BEYOND SKILLS TO PAY THE BILLS: EFFECTS OF A VOCATIONAL SOCIAL SKILLS
INTERVENTION ON PSYCHOSOCIAL FUNCTIONING AMONG YOUNG ADULTS WITH
AUTISM
By
Annemarie Connor

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Rehabilitation Counselor Education - Doctor of Philosophy
2017

ABSTRACT
BEYOND SKILLS TO PAY THE BILLS: EFFECTS OF A VOCATIONAL SOCIAL SKILLS
INTERVENTION ON PSYCHOSOCIAL FUNCTIONING AMONG YOUNG ADULTS WITH
AUTISM
By
Annemarie Connor
The transition from secondary education to higher education, work, and independent
living is a critical period in the development of all youth, and one that elicits unique challenges
for young adults with autism spectrum disorder and average to above-average intelligence.
Transition service systems, including secondary and postsecondary education, state-federal
vocational rehabilitation (VR), and community mental health (CMH) have struggled to address
the unique needs of this increasingly identified, yet historically underserved, group. Educational
and employment outcomes have been disappointing and disproportionately poor in comparison
to other disability groups. Recognizing young adulthood as a critical developmental period, and
noting the bidirectional relationships between social functioning, self-efficacy, and psychological
wellness, this mixed methods study examined the efficacy of an empirically-derived,
manualized, vocational social skills intervention for young adults with high-functioning autism
spectrum disorder (HFASD). Results indicated that the Assistive Social Skills and Employment
Training (ASSET) program is associated with: (1) immediate, small to medium-sized
improvements in social cognition, (2) long-term, small to medium-sized improvements in social
communication among participants with higher socioeconomic status, (3) immediate, very large
gains in self-efficacy, and (4) long-term, medium to large declines in reported anxiety.
Furthermore, results of the qualitative analysis suggested varying degrees of social skills
generalization in employment, social, and family contexts up to two years after completing the

program. This study is a unique contribution to the state of the science as it provides validation
that vocational social skills training positively impacts psychosocial functioning, including social
functioning, self-efficacy, career outcome expectations, and psychological wellness, among
young adults with HFASD. Furthermore, these results validate the use of Social Cognitive
Career Theory to guide research and interventions with this population. In sum, this study
provides level-three (Chan, Bezyak, Ramirez, Chiu, & Sung, 2010) evidence that ASSET can:
(1) partially meet the need for manualized, evidence-based strategies to address work-related
social skill development among young adults with HFASD, (2) help young adults with HFASD
capitalize on the iterative power of bolstered self-efficacy and optimistic career expectations
when facing the social challenges inherent to HFASD, (3) mitigate barriers to delivering such
services by making them community-based and interdisciplinary, and (4) provide an innovative
and efficient mechanism for addressing psychological wellness while targeting transferable
skills.

Copyright by
ANNEMARIE CONNOR
2017

This dissertation is dedicated to my husband, Kenneth James Tsang, and children, James Connor
Tsang and Cassidy Connor Tsang, who bore sacrifice with grace, love, patience, and
determination. The journey was not possible without you.

v

ACKNOWLEDGMENTS

I would like to acknowledge each member of my committee for their unique
contributions to this study and my development as researcher, scholar, and educator. First and
foremost, I owe a debt of gratitude to my advisor, Dr. Connie Sung, who pushed me to achieve
more than I thought possible, while also showing me how to work “smarter, not harder.” Your
unyieldingly positive attitude, boundless energy, and enthusiasm sustained me, while your
excellence and dedication inspired me. Thank you, Dr. Sung, for so generously providing
mentorship and guidance, while also valuing me as a colleague.
I am deeply grateful to have had the opportunity to study under and collaborate with a
leader in the field, Dr. Michael Leahy, who has modeled effective leadership by achieving great
things without neglecting small kindnesses. Dr. Leahy, thank you for helping this occupational
therapist cultivate counselor knowledge, affect, and skills. What I have learned about the
working alliance informs everything I do. As you say, “it’s all about relationships.”
I have also been most fortunate to study under Dr. John Kosciulek whose scholarly work
and teaching has been a tremendous influence. Dr. Kosciulek, I thank you for your willingness
to provide constructive criticism while also showing enthusiasm for my work. You have been
instrumental in shaping my interests in career development and vocational rehabilitation. Your
teaching has me helped to grow from a doctoral student who “knows nothing about work” to a
researcher with passion for examining the meaning and purpose that work brings to all people.
As a doctoral student tackling an intervention study, I have been most fortunate to have
been afforded the example set by Dr. Gloria Lee. Dr. Lee, your research on the needs of families
of children with autism is vital. In my own nascent work as a researcher, it seems that much of

vi

what we conclude points to the need for family education and support. Thank you for your
continued leadership and willingness to share what you have worked so hard to achieve.
I am also thankful to Dr. Kimberly Maier, who graciously agreed to serve on this
committee as a quantitative expert. Dr. Maier, your willingness to consult on methodology and
ensure the quality of my results, despite a pending leave and multiple students vying for your
attention, was essential to the success of this project.
In addition to the committee, I offer heartfelt thanks to my doctoral student colleagues
who have been consistently supportive, collaborative, insightful, and a true joy to labor
alongside. Special thanks to Alicia Strain who cross-checked my qualitative data and took
special interest in this project and the ASSET program. Finally, none of this work would be
possible without our ASSET participants. Thank you for your commitment to the program, your
input on curriculum development, willingness to endure multiple questionnaires and interviews,
and, most importantly, the opportunity to share in your successes and struggles.

vii

TABLE OF CONTENTS

LIST OF TABLES

xii

LIST OF FIGURES

xiii

KEY TO ABBREVIATIONS

xiv

CHAPTER ONE: Introduction
Statement of the Problem
Sparse research on the unique needs of young adults with ASD.
Growing population with disparate postsecondary outcomes.
Barriers to work.
The importance of social functioning.
Need for evidence-based interventions.
Summary of the problem.
Theoretical Framework
Statement of Purpose
Research Questions and Hypotheses
Significance of the Study

1
2
2
2
4
5
5
6
6
8
9
10

CHAPTER TWO: Literature Review
Current Understanding of Autism
Relationship of Social Skills, Self-efficacy, Anxiety and Depression
Employment as the Proving Ground for Psychosocial Functioning
Service System Limitations and Barriers
Special Education.
Vocational Rehabilitation.
Need for Evidence-based Transition Interventions
Current Social Skills Interventions
Social Cognitive Career Theory
Self-efficacy (SE).
Outcome expectations.
Goals.
SCCT and disability.
Innovation
Literature Summary

11
11
13
16
18
18
19
21
24
26
26
27
27
27
29
31

CHAPTER THREE: Methods
Design
Sampling
Participants.
Procedures

33
33
33
33
34

viii

Intervention
Instrumentation
Screening instruments.
Diagnostic screening.
Cognitive assessment.
Reading assessment.
Primary outcome measures.
Social functioning.
Perceived self-efficacy.
Empathic self-efficacy.
Social self-efficacy.
Secondary outcome measures.
Depression.
Anxiety.
Data Analysis Plan
Preliminary quantitative analysis plan.
The need for multilevel modeling.
Applying multilevel modeling to the ASSET dataset.
Selection of covariates.
Multilevel modeling procedures.
(1) Choice of appropriate parameter estimator.
(2) Confirmation of need for multilevel modeling.
(3) Building the Level 1 model.
(4) Building the Level 2 model.
(5) Multilevel effect size reporting.
(6) Likelihood ratio testing.
Quantitative analysis summary.
Qualitative data analysis plan.
Summary of Analysis by Research Questions
To answer research questions 1-3.
To answer research question 4.
To answer research question 5.
CHAPTER FOUR: Results
Demographics and Characteristics of the Sample
Preliminary Data Screening and Analyses
Descriptive statistics.
Scale reliability.
Normality.
Screening potential covariates.
Hypothesis Testing
Research question 1: Is ASSET associated with improved social
functioning?
Preliminary analyses, social functioning.
LMM results, social cognition.
Final model.

ix

35
38
38
38
38
39
39
40
40
40
41
41
41
42
42
42
43
45
49
50
50
50
51
51
52
53
53
54
55
55
56
56
57
57
59
59
60
60
61
62
63
63
64
68

LMM results, social communication.
Final model.
Research question 2: Is ASSET associated with improved self-efficacy?
Preliminary analyses, self-efficacy.
LMM results, empathy self-efficacy.
Final model.
LMM results, social self-efficacy.
Final model.
Research question 3: Is ASSET associated with improved psychological
wellness?
Preliminary analyses, psychological wellness.
LMM results, anxiety.
Final model.
LMM results, depression.
Final model.
Research question 4: Are self-efficacy and psychological wellness
correlated?
Research question 5: Is ASSET associated with improvements in
participants’ self-perceived psychosocial functioning (social
functioning, self-efficacy, and psychological wellness) and outcome
expectations over time?
Social functioning.
Self-efficacy.
Psychological wellness.
Outcome expectations
Fit of qualitative results with quantitative findings.
Results Summary
CHAPTER FIVE: Discussion
Major Findings and Interpretations
Social cognition.
Social communication.
Social implications.
Empathic self-efficacy.
Social self-efficacy.
Self-efficacy implications.
Anxiety.
Depression.
Psychological wellness implications.
Integration of Findings with the Literature
Theoretical alignment.
Limitations
Threats to external validity.
Measurement biases.
Threats to internal validity.
Implications for Practice

x

70
73
76
76
77
80
81
84
85
85
86
89
90
93
94
94

95
100
103
104
105
106
107
108
109
110
112
113
114
114
115
115
116
119
125
126
126
127
128
129

Implications for Research
Future directions.
Conclusion

133
135
138

APPENDICES
APPENDIX A: Consent Form
APPENDIX B: Demographics
APPENDIX C: Questionnaire (Pre-Intervention)
APPENDIX D: Psychosocial Questionnaire (Post-Intervention)
APPENDIX E: Focus Group Protocol (1-2 years Post-Intervention)
APPENDIX F: Human Subjects Institutional Review Board Approval

140
141
144
145
149
154
155

REFERENCES

156

xi

LIST OF TABLES

Table 3.1 Summary of ASSET curriculum and examples of key points

36

Table 3.2 Features of the analytic approaches considered

54

Table 4.1 Participant demographics, baseline status, and disability characteristics

58

Table 4.2 Dependent variable means and standard deviations at baseline, postintervention, and two-month follow-up

60

Table 4.3 Tests of scale reliability

60

Table 4.4 Pearson product moments

61

Table 4.5 Social functioning dependent samples t tests, mean differences, standard
deviations, and effect sizes

64

Table 4.6: Model summaries: social cognition

66

Table 4.7 Model summaries: social communication

72

Table 4.8 Self-efficacy dependent samples t tests, mean differences, standard
deviations, and effect sizes

77

Table 4.9 Model summaries: empathic self-efficacy

79

Table 4.10 Model summaries: social self-efficacy

83

Table 4.11 Psychological wellness dependent samples t tests, mean differences,
standard errors, and effect sizes

86

Table 4.12 Model summaries: anxiety

88

Table 4.13 Model summaries: depression

92

Table 4.14 Pearson product-moment correlations among self-efficacy and
psychological wellness variables

94

xii

LIST OF FIGURES

Figure 1.1 Theoretical Framework: Modified from Social Cognitive Career Theory

8

Figure 2.1 Social Cognitive Career Theory

27

Figure 3.1 Multilevel nature of ASSET dataset

46

Figure 3.2 Example of individuality in SRS baseline and outcome

47

Figure 3.3 Data structure in long format

48

Figure 4.1 Social cognition scores by individual

65

Figure 4.2 Social cognition scores by time

65

Figure 4.3 Social cognition scores by time and socioeconomic status

69

Figure 4.4 Social communication scores by individual

71

Figure 4.5 Social communication by time

72

Figure 4.6 Social communication by time and socioeconomic status

75

Figure 4.7 Perceived empathic self-efficacy by individual

78

Figure 4.8 Perceived empathic self-efficacy by time

78

Figure 4.9 Perceived social self-efficacy by individual

82

Figure 4.10 Perceived social self-efficacy by time

82

Figure 4.11 Anxiety by individual

87

Figure 4.12 Anxiety by time

87

Figure 4.13 Depression by individual

90

Figure 4.14: Depression by time

91

xiii

KEY TO ABBREVIATIONS

ASD

Autism Spectrum Disorder

ASSET

Assistive Social Skills and Employment Training

CDC

Centers for Disease Control and Prevention

CMH

Community Mental Health

HFASD

High Functioning Autism Spectrum Disorder

LMM

Linear Mixed-effects modeling

NLTS-2

National Longitudinal Transition Study

PDD-NOS

Pervasive Developmental Disorder-not otherwise specified

SCCT

Social Cognitive Career Theory

RSA

Rehabilitation Service Administration

SE

Self-efficacy

SSI

Supplemental Security Income

SSDI

Social Security Disability Insurance

VR

Vocational Rehabilitation

WIOA

Workforce Investment and Opportunities Act

xiv

CHAPTER ONE
Introduction
Individuals with autism and without intellectual disability, referred to here as highfunctioning autism spectrum disorder (HFASD), present with unique service needs within the
educational, vocational, and mental health systems that serve them. While exhibiting average to
above average intelligence, impaired social functioning and psychological factors, including high
levels of co-occurring anxiety and depression, have a profound impact on the social and
vocational engagement of this group (Cloutier, Kurtz, May, & Hagner, 2014). In fact, older age
and higher intelligence (IQ) predict lower self-perceived social competence and higher levels of
depressive symptoms for individuals with ASD (Vickerstaff, Heriot, Wong, Lopes, & Dossetor,
2007). Notably, young adults with HFASD are among the least likely of their peers to gain
employment (Westbrook et al., 2013) or live independently (Anderson, Shattuck, Cooper, Roux,
& Wagner, 2014). Moreover, persistent struggles with social skills (Howlin & Moss, 2012), and
anxiety and depression (White & Roberson-Nay, 2009) shape the social and vocational interests,
goals, and behaviors of this population, particularly in the critical first two years after high
school (Anderson et al. 2014: Shattuck et al., 2012). Both singularly and combined, these factors
complicate participation in transition activities and negatively impact post-secondary outcomes.
The research literature enumerates disparate outcomes and elaborates multiple unmet
needs related to autism. However, intervention studies, particularly those targeting adults with
HFASD, are scant. While tremendous time, money, and resources have been devoted to
addressing the medical and behavioral needs of children with autism, the most recent analysis of
autism research funding from the Government Accounting Office (Federal Autism Research,
2015) reveals that funding to support research on the needs of adults with autism actually

1

declined from 2008 to 2013. The lack of research and resources has led some to describe school
to work transition as a “confusing maze of misinformation, limited opportunities, insufficient
resources and inappropriate or time-limited services” (Gerhardt & Holmes, 1997 as cited by
Eaves & Ho, 2007, p. 739). Federal initiatives, including the Workforce Innovation and
Opportunity Act (WIOA, 2014) and the National Longitudinal Transition Study-2 (NLTS-2,
2011) have called for improved transition outcomes. In response, service systems and
researchers will need to look beyond outcomes, such as degree attainment and employment, and
consider innovations in service delivery. Accordingly, this study examined how bolstering
work-related social functioning may help students with HFASD go beyond gaining “skills to pay
the bills,” and ultimately experience improved psychological wellness.
Statement of the Problem
Sparse research on the unique needs of young adults with ASD. Most of the existing
research and book literature is focused on toddlers and young children, and the use of applied
behavior analysis (ABA) as an evidence-based intervention (Wehman, Smith, & Schall, 2009).
While ABA is an appropriate early intervention for children with ASD, and an effective support
for those with severe behavioral issues, it is arguably less appropriate for young adults with
HFASD whose needs are more behaviorally complex and socially sophisticated. Moreover,
individuals with HFASD are more self-aware of their social deficits (Vickerstaff et al., 2007)
and, therefore, show significantly higher rates of anxiety (Kuusikko et al., 2008) and depression
than individuals with ASD and lower levels of functional behavior (Hillier, Fish, Siegel, &
Beversdorf, 2011).
Growing population with disparate postsecondary outcomes. As rates of ASD
diagnosis have continued to rise (CDC, 2016), so too will the need to address the psychological

2

wellness and social integration of children with ASD as they transition to adult life. While social
and psychological difficulties are central to ASD (Carpenter Rich, Loo, Yang, Dang, & Smalley,
2009; Hillier et al., 2011), work related social skills, or soft skills, are central to workplace
success (Pritchard, 2013). Due, in part, to social skills deficits and mental health challenges,
individuals with HFASD have disappointing postsecondary educational and employment
outcomes compared to their nondisabled peers (Shattuck et al., 2012). Furthermore, adults with
ASD who actually obtain employment are often under employed, working less hours and with
fewer responsibilities than their hard skills and educational backgrounds would indicate (Müller,
Schuler, Burton, & Yates, 2003; Taylor & Seltzer, 2011).
The National Longitudinal Transition Study (NLTS-2) data paints a similarly bleak
picture of postsecondary independent living outcomes for young adults with ASD, indicating that
only 17% were living outside of the parental home, a rate lower than other disability groups
including emotional, intellectual, and learning disabilities (Newman et al., 2011). These
disparities also extend to post-secondary education for young adults with ASD who enroll in
college at rates much lower than their cognitive abilities might predict (Shattuck et al., 2012).
Such disruptions to the normal acquisition of developmental milestones in young adulthood
further isolate individuals with HFASD who may excel cognitively, but struggle socially.
Furthermore, transition is a critical period, particularly the first two years following high school,
in which the shift from the entitlement services of special education to the eligibility services of
adult, community-based services, leaves individuals with ASD vulnerable to service disruption,
discontinuation, or disengagement (Anderson et al., 2014; Shattuck et al., 2012). Fittingly, the
Interagency Autism Coordinating Committee (IACC, 2011) and multiple authors (Anderson et
al., 2014; Shattuck et al., 2012) have called for more research on the transition from special

