require sphericity, and appears to yield higher statistical power (Quené & Van den Bergh, 2004). Most importantly, being able to define intra-individual and inter-individual variability allows researchers to simultaneously test within-person and between-person relations in a repeated measures framework.
We tested three models to address our research questions. Model 1 includes the DUACS sum score with only time as a predictor. Model 2 includes all time invariant predictors including gender, language level, grade, and age. Model 3 includes time variant predictors including pre- and post-measures on PDDBI subscales of interest. See Table 2 for the means and standard deviations for all items included in at least one of the models.
Descriptive Statistics for Music Therapy Study.
Note. Elang = expressive language competence; Express = expressive language; Socpp = social pragmatic problems; Socaware = social awareness problems; Empathy = empathy behaviors; Aslearn = associative learning skills; Pragtalk = pragmatic conversational skills; DUACS = Duke University Autism Communication and Socialization.
1 = Male.
Results
Model 1 focuses on whether time plays an important role in understanding changes in the DUACS scores and whether changes in the intercept (i.e., mean scores over time) and slope (i.e., variability between scores) were linear. The MLM procedures provide random-effects and fixed-effects scores. The random effects include the following: the variance of the intercepts, the variance of the slopes, the covariance between intercepts and slopes, and residual variances. The variance of the intercepts (γ = 144.07), covariance between the intercepts and slopes (γ = 12.61), and residual variances (γ = 56.34) were all statistically different from zero (z-scores above 1.97). The significant intercept variance indicates significant differences in students’ first DUACS scores in the model. This can also be verified through the sum score means and standard deviations found in Table 2. Therefore, students were at different levels of communication and socialization at the baseline assessment. In addition, a positive correlation found between the intercept and the slope indicates that students’ higher initial scores on the DUACS were associated with a more rapid slope change (i.e., improvement) over time. In other words, students with higher DUACS scores at the beginning had greater gains in the DUACS over time. Finally, significant random residual effects indicate individual variability around the mean intercept and slope. The fixed effects showed a significant effect on the intercept, with an estimate of 22.65 with a linear slope increasing modestly at 1.52. Both the quadratic time term and the likelihood ratio test were non-significant, indicating that a quadratic term did not improve the model and that the assumption of homoscedasticity on the Level 1 residual variance over time is met. See Table 3 for details on Model 1 fixed estimates and fit statistics.
Multilevel Models Predicting DUACS Scores in Music Therapy Study.
Note. 1 = male; Elang = expressive language competence; Express = expressive language; Socpp = social pragmatic problems; Socaware = social awareness problems; Empathy = empathy behaviors; Aslearn = associative learning skills; Pragtalk = pragmatic conversational skills.
p < .05.
We further examined DUACS data using a series of paired t-tests between baseline and treatment. No differences were found at Time 1, but there were significant improvements between DUACS scores at Time 2, t(49) = 3.48, p < .05, and Time 3, t(53) = 2.89, p < .05, with treatment gains compared with baseline scores.
In Model 2, we included time invariant demographics (age, grade, sex, and language level) to predict changes in DUACS scores over time. With the inclusion of the demographic variables, the variances between the intercepts, slopes, and the covariance between the intercepts and slopes were no longer significant. However, a significant random residual effect (γ = 63.27) indicated individual variability around the mean intercept and slope. Including the demographic variables appeared to improve model fit as seen in the Akaike information criterion (AIC) scores in Table 3. The fixed effects showed a significant effect on time and beginning language level. As before, there was a modest gain in DUACS scores over time of 1.60. More notably, there was a sizable estimate of 10.72 on beginning language level, indicating that higher language scores had higher DUACS scores. None of the other fixed effects on demographics were significant.