3

education and higher education to adult life for youth with ASD. As impaired social
communication is a hallmark of the disorder, social skills interventions may be particularly
relevant to addressing the transition needs of young adults with HFASD.
Barriers to work. Knowing that social deficits, anxiety, and depression are challenges
to the wellness, functional independence, and transition needs of this population, an
understanding of what constructs mediate their effects is imperative. Throughout transition,
students with ASD, particularly at the high school level, experience limited or absent schoolbased counseling services for psychosocial adjustment or career counseling (Connor, 2017), and
may also face barriers to obtaining and benefiting from state VR counseling services due to lack
of student/family involvement and ineffective interagency collaboration (Riesen, Schultz,
Morgan, & Kupferman, 2014). At the college level, out-of-state students may have no access to
state VR due to residency issues. Other barriers to addressing the psychosocial needs of this
population are secondary education curricular content limitations imposed by diploma
requirements, rationing of special education interventions or community mental health services
for those with the most significant behavioral needs (Connor, 2017), and passive learning styles
among students with ASD (Newman, 2007).
Moreover, increased academic opportunities for students with ASD have not translated
into improved employment and economic self-sufficiency for the population (Eaves & Ho,
2008). Outcomes are marginally better within the state VR context, but consumers with ASD
use more services and incur higher costs, yet receive lower wages and fewer work hours than
other disability groups (Cimera & Cowen, 2009). At the individual level, difficulties with social
communication, anxiety, stress, and sensory processing have been reported as barriers to finding
work appropriate to skill level (Hurbutt & Chalmers, 2004). These barriers should compel

4

researchers to create and validate innovative solutions to meeting both the soft skill development
and psychological wellness needs of this growing population. Tantamount is the need to address
these issues while young adults with ASD are still connected with schools and, therefore, more
accessible to service providers.
The importance of social functioning. Work-related social skills or “soft skills,” have
been identified as essential to both the hireability and retention of all workers, particularly as
industries expand service roles and adopt more team-based workforces (Lawler & Finegold,
2000). Social competence, or the expression of prosocial behaviors such as empathy and
perspective taking, is associated with positive workplace relationships, job tenure, and social
involvement (Lane, Menzies, Barton-Arwood, Doukas, & Munton, 2005). Individuals with
HFASD are acutely aware of their social difficulties, and experience a widening gap between
social capabilities and contextual demands as they transition to work and higher education
(Vickerstaff et al., 2007). When further considering that anxiety and depression negatively
impact peer relationships (Kim, Szatmari, Bryson, Streiner, & Wilson, 2000), it is apparent that
individuals with HFASD often become stuck in a vicious cycle: social skill deficits generate poor
outcomes, poor outcomes exacerbate anxiety and depression, and anxiety and depression lead to
impaired social interactions.
Need for evidence-based interventions. While interventions targeting social cognition
and social functioning in the larger ASD population have become increasingly prevalent, most
studies are atheoretical and descriptive, lack experimental control (Laugeson, Frankel, Gantman,
Dillon, & Mogil, 2012; Laugeson & Park, 2014), and focus on the needs of young children (Jang
et al., 2014). The special education and rehabilitation literature, while identifying the need for
interventions that target career self-efficacy, expectations, and job-related social skills, lacks

5

rigor in identifying manualized interventions for addressing these pressing needs (Alwell &
Cobb, 2009). To date, research has not systematically examined work-related social skills
interventions as evidence-based practices for young adults with HFASD. Additional
interventions that target the work-related social skills and mental health needs of adults with
HFASD are needed, and could begin to improve the work, independent living, and mental health
disparities experienced by this population. Additionally, rigorous research on the efficacy and
effectiveness of novel interventions is essential to ensuring that programs are evidence-based,
suited to the population, and fundable.
Summary of the problem. The transition from secondary education to higher education,
work, and independent living is a critical period in the development of all youth, and one that
elicits unique challenges for young adults with HFASD. Transition service systems, including
secondary and postsecondary education, state-federal vocational rehabilitation (VR), and
community mental health (CMH) have struggled to address the unique needs of this increasingly
identified yet historically underserved group. Educational and employment outcomes have been
disappointing and disproportionately poor in comparison to other disability groups. Recognizing
young adulthood as a developmentally sensitive window, and noting the bidirectional
relationships between social functioning, self-efficacy, and psychological wellness, evidence is
needed to inform how policy and practice will address psychosocial adaptation and functional
outcomes for young adults with HFASD.
Theoretical Framework
Steeped in constructivism and based on Bandura’s (1986) theory of social cognition, the
purpose of Social Cognitive Career Theory (SCCT) is to explain three interrelated aspects of
career development: how interests are shaped, how educational and career choices are made, and

6

how academic and career success is achieved (Greenhaus & Callahan, 2006). SCCT provides a
framework to describe how social learning constructs in association with personal and
environmental factors influence not only career choice, but also work performance.
SCCT has been applied widely to understand how theory links to practice in addressing
the vocational rehabilitation needs of individuals with various disabilities, including adolescents
(Betz, 2007; Fabian, 2000). SCCT tenets state: (1) career choice, goals, and performance are
influenced by the reciprocal interaction of personal, behavioral, and environmental factors, and
(2) career development may be helped or hindered by environmental factors including
differential socialization. Recognizing that ASD generates differential socialization, adolescents
are primed for cognitive and behavioral shaping in relation to career expectancy and skill
development, and further noting that SCCT accounts for not only intrapersonal, but also social
factors that influence career development, SCCT is the logical choice in framing this study of the
employment-related social skills training needs of youth with ASD.
According to SCCT, social skills training can be conceptualized as a learning experience
that considers personal inputs (ASD related social functioning), and provides a learning
experience (social skills intervention) that impacts self-efficacy and outcomes expectations,
which subsequently drive improved social skills and psychological wellness. While Lent and
colleagues (2002) proposed interests, goals and tasks as the outcomes of SCCT, this study
adapted their framework to include anxiety and depression as psychological outcomes and novel
constructs to SCCT (see Figure 1.1).

7

Figure 1.1
Theoretical Framework: Modified from Social Cognitive Career Theory (Lent, Brown, &
Hackett, 1994, 2000, 2002.)

Statement of Purpose
The purpose of this study is fourfold: (1) contribute to the research literature on evidencebased work-related social skills interventions for young adults with HFASD, (2) highlight the
multi-disciplinary, multi-outcome, community-based and manualized nature of the intervention
as an innovation in service delivery, (3) better understand the relationships between social
functioning, self-efficacy, and psychological wellness in this population, and (4) explore
participants’ awareness of individual changes in psychosocial functioning and outcome
expectations over time. Pragmatically, this study seeks to test the preliminary efficacy of
ASSET as an innovative and efficient method for impacting work-related social skills among
young adults with HFASD. Theoretically, the proposed study seeks to quantitatively describe
the relationship between social functioning, self-efficacy, and anxiety and depression, while
qualitatively elaborating on these relationships and exploring the influence of self-efficacy on

8

psychological wellness. Aspirationally, this study seeks to encourage innovative service
delivery, multi-disciplinary collaboration, and greater attention to the psychosocial needs of
young adults with HFASD and their families.
Research Questions and Hypotheses
1.

Is ASSET associated with improved social functioning in the identified population? It is

hypothesized that when compared to baseline (T1), students with HFASD who complete the
ASSET program will show significant improvements in social functioning (a) immediately post
intervention (T2) and (b) at two-month follow-up (T3).
2.

Is ASSET associated with improved self-efficacy in the identified population? It is

hypothesized that when compared to baseline (T1), students with HFASD who complete the
ASSET program will show significant improvements in self-efficacy (a) immediately post
intervention (T2) and (b) at two-month follow-up (T3).
3.

Is ASSET associated with improved psychological wellness (anxiety and depression) in

the identified population? It is hypothesized that when compared to baseline (T1), students with
HFASD who complete the ASSET program will show significant improvements in
psychological wellness (anxiety and depression) (a) immediately post intervention (T2) and (b)
at two-month follow-up (T3).
4.

Are self-efficacy and psychological wellness correlated? It is hypothesized that self-

efficacy is inversely correlated with anxiety/depression.
5.

Is ASSET associated with improvement in self-perceived, work-related psychosocial

functioning and outcome expectations over time? It is hypothesized that ASSET will improve
self-perceived psychosocial functioning as evidenced by alignment of the four outcome measures
with qualitative statements. Further, it is hypothesized that participants will report development

9

in career and independent living goals as captured by interview and focus group data.
Significance of the Study
The research literature is replete with descriptions of disparate post-secondary outcomes
for young adults with ASD, yet evidence-based soft skills interventions are lacking. Transition is
a critical period for all youth and may be especially pivotal for youth with HFASD. Social
isolation and co-occurring mental health problems make this group particularly at risk for service
disengagement after high school. Moreover, these students have average to above average
intellectual functioning and often possess strong splinter skills, making them less likely to access
a functional school curriculum. Consequently, both students and families may lack access to
special education and community resources for persons with disabilities.
Accordingly, the results of this study may contribute to the evidence base of transition
practice by validating an intervention that can be easily embedded into multidisciplinary
community contexts, and made accessible to young adults with HFASD who may otherwise not
receive services to address their social and emotional needs. Finally, if validated, the ASSET
intervention will provide practitioners with a manualized program that utilizes group process and
principles of social cognition to impact work-related social skills and the bidirectional
relationships between social functioning, self-efficacy, and anxiety and depression.

10

CHAPTER TWO
Literature Review
Current Understanding of Autism
Autism was previously classified (DSM-IV-TR) as four separate but related
neurodevelopmental diagnoses: autistic disorder, Asperger’s disorder, childhood disintegrative
disorder, and pervasive developmental disorder not otherwise specified – a blanket diagnosis
applied when the other three could not be definitively determined (American Psychiatric
Association [APA], 2013; Levy, Mandel, & Schultz, 2009; Pellicano & Stears, 2011). While
behaviorally descriptive, these four diagnoses were inconsistently applied across treatment
settings, thereby creating confusion among individuals, practitioners, and caregivers (APA,
2013). Recent changes in the Diagnostic and Statistical Manual of Mental Disorders (5 th ed.;
DSM-5; APA, 2013) eliminated Asperger’s disorder, the diagnostic group with the highest
functional levels. Autism is now characterized along a continuum, or spectrum, that accounts for
variations in severity but not type of autism. Howlin and colleagues (2005) estimated that
individuals with autism and without intellectual disability (HFASD) represent approximately 2025% of all diagnosed cases.
Autism spectrum disorder (ASD) is currently understood as a heritable, neurobehavioral
disorder of unknown etiology that is characterized by persistent impairments in social
communication and social interaction; repetitive patterns of behaviors; restrictive interests or
activities; and sensory processing abnormalities (APA, 2013). Due to the lifelong nature of
ASD, early diagnosis and sustained intervention are important to effective management. While
ASD is developmental, it is not always identified early. HFASD, in particular, may be difficult
to diagnose early, as signs of social impairment may not be observable until social demands

11

exceed capacities (APA, 2013). Consequently, individuals with HFASD may not receive a
diagnosis until adolescence or early adulthood, when social demands become increasingly
complex.
Under the new umbrella of ASD, individuals exhibit a broad range of functioning,
symptom severity, and intelligence. Impaired social communication and interaction are
hallmarks of the disorder, and account for the primary disabling effects of autism, regardless of
the individual’s intelligence or verbal ability (Carter, Davis, Kiln, & Volkmar, 2005). Functional
implications include difficulties understanding non-verbal and abstract verbal interactions with
resultant problems in forming friendships and social networks (APA, 2013). Additionally,
mental health disorders, particularly anxiety and mood disorders, are common comorbid
conditions among youth with ASD (deBruin et al., 2007; Joshi et al., 2010). While estimates of
the prevalence of anxiety and depression in the general population vary from as low as 5% to as
high as 32% (Merikangas et al., 2010), White & Roberson-Nay (2009) estimated that nearly half
of all youth with ASD struggle with anxiety and/or depression. More recently, Croen and
colleagues (2015) found that adults with ASD have a 117% higher incidence of anxiety; a 123%
higher incidence of depression, and an astounding 433% higher incidence of suicide attempts
than their peers when controlling for age, sex, and race/ethnicity. Over the past two decades the
reported prevalence of autism has escalated dramatically and so frequently that reporting a
proportion this year may be inaccurate the next. Nevertheless, the most recent estimated
prevalence of ASD at the time this manuscript was written was one in 68 as reported by the
Centers for Disease Control (CDC, 2016), a striking contrast to the one in 1,000 rate reported in
1998 (Bryson & Smith, 1998) when the author of the current study was entering the field. While
this may seem like a dramatic rise, it should be noted that the 1998 estimates excluded those with

12

Asperger’s disorder (i.e., the group with the highest incidence). Furthermore, changes in
diagnostic criteria as specified in the DSM-5 as well as regional sampling methods may at least
partially account for the dramatic rise in prevalence, or proportion of the population that has the
diagnosis. In conclusion, the trends do not necessarily indicate that the incidence, or the rate at
which new cases occur each year, is growing exponentially. Moreover, medical and behavioral
health providers, educators, and families may be doing a better job of early and more accurate
identification.
What is well understood is that impaired social functioning is a hallmark of ASD.
Evidence is mounting to suggest that difficulty regulating the autonomic, physiological
symptoms of heightened arousal cause discomfort and avoidance of social situations in
individuals with ASD (Joseph, Ehrman, McNally, & Keehn, 2008). This interplay of
physiological symptoms and behavioral reactions is similar to the features of anxiety disorders
and may contribute to difficulty reading social cues and responding appropriately to social
interactions. Behaviorally, anxiety and depression may be closely related to some of the central
difficulties of ASD including poor social communication, difficulty with unexpected changes in
routine, and narrow and repetitive interests (Hillier et al., 2011). As a result of repeatedly
uncomfortable and frequently failed interactions, the individual may withdraw from social
engagement, thereby experiencing fewer opportunities to practice social skills and suffering from
increasing anxiety and depression.
Relationship of Social Skills, Self-efficacy, Anxiety and Depression
While essential, social skill development is not innate. Social competence is culturally
contextual and is influenced by disability. Developmental disabilities have been shown to
impact social functioning in both children and adults. In particular, some of the core features of

13

HFASD directly impact social functioning. For example, socially awkward behaviors, poor eye
contact, sensory processing disturbances, difficulty with perspective taking, literalism, repetitive
behaviors, restricted interests, and difficulty reading nonverbal communication each have the
potential to disrupt social interaction, particularly in workplace settings where awareness of ASD
may be limited or misunderstood
Extrapolating from the work of Moe and Zeiss (1982) and Segrin (1996; 2000), the
relationship between social skills, self-efficacy, and psychological wellness can be seen as
cyclical. Poor social skills foster low self-efficacy in social situations leading to anticipatory
anxiety, increased social isolation, depression, limited motivation to practice social competence
and, in turn, meager opportunities to strengthen future social performance. Moreover,
individuals with HFASD are cognizant of their social difficulties. For example, Vickerstaff and
colleagues (2007) replicated research documenting a negative correlation between intelligence
(IQ) and self-perceived social competence among children with HFASD and found that older age
and higher IQ predicted lower self-perceived social competence and higher levels of depressive
symptoms. Furthermore, as children with HFASD transition from school to work or higher
education, social demands become more complex and social deficiencies more noticeable
(Hillier et al., 2011; White & Roberston-Nay, 2009). Consequently, young adults with HFASD
may choose to socially withdraw rather than risk repeated social failure.
In addition to the social skills deficits inherent to the diagnosis, individuals with ASD
experience disproportionate levels of anxiety (Brereton, Tonge, & Einfeld, 2006; Leyfer et al.,
2006) and depression (Brereton et al., 2006; Leyfer et al., 2006; Sterling, Dawson, Estes, &
Greenson, 2008; Stewart, Barnard, Pearson, Hasan, & O’Brien, 2006; Vickerstaff et al., 2007),
two conditions that negatively impact peer relationships (Kim et al., 2000). A review of the

14

counseling and career development literature indicates that self-efficacy is tied to both work and
social performance, and is inversely proportional to affective disorders, such as anxiety and
depression. Specifically, low self-efficacy is strongly negatively correlated with high levels of
social anxiety among children, adolescents, and adults (Leary & Atherton, 1986; Muris, 2001),
and also predicts depression (Bandura, 1997; Muris, 2001). Bandura (1997) posits that low selfefficacy related to feeling unable to effectively cope with stressors, primarily leads to anxiety;
whereas, low self-efficacy related to feeling powerless to affect a desired result tends to predict
depression. Moreover, self-efficacy has long been known to be a key factor in the effectiveness
of behavioral and cognitive interventions for depression (Zeiss, Lewinsohn, & Munoz, 1979).
Although the research literature has not yet identified causal pathways, it is likely that the
effects of social skills and psychological factors are bidirectional (White & Roberson-Nay, 2009)
and, therefore, interventions targeting any one aspect may impact others. However, it should also
be noted that young adults with HFASD often have lifelong struggles with both social skills
(Howlin & Moss, 2012), and anxiety and depression (White & Roberson-Nay, 2009), creating a
tremendous extended burden on their psychological well-being. Furthermore, anxiety and
depression may further impair social function by prompting avoidance behaviors or socially
misaligned thinking and behavior. Thus, the current study examines the relationship between
social functioning, self-efficacy, and psychological wellness in order to better understand how
targeting social functioning may lead to decreased symptoms of anxiety and depression.
Finally, the mental health symptoms that young adults with HFASD experience are not
only prevalent but can also be alarmingly severe. For example, one recent cohort study in the
United Kingdom found that a staggering 66% of a sample (n = 367) of adults with HFASD selfreported suicidal ideation, 35% had plans or attempts at suicide, and 31% reported depression

15

(Cassidy et al., 2014). Moreover, those with HFASD and depression have a significantly
elevated risk for suicidal ideation and behavior when compared to peers with HFASD and
without depression. Knowing that social isolation, exclusion, and unemployment are risk factors
for secondary depression, the cyclical relationships of these constructs highlights the need for
interventions that not only address mental health, but also social functioning and employability.
Employment as the Proving Ground for Psychosocial Functioning
Results of the National Longitudinal Transition Study ([NLTS-2]; Newman et al., 2011)
indicate that youth with ASD experience lower employment rates (37.2%) than any other
disability category, and markedly disparate postsecondary employment outcomes when
compared to their same age peers (Baldwin et al., 2014; Shattuck et al., 2012). Results of a
smaller study of 66 young adults with ASD found that post-secondary employment was a meager
18% for this group, with only 33% of that subset obtaining competitive employment (Taylor &
Seltzer, 2011). In contrast, the employment rate of non-disabled young adults, ages 18-24, is
dramatically higher (54%) and accounts for a sizable number of individuals who are unemployed
due to the pursuit of higher education rather than barriers to work (Pew Research Group, 2012).
Within the state VR context, analysis of the Rehabilitation Service Administration’s
national data (RSA-911) indicated that VR services improved employment outcomes for
consumers with ASD but entailed more services, higher costs, lower resultant wages, and fewer
hours worked than other disability groups (Cimera & Cowen, 2009). Another analysis of RSA911 data, found that participation in sheltered workshops increased service costs, while not
significantly impacting employment outcomes for consumers with ASD (Cimera, Wehman,
West, & Burgess, 2011). Hurbutt and Chalmers (2004) interviewed a small sample of adults
with ASD who had difficulty finding work appropriate to their ability levels, and identified

16

difficulties with social communication, anxiety, stress, and sensory processing as barriers to
gaining and maintaining employment.
Furthermore, the majority of young adults with ASD are not only unemployed, but are
also not participating in employment training or post-secondary education in the critical first two
years after high school (Shattuck et al., 2012). For example, a secondary analysis of NLTS-2
data indicated that, among 680 youth with ASD, 28% had attended a 2-year college; 1%, a 4year college; 9.3%, a vocational or technical education program; and 33% had not participated in
any post-secondary training (Shattuck et al., 2012). These numbers are disappointing, and
represent only slightly higher rates of 2-year and 4-year college attendance when compared to all
students with disabilities who enrolled in 2-year and 4-year institutions (12.8% and 7.7%,
respectively).
Finally, when employment is achieved, individuals with HFASD are prone to short tenure
as a result of poor social functioning. In fact, 90% of all job loss can be explained by deficient
social functioning (Elksnin & Elksnin, 2001). The damage to self-efficacy and mental health
implications that result from unemployment and underemployment can be devastating. Selfefficacy is an action oriented, future focused construct that gives individuals the confidence to
pursue interests, move forward with goals, and manage tasks (Lorenz, Frischling, Cuadros, &
Heinitz, 2016). Career success leads to increased self-efficacy and, ultimately life satisfaction
(Weber, Ruch, Littman-Ovadia, Lavy, & Gai, 2013), whereas repeated job failures negatively
impact career self-efficacy and, consequently, life satisfaction. Pragmatically, poor employment
outcomes present a barrier to independent living and long-term wellbeing as illustrated by
estimates indicating that only 15% of individuals with ASD achieve independence (Howlin et al,
2005).

17

While the psychological effects of long-term under-employment, unemployment, and job
loss have not been extensively studied in the ASD population, substantial research, including
individual longitudinal studies, on the general population identifies high social costs, including
depression, psychiatric hospitalizations, substance abuse, and violence (Dooley, 2003). Poor
employment outcomes foster not only income loss, but also loss of social status, structure, and
routine (Jahoda, 1982). Thus, the need for training and supports is critical in the transition years.
Specifically, social skills development is vital, as poor social communication has been identified
as the most significant barrier to employment for persons with ASD (Chen, Leader, Sung, &
Leahy, 2014; Hurlbutt & Chalmers, 2004).
Service System Limitations and Barriers
Despite federal mandates, and the knowledge, skill, and dedication of professionals
employed within educational, state-federal, and community settings, several systemic barriers to
more fully meeting the unique, complex, and sometimes subtle needs of young adults with
HFASD have been identified.
Special Education. Transition-age students with HFASD, depending on their ability to
fully participate in the general education curriculum and environment, may or may not receive
special education services or Section 504 disability-related accommodations. Even among those
who are eligible for services, diploma requirements in some districts limit opportunities for these
students to enroll in classes that address life skills, such as social and vocational skills (Connor,
2017). Although early work experience has been identified as an evidence-based transition
practice with a moderate level of empirical support, transition youth are less likely than their
non-disabled peers to gain access to work experiences (Wagner, Newman, Cameto, Garza, &
Levine, 2005). Landmark and colleagues (2010) identified several barriers to early work

18

experience for transition youth, including a disconnect between academic and vocational
opportunities, limited funding for vocational and community-based learning, and limited time to
address career development in programs where academic development takes precedence. The
emphasis on academic over vocational, social, and life skills development is particularly
troublesome for youth with HFASD, who may mirror their peers in terms of academic
development, but struggle with soft skills and generalization of learning to adult living contexts
(McKenzie, Evans, & Handley, 2010).
Vocational Rehabilitation. The most recent reauthorization of the Rehabilitation Act,
the Workplace Innovation and Opportunity Act (WIOA, 2014), mandates that state-federal
vocational rehabilitation (VR) offices direct 15% of their budgets to pre-employment training
services, or job readiness training, for transition-age youth. Further, the Act calls for increased
collaboration between schools, VR, and community agencies in the delivery of service. While
the spirit of this legislation is to address disparate outcomes by bolstering service delivery, VR
administrators will be challenged to staff such imperatives. The National Council on Disability
([NCD]; 2008) has recognized that far fewer youth who could benefit from VR services are
actually receiving such interventions due, in part, to a lack of access to services. Furthermore, a
projected shortage of qualified VR counselors (Drew & Peters, 2002; O’Brien & Graham, 2009)
in conjunction with historically high counselor caseloads (Cloutier et al., 2014; Kierpiec,
Phillips, & Kosciulek, 2010) inhibits the ability of local VR offices to staff and deliver transition
services in the schools. Yet, while systems struggle to meet demand, legislation continues to
mandate expansion of VR services to greater numbers of students and those with the most
significant disabilities (Everson & Moon, 2007; Oertle & Trach, 2007; Plotner, 2009; WIOA,
2014). The increased demand for service to the most significantly disabled in combination with

19

staff shortages creates a significant barrier to individuals with HFASD who may not meet
eligibility requirements of state agencies and concurrently may only receive academic
programing in school. Thus, youth with HFASD are at risk of falling through the cracks. In fact,
previous research (Connor, 2017) has indicated that some youth with HFASD have been told
they are “not disabled enough” to qualify for programming through state agencies such as
Community Mental Health (CMH).
In addition to the lack of services, systems also face barriers to hiring qualified providers.
Critical shortages of certified rehabilitation counselors within VR systems have resulted in the
relaxing of standards for qualified providers of vocational counseling. Concurrent with the
passing of WIOA, the Commission on State Vocational Rehabilitation (CSVR) weakened the
Comprehensive System of Personnel Development (CSPD) and no longer requires master’s-level
preservice training and counselor certification in the delivery of vocational rehabilitation
services. Ironically, these legislative and policy changes threaten to perpetuate rationing and
watering down of services, particularly for students labeled as high functioning. Finally, the 2008
NCD brief summarizes: “both the quantitative and qualitative data reviewed indicated that lack
of personnel, service unit credit policies, and dedicated transition units in local rehabilitation
agencies limit the impact and effectiveness of VR collaboration with other agencies involved in
service delivery” (p. 12).
In addition to staff shortages, VR agencies may lack staff with expertise in autism, and
also lack capacity to provide the level of supports necessary for successful employment. In
effect, traditional VR approaches may not meet the unique needs of individuals with autism. For
instance, Wehman and colleagues (2014) observed that there is a growing demand for VR
services among individuals with ASD, yet the effectiveness of VR services in placing such

20

individuals in competitive employment has actually declined in recent years. Clearly,
educational and VR systems will need to find new and empirically validated approaches to
serving this growing population.
Need for Evidence-based Transition Interventions
Evidence based practices are defined as effective practices based on rigorous research
designs evaluated and ranked through systematic review using quality indicators, such as use of
randomization and control (NSTTAC, 2014). Similarly to health care, contemporary
rehabilitation policy and service provision will increasingly be driven by empirical validation of
intervention feasibility, efficacy, and effectiveness, particularly for subpopulations with the
worst outcomes (Chan et al., 2009; Leahy, Tarvydas, Connor & Landon, in press). Accordingly,
researchers will need to develop, extend, and validate models that guide the systematic delivery
of interventions and drive future research efforts (Chan et al., 2009). Several evidence-based
practices (EBPs), including supported employment and motivational interviewing, have been
identified within the context of state-federal VR (Leahy et al., 2014); however, models that
address the unique work-related social and mental health needs of youth with HFASD have yet
to be developed. For instance, the place and train model espoused by supported employment has
been shown to be effective for individuals with more severe symptoms and lower adaptive
functioning (Wehman et al., 2014), but there are few evidence-based employment interventions
targeting adults with HFASD. In fact, only one study of social skills interventions targeting
individuals with disabilities used randomized control and attention to fidelity in assessing
effectiveness (Murray & Doren, 2013). Furthermore, this study included youth with a number of
disability types and was not specifically designed to meet the unique needs associated with ASD.
Moreover, state-federal VR has struggled with serving this population, particularly at the

21

high school level. For example, person-centered planning has been identified as an evidencebased intervention that is well-suited to effective transition planning (NSTTAC, 2015), yet VR
counselors have such limited time to provide school-based interventions that it is unrealistic to
expect them to facilitate meetings much less provide ongoing, direct interventions with youth
(Cloutier et al., 2014). In fact, Cloutier and colleagues (2014) go so far as to recommend that
state offices refer transition youth to private counselors for comprehensive, evidence-based
services such as person-centered planning. Thus, both the quality and quantity of vocational
intervention during the transition years is lacking at the state agency levels.
Furthermore, within school contexts, transition needs tend to be segregated, and
interventions are largely practice-based and rather than evidence-based (Cobb & Allwell, 2009).
Despite federal mandates for evidence-based practices in special education programming,
including transition-related services, Westbrook and colleagues (2013) described a “remarkable”
lack of intervention efficacy documented in the research literature. The Individuals with
Disabilities Education Act (IDEA) mandates evidence-based practices (EBPs) in all Special
Education programming (Turnbull, Wilcox, & Stowe, 2002). Yet, EBP is not widely
substantiated in transition. Cobb and Alwell (2009) described the transition literature as merely
descriptive, while Landmark et al. (2010) identified “many opinion-based practices” (p. 174).
Test, Fowler, and Kohler (2013) reviewed the transition literature and found only two transition
EBPs with a high level of evidence. Landmark, Song Ju, & Dalun Zhang (2010) observed that
even “best practices,” or those with the greatest number of studies, are not considered EBP.
Consequently, they called for the use of effect sizes to make meaningful comparisons feasible.
Gersten, Fuchs, Compton, Coyne, Greenwood, and Innocenti (2005) promoted universal quality
indicators as benchmarks of the rigor of studies, yet much of the extant literature points to “best

22

practices”, “promising practices,” or “substantiated practices,” and fails to establish evidencebased practice. NSTTAC (2014) continues to encourage increased research on EBPs addressing
barriers to successful transition for youth with ASD.
Ashley, Kriner, Stehle, and Sewell (2012) described weak evidence as one barrier to the
development of needed transition programs. This lack of robust evidence regarding transition
services has been discussed now for decades. Both the National Council on Disability (1999)
and Halpern (1994) identified the segregation of transition planning and services from the
broader context of school services as a barrier to the inclusion of transition in policy initiatives.
Benz & Halpern (1993) identified the overuse of generalized transition goals as prohibitive,
while multiple studies found limited options, supports (Furney, Hasazi, & Destefano, 1997;
Wehman & Revell, 1997), knowledge, and involvement in transition (Wehman & Revell, 1997)
to be problematic. Furthermore, attitudes can present barriers or create opportunities for the
implementation of EBP. Professionals may perceive EBP to be time consuming, academic, and
impractical but, when translated effectively, EBPs can be pragmatic. Stahmer & Aarons (2009)
found increasing resistance to new EBPs among mature professionals, but general acceptance of
EBPs perceived as intuitive, logical, and accepted by colleagues. They also noted the influence
of family attitudes, including both advocacy and litigation, on the call for more EBPs in school
settings.
However, in a systematic review of the ASD literature, Spain & Blainey (2015) did find
preliminary but promising evidence that group social skills interventions for high-functioning
adults with ASD are “effective for enhancing social knowledge and understanding, improving
social functioning, reducing loneliness and potentially alleviating comorbid psychiatric
symptoms” (p. 874). For example, Hillier and colleagues (2011) recognized anxiety and

23

depression as factors related to social difficulties in individuals with ASD and found that social
and vocational skills training was associated with significantly lower anxiety and depression
among young adults with ASD. Similarly, Laugeson and colleagues (2012) developed one of the
few evidence-based social skills interventions for adolescents with HFASD and reported
significant, durable improvements in social functioning over time. While promising, such studies
are few in number, and only one (Hillier et al., 2011) addressed work-related social functioning.
Despite the challenges and current limitations, high-level EBPs provide school districts
and state agencies with the best assurances of program effectiveness. Cooney, Huser, Small, and
O’Connor (2007) extolled EBPs for their ability to limit the use of unproven resources, mitigate
the need for development of in-house programs, and increase the likelihood of success. Given the
paucity of empirical evidence and the significant needs of young adults with HFASD, it is
imperative that high-quality experimental and multivariate correlational research is conducted
(Test et al., 2013). Through the establishment of an evidence base of practice for work-related
social skills programs, young adults with HFASD will not only gain access to effective social
training, but can also safe, effective, non-pharmaceutical treatments for the symptoms of social
anxiety and depression. Therefore, replication and extension among multiple researchers are
necessary in order to build the evidence base.
Current Social Skills Interventions
In a systematic review of transition planning and coordinating activities for youth with an
array of disabilities, Cobb and Alwell (2009) identified 156 intervention studies targeting
improvements in four outcome areas: social and communicative skills, functional/life skills,
school-based and community-based employment training, and self-determination skills. Among
this group, 30 interventions focused on social and communication skill development in

24

adolescents with disabilities. Although none of the studies focused exclusively on ASD, gains
were observed in work-related behaviors (Kohler, 1994), career self-efficacy (Hua, 2002), and
social competence (Chadsey-Rusch, 1990). While promising, these studies offer only
preliminary evidence of the efficacy of group-based social skills trainings due to issues with
study design, and use of diverse samples from multiple disability groups. While helpful in
demonstrating the need for further research, many were qualitative designs and one was a single
case study. A more recent review by Cappadocia and Weiss (2011) provided ASD-specific, yet
still preliminary, evidence on the efficacy of group social skills interventions for children with
ASD. However, given the focus on children, none of these interventions targeted workplace
social skills.
Perhaps the most targeted and compelling results thus far were reported by Laugeson and
colleagues (2014) who described a classroom, group-based, social skills intervention targeting
middle school students with HFASD. This teacher facilitated, school-based intervention resulted
in gains in social functioning across multiple measures. While the gold standard of
randomization and control were not feasible in this classroom intervention, the program does
offer an evidence-based, teacher-led intervention for middle school students with HFASD. Only
one study identified to date has focused on changes in anxiety and depression following groupbased social and vocational skill training (Hillier et al., 2011), and while significant
improvements were noted, effect sizes were small. Accordingly, the authors call for increased
duration and/or frequency of sessions in order to potentially see greater effect sizes.
Furthermore, the intervention appears to be compartmentalized (e.g., employment addressed in
week seven) rather than an integrated focus on work-related social skills. Finally, previous
related studies have not tested theoretical constructs. Clearly, a contemporary and more rigorous

25

research focus on group-based social skills interventions is needed to extend the promising
results achieved to date, and provide targeted, theory-based information on the work-related
social skills needs of young adults with HFASD.
Social Cognitive Career Theory
Social Cognitive Career Theory (SCCT) is an extension and adaptation of social
cognitive theory to career development applications. SCCT provides constructs to aid
understanding of how individuals develop career and academic interests, make choices, and
achieve career and educational successes (Lent et al., 1994). The primary mechanisms that drive
interests and choices are: self-efficacy, outcomes expectations, and goals. SCCT posits that
skills and abilities influence one’s self-efficacy and outcome expectations, which in turn shape
performance, goals and, ultimately, career outcomes.
Self-efficacy (SE). SE refers to personal beliefs about one’s ability to perform particular
behaviors or courses of action, and has been defined as the level of confidence that one has in
being able to act in order to affect a desired outcome (Brown et al., 2014). It should be noted
that self-efficacy is domain-specific and contextual, i.e. self-efficacy is not a durable trait but,
rather, a perception or belief about a particular task. Furthermore, self-efficacy need not align
with abilities. In fact, Brown and colleagues (2000) found that self-efficacy was more predictive
of career interests and perceived career options than self-assessed abilities. SE has been further
divided and measured by subtypes, including general, empathic, and social. Furthermore,
because SE is task specific, many other subtypes have and can be developed.
Outcome expectations. Outcome expectations (not to be confused with outcome
expectancy, or placebo effect) have been defined as “a person’s estimate that a given behavior
will lead to certain outcomes” (Bandura, 1977, p. 193). According to Bandura, outcome

26

expectations are related but distinct from self-efficacy. Specifically, self-efficacy is the
“perceived ability to do a behavior, whereas outcome expectations are judgments about the
likelihood of outcomes that flow from behavior” (Williams, 2010, p. 418). It should also be
observed (see Figure 1.2) that outcome expectations do not impact self-efficacy, while selfefficacy does influence outcome expectations.
Figure 2.1
Social Cognitive Career Theory (Lent, Brown, & Hackett, 1994, 2000, 2002.)