Fixed effects reflect intercept differences while interaction terms examine differences in slope. We conducted a model including all of the same fixed effects and adding an interaction term (i.e., demographic variable by time) to discern differences in slopes. Including the demographic variable interaction terms appeared to marginally improve model fit (AIC in original Model 2 = 2049.00; AIC in slope model = 2044.00). The fixed effects showed a marginally significant interaction effect between beginning language level and time (b = 0.50, t = 1.92, p = .06). This interaction indicated that students with a higher beginning language level were more likely to exhibit a higher DUACS score more rapidly over time (i.e., higher slopes/gains for students with higher initial language levels). None of the other interaction effects on demographics were significant.
In Model 3, we included time variant pre- and post-behavior elements from the PDDBI to both the demographic variables and the time variable to predict changes in DUACS scores. This included a subset of scales and subscales including the expressive language scale, expressive language subscale, social pragmatic problems, social awareness problems, empathy behaviors, pragmatic conversational skills, and associative learning skills. These scales were selected for their alignment with VT intervention elements. With the inclusion of these time-varying behavioral components, the variances between the intercepts and slopes, much like Model 2, were not significant. Both the covariance between the intercepts and slopes (γ = 6.65) and the residual effect (γ = 63.27) were significant. Inclusion of the behavior variables did not appear to improve model fit as seen in the AIC scores in Table 3. As before, the fixed effects showed a significant effect on time and beginning language level. The only behavior element that was statistically significant was the empathy behaviors score (b = 0.45, t = 2.97, p < .05), which indicated that students with higher scores on empathy behaviors also had higher DUACS scores. None of the other fixed effects on demographics were significant.
We conducted a model including all of the same fixed effects and adding an interaction term (i.e., behavior variable by time) to discern differences in slopes. As in the model with only fixed effects, there were significant effects on time, beginning language levels, and empathy scores. In addition, the fixed effects for social pragmatic problems (b = 0.45, t = 2.10, p < .05) and social awareness problems (b = −0.95, t = 2.02, p < .05) were significant in this model. However, only one interaction term was significant, indicating a difference in slope over time. The interaction between social pragmatic problems and time indicated that students with higher levels of social pragmatic problems showed less growth in DUACS scores compared with students with lower social pragmatic problems (b = −0.14, t = 2.56, p < .05).
Discussion
In a field of burgeoning interest in but scarce robust evidence on effective music therapy interventions for children with ASD, these results provide promising evidence for VT, particularly for students with moderate baseline social communication skills Overall, participation in the VT program was associated with gains in language, communication, and social-emotional learning among elementary school-aged children with ASD. These findings not only support the use of VT but contribute to the evidence for the effectiveness of group-based music therapy for children with ASD (Geretsegger et al., 2014; James et al., 2015).
Our first research question focused on identifying differences in student outcomes at baseline. Students in this sample varied significantly in their baseline communication skills as shown in Model 1. Some children were able to answer almost all questions appropriately in this assessment, whereas others were not able to verbalize or communicate at all in response to the prompts. This expected variability reflects the well-documented spectrum of differences in communication, social skills, and language seen in those with ASD (Baio et al., 2018). In addition to describing the sample, this diversity underscores the need for measurement that accounts for this variability and can examine changes in relation to each student’s baseline.
Our second research question examined whether the rate of change or improvements based on the VT intervention occur differently for different students. Baseline differences between students appear to explain individual change trajectories and improvements due to the intervention over time. Children with higher baseline scores on the DUACS showed greater gains over time compared with children with lower baseline scores, which showed more modest rates of change over time.
Next, we were interested in the role that student demographic characteristics played in understanding both the baseline scores and the rate of change over time. Fixed predictors of gender, grade, and age were not associated with baseline scores or rates of change over time, but language level was associated with both. As expected, students with higher baseline language scores had higher baseline DUACS scores because both a measuring communication. In addition, students with higher baseline language levels were more likely to have faster growth in the DUACS across the intervention, which indicate greater gains across the intervention. Conversely, students with lower baseline language scores were less likely to show change over time. Together, these results appear to indicate that children with moderate baseline social communication skills were most likely to benefit from VT participation.