For example, if a young adult believes that teaching is a rewarding career (outcome expectation)
but has low self-efficacy beliefs regarding ability to manage students, then the individual may
not consider education as a viable career choice. This scenario illustrates how self-efficacy is the
primary driver of goals, interests, and activities.
Goals. Goal setting helps individuals to organize and direct behavior over time in order
to meet predetermined parameters (Lent et al., 1994). In the career development literature, goals
are sometimes referred to as career plans or aspirations.
SCCT and disability. Self-efficacy (SE) and outcome expectations (OE) are seen as key

27

drivers of career success, but the interactive nature of each factor is also inherent to the theory,
and points to the utility of skill development, practice, and feedback as a means not only to
improve performance, but also to drive self-efficacy and career expectancy. Specifically, workbased skills, such as soft skills, require development through introduction, practice, and
feedback; in turn, skill mastery leads to bolstered self-efficacy and more optimistic career
expectancy (Lent & Brown, 1996, 2013). Moreover, this cycle is both multidimensional and
iterative; that is, as new skills are targeted and mastered, self-efficacy is continually bolstered,
and career expectancy becomes more nuanced and skills based.
SCCT has been used to show how vocational rehabilitation practices can be linked to
theory so that rehabilitation researchers can use consistent constructs in examining common
factors and underlying mechanisms for change (Fabian, 2000). In addition, SCCT has been
applied to examining the rehabilitation needs of individuals with various physical, emotional,
and developmental disabilities, including adolescents (Betz, 2007; Fabian, 2000). Pragmatically,
group interventions targeting SCCT constructs have been shown to be helpful in bolstering
career-related self-efficacy among college students with disabilities (Corrigan, Jones, &
McWhirter, 2001). Recognizing that adolescents are prime for cognitive and behavioral shaping
in relation to career expectancy and skill development, and further noting that SCCT accounts
for both intrapersonal and social factors that influence career development, SCCT is the logical
choice in framing this study of the relationship between social functioning, and self-efficacy, and
psychological wellness.
Furthermore, Fabian (2000) posits that individuals with disabilities are particularly at risk
due to environmental factors, such as under-stimulating or segregated learning environments,
and lack of access to quality community-based rehabilitation supports and services. Although

28

expectancy may be negatively influenced by stigma and stereotypes toward persons with
disabilities, cognitions are shapeable and may be particularly malleable during adolescence, a
developmental period in which physical, cognitive, and emotional growth is pronounced. Thus,
targeted approaches to shape career expectancy among transition-age youth are necessary to
improve postsecondary outcomes (Fouad & Guillen, 2006; Murray & Doren, 2016). In addition
to positive career expectancy, transition-age youth need to develop job-related social skills in
order to both obtain and retain employment (Elksnin & Elksnin, 2001). Recognizing that the
social, psychological, and cognitive ramifications of disability can be barriers to employment and
social integration, this proposal suggests that SCCT be used to guide the research questions and
analyses.
High comorbidity of anxiety and depression in the ASD population further complicates
the relationship between personal factors, self-efficacy, and outcome expectations. In fact,
depression and anxiety are inversely related to self-efficacy, such that strong self-efficacy tends
to reduce stress, decrease vulnerability to depression, and drive accomplishment (Tahmassian &
Jalali Moghadam, 2011). Conversely, significant levels of anxiety and depression reduce
interests, goals, and actions, thereby weakening self-efficacy. Given the seemingly bidirectional
relationship between anxiety/depression and self-efficacy, it is apparent that improving
functional outcomes relies on addressing mental health, intervening to improve self-efficacy, or
some combination of the two.
Innovation
The ASSET social skills intervention is unique in that it addresses work-related social
skills and is focused on young adults, rather than children with ASD. Furthermore, the program
employs increased duration (1.5-hour sessions vs. 1-hour sessions used by Hillier et al., 2011),

29

and was designed to spur future focused, goal directed tasks that improve career self-efficacy.
For example, the curriculum engages participants in discussion, video review, role-play, and
feedback that allows for scaffolding of learning and practice of workplace social behaviors in a
safe, group context. With a focus on the development of work-related social skills, adaptive
functioning, and self-efficacy to foster successful school-to-work outcomes, the mission and
activities of the ASSET program align with the spirit of legislation mandating expanded services
for transition youth (WIOA, 2014). Furthermore, while the law is aspirational, ASSET is both
pragmatic and innovative in designing an empirically based, disability specific intervention that
can be delivered by a variety of professionals in a group format within community contexts.
ASSET’s collaborative, interdisciplinary origin and iterative, developmental program evaluation
ensure that the intervention is well suited to community contexts where service providers
struggle with high caseloads (Cloutier et al., 2014; Kierpiec, Phillips, & Kosciulek, 2010), and
have limited access to evidence-based and manualized interventions (Westbrook et al., 2013).
Lastly, by offering a program design that is easily embedded within school and/or community
contexts, ASSET minimizes barriers to service common to youth with disabilities, such as poor
social functioning, limited student/family involvement, limited access to transportation, poor
interagency collaboration, unrealistic expectations, and difficulty navigating eligibility
requirements (Riesen et al., 2014).
In addition, ASSET seeks to partially fulfill family needs. ASD is a highly heritable
neurodevelopmental disorder (Levy et al., 2009) and, therefore, students with ASD may have
parents and other family members who struggle with similar symptoms and functional deficits.
Given the tremendous stress that caregivers face and the lack of a systematic approach to
addressing autism related mental health needs, it is unrealistic to expect that families can fully

30

meet the social skills training needs of their young adult children. Therefore, integrated delivery
of service is particularly important for youth and their families as they are accustomed to the
entitlement services of special education and may struggle to identify and connect with the
multiple agencies that provide adult services.
While negative or unrealistic self-appraisal is associated with disability, it is not innate to
disability but, rather, socially constructed. In fact, Fabian (2000) posits that rich, supportive and
nonjudgmental learning contexts can foster realistic self-appraisal. In contrast to traditional
classroom settings, the ASSET school-based group intervention seeks to disrupt traditional
power differentials between the teacher and student, thereby fostering supportive relationships
without judgment. In essence, the ASSET group context is designed to mimic professional
settings rather than classroom contexts. The professional milieu offers participants a safe and
natural environment in which to practice skills, give and receive peer feedback, exercise healthy
competition, and enact professional role-play. Furthermore, the ASSET group context minimizes
disability-related stigma that may affect performance in fully immersive contexts, thereby
creating a favorable environment for successful scaffolding of soft skill development from
practice to real-world settings. Finally, the group interaction provides opportunities for vicarious
learning (Bandura, 1986) through live and video modeling of positive social behaviors by peers
and guest experts in order to bolster participants’ self-efficacy.
Literature Summary
As our current understanding of ASD as a heritable and persistent condition of
questionable etiology becomes more nuanced and sophisticated, so too must our understanding
of the unique needs of individuals with HFASD. The need for intervention research relevant to
adults with HFASD and focused on functional issues, such as work and higher education, and

31

psychological wellness is great. Examining the close and cyclical relationship between social
skills, self-efficacy, and psychological wellness, particularly within the critical period around
transition, can aid in the development of interventions that address some of the unmet
employment and mental health needs of this population. SCCT provides a relevant and welldeveloped framework for examining these issues and growing the evidence base of practice
through use of common constructs. Lastly, although systemic barriers are numerous, recent
legislation has brought new attention to the need for pre-employment services for our most
vulnerable youth. Consequently, researchers will need to not only generate new knowledge and
effectively translate research into practice, but also innovate. Community-based,
interdisciplinary, empirically validated, and manualized interventions, like ASSET, will be
needed if we are to more fully realize the intent of the latest federal legislation.

32

CHAPTER THREE
Methods
This chapter provides details on the research design, sample, procedures, intervention,
instrumentation, and data analysis plan.
Design
This mixed-method, quasi-experimental, single pooled group, repeated measures time
series study uses quantitative data to measure program effects over time (baseline, postintervention, and 2-month follow-up) on work-related social skills, self-efficacy, depression, and
anxiety. Exploration of qualitative interview data was used to elaborate on participant
perceptions of present levels and change over time in dependent variables.
Sampling
Participants. A purposive sample of 18 students, comprising four intervention groups
was recruited across two cohorts. Seventeen young adults ranging from 19 to 23 years of age (M
= 20.5; SD = 1.28) with HFASD and without intellectual disability (IQ mean = 95.11; SD =
18.08) participated in and completed the ASSET program in the fall of 2014 and 2015. Basic
reading ability (a composite of word fluency and attack subscales) was in the average range, (M
= 90.45; SD = 24.03). All participants had a previous diagnosis of Autistic Disorder, Asperger’s
Disorder, or Pervasive Developmental Disorder – Not Otherwise Specified (PDD-NOS)
according to DSM-IV-TR criteria (APA, 2013). Two participants (11.76%) were female and the
remaining 15 (88.24%), male. Nine participants (52.94%) identified themselves as Caucasian;
six (35.29%) as Black or African American; and two (11.76%) as Asian. All were enrolled as
students in college, high school, or special education career services, and resided with parent/s or
caregiver/s at the time of intervention.

33

While a comparison of data from independent raters on the SRS (self, parent, teacher)
would have been ideal, this analysis was not suited to the ASSET dataset. College participants
completed self-report exclusively, so no comparative data was available for them. Secondly, few
high school students were able to complete the SRS without one-on-one assistance from
researchers or parents, and this process was not only unduly time consuming but also introduced
potential bias in reporting. For example, one parent was observed telling his son not to answer in
the way he had when verbally asked for a response. Finally, very few teacher reports were
collected. In fact, when cohort three data collection began, two teachers wrote to say that the
data process was too lengthy and that they would no longer complete the SRS and two other
measures. For these reasons, it was not possible to generate meaningful correlation data between
all three types of reporters and, consequently, averaged scores across all three types of raters
were used in all relevant analyses.
Procedures
After obtaining research approval from the Michigan State University Institutional
Review Board, participants were recruited from two locations: a disability resource center in a
university located in a mid-Western state and a local intermediate school district. The informed
consent process included an explanation of the purpose, timeframe, potential risks and benefits
of the study. Participation was voluntary and offered at no cost. Potential participants were
required to attend an in-person screening interview and to complete descriptive self- and/or
other-report (parent, teacher) measures in order to become eligible for the program. Eligibility
criteria included: (a) 18–24 years of age; (b) previous diagnosis of autistic disorders, Asperger’s
disorder, PDD-NOS according to DSM-IV-TR; (c) social problems as reported by caregiver
and/or pre-intervention self-assessment measures; (d) no current participation in other work-

34

related social skills treatment; (e) no history of other neurological disorders; (f) no history of
major mental illness (e.g., schizophrenia), (g) composite IQ score > 70 on the Wechsler
Abbreviated Scale of Intelligence: Second Edition (WASI-2; Wechsler, 2011), (h) minimum
score of 6 on the Autism Spectrum Quotient (AQ-10; Baron-Cohen, Wheelwright, Skinner,
Martin, & Clubley, 2001), and (i) minimum basic reading composite of 80 or above [(low
average to very superior); WJ-ACH-III; Schrank, Mether, & McGrew, 2014).
After baseline assessment, eligible participants were offered an 8-week group
intervention followed by three post assessments: one within two weeks of program completion, a
second follow-up assessment 8-10 weeks post, and a final 1-2 year assessment and focus group.
Each interview/assessment meeting was approximately 120 minutes in length.
Intervention
ASSET targeted work-related social skills (soft skills) development among three cohorts
of students with HFASD, ages 18-24, in both special education and college contexts. The
ASSET program consisted of eight weekly 90-minute sessions offered in a community-based
conference room setting. In fitting with group counseling conventions (Corey, Corey & Corey,
2014), each ASSET group was comprised of 4-5 students from similar educational backgrounds
(college students with a career focus, or transition-age youth with an entry-level employment
focus). Group sessions were led by a group facilitator with standby support from a research
assistant who modified curriculum and procedures based on observations of participants,
facilitators, as well as participant and facilitator feedback collected at the end of each session.
Facilitators and research assistants had a minimum of a master’s degree, and came from the
fields of rehabilitation counseling, occupational therapy, and special education. Facilitators were
trained and supervised by rehabilitation counseling faculty researcher who attended sessions and

35

conducted weekly team meetings with all facilitators, research assistants, faculty consultants, and
technical support assistants. The intervention team met weekly for 2-4 hours to debrief and
ensure fidelity. Two doctoral students and one post-doctoral fellow were involved in curriculum
development, group facilitation, and monitoring for treatment fidelity. All lessons were guided
by a lesson plan and supplementary materials, including PowerPoint presentations, handouts,
agenda, and various supplies related to group activities. An unpublished manual was created and
continually revised throughout the pilot. Table 3.1 summarizes curriculum content by week.
Table 3.1
Summary of ASSET curriculum and examples of key points
Topic

Examples of Key Points

1 Introduction

Introduction & set up ground rules

2 Communication

Oral expressions & body language

3 Attitude & Enthusiasm

Ways to convey such attitudes

4 Teamwork

Roles in team activity & essentials of good team player

5 Networking

Build contacts & develop elevator pitch

6 Problem-solving &
Critical Thinking

IDEAL problem solving model

7 Professionalism

Professional behaviors & manners

8 Graduation

Review, graduation ceremony & presentation of group project

The original training curriculum was developed by the U.S. Department of Labor’s Office of
Disability Employment Policy (ODEP) and published in 2012, namely “Skills to Pay the Bills:
Mastering Soft Skills for Workplace Success” (SPB), which was intended to meet the basic soft
skills training needs for all youth, including those with disabilities. The SPB curriculum covers
six key skill areas of social skills: (a) Communication; (b) Networking; (c) Enthusiasm and
Attitude; (d) Teamwork; (e) Problem Solving and Critical Thinking; and (f) Professionalism.

36

Based on the Skills to Pay the Bills framework, each ASSET lesson was customized for the ASD
population, and new content was developed based on multidisciplinary team collaboration, the
research literature, and other available and relevant resources. The SPB curriculum was a
starting point for development of the ASSET curriculum. While the six key skill areas remain
the same, the curriculum was adapted to target ASD behaviors and accommodate ASD
characteristics. Through an iterative developmental process, the modules and pedagogy have
been subtly refined with expert input from professionals in rehabilitation counseling, special
education, and occupational therapy.
Pedagogically, each session employed a combination of warm up activities, didactic
lessons, role-play, peer and facilitator performance feedback, and socialization, which included a
dinner hour. Facilitators used individually directed questions, probing and clarifying statements,
and summaries to engage participants in group discussion. Knowing that difficulty generalizing
learning across contexts is problematic for this population (McKenzie et al., 2010), topics were
covered using a didactic-discussion-practice sequence with explicit instruction about the
importance of using learned skills outside of the group context. In addition, each session began
with a warm-up activity that not only illustrated the topic, but also got students moving, engaged,
and collaborating with peers. In order to provide opportunity to practice work-related social
skills in a natural context (White, Koenig, & Scahill, 2007), sessions took place in a professional
conference room, not a classroom, and homework was assigned to provide opportunities for
practice of skills such as networking, and leaving a professional voicemail/email outside of the
group context. Furthermore, sessions were followed by a social gathering, which included
sharing a meal and conversation among students, facilitators and, on occasion, parents, high
school teacher and graduates of the program.

37

Instrumentation
Implementation of the ASSET social skills intervention program was the independent
variable while specific research hypotheses were developed to measure three dependent
variables/constructs: 1) social functioning (social cognition and communication, 2) self-efficacy
(social and empathy), and 3) psychological wellness (anxiety and depression); together, these
constructs are defined as psychosocial functioning. All participants completed a demographic
form identifying information related to age, education, gender, ethnicity, social security benefits,
diagnosis, medications, and work related activities and experiences. Appropriate testing
accommodations, including providing a reader and scribe, were provided to help participants
complete the questionnaires as needed.
Screening instruments. All diagnostic screening assessments were administered,
scored, and interpreted by trained doctoral level psychology students with supervision from a
licensed psychologist.
Diagnostic screening. ASD diagnosis was confirmed using the Autism Spectrum
Quotient (AQ; Baron-Cohen et al., 2001) at baseline and, if identified, those with scores <26
would have been excluded. The AQ is a 50-item self/parent report rating scale that measures
autistic traits across five subscales: social skills, attention shifting, attention to detail,
communication, and imagination. Test-retest reliability (.70) and internal consistency (.82) was
acceptable in a validation study using neurotypical college students, and adults and adolescents
with ASD (Laugeson, Gantman, Kapp, Orenski, & Ellingsen, 2015). The threshold score is 26
and higher scores indicate stronger autistic traits.
Cognitive assessment. The Wechsler Abbreviated Scale of Intelligence: Second Edition
(WASI-II; Wechsler, 2011) was used to assess cognitive functioning. The WASI-II is a norm-

38

referenced instrument designed to assess cognitive functioning of individuals ages 6-90 years
old, and contains four subtests: Block Design, Vocabulary, Matrix Reasoning, and Similarities
that when combined form a Full-Scale IQ, Verbal Comprehension Index, and a Perceptual
Reasoning Index (McCrimmon & Smith, 2013). Test-retest reliability for the full, verbal, and
perceptual scales ranges from .86 to .90 (Irby & Floyd, 2013). Those with WASI-II composite
scores of < 70 were excluded with the exception of one individual who demonstrated positive
work behaviors and was able to fully participate given accommodations for reading and written
output.
Reading assessment. The Woodcock Johnson Tests of Achievement: Third Edition (WJACH-III; Schrank, Mether, & McGrew, 2014) was used to measure individuals’ ability to read
and comprehend English. The median reliability coefficient alphas for all age groups for the
standard battery of the WJ III ACH for tests 1 through 12 ranged from .81 to .94 (McGrew,
LaForte, & Schrank, 2014). The WJ III demonstrates construct, convergent, and divergent
validity in comparison to other measures of academic achievement (McGrew, LaForte, &
Schrank, 2014). For the current study, participants with WJ III composite (combined reading
fluency and word attack) scores less than 80 were ineligible to participate, with the exception of
the one individual (see above) who scored below 70 on the WASI-II and received a no score on
the WJ fluency subtest due to poor decoding skills.
Primary outcome measures. Due to the repeated measures design of the study,
outcome measures with strong temporal consistency (test-retest reliability) were preferred. In
addition, because all outcome measures in this study are self-report, good discriminant validity
was desired (Cone and Foster, 2006). When available, multiple raters (parent, teacher, self) were
recruited to assess social functioning.

39

Social functioning. The Social Responsiveness Scale, Second Edition (SRS-2) was used
to measure severity of ASD symptoms and social skill functioning (Constantino, 2012). The
SRS-2 is a 65-item rating scale measuring social functioning categorized into five subscales:
Social Awareness, Social Cognition, Social Communication, Social Motivation and Restricted
Interests, and Repetitive Behavior. Each item is scored on a 4-point Likert-type scale, ranging
from 1 (not true) to 4 (almost always true). A total raw score above 70 is diagnostically
significant, with sensitivity value of .78 and specificity value of .94 (Constantino, 2012).
Reliability and validity of the SRS-2 is supported by internal consistency reliability ranges from
.93 to .97, temporal stability estimates of r = .88, and convergent validity between parent and
teacher ratings of r = .72 (Bolte, Poustka, & Constantino, 2008; Constantino et al., 2007).
For the purposes of this study, the social cognition, and social communication subscales
of the SRS were selected as the subscales most relevant to the first research question. The RRB
(restrictive interests and repetitive behavior) subscale was eliminated as it measures disabilityrelated characteristics that were not targeted by the ASSET intervention. The social motivation
subscale was also eliminated from analysis as it too captures a construct that ASSET does not
address. Further, given that participants were adult volunteers who sought out the intervention,
social motivation bias was introduced into the study as those who chose to participate likely had
inherently greater social motivation than those who did not express interest in ASSET. Thus, the
social cognition and social communication subscales were selected as the most appropriate
outcome measures because the intervention directly targets those specific social skill subsets.
Perceived self-efficacy. Self-efficacy was measured using two scales targeting different
aspects the of workplace social interactions as described below.
Empathic self-efficacy. The Perceived Empathic Self-Efficacy Scale (PESE), a 6-item

40

self-report inventory, was used to measure participants’ perceived ability to exercise theory of
mind in understanding the feelings of others and to be both sensitive and responsive to the
emotional states of others (Caprara & Steca, 2005). The PESE uses a five-point Likert-type
scale, ranging from 1 (not well at all) to 5 (very well). Internal consistency reliability estimates
have been found to be .78, .80, .81 across three different samples (Giunta et al., 2010).
Social self-efficacy. The Perceived Social Self-Efficacy Scale (PSSE), a 5-item inventory
was used to measure participants’ self-perceived ability to express opinions, share personal
experiences, work cooperatively, and manage interpersonal conflict (PSSE; Caprara & Steca,
2005). Respondents were asked to rate each item on a five-point scale from 1 (not well at all) to
5 (very well). Internal consistency reliability was .69, .76, and .66 in three different samples
(Giunta et al., 2010).
Secondary outcome measures. These measures were used to assess the psychological
wellness construct.
Depression. Due to the scarcity of validated and normed measures that specifically
assess psychosocial adaptation among young adults with ASD (Gantman et al., 2011),
assessment instruments developed for neurotypical populations were employed. The Patient
Health Questionnaire-9 (PHQ-9; Spitzer, Williams, & Kroenke et al., 2000), a nine-item selfreport inventory, was used to assess depressive symptoms. The PHQ-9 uses a four-point Likerttype scale, ranging from 0 (not at all) to 4 (nearly every day) with total scores ranging from 0 to
27, and internal reliability estimates of .89 and .96 (Kroenke, Spitzer, & Williams, 2001).
Sensitivity is 73%, with high specificity (98%) for diagnosing depression in adults, and score cut
points of 5 indicate mild depressive symptoms; 10, moderate; 15, moderately severe; and 20,
severe (Cassin et al., 2013).