A key aim of this study was to determine whether and how gains during VT generalize to everyday behaviors. It is one thing to find improvements in response to specific prompts or within an intervention setting, but do individuals acquire new skills to help them navigate the world? We examined teacher-reported domains involved in expressive language, social processing problems, social awareness, empathy, associative learning, and pragmatic communication in relation to the DUACS to determine whether there were associations with teacher-reported gains and observational behavioral gains. Empathy and social pragmatic problems emerged as significant. Students with higher empathy scores at baseline also had higher DUACS scores, and those with higher empathy scores also had higher rates of change in social communication skills, compared with those with lower baseline empathy scores. Children with higher levels of social pragmatic problems, on the contrary, fared worse on outcomes compared with students with fewer social pragmatic problems.
These results add to the research literature by more precisely understanding changes in communication and socialization associated with the VT intervention, without relying on a RCT. Given methodological and ethical concerns, RCT may not prove appropriate or yield reliable results for children with ASD. By constructing a quasi-control group, we assessed intervention effects by matching students to individuals similar to themselves (i.e., themselves prior to the intervention). This method allowed us to better assess whether changes over time were due to the intervention, rather than differences between group members. Results are promising, but we need more data using this method to establish its use broadly within the ASD population.
This study also advances the field by introducing a new measure, DUACS, to assess social communication behaviors. DUACS was developed in response to the paucity of brief assessments of children with ASD behaviors in social communication (Schmid et al., 2019). Although full assessment of the DUACS is beyond the scope of this article (see Schmid et al., 2019), it appears that this measure can directly assess student behavioral changes in communication and social-emotional learning. Generally, the DUACS provides a feasible and reliable way to directly assess social communication in children with ASD.
These findings should be considered in light of limitations that may constrain the generalizability of the findings. First, this is a relatively small sample from self-contained, elementary school special education classrooms for children with ASD in a one local education area (LEA) in the southeastern region of the United States. Within this sample, children vary widely in their levels of language, communication, and social skills. We cannot assume our findings would generalize to other samples in other settings without replication.
Second, one of the primary measurement tools, the DUACS, was created specifically for children with ASD who participated in the VT intervention. Further investigation of the measure is underway (see Schmid et al., 2019). Although we surmise the measure would work similarly with other samples of elementary-aged students with ASD and other developmental disabilities, its broader utility is, for now, unknown. In this study, the DUACS successfully detects changes in communication and socialization associated with participation in VT, and we are cautious in generalizing this success to a wider ASD population or to other disabilities or to evaluate other interventions without further study.
Third, the DUACS is primarily a language-based measure and most items require a verbal response. As shown, the DUACS was correlated to language level and other behaviors related to communication and socialization. Differences in baseline scores were somewhat explained by baseline language abilities and teacher-reported empathy. Students with higher baseline scores in language and empathy also had higher improvements in DUACS over time. Since the DUACS is our primary proxy for the VT program, this might reflect that the program is optimal for students who have functional language skills (Tager-Flusberg et al., 2009). Further research is necessary to determine whether and how the focus on verbal responses in the DUACS affect the findings.
Evidence-based classroom interventions that help children with ASD to develop and use communication and social-emotional strategies and skills are in high demand. Schools struggle to find effective, easily implemented interventions that work with a broad array of student abilities that are typically found in self-contained classrooms. The growing body of evidence on the effectiveness of music-based approaches among children with ASD supports their adoption. Furthermore, this group-based intervention is designed to be more efficient than one-on-one therapy and can therefore be implemented on a larger scale, reaching more of the target population (Eren, 2015). In addition, the peer-to-peer interaction fostered in the VOICSS model supports expressive and pragmatic communication skills. More research is needed to determine which strategies work best with which types of students and under what conditions. This study provides evidence for the effectiveness of VT in growing and enhancing communication and social-emotional skills among elementary school children with ASD.
Footnotes
Appendix
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This team, supported by Duke University’s Bass Connections, included researchers from the Duke Center for Autism and Brain Development, the Social Science Research Institute, Voices Together, and Duke students.