41

Anxiety. The General Anxiety Disorder Questionnaire (GAD-7; Spitzer, Williams, &
Kroenke et al., 2006), a seven-item self-report measure, was used to measure anxiety. The
GAD-7 uses a four-point Likert-type scale to assess frequency of symptoms, ranging from 0 (not
at all) to 4 (nearly every day). Total scores range from 0 to 21, and internal consistency and testretest reliability were .92 and .83, respectively (Spitzer et al., 2006). Score cut points of 5, 10,
and 15 respectively indicate mild, moderate, and severe anxiety. Ten is the threshold of clinical
significance.
Data Analysis Plan
Preliminary quantitative analysis plan. As recommended by Cone and Foster (2006),
the following preliminary analyses were conducted to examine: (1) characteristics of the data, (2)
properties of the measures used, (3) possible confounding variables, and (4) adherence of the
data to statistical test assumptions.
Data entry for each variable was audited at random to reconcile paper, Excel, and SPSS
entries. Before analysis, all data were checked for errors. Specifically, SPSS 22.0 was used to
generate frequency tables and case summaries in order to identify out of range scores, entry
errors, and missing data. When missing values were discovered on self-report measures, means
of other available indicators were used to replace these values. For example, if two out of 12
responses for the SRS social cognition subscale were missing completely at random, then these
missing item values were replaced with the mean value for that subtest. Cone and Foster (2006,
p. 243) caution that such use of “unweighted means estimation” (Winer, Brown, & Michels,
1991) can produce bias. However, they go on to note that bias is typically small when the
remaining items have high internal consistency. Thus, unweighted means estimation was used
only within subscales, not across, and only in instances when the majority of items were

42

complete.
In order to determine whether the data met the underlying assumptions for parametric
statistical analysis, all continuous dependent variables were examined for approximate normality
of distribution, i.e. skewness and kurtosis, using histograms to inspect the shape of the
distributions (Tabachnick & Fidell, 2013, p. 81). In addition, the Shapiro-Wilk test was applied
to all measures to ensure approximate normality of distribution. Acceptable limits of skewness
and kurtosis were set at Âą2 (Field, 2000, 2009; Gravetter & Wallnau, 2014; Trochim &
Donnelly, 2006), and all variables were within these limits. Further, trimmed means (5%) were
compared with original means to ensure that extreme scores were not strongly influencing group
means for each variable (Pallant, 2013). As recommended by the American Psychological
Association Task Force on Statistical Inference (Wilkinson et al., 1999), graphical analysis was
used in addition to statistical analysis to screen for normality, linear relationships, and extreme
scores. Specifically, box plots were examined to identify potential outliers, while histograms,
Normal Q-Q plots, and Detrended Normal Q-Q plots were inspected for normality of
distributions. Scatterplots were used to visually assess for linear relationships between potential
covariates and dependent variables. Based on visual screening, the similarity of mean scores to
the 5% trimmed means, and acceptable skewness and kurtosis, four outliers were identified and
removed. Mean scores on all measures were calculated by downloading the raw data into
Microsoft Excel, restructuring to case format, and computing values. The IBM Statistical
Package for Social Sciences (SPSS) 24.0 for Mac and Excel were used to manage raw data and
perform all analyses.
The need for multilevel modeling. While dependent samples t-tests revealed some
noteworthy preliminary results, the data analysis plan needed to account for: (1) between-

43

participant factors, such as socioeconomic status, and (2) within-participant comparison of mean
differences on the dependent variables as a result of the intervention. Optimally, repeated
measure multivariate analysis of variance (RM-MANOVA) would be conducted to assess
change over time (T1, T2, T3) using grouped primary and secondary outcome variables. RMMANOVA allows for analysis of differences in-group means over time while accounting for
within-group differences and minimizing the Type I error inherent to multiple ANOVAs
(Raykov & Marcoulides, 2010). However, the current dataset (n=17) was small and unbalanced,
precluding reliable results from RM-MANOVA (Pallant, 2013).
Although generalized linear models (GLM), such as RM-ANOVA and analysis of
covariance (ANCOVA), have been traditionally used to measure behavioral change over time,
GLM is most accurate with balanced datasets, i.e. those with equal time intervals, independent
samples, equal group sizes, and no missing data (Shek & Ma, 2011). Unfortunately, the use of
GLM with unbalanced datasets may result in increased Type I errors and potentially invalid
conclusions (Peugh, 2010; Shek & Ma, 2011). Application of GLM to the ASSET dataset
proved problematic, as longitudinal data are characteristically unbalanced, i.e., prone to missing
data (Davis, 2002). Moreover, longitudinal data tend toward correlation of individual
observations over time, thereby violating the assumption of independence imposed by GLM
(Diggle, Heagerty, Liang, & Zeger, 2002; Peugh, 2010). Furthermore, t-test methods proved to
be problematic as they resulted in listwise deletion with substantial loss in effective sample size.
When RM-ANOVAs were run for the primary (self-efficacy, social functioning) and secondary
(anxiety, depression) outcomes across the three time intervals, poor fit was achieved as the data
was not balanced and the general linear model resulted in exclusion of cases. For example, from
a total sample size of 17, 42.2% of all cases were excluded in the analysis of general self-efficacy

44

and depression; 35.3%, in the analysis of perceived empathy and social self-efficacy; and 29.4%
in the analysis of social functioning.
In effect, use of GLM would have been better suited to a cross-sectional, rather than
longitudinal, examination of the ASSET dataset. In response to these constraints, both
statisticians (Field, 2013; Hayes, 2006; West, Welch, & Galecki, 2014) and applied researchers
(Peugh, 2010; Shek & Ma, 2011) emphasize the utility of using multi-level, mixed-effects
modeling in longitudinal studies to overcome limitations of conventional statistical methods,
including general linear models (GLM) and ordinary least squares (OLS). According to the
SPSS Technical Report on Mixed-effects Linear Modeling (SPSS, 2005; p. 12) the mixed
approach is “generally preferred because it is asymptotically efficient (minimum variance),
whether or not the data are balanced. GLM, however, only achieves its optimum behavior when
the data are balanced.” Given the longitudinal designs, relatively small sample sizes,
heterogeneity of populations, and missing data inherent to repeated measures in the social
sciences, multilevel, linear, mixed-effects models are becoming increasingly popular with
applied researchers (Peugh & Enders, 2005).
Applying multilevel modeling to the ASSET dataset. Longitudinal data, by their nature,
are multilevel as they include clusters of data over multiple observations (Skrondal & RabeHesketh, 2008). Peugh (2010) offers the example of students from the same school having more
correlated scores than students form different schools due to the effects of environment. In the
ASSET dataset, clustering of scores occurs within individuals, making the likelihood of
correlated scores even greater. Thus, the application of traditional analysis models (GLM),
appropriate for cross-sectional data, would likely produce excessive Type I errors and biased
parameter estimates, significant threats to statistical validity.

45

More specifically, longitudinal data are considered two-level data, with Level 1 capturing
the repeated observations within individuals, and Level 2 predicting the outcomes observed in
Level 1. West and colleagues (2014) describe this succinctly by stating: “Repeated-measures
data sets can be considered to be a type of two-level data, in which Level 2 represents the
subjects and Level 1 represents the repeated measurements made on each subject” (p. 175).
Figure 3.1 illustrates the multilevel nature of the ASSET dataset.
Figure 3.1
Multilevel nature of ASSET dataset

This type of multi-level model is termed a “mixed model” or, more specifically, a “mixed
effects” model because it includes the usual fixed effects captured by linear regression, and one
or more random effects. The random effects terms in the model provide additional consideration
of the error term by attributing some variation to a random factor or covariate (West et al., 2014).
Whereas ANOVA would address the differences in group mean scores between time one, two,
and three, a mixed-effects model also considers how individual variation within the sample
impacts assessment of the intervention. For example, Figure 3.2 provides an example of the
marked individual variability in baseline scores (intercepts) and rate/direction of change (slopes)
in social cognition outcomes: some start high and decline dramatically, others start in the
midrange and change modestly, while other start low and increase or decline. On average, scores
decline.

46

Figure 3.2
Example of individuality in SRS baseline and outcome

A mixed-effects approach seeks to account for individuality by assessing variance for each
individual rather than reducing all variance to a mean score. This is particularly helpful in
repeated measures data wherein the influence of this individuality is repeated over time through
correlated scores. The inclusion of a random intercept for each participant effectively eliminates
any issues associated with the non-independence introduced when taking repeated measures on
the same subjects over time (Winter, 2013).
Further, an intraclass correlation, or ICC, can be calculated by dividing the between
subject variance by the total variance (Twisk, 2006). The ICC indicates how much variance is
distributed within the repeated observations versus how much is distributed between subjects.
Peugh (2010) describes the ICC as conceptually similar to the R2 and eta-squared indicators of
effect size values in regression and ANOVA, respectively. The ICC is an important
consideration in LMM because an ICC of zero indicates that there is no Level 2, or individual,
contribution to the model variance and, therefore, a simpler, non-hierarchical model is called for
(Twisk, 2006). In effect, any amount of intraclass correlation is justification for using LMM.

47

One substantial benefit of LMM was that the long format data structure effectively
increased the study’s sample size from 17 to 51 by using the measurement occasions, rather than
the number of people, as the units of analysis. This approach allowed for the inclusion of
demographic variables as controls for change over time. Mixed effects data must be configured
into long format as opposed to the wide format traditionally used with generalized linear models.
For the ASSET dataset, each case (ID) was “stacked” so that each case had three observations,
one for each time point (Peugh & Enders, 2005). In this data structure, repeated measurements
are collapsed into one dependent variable. Figure 3.3 provides an illustration of this data
structure.
Figure 3.3
Data structure in long format

While linear regression assumes that each observation is independent, the single group,
repeated measures design of this study assumes correlation of scores nested within individuals
(Heck, Thomas, & Tabata, 2014). Using a general linear model approach, such as repeated
measures ANOVA, with nested data, as in the current study, violates the assumption of
independence imposed by “single level,” non-nested, data (Bliese & Hanges, 2004; Hayes,
2006). Mixed approaches account for this by allowing for inclusion of random effects into the

48

model. These random effects account for the random intercepts (or baseline scores) and slopes
(change over time) of each individual in the sample. The linear mixed model (LMM) can be seen
as an extension of the general linear model (GLM). It differs from the GLM in that the LMM
allows for correlated and non-constant variability, thereby providing modeling of variance and
covariances, not just means (Albright & Marinova, 2015). In sum, the mixed approach includes
both the fixed effects (time) discerned by linear regression and the random effects introduced
through individual variability across repeated observations. Consequently, mixed effects offer a
“more comprehensive picture of developmental change across time” (Shek & Ma, 2011; p. 43).
Selection of covariates. A series of Pearson product moment correlations between
relevant demographic characteristics and dependent variables was conducted to explore potential
demographic covariates. The following constructs were examined as potential confounders on
the primary and secondary outcomes: socioeconomic status (SES), and clinically comorbid
anxiety, ADHD, and/or depression as identified on the demographic form at intake.
In order to estimate SES, participation in the Medicaid program was explored for use as a
proxy, but only one participant reported Medicaid enrollment. However, eight participants
(47.1%) reported receiving some form of social security income (disability insurance or
supplemental security), and recipients of such benefits are automatically eligible for Medicaid in
the majority of states, including Michigan (U.S. Department of Labor, 2017). Moreover, the
Center for American Progress (Fremstad & Vallas, 2013) reports that, for most beneficiaries, this
supplemental income actually accounts for most or all of their income. Therefore, participants
who indicated receiving Medicaid and/or social security income were considered to have low
socioeconomic status. Additionally, correlations were used to examine how social functioning,
self-efficacy, anxiety, and depression might be interrelated.

49

Multilevel modeling procedures. Model consideration, building, and testing was
informed by strategies suggested by Singer and Willet (2003), elucidated by Hayes (2006) and
Peugh (2010), and applied by Shek and Ma (2011). After clarification of the research
question(s), Peugh recommends: (1) choosing the appropriate parameter estimator,
(2) determining if multilevel modeling is necessary, (3) building the Level 1 model, (4) building
the Level 2 model, (5) multilevel effect size reporting, and (6) likelihood ratio testing.
(1) Choice of appropriate parameter estimator. Authors vary on their suggestions
regarding the use of maximum likelihood estimators (Peugh, 2010; Twisk, 2006). However, this
study is primarily focused on examining fixed parameters rather than estimating variances of the
random effects; therefore, the choice of restricted versus full information estimation was
inconsequential. In conclusion, full information maximum likelihood (FIML; ML in SPSS 24.0)
was chosen as it is necessary for model comparison.
(2) Confirmation of need for multilevel modeling. Due to this study’s repeated-measures
design, it is apparent that participants are the level-2 analysis unit and that outcomes on the
dependent variable measures can be intuitively expected to vary significantly across individuals.
In order to confirm this, null, “unconditional means” (Peugh, 2010, p. 101), or “random intercept
only” (Hayes, 2006, p. 7) models were estimated. The null model is equivalent to a random
effect ANOVA (Peugh, 2010). As applied to the ASSET dataset, the null model parses out the
total outcome score into variation within participants over time, and idiosyncratic variation
between participants in order to give a better estimate of outcomes. Results of the null model are
used to calculate the intraclass correlation (ICC). In essence, the ICC is a measure of the
dependency of the data (Field, 2013) and, therefore is a key consideration in deciding whether or
not to use multilevel modeling. The individuality of participants is quantified by the ICC. An

50

ICC = 0 indicates no level-2 individual contribution to the model variance and, therefore, no
need for LMM (Twisk, 2006). The ICC is defined as:
ICC = τ00/ (τ00 + σ2).
(3) Building the Level 1 model. Peugh (2010) notes that the confirmatory null model,
described above, captures the change in each participant’s outcomes using a flat regression line
with a slope of zero. This feature of the null model makes it difficult to interpret. Without
including Time as a predictor, the null model simply tells us that individuality is significant in
our sample and that multilevel modeling can capture this variance. Adding a level-1 Time
predictor to the null model allows for modeling of each participant’s change in the dependent
variable over time using a regression line with a non-zero slope (Peugh, 2010), thereby capturing
change within subjects.
Unlike the null model, the new model is now conditional on time. Time is the key
variable in this study as it represents the ASSET intervention. The combined equation captures
how each participant’s outcome score changes over time by modeling a grand-mean rate of
change plus a residual term that captures individual differences in the outcome change about the
grand mean. Slopes from Level 1 were tested to determine whether they were random, and if the
covariance between random intercepts and slopes contributed to the models. For models that
converged with random intercepts and slopes, covariance structures were also compared. As
recommended by Twisk (2006), unstructured and autoregressive (AR[1]; first order process;
current value based on value immediately preceding) structures were tested as they assume
correlation between measurement times, a common characteristic of repeated measures data.
(4) Building the Level 2 model. Building on Level 1, Level 2 models retained any level-1
fixed and random effects occurring within-participant, and allowed for the addition of level-2

51

covariates that influence between-participant variation and produce main effects. Time, as
measured in the Level 1 models, was the primary variable of interest in this study; therefore,
Level 2 models incorporating the covariate SES were only relevant for outcomes significantly
correlated with SES, i.e. the social function and self-efficacy outcomes. No relationship between
SES and the psychological wellness outcomes was identified.
(5) Multilevel effect size reporting. Peugh (2010) explains that multilevel effects can be
parsed out in terms of global effects (change in variance explained by the all independent
variables in the model), and local effect (change in the dependent variable associated with
individual predictor variables). Since changes in social function, self-efficacy, and psychological
wellness as predicted by the ASSET intervention, as opposed to global changes in variance, are
the outcomes of interest in this study, local effect sizes were reported to provide standardized
estimates of intervention effects. Readers interested in learning about global effects in LMM are
referred to Snijders and Bosker (1999). In order to compute the local effects, a proportional
reduction in variance statistic (PRV) was computed based on the following equation described by
Peugh (2010):
PRV = (var null - varMainEffect) / varnull
where PRV represents the proportion of variance reduction, and varnull is the variance estimate
from the null (no predictor) model, and var MainEffect is the variance estimate from the Level 2
model, conditional on Time and including any relevant covariates.
In addition, as detailed by Hedges (2007), LMM effect size estimates comparable to
Cohen’s d, were computed by converting standard error estimates from LMM output generated
in SPSS to standard deviations for each time point (post, baseline) and predictor. Conversions
were calculated in Excel by using the following:

52

SD = SE * √𝑛
LMM effect size estimates were then simply computed using Cohen’s (d = 𝑏 / SD). This allows
for meaningful comparison of ASSET with other social skills interventions in the literature.
(6) Likelihood ratio testing. Likelihood ratio testing (LRT) was used to statistically test
whether the addition of Time as a main effect significantly improved explanation of variance
when comparing null and Level 1 models. Likewise, LRT was used to assess the significance of
adding covariates and changing covariance assumptions. As recommended by Field (2013),
deviance values (changes in -2 log likelihood between models) and changes in the number of
estimated parameters were computed for each change in the model. The deviance values were
then assessed for significance based on the χ2 distribution.
Quantitative analysis summary. Due to the unbalanced, dependent nature of the
ASSET dataset, along with substantial intraclass correlation of within and between variance, a
linear, multilevel, mixed approach proved to be ideal. Multilevel modeling accounts for the
correlated errors that result from repeated measures designs and, therefore, is better suited to the
ASSET dataset. While repeated measures ANOVA would have accounted for correlated errors, it
would have resulted in listwise deletion of entire cases when data from one time point were
missing. Conversely, the mixed approach resulted in retention of cases and deletion of only time
points with missing data. Due to the tendency of participants to complete fewer assessments
over time, unbalanced data is often an issue; therefore, LMM is particularly suited to longitudinal
designs (Heck et al., 2014). Furthermore, the ASSET dataset exhibited high intraclass
correlations for each dependent variable; therefore, controlling for individuality was important.
LMM allowed for the use of participant as a random factor in the model, in effect, redefining the
residuals as the distance between each data point and the mean for that participant, rather than

53

the distance between each data point and the grand mean. By modeling participant as a random
factor rather than entering it as a fixed factor, degrees of freedom are retained, non-independence
is accounted for, and individuality is controlled for (Albright & Marinova, 2015; Singer &
Willet, 2003). In effect, by using both within and between-participant variance, LMM gave a
better estimation of how the intervention influences individual performance, not just group
means. Table 3.2 offers a comparison of features of the analytic approaches considered for this
study.
Table 3.2
Features of the analytic approaches considered
Allowances

Examination of group mean differences
Hierarchical structuring of data effectively increases power
Examination of within and between-participant differences
Examination of covariates
Handles missing data
Specification and computation of random effects increases
generalizability to population
Specification of non-constant variability and correlation of
error terms inherent to repeated measures designs with
single pooled group
Examination of intraclass correlation (ICC)

Dependent
Samples tTest

Repeated
measures
ANOVA

✓

✓
✓
✓
✓

Linear
Mixed
Effects
Models
✓
✓
✓
✓
✓
✓

✓
✓

Qualitative data analysis plan. Based on pragmatic philosophy and action research
orientation, this qualitative phase is designed to provide insight into the relationship between
outcome expectations, self-efficacy, and psychological wellness with the overarching aim of
better understanding how and for whom the ASSET intervention works. Using an intervention
mixed methods framework, the qualitative data were used to expand, clarify, and deepen
understanding of the quantitative results and, fundamentally, to help explain the results of the
intervention (Fetters, Curry, & Creswell, 2013). The quantitative and qualitative datasets were

54

integrated at the design level through convergence, as opposed to sequential approaches
(Creswell, 2009). Specifically, quantitative and qualitative data were collected in parallel and
analyses for both were conducted after the cohort one and two interventions ended. This is in
contrast to sequential designs, in which collection and/or analysis of one dataset influences that
of the other (Creswell, 2009; Fetters et al., 2013).
Data were generated through follow-up focus groups with graduates of ASSET who
entered the program in the first two cohorts (2014, 2015). To facilitate discussion and maximize
familiarity, two groups were interviewed: one comprised of ASSET college students (n = 3), the
other of ASSET high school students (n = 2). The aim of the qualitative analysis was to expand,
clarify, and deepen understanding of the quantitative results. Accordingly, theory-based coding,
derived from SCCT constructs, was used in the first layer of qualitative analysis. A second layer
used pragmatic codes, including the identification of program and career outcome expectations
themes that deepened understanding of the change in self-efficacy scores over time. Further
layering occurred at the group level, with coding completed by stakeholder group, and then
holistically. Layers were compared and contrasted in the final analysis. Cross-checking with
faculty and doctoral students with experience in qualitative analysis was used to assure
credibility of the analysis and to promote transferability of the results (Creswell, 2009).
Results of the qualitative analysis are reported in chapter four using what Fetters and
colleagues (2013) describe as a contiguous approach to integration, meaning that the qualitative
results are reported separately from the quantitative results. Appendix E contains the focus
group protocol and questions.
Summary of Analysis by Research Questions
To answer research questions 1-3. Is ASSET associated with improved work-related social

55

functioning, self-efficacy, and anxiety and depression in the identified population? The efficacy
of ASSET to improve social functioning, self-efficacy (primary outcomes), and psychological
wellness (secondary outcomes) will be tested using linear mixed-effects modeling to examine
changes over time and the influence of demographic covariates.
To answer research question 4. Are self-efficacy and psychological wellness correlated? The
relationship between self-efficacy (empathic and social), anxiety, and depression was tested using
correlation analysis. The relationships between empathy and social self-efficacy (as measured by
the PESE and PSSE), anxiety (as measured by the GAD-7) and depression (as measured by the
PHQ-9) was investigated using Pearson product-moment correlation coefficients. Preliminary
analyses were performed to ensure no violations of the assumptions of normality, linearity, and
homoscedasticity.
To answer research question 5. How does ASSET improve participants’ self-perceived
psychosocial functioning (social functioning, self-efficacy, and psychological wellness) and
outcome expectations over time? Focus group transcripts were analyzed using layered, content
analysis with theory-based coding derived from SCCT constructs. Resultant themes were then
compared with quantitative results and alignment or discordance was assessed to triangulate the
data. Finally, cross-checking was used to help ensure representativeness.

56

CHAPTER FOUR
Results
The purpose of this study was to investigate the effects of a vocational social skills
training intervention on the perceived work-related social functioning, self-efficacy, and
psychological wellness of a sample of young adults with HFASD. In order to measure these
effects and potentially generalize findings to the population represented by this sample,
multilevel, linear mixed-effects models (LMM) were applied so that the fixed effects (group
mean differences) of the intervention could be described, while the random effects (variance of
the distributions of means) could be modeled to represent the theorized normal distribution of
scores of the population of young adults with HFASD. LMM was specifically chosen as the
preferred method of analysis not only for its ability to model population variance and thereby
improve generalizability of findings, but also for its fundamental suitability to the ASSET
dataset, i.e. the ability to retain cases in unbalanced datasets, and compatibility with correlated
repeated measures data (West et al., 2014). Following the sequence outlined by Smith and
colleagues (2007), these results: (1) complete the pilot phase of the validation of the ASSET
curriculum, (2) provide preliminary level-three, quasi-experimental evidence of associated
outcomes (Chan et al., 2009), and (3) set the stage for multi-site testing of ASSET using
randomized and controlled trials.
Demographics and Characteristics of the Sample
A total of 19 individuals were screened and 17 (89.47%) were eligible to participate.
One individual was screened out due to a schedule conflict with the group meeting times, and the
second exhibited disruptive behaviors (e.g., spitting), acculturation issues, and relative
immaturity (freshman college student) for participation at the time of the intervention. This

57

individual will be invited to be re-screened in a future cohort. One participant dropped out after
two sessions, making the retention rate 94.4%.
Seventeen young adults ranging from 19 to 23 years of age (M = 20.5; SD = 1.28) with
HFASD and without intellectual disability (IQ mean = 95.11; SD = 18.08) participated in and
completed the ASSET program in the fall of 2014 and 2015. Basic reading ability (a composite
of word fluency and attack subscales) was in the average range, (M = 90.45; SD = 24.03). All
participants had a previous diagnosis of Autistic Disorder, Asperger’s Disorder, or Pervasive
Developmental Disorder – Not Otherwise Specified (PDD-NOS) according to DSM-IV-TR
criteria (APA, 2013). Two participants (11.76%) were female and the remaining 15 (88.24%),
male. Nine participants (52.94%) identified themselves as Caucasian; six (35.29%) as Black or
African American; and two (11.76%) as Asian. All were enrolled as students in college, high
school, or special education career services, and resided with parent/s or caregiver/s at the time
of intervention. Table 4.1 summarizes the sample characteristics.
Table 4.1
Participant demographics, baseline status, and disability characteristics
Demographic Covariates (n=17)
n (%)
Mean (SD)
Age
20.50 (1.28)
Gender
Male
15 (88.24)
Female
2 (11.76)
Race
White
9 (52.94)
Black
6 (35.29)
Asian
2 (11.76)
Education
Some High School
9
Post-secondary
8
Employment Status
Full-time
0
Part-time
5 (29.4)
Unemployed, looking
2 (11.8)
Unemployed, not looking
8 (47.1)
Work-based learning
1 (5.9)

58

Table 4.1 (cont’d)
Volunteer
Average hours worked per week
Cash benefits (SSI/SSDI/TANF)
Yes
No
Baseline Neurocognitive Status and
Disability-related Variables
Diagnosis
Autistic Disorder
Asperger’s Syndrome
Pervasive Developmental
Disorder-NOS
Age, in years, at onset
pre-school
6-10
11-18
post-secondary
unknown
Comorbidity
Physical and/or learning
Mental health
WASI-II
WJ-III Reading Level
(fluency/attack)

1 (5.9)
3
9
8

10 (58.8)
6 (35.3)
1 (5.9)

3 (17.6)
5 (29.4)
2 (11.8)
0
7 (41.2)
9 (52.9)
8 (47.1)
95.11 (18.08)
90.45 (24.03)

Preliminary Data Screening and Analyses
Descriptive statistics. Table 4.2 summarizes mean scores and standard deviations
organized by construct (social function, self-efficacy, psychological wellness) and associated
outcome measures. It should be noted that the social cognition and social communication
subscales of the SRS measure dysfunction. Therefore, a decline in scores indicates improved
social functioning. Moreover, multi-rater (parent, teacher, self) was included as available among
participants enrolled in high school. Multi-rater data was not recruited among college-enrolled
participants.

59

Table 4.2
Dependent variable means and standard deviations at baseline, post-intervention, and twomonth follow-up
Construct
Baseline
Post
Follow-up
(n = 16)
(n = 9)
(n = 11)
M
SD
M
SD
M
SD
Social function
SRS social cognition
60.11 6.83
52.31
9.10
55.82
7.11
Social communication 63.38 7.53
57.81
10.74
59.00
11.77
Self-efficacy
PESE (empathy)
2.79
0.68
3.85
0.66
2.88
0.92
PSSE (social)
2.80
0.71
3.89
1.01
3.02
0.73
Psychological Wellness
GAD-7 (anxiety)
12.81 4.05
9.67
2.00
8.09
1.30
PHQ-9 (depression)
16.38 4.26
14.89
5.21
13.64
2.29

Scale Reliability. Tests of scale reliability (Table 4.3) were conducted on each outcome
measure in order to ensure that they consistently reflected the constructs of interest in the ASSET
dataset. All results indicated good reliability (𝛼 >.70) as defined by Kline (1999).
Table 4.3
Tests of scale reliability
Instrument
Cronbach’s 𝛼
SRS
PESE
PSSE
GAD-7
PHQ-9

.92
.82
.86
.85
.83

Number of
subscales/items
6
6
5
7
10

N
17
17
17
17
17

Normality. The assumptions for linear mixed effects models are: (1) independent
variables are linearly related to dependent variables, (2) errors have constant variance, (3) errors
are independent, and (4) errors are normally distributed (Heck et al., 2014). Scatterplots of each
outcome were plotted over time to visualize the overall direction of the raw data and assess
linearity. Plots with similar trends across individuals are often indicative of data that can be
readily fitted to a model (Twisk, 2006). However, as suggested by Diggle and colleagues
(2002), visualization of the variation between individual responses may be difficult. Therefore,

60

standardized residuals were plotted against predicted values for each dependent variable. Finally,
visual inspection was performed in order to identify any potential heteroscedasticity, which is a
threat to the statistical validity of LMM. Heteroscedasticity, a violation of the assumption that
the spread of outcome scores is approximately equivalent at each observation (Field, 2013), was
suspected in the anxiety variable. Consequently, a Levene’s test was performed with significant
results, F (2, 33) = 11.74, p < .01. Thus, the assumption of homoscedasticity was violated and
robust analysis techniques (bootstrapping) were applied to all anxiety analyses. Levene’s tests
for all other dependent variables were non-significant: social cognition, F (2, 33) = 0.33, p = .72,
social communication, F (2, 33) = 2.06, p = .14, empathic self-efficacy, F (2, 33) = 0.47, p = .63,
social self-efficacy, F (2, 33) = 0.57, p = .62, depression, F (2, 33) = 1.95, p = .16. Thus,
homoscedasticity was assumed for these outcomes.
Screening potential covariates. Based on clinical observations of participants, SES
was suspected as a Level 2 covariate. As both IQ and comorbid mental health diagnosis could
also be potential confounders, they too were tested for correlation with the dependent variables.
As illustrated in Table 4.4, a series of Pearson product moments was generated to test the
suitability of potential covariates.
Table 4.4
Pearson product moments
Cog
Com
Cog
Com
Emp SE
Soc SE
Anxiety
Depression
SES
CMD

-

.75**
-

Emp
SE
-.37**
-44*
-

Soc
Anxiety Depression
SE
-.44** .43**
.24
-.47**
.36*
.27
.80**
-.09
.14
.01
-.01
.54**
-

SES

CMD

IQ

.57**
.52**
.52**
.58**
-.17
.13
-

.11
.08
-.17
.07
.29
.05
.17

.43**
-.16
-.003
.12
.001
-.33*
.59**
.51**
-

-

61

Table 4.4 cont’d
*p < .025, **p < .005 (multiple testing correction 𝛼/2); Cog = social cognition, Com = social
communication, Emp SE = empathic self-efficacy, Soc SE = social self-efficacy, CMD =
comorbid mental disorder, IQ = intelligence quotient.
Both SES and IQ were demographic factors identified as potential covariates (see Table 4.4).
Specifically, SES was significantly correlated with social cognition, social communication,
empathy self-efficacy, and social self-efficacy, and IQ was significantly correlated with social
cognition and depression. Comorbid mental health disorders (CMD) was also screened as a
potential confounder, but it did not show a significant linear relationship with any of the
dependent variables. However, in the final LMM models, SES proved to be the only significant
covariate, acting as a significant predictor of social functioning. Therefore, SES is the only
covariate reported in the model summaries (Tables 4.6 and 4.7) found in chapter four.
Hypothesis Testing
Dependent samples t-tests and linear mixed-effects modeling was used to test quantitative
hypotheses, and both qualitative and quantitative results were synthesized to answer research
question five as a mixed methods question. Use of paired t-tests was considered a preliminary
analysis, as the use of multiple t-tests results in increased Type I error when considering the
relationships between dependent variables and Time. Best fit of LMM was determined using
likelihood ratio testing against a chi-square critical value. In addition, covariance structures,
including the SPSS system default and most parsimonious structure (identity), as well as
autoregressive (AR[1]) and compound symmetry were compared. AR(1) and compound
symmetry structures were selected for their suitability to repeated measures data; i.e., each
assumes correlation of scores between measurement occasions. Specifically, AR(1) assumes that
the correlation of scores between measurement occasions becomes less over time, while

62

compound symmetry assumes constant correlation. Selection of a covariance structure was only
relevant in models that converged with both random intercepts and slopes, which occurred only
in the social self-efficacy model (see Table 4.10).
Research question 1: Is ASSET associated with improved social functioning? A
primary aim of this study was to first determine whether or not ASSET improved work-related
social skills. It was hypothesized (H1) that, when compared to baseline (T1), students with
HFASD who complete the ASSET program will show significant improvements in social
functioning (a) immediately post intervention (T2) and (b) at two-month follow-up (T3). H1 was
tested using dependent samples t tests and LMM on the social cognition and social
communication subscales of the SRS (composite scores of self, teacher, and parent report). As
detailed in chapter three, the SRS can be used as one component in the diagnosis of
autism. Scores at or above 70 are considered clinically significant. Therefore, a decrease in SRS
scores is indicative of improved social function in the current study.
Preliminary analyses, social functioning. Separate dependent samples t tests (“paired
samples” in SPSS) were performed to explore group mean differences between baseline and
post, and baseline and follow-up. Dependent variables were later more comprehensively
analyzed using linear mixed-effects modeling (LMM). Table 4.5 reports results of t tests
displayed by baseline and post, and baseline and follow-up averages on the social cognition and
social communication subscales.

63

Table 4.5
Social functioning dependent samples t tests, mean differences, standard deviations, and effect
sizes
Baseline to post

Baseline to follow-up

M

SD

df

t

d

M

SD

df

t

d

Social cognition

-3.98

5.88

13

-2.53*

-0.7

-2.17

7.22

11

-1.04

-0.3

Social
communication

-3.48

9.51

13

-1.37

-0.4

-4.29

9.79

11

-1.52

-0.4

*p < .025, **p < .005 (multiple testing correction 𝛼/2); Cohen’s d = 0.2 (small), d = 0.5
(medium), d = 0.8 (large).
On average, ASSET participants showed improved social cognition scores after the
intervention (post). This difference, -3.98, 95% CI [-7.37, -.58], was statistically significant, t
(13) = -2.53, p = .03, and represented a medium to large effect, d = -0.7. While improvements in
social cognition continued through the follow-up period, -2.17, 95% CI [-6.76, 2.42], the overall
effect, from baseline to follow-up, was small (d = 0.3) and did not reach statistical significance, t
(11) = -1.04, p = .32.
Although participants’ thinking about social interactions improved significantly, gains in
social communication were small (d = 0.4) both at post, -3.48, 95% CI [-6.75, 2.42], and followup, -4.29, 95% CI [-10.51, 1.93]. While clinically relevant, these improvements were not
statistically significant, t (13) = -1.37, p = .20; t (11) = -1.52, p = .16.
LMM results, social cognition. LMM was used in order to account for individual
variance in the sample and to examine SES as a potential confounder. As indicated by Figure
4.1, there was notable between-participant variation in individual baselines and responses over
time in social cognition. This individuality was later captured through the inclusion of random
intercepts and SES as a covariate in the Level 2 model.

64

Figure 4.1
Social cognition scores by individual

It is noteworthy that, despite having confirmed ASD diagnoses, none of the participants’ scores
fell within the clinically significant range (>70).
Figure 4.2 represents the level-1, within-participant variation, or variation over time, which is
notably less varied. This variation over time is the main effect or primary outcome of interest in
Figure 4.2
Social cognition scores by time

65

the study. Noting that the between-participant variability is greater over time, it is visually clear
that a multilevel model can more comprehensively assess the overall impact of the intervention.
Table 4.6 shows the continuum of models tested: (1) a null, random-intercepts-only,
model used to establish the need for LMM, (2) a Level 1, main effects, model conditioned on
Time, (3) a final Level 2 model that include SES as a potential covariate, and (4) a Level 2 model
with interaction that proved to be non-significant.
Table 4.6
Model summaries: social cognition
Parameters

Null
Level 1: main
(unconditional) effects
Regression coefficients (fixed effects)
Intercept
57.23** (1.63) 59..49** (1.80)
(Îł00)
Time (b10) -4.16* (1.63)
post
Time (b20) -3.19 (1.71)
follow-up
SES (b01)
Interaction (b11)
Interaction (b21)
Variance components (random effects)
Residuals
23.53** (6.38) 19.36** (5.24)
(σ2)
Intercepts
35.26* (15.63) 33.95* (14.48)
(τ00)
Deviance
290.260
283.809
statistic
Estimated 3
5
parameters

Level 2:
covariates

Level 2: interaction

63.81**(1.98)

63.50* (2.14)

-3.91* (1.62)

-2.46 (2.42)

-2.94 (1.71)

-3.45 (2.59)

-8.14** (2.47)
-

-7.52* (3.00)
-2.48 (3.22)
.731 (3.40)

19.48** (5.29)

18.76** (5.12)

17.45* (9.00)

17.75* (9.01)

275.523

274.526

6

8

Likelihood ratio testing
Δdf
2
1
2
χ Test
6.45*
8.29*
Statistic
Note: Parameter estimate standard errors listed in parentheses
*p < .05, **p < .01
66

2
4.00

The Level 1 model that included both random slopes and intercepts failed to converge.
While the Level 1 model with random intercepts and conditioned on Time was significantly
better than the null, χ2 (2) = 6.45, p = .04, a Level 2 model with random intercepts, conditioned
on Time, including SES as a categorical covariate, demonstrated the best fit with the data, χ2 (1) =
8.29, p < .001. As expected, given the relatively small sample size, a model with random slopes
and intercepts failed to converge. Therefore, a detailed analysis of the random effects, which
might yield greater description of the larger population, was not possible in this sample.
However, the main effects were of primary interest in this study, and random intercepts were
accounted for in all models run. Additionally, no significant cross-level effects (interaction)
were observed. Again, this can be due to sample limitations (Twisk, 2006). Thus, while some
models in this study captured significant demographic covariates, none captured significant
interaction. Results of the entire range of models, from null to Level 2 interaction, are reported
to demonstrate how the final model was chosen for each dependent variable.
Results of the null model were used to compute an intraclass correlation (ICC), which
showed that 60% (i.e., ICC = τ00/ (τ00 + σ2) = 35.26 / [35.26 + 23.53] = .60) of the variation in
social cognition occurred between participants. While the ICC (.60) for social cognition is high,
ICCs exceeding .40 are common in longitudinal social research (Peugh, 2010; Spybrook,
Raudenbush, Liu, Congdon, & Martinez, 2008). Thus, the need for the need for multilevel
modeling was supported. In conclusion, the use of GLM methods to examine this dependent
variable would have led to substantial Type I error and threats to inferential validity.
Specifically, ignoring the dependency among observations of individuals across time in this
repeated-measure design could have resulted in underestimation of standard errors of regression
coefficients and overestimation of the significance of the independent variables (Cohen, Cohen,

67

West, & Aiken, 2003). Therefore, multilevel modeling of social cognition was continued to
include Level 1 and Level 2 models.
Final model. The final, best-fitting model was a Level 2 model with Time and SES
included as predictors. The Level 1 model conditioned on Time, indicated significant overall
change in social cognition, F (2, 28.37) = 3.52, p < .05. Specifically, ASSET was associated
with statistically significant improvements in social cognition, b = -3.91, t (28.60) = -2.41, p <
.05, at post; however, improvements at follow-up were no longer significant, b = -2.94, t (29.21)
= -1.72, p = .10. Results further showed a significant grand-mean social cognition score at
baseline (Îł = 63.81, p < .01) that, on average, significantly improved at post, b = -3.91, p <
00

10

.05. However, this improvement was not significant at follow-up, b = -2.94, p = .10. Variance
20

component estimates showed: (a) significant variance in observed versus predicted social
cognition scores within participants (level-1 residual; σ = 19.48, p < .01), and (b) significant
2

variation in social cognition outcomes between participants at baseline (τ = 17.45, p <
00

.05). Converting the intercepts variance estimate to a standard deviation (i.e., √17.45 =4.18),
and assuming normal distribution of intercepts residuals, suggests that 95% of participants’
social cognition scores at baseline were likely to fall between 55.62 and 72.00 (i.e.,1.96
deviations from the mean intercepts estimate; 63.81 Âą1.96[4.18]), indicating that some
participants did have clinically significant social cognition scores at baseline.
By extending the final analysis from paired t tests to LMM, 17% more variance was
captured, [PRV = (23.53-19.48)/23.53 = .17], and consequently the LMM results were able to
examine the influence of ASSET and SES. In order to help explain some of the significant
variance between individuals at baseline and around the grand mean of the Level 1 model, SES
was added to build a Level 2 model, and fit was significantly improved, χ2 (1) = 8.29, p < .01. In

68

this subgroup analysis, SES showed a highly significant relationship with social cognition, F (1,
16.91) = 10.83, p < .01. Figure 4.3 details the comparative subgroup analysis between higher
and lower SES participants. This supplemental analysis revealed a significant overall association
between ASSET and social cognition in the higher SES group, F (2, 15.11) = 3.57, p = .05, but
an insignificant overall association in the lower SES group, F (2, 12.36) = 1.33, p = .30.
Figure 4.3
Social cognition scores by time and socioeconomic status

When examining these changes over time, ASSET was associated with a statistically significant
improvement, b = -4.68, t (14.94) = -2.54, p < .05, in social cognition in the higher SES group at
the post-intervention evaluation. This medium sized (d = -0.5) effect was no longer significant at
follow-up, b = -1.10, t (15.34) = -.55, p = .59. In contrast the lower SES group, on average,
showed small to medium improvements (d = -0.3) at post, b = -4.48, t (12.56) = -1.49, p = .16,
and follow-up, b = -3.40, t (12.56) = -1.13, p = .28, but these changes were not statistically
significant.
Based on the significant influence of SES as a categorical covariate, b = -8.14, t (16.91) =
-3.29, p < .001, on average, participants with lower SES were described by self and others as

69

having clinically significant social cognition dysfunction (63.81 + 8.14 = 71.95), while
participants with higher SES, on average, showed subclinical symptoms (63.81 - 8.14 = 55.6).
This difference in social cognition as predicted by SES represents a medium effect (d = 0.5).
Finally, it should be noted that, although SES clearly showed a moderating effect on social
cognitive outcomes, multilevel modeling failed to detect a significant interaction at post, b = 2.48, t (28.61) = -0.77, p = .45, or follow-up, b = 0.73, t (29.22) = 0.22, p = .83. Based on
consultation and review of the LMM literature (Bell, Ferron, & Kromrey, 2009; Twisk, 2006;
Wang & Ware, 2013), this failure to detect an interaction is likely due to a lack of power as a
result of testing an eight-parameter, nine-level model with a relatively small sample. Yet, Wang
and Ware (2013) provide reassurance that, while the use of multilevel models with relatively
small samples like the ASSET dataset may fail to fully explain random variances or effects
beyond two levels, they are well-suited, if conservative, in examining fixed effects.
LMM results, social communication. To further explore hypothesis one regarding
associations between ASSET and social function, the social communication subscale of the SRS
was examined. Like the social cognition subscale, the cut score for clinical significance is 70,
with scores of 70 or greater indicating disability related dysfunction in social communication.
Thus, declines in social communication scores are considered positive outcomes.
As indicated by preliminary dependent samples t tests (Table 4.4), participants’ thinking
about social interactions improved significantly, but gains in social communication were small (d
= 0.4) both at post, -3.48, 95% CI [-6.75, 2.42], and follow-up, -4.29, 95% CI [-10.51, 1.93].
While clinically relevant, these improvements were not statistically significant, t (13) = -1.37, p
= .32; t (11) = -1.52, p = .16
To examine whether accounting for individual variability would give a clearer picture of

70

the influence of ASSET on social communication, combined parent, teacher, and SRS social
communication subscale scores were evaluated. First, a null model was examined to determine if
LMM was appropriate. As indicated by Figure 4.4, there was notable between-participant
variation in individual baselines and responses over time in social communication.
Figure 4.4
Social communication scores by individual

This individuality was later captured through the random intercepts equation of the model. It is
noteworthy that, despite having confirmed ASD diagnoses, most scores fell below the clinically
significant range (>70).
Figure 4.5 (below) represents the level-2, within-participant variation, or variation over
time, which is notably less varied. This variation over time, which can be seen to increase, is the
main effect or primary outcome of interest in the study, and is later captured by the Level 2
model.

71

Figure 4.5
Social communication by time

Table 4.7 shows the continuum of models tested: (1) a null, random-intercepts-only
model used to establish the need for LMM, (2) a Level 1, main effects model conditioned on
Time, (3) a Level 2 model that include SES as a potential covariate, and (4) a Level 2 model with
interaction that proved to be non-significant. A Level 1 model that included both random slopes
and intercepts failed to converge, and the Level 1 model with random intercepts and conditioned
on Time was just below the threshold of significance for improved fit, χ2 (1) = 3.80, p = .05.
Therefore, a Level 2 model with random intercepts, conditioned on Time, and including SES as a
categorical covariate was used to examine outcomes.
Table 4.7
Model summaries: social communication
Parameters Null
Level 1: main
(unconditional)
effects
Regression coefficients (fixed effects)
Intercept
60.92** (1.96)
63.09** (2.24)
(Îł00)
Time (b10) -3.23 (2.05)
post

72

Level 2: covariates

Level 2:
interaction

67.50** (2.62)

66.31** (2.77)

-2.82 (2.07)

-1.55 (3.06)

Table 4.7 (cont’d)
Time (b20)

-

-4.03 (2.16)

-3.59 (2.18)

-.04 (3.28)

follow-up
SES (b01)

-

-

-8.45* (3.31)

-6.02 (3.88)

-

-

-

-2.42 (4.08)

-

-

-

-6.13 (4.31)

31.46** (8.56)

30.02** (8.15)

32.52* (16.07)

31.23* (15.31)

Interaction
(b11)
Interaction
(b21)

Variance components (random effects)
Residuals
35.25** (9.45)
30.84** (8.29)
(σ2)
Intercepts
50.47*(22.06)
51.60* (21.90)
(τ00)
Model
summary
Deviance
statistic
Estimated
parameters

307.453

303.651

298.328

296.341

3

5

6

8

Likelihood ratio testing
Δdf
1
1
χ2 Test
3.80
5.32*
Statistic
Note: Parameter estimate standard errors listed in parentheses.
*p < .05, **p < .01

2
1.98

Results of the null model were used to compute an intraclass correlation (ICC), which
showed that 59% (i.e., ICC = τ00/ (τ00 + σ2) = 50.47 / [50.47 + 35.35] = .59) of social
communication variation occurred between participants. Thus, the need for multilevel modeling
was supported. In conclusion, the use of GLM methods to examine this dependent variable
would have led to substantial Type I error and threats to inferential validity. Therefore,
multilevel modeling of social communication was continued to include Level 1 and Level 2
models.
Final model. The final, best-fitting model was a Level 2 model with Time and SES
73

included as predictors. The Level 1 model, with Time as the only predictor, failed to show a
significant relationship between ASSET and social communicative gains at post, b = -3.23, t
(28.97) = -1.58, p = .13, or follow-up, b = -4.03, t (29.37) = -1.87, p = .07. In order to help
explain some of the significant variance between individuals at baseline and around the grand
mean of the Level 1 model, SES was added to the Level 2 model, and fit was significantly
improved, χ2 (1) = 5.32, p < .05. This new Level 2 model indicated significant change in social
communication over time as predicted by SES, F (1, 16.88) = 6.54, p < .05. However, when
tested for interaction, the model was not able to capture a significant effect, F (2, 28.61) = 1.02, p
= .38.
Results further showed a significant grand-mean social communication score at baseline
(Îł = 67.50, p < .01) that, on average, improved non-significantly at post, b = -2.82, t (28.43) = 00

1.36, p = .18, and follow-up, b = -3.59, t (28.98) = -1.65, p =.11. On average, these
improvements were small (d = -0.2). Variance component estimates showed: (a) significant
variance in observed versus predicted social communication scores within participants (level-1
residual; σ = 31.46, p < .01), and (b) significant variation in social communication outcomes
2

between participants at baseline (τ = 32.52, p < .05). Converting the intercepts variance estimate
00

to a standard deviation (i.e., √32.52 = 5.70), and assuming normal distribution of intercepts
residuals, suggests that 95% of participants’ social communication scores at baseline were likely
to fall between 56.34 and 78.67 (i.e.,1.96 deviations from the mean intercepts estimate; 67.50
Âą1.96[5.70]), indicating that only a portion of the sample reported clinically significant social
communication dysfunction at baseline.
Extending the final analysis from a null to Level 2 model captured an additional an
additional 11% of the residual variance, [PRV= (35.25-31.46)/35.25 = .11], and enabled analysis

74

of SES as a categorical covariate that offered some meaningful results. To further explore the
influence of SES on these outcomes, a comparative subgroup analysis was performed between
higher and lower SES groups. Figure 4.6 details the subgroup analysis.
Figure 4.6
Social communication by time and socioeconomic status

This supplemental analysis showed that ASSET was associated with a statistically significant
improvement, b = -6.21, t (17.18) = -2.12, p < .05, in social communication in the higher SES
group at follow-up. This improvement represented a small to medium sized (d = -0.4) effect.
However, improvements measured at post among the higher SES group, while small to medium
in effect size (d = -0.3), had not yet reached statistical significance, b = -3.99, t (16.77) = -1.42, p
= .17. In contrast, the lower SES group, on average, made only marginal and statistically nonsignificant improvements at post, d = -0.1, b = -1.81, t (11.68) = -.64, p = .53, and follow-up, d =
0.0, b = -.32, t (11.81) = -.11, p = .92.
Based on the significant influence of SES as a categorical covariate, b = -8.45, t (16.88) =
-2.56, p < .05, on average, participants with lower SES were described by self and others as
having clinically significant social communicative dysfunction (67.50 + 8.45 = 75.95), while

75

participants with higher SES, on average, showed subclinical symptoms (67.50 - 8.45 = 59.05).
This difference in social communication as predicted by SES represents a medium to large effect
(d = -0.6). Finally, as discussed in the social cognition section of this manuscript, it should be
noted that, although subgroup analysis clearly showed a moderating effect of SES on social
communicative outcomes, multilevel modeling failed to detect a significant interaction at post, b
= -2.42, t (28.78) = -0.60, p = .56, or follow-up, b = -6.13, t (29.35) = -1.42, p = .17.
Research question 2: Is ASSET associated with improved self-efficacy? H2: It is
hypothesized that when compared to baseline (T1), students with HFASD who complete the
ASSET program will show significant improvements in empathic and social self-efficacy (a)
immediately post intervention (T2) and (b) at two-month follow-up (T3).
H2 was tested using paired t tests and LMM on the perceived empathic self-efficacy
(PESE) and perceived social self-efficacy (PSSE) measures. As detailed in chapter three, PESE
measures ability to exercise theory of mind in understanding the feelings of others and to be both
sensitive and responsive to the emotional states of others (Caprara & Steca, 2005), while the
PSSE measures self-perceived ability to express opinions, share personal experiences, work
cooperatively, and manage interpersonal conflict (PSSE; Caprara & Steca, 2005).
Preliminary analyses, self-efficacy. Separate dependent samples t tests were performed
to explore group mean differences between baseline and post, and baseline and followup. Dependent variables were later more comprehensively analyzed using linear mixed-effects
modeling (LMM). Table 4.8 reports results of t tests displayed by baseline and post, and baseline
and follow-up averages on the empathy self-efficacy and social self-efficacy subscales.

76

Table 4.8
Self-efficacy dependent samples t tests, mean differences, standard deviations, and effect sizes
Baseline to post
M

Baseline to follow-up

SD

df

t

d

M

SD

df

t

d

Empathic self-efficacy 0.93

0.57

14

6.35**

1.6

0.13

0.68

11

0.64

0.2

Social self-efficacy 1.16

0.89

14

5.07**

1.3

0.35

0.69

11

1.78

0.5

*p < .025, **p < .005 (multiple testing correction 𝛼/2);.Cohen’s d = 0.2 (small), d = 0.5
(medium), d = 0.8 (large).
On average, ASSET participants showed significantly improved empathic self-efficacy, 0.93,
95% CI [-1.25, -.618], and social self-efficacy, 1.16, 95% CI [-1.65, -.670], after completing the
ASSET intervention (post). These improvements in empathic, t (14) = 6.35, p < .001, and social
self-efficacy, t (14) = 5.07, p < .001, were statistically significant. Specifically, ASSET was
associated with a very large effect on empathic self-efficacy, d = 1.63, and social self-efficacy, d
= 1.30, at post. While improvements in empathic, 0.13, 95% CI [-.56, .31], and social selfefficacy, .35, 95% CI [-.78, .08] were still measured at follow-up, these effects were small (d =
0.3) and did not reach statistical significance, t (11) = .64, p = .54; t (11) = 1.78, p = .10,
respectively.
LMM results, empathy self-efficacy. As detailed in chapter three, The PESE uses a fivepoint Likert-type scale, ranging from 1 (not well at all) to 5 (very well). First, PESE scores were
evaluated with a null model to determine the appropriateness of using LMM. As indicated by
Figure 4.7 (below), there was notable between-participant variation in individual baselines and
responses over time.

77

Figure 4.7
Perceived empathic self-efficacy by individual

This individuality was later captured through the random intercepts and/or slopes
equation of the model. Figure 4.8 represents the level-2, within-participant variation, or
variation over time, which is notably less varied. This variation over time is the main effect or
primary outcome of interest in the study, and is later captured by the Level 1 model.
Figure 4.8
Perceived empathic self-efficacy by time

78

Noting that the between-participant variability is greater than the within-participant variability
over time, it is visually clear that a multilevel model can more comprehensively assess the
overall impact of the intervention.
Table 4.9 shows the continuum of models tested: (1) a null, random-intercepts-only,
model used to establish the need for LMM, (2) a Level 1 model conditioned on Time, (3) a Level
2 model that included SES as a main effect, χ2 (1) = 3.56, p = .06. and (4) a Level 2 model with
interaction that proved to be non-significant, χ2 (2) = 4.73, p = .19. A random slopes and
intercepts model failed to converge. The Level 1 model conditioned on Time had the best fit to
the ASSET data, χ2 (2) = 25.08, p < .001, as it was the only model to reach significance.
Table 4.9
Model summaries: empathic self-efficacy
Parameters
Null
Level 1: main
(unconditional) effects

Level 2:
covariates

Level 2: interaction

Regression coefficients (fixed effects)
Intercept (Îł00) 3.13** (.18)
2.79** (.18)

2.48** (.23)

2.56** (.24)

Time (b10)
post
Time (b20)
follow-up
SES (b01)
Interaction
(b11)
Interaction
(b21)

-

.89** (.15)

.88** (.15)

.73** (.23)

-

.15 (.16)

.13 (.16)

-.04 (.25)

-

-

.60 (.30)
-

.44 (.33)
.27 (.30)

-

-

-

.30 (.33)

.17** (.05)
.31* (.13)

.17** (.05)
.30* (.13)

76.622

75.452

Variance components (random effects)
Residuals (σ2) .41** (.11)
.17** (.05)
Intercepts (τ00) .35* (.18)
.40* (.16)
Model
summary
Deviance
statistic

105.255

80.179

79

Table 4.9 (cont’d)
Estimated
parameters

3

5

6

Likelihood ratio testing
Δdf
2
1
2
χ Test
25.08**
3.56
Statistic
Note: Parameter estimate standard errors listed in parentheses.
*p < .05, **p < .01

8

3
4.73

Results of the null model were used to compute an intraclass correlation (ICC), which
showed that 46% (i.e., ICC = τ00/ (τ00 + σ2) = .35 / [.35 + .41] = .46) of empathic self-efficacy
variation occurred between participants. Thus, the need for multilevel modeling was
supported. In conclusion, the use of GLM methods to examine this dependent variable would
have led to substantial Type I error and threats to inferential validity. Therefore, multilevel
modeling of empathic self-efficacy was continued to include Level 1 and Level 2 models.
Final model. The final, best-fitting model was a Level 1 model with Time as the only
significant predictor. In building the Level 1 model, both random slopes and intercepts were
considered. However, the only Level 1 model that reached convergence was a random intercepts
model conditioned on Time. In order to help explain some of the significant variance between
individuals at baseline and around the grand mean of the Level 1 model, SES was added to create
a Level 2 model, but fit was not significantly improved, χ2 (1) = 3.56, p = .06. As shown in
Table 4.8, the Level 1 model captured an additional 59% of the residual variance, [PRV = (σ2null σ2final) / σ2null = (.41-.17)/.17 = .59] in comparison to the null model, thus explaining most of the
variance in the data. This Level 1 model had the best fit, χ2 (2) = 25.08, p < .01, and indicated
significant improvement in empathic self-efficacy over time, F (2, 28.08) = 19.67, p < .01.
Results further showed a significant grand-mean empathic self-efficacy score at baseline

80

(Îł00 = 2.79, p < .01), which indicated that, on average, participants felt that they were able to
recognize the needs of others “moderately well” before receiving the intervention. Despite
already starting at a relatively high average, empathic self-efficacy significantly improved at post,
b = .89, t (27.83) = 6.00, p < .01. This improvement represented a very large effect size (d = 0.9)
at post. In contrast, changes in empathic self-efficacy between baseline and follow-up, while
positive, were very small (d = 0.1) and statistically insignificant b = .15, t (28.26) = .91, p = .37.
Variance component estimates showed: (a) significant variance in observed versus predicted
empathic self-efficacy scores within participants (level-1 residual; σ2 = .17, p < .01), and (b)
significant variation in empathic self-efficacy outcomes between participants at baseline (τ00 =
.40, p < .05). Converting the intercepts variance estimate to a standard deviation (i.e., √. 40
=.63), and assuming normal distribution of intercepts residuals, suggests that 95% of
participants’ empathic self-efficacy scores at baseline were likely to fall between 1.56 and 4.02
(i.e.,1.96 deviations from the mean intercepts estimate; 2.79 Âą1.96[.63]), indicating that empathic
self-efficacy perceptions ranged from “slightly well” to “well” at baseline.
LMM results, social self-efficacy. As detailed in chapter three, the PSSE uses a fivepoint Likert-type scale, ranging from 1 (not well at all) to 5 (very well). First, PSSE scores were
evaluated with a null model to determine the appropriateness of using LMM. As indicated by
Figure 4.9 (below), there was notable between-participant variation in individual baselines and

81

Figure 4.9
Perceived social self-efficacy by individual

responses over time. This individuality was later captured through the random intercepts and/or
slopes equation of the Level 1 model.
Figure 4.10 represents the level-2, within-participant variation, or variation over time,
which is notably less varied. This variation over time is the main effect or primary outcome of
Figure 4.10
Perceived social self-efficacy by time

82

interest in the study, and is later captured by the Level 1 model. Noting that the betweenparticipant variability is greater than the within-participant variability over time, it is visually
clear that a multilevel model may more comprehensively assess the overall impact of the
intervention.
Table 4.10 shows the continuum of models tested: (1) a null, random-intercepts-only,
model used to establish the need for LMM, (2) a Level 1 model conditioned on Time, (3) a Level
2 model that includes SES as a covariate, and (4) a Level 2 model with interaction failed to
converge. The Level 1 model conditioned on Time had the best fit to the ASSET data, χ2 (3) =
107.34, p < .001, as it was the only one to reach significance.
Table 4.10
Model summaries: social self-efficacy
Parameters
Null
Level 1: main
(unconditional) effects
Regression coefficients (fixed effects)
Intercept (Îł00)
3.24** (.16)
2.80** (.17)

Level 2: covariates

2.53** (.22)

Time (b10) post

-

1.11** (.18)

1.10** (.18)

Time (b20)
follow-up
SES (b01)
Interaction (b11)
Interaction (b21)

-

.28 (.18)

.26 (.18)

-

-

.50 (.28)

Variance components (random effects)
Residuals (σ2)
.61** (.16)
.28** (.16)
Intercepts (τ00)
.17 (.14)
.29* (.13)
Table 4.10 (cont’d)
Model summary
Deviance
statistic
Estimated
parameters

112.594

85.251

82.439

3

6

7

Δdf
3
1
2
χ Test Statistic
27.34**
2.81
Note: Parameter estimate standard errors listed in parentheses; *p < .05, **p < .01

83

Results of the null model were used to compute an intraclass correlation (ICC), which
showed that 22% (i.e., .17 / [.17 + .61] = .22) of the variation in social self-efficacy occurred
between participants. Thus, the need for multilevel modeling was supported. In conclusion, the
use of GLM methods to examine this dependent variable would have led to substantial Type I
error and threats to inferential validity. Therefore, multilevel modeling of social self-efficacy
was continued to include Level 1 and Level 2 models.
Final model. The final, best-fitting model was a Level 1 model with Time as the only
significant predictor. In building the Level 1 model, both random slopes and intercepts were
considered. Extending the final analysis from a null to Level 1 model captured an additional
87% of the residual variance, [PRV = (.61-.08)/.61], thus explaining most of the variance in the
data. As shown in Table 4.9, this Level 1, main effects model had the best fit, χ 2 (3) = 27.34, p <
.01, and indicated significant improvement in social self-efficacy over time, F (2, 24.80) = 19.82,
p < .01. Results further showed a significant grand-mean social self-efficacy score at baseline
(γ00 = 2.80, p < .01), which indicated that, on average, participants felt “moderately well” that
they were able to express opinions, share personal experiences, work cooperatively, and manage
interpersonal conflict before starting the ASSET program.
Despite already starting at what seems like a relatively high average for the population,
social self-efficacy significantly improved at post, b = 1.11, t (35.98) = 6.11, p < .01,
representing a very large effect size (d = 0.9). However, improvement at follow-up was very
small (d = 0.2) and no longer statistically significant, b = .28, t (20.99) = 1.59, p = .13. Variance
component estimates showed: (a) non-significant variance in observed versus predicted social
self-efficacy scores within participants (level-1 residual; σ2 = .08, p = .62), and (b) nonsignificant variation in social self-efficacy outcomes between participants at baseline (τ00 = .20, p

84

= .12). Converting the intercepts variance estimate to a standard deviation (i.e., √. 20 =.45), and
assuming normal distribution of intercepts residuals, suggests that 95% of participants’ social
self-efficacy scores at baseline were likely to fall between 1.92 and 4.31 (i.e.,1.96 deviations
from the mean intercepts estimate; 2.80 Âą1.96[.45]), indicating that social self-efficacy
performance perceptions ranged from “slightly well” to “well” at baseline.
Research question 3: Is ASSET associated with improved psychological wellness?
H3: It is hypothesized that when compared to baseline (T1), students with HFASD who complete
the ASSET program will show significant improvements in psychological wellness (a)
immediately post intervention (T2) and (b) at two-month follow-up (T3).
H3 was tested using paired t tests and LMM on the General Anxiety Disorder
Questionnaire (GAD-7) and Patient Health Questionnaire-9 (PHQ-9) measures. As detailed in
chapter three, the GAD-7 is a self-report measure of anxiety with cut points of 5, 10, and 15
respectively indicative of mild, moderate, and severe anxiety (Spitzer et al., 2006). Ten is the
threshold of clinical significance. The PHQ-9 is a self-report measure of depression with cut
scores of 5 indicative of mild depressive symptoms; 10, moderate; 15, moderately severe; and
20, severe (Cassin et al., 2013). Again, 10 is the threshold of clinical significance.
Preliminary analyses, psychological wellness. Separate dependent samples t tests using
multiple testing correction (𝛼/2) were performed to explore group mean differences between
baseline and post, and baseline and follow-up. These dependent variables were later more
comprehensively analyzed using linear mixed-effects modeling (LMM). Table 4.11 (below)
reports results of t tests displayed by baseline and post, and baseline and follow-up averages on
the GAD-7 and PHQ-9 measures.
Table 4.11
Psychological wellness dependent samples t tests, mean differences, standard errors, and effect

85

sizes
Baseline to post
M
SD
df

t

d

Baseline to follow-up
M
SD
df t

d

Anxiety
-0.08 0.29 14 -1.02 -0.3
-0.37 0.58 11 -2.22 -0.6
Depression
-0.05 0.46 10 -0.40 -0.1
-0.09 0.47 11 -0.67 -0.2
*p < .025, **p < .005 (multiple testing correction 𝛼/2). BCa bootstrap GAD-7; Cohen’s d = 0.2
(small), d = 0.5 (medium), d = 0.8 (large).

Although, on average, ASSET participants showed a trend of decreasing anxiety between
baseline and post, -0.08, 95% CI [-0.22, 0.07], and between baseline and follow-up, -0.37, 95%
CI [-0.74, -0.03], none of these changes were statistically significant when evaluated with
dependent samples t tests. However, the improvement in anxiety between baseline and followup approached significance, t (11) = -2.22, p = .05, 𝛼 = .025, while changes between baseline
and post, t (14) = -1.02, p = .33, were decidedly non-significant.
Similarly, ASSET was associated with declining depression between baseline and post, .05 95% CI [-3.61, 0.25], and between baseline and follow-up, -.37, 95% CI [-0.74, -.003], yet
neither the post assessments, t (10) = -.40, p = .70, nor the follow-up assessments, t (11) = -.67, p
= .52, were statistically significant. In terms of clinical significance, ASSET was associated with
small improvements in anxiety at post (d = -0.3) and medium improvements (d = -0.6) at followup. This trend differed for depression outcomes, in which ASSET was associated with very
small improvements both at post (d = -0.1) and follow-up (d = -0.2).
LMM results, anxiety. As detailed in chapter three, the GAD-7 (Spitzer et al., 2006) uses
a four-point Likert-type scale to assess frequency of symptoms, ranging from 0 (not at all) to 4
(nearly every day). First, GAD scores were evaluated with a null model to determine the
appropriateness of using LMM. As indicated by Figure 4.11, there was notable betweenFigure 4.11
Anxiety by individual

86

participant variation in individual baselines and responses over time. This individuality was later
captured through the random intercepts and/or slopes equation of the Level 1 model.
Figure 4.12 represents the level-2, within-participant variation, or variation over time, which is
Figure 4.12
Anxiety by time

notably less varied. This variation over time is the main effect or primary outcome of interest in
the study, and is later captured by the Level 1 model. It is noteworthy that although at least 75%

87

of all scores were clinically significant at baseline, at least 75% off all scores fell below the level
of clinical significance by follow-up. Noting that the between-participant variability is greater
than the within-participant variability over time, it is visually clear that a multilevel model can
more comprehensively assess the overall impact of the intervention.
Table 4.12 shows the continuum of models tested: (1) a null, random-intercepts model
used to establish the need for LMM, (2) a Level 1 model conditioned on Time, (3) a Level 2
model that includes SES as a covariate, and (4) a Level 2 model with interaction. A random
slopes and intercepts model failed to converge. The Level 1 model conditioned on Time had the
best fit to the ASSET data, χ2 (2) = 14.74, p < .01.
Table 4.12
Model summaries: anxiety
Parameters
Null
(unconditional)

Level 1:
main effects

Level 2:
covariates

Level 2: interaction

Regression coefficients (fixed effects)
Intercept (Îł00)
11.48** (.87)

12.71** (.87)

13.50** (1.26)

Time (b10) post

-

-1.27 (.71)

12.94**
(1.21)
-1.26 (.71)

Time (b20) followup
SES (b01)
Interaction (b11)
Interaction (b21)

-

-3.31** (.78)

-3.30** (.78)

-5.29** (1.15)

-

-

-.44 (1.6)
-

-1.50 (1.73)
1.03 (1.33)
3.29 * (1.45)

Variance components (random effects)
Residuals (σ2)
5.80** (1.68)
3.64** (1.05)
Intercepts (τ00)
10.27* (4.68)
9.08* (3.85)
Model summary
Deviance statistic
Estimated
parameters

221.058
3

206.316
5

Table 4.12 (cont’d)

88

-1.81 (1.05)

3.64** (1.05) 2.97** (.85)
9.05* (3.83) 9.69* (3.92)

206.241
6

201.661
8

Likelihood ratio testing
Δdf
2
1
2
χ2 Test Statistic
14.74**
0.08
4.58
Note: Parameter estimate standard errors listed in parentheses. VC = variance components.
*p < .05, **p < .01
Results of the null model were used to compute an intraclass correlation, which showed
that 29% (i.e., ICC = τ00/ (τ00 + σ2) = 3.64 / [3.64 + 9.08] = .29) of anxiety variation occurred
between participants. Thus, the need for multilevel modeling was supported. In conclusion, the
use of GLM methods to examine this dependent variable would have led to substantial Type I
error and threats to inferential validity. Therefore, multilevel modeling of anxiety was continued
to include examination of Level 1 (within-participant) and Level 2 (between-participant) models.
Final model. The final, best-fitting model was a Level 1 model with Time as the only
significant predictor. In order to help explain some of the significant variance between
individuals at baseline and around the grand mean of the Level 1 model, SES was added to build
a Level 2 model, but fit was not significantly improved, χ2 (1) = 0.08, p = .78. Thus, the Level 1
model showed the best fit, χ2 (2) = 14.74, p < .05, and indicated significant change in anxiety
over time, F(2, 24.87) = 9.08 , p < .01. By extending the final analysis from paired t tests to
LMM, 32% more variance was captured, [PRV = (5.80-3.64)/5.80 = .32], and consequently the
model was able to identify a statistically significant association between ASSET and improved
anxiety at follow-up that was not revealed through dependent samples t tests. Results further
showed a significant grand-mean anxiety score at baseline (Îł = 12.71, p < .01) that, on average,
00

approached significant improvement at post, b = -1.27, t (24.75) = -1.80, p = .08, and was
definitively significant at follow-up, b = -3.31, t (25.13) = -4.26, p < .01. These improvements
represented a medium to large effect (d = -0.6) at follow-up and a small (d = -0.3), but
statistically nonsignificant, reduction in anxiety at the earlier post-intervention evaluation.

89

Variance component estimates showed: (a) significant variance in observed versus
predicted social cognition scores within participants (level-1 residual; σ = 3.65, p < .01), and (b)
2

significant variation in anxiety between participants at baseline (τ = 9.08, p < .05). Converting
00

the intercepts variance estimate to a standard deviation (i.e., √9.08 =3.01), and assuming normal
distribution of intercepts residuals, suggests that 95% of participants’ anxiety scores at baseline
were likely to fall between 6.81 and 13.76 (i.e., 1.96 deviations from the mean intercepts
estimate; 12.71 Âą1.96[3.01]), which ranges from mild to moderate.
LMM results, depression. As detailed in chapter three, the PHQ-9 uses a four-point
Likert-type scale, ranging from 0 (not at all) to 4 (nearly every day) with total scores ranging
from 0 to 27 (Kroenke, Spitzer, & Williams, 2001). While results of preliminary analysis using
multiple t tests revealed small effects and non-significant results, the analysis was extended to
LMM in order to explore if additional variance could be explained. As indicated by Figure 4.13,
Figure 4.13
Depression by individual

there was notable between-participant variation in individual baselines and responses over time.
This individuality was later captured by including random intercepts and slopes in the Level 1

90

model. Figure 4.14 represents the level-2, within-participant variation, or variation over time,
which is notably less varied than the between-participant variation. This variation over
Figure 4.14
Depression by time

time is the main effect or primary outcome of interest in the study, and is later captured by the
Level 1 model. It should be noted that, while the upper range of depression can be seen to
decrease over time, the lower range scores all remain at or above the level of clinical significance
(10 or greater). Noting that the between-participant variability is greater than the withinparticipant variability over time, it is visually clear that a multilevel model can more
comprehensively assess the overall impact of the intervention.
Table 4.13 shows the continuum of models tested: (1) a null, random-intercepts model
used to establish the need for LMM, (2) a Level 1 model conditioned on Time, (3) a Level 2
model that includes SES as a covariate, and (4) a Level 2 model with interaction. A random
slopes and intercepts model failed to converge. As indicated by likelihood ratio testing (see
Table 4.13), none of the models improved fit in comparison to the null. Therefore, results are
discussed only for the null LMM model and t tests.

91

Table 4.13
Model summaries: depression
Parameters
Null
(unconditional)

Level 1: main
effects

Regression coefficients (fixed effects)
Intercept
15.58** (.98)
16.00** (1.03)
(Îł00)
Time (b10)
-.51 (1.18)
post
Time (b20)
-1.07 (1.04)
follow-up
SES (b01)
Interaction
(b11)
Variance components (random effects)
Residuals
7.89** (2.39)
1.11 (7.41)
2
(σ )
Intercepts
12.57* (5.81)
8.22 (5.88)
(τ00)
Model
summary
Deviance
statistic
Estimated
parameters

Level 2:
covariates

Level 2: interaction

15.36** (1.45)

15.00** (1.45)

-.55 (1.18)

2.22 (2.13)

-1.10 (1.03)

-.84 (1.66)

1.19 (1.94)
-

1.89 (1.99)
-3.90 (2.55)

1.22 (6.90)

1.20 (7.75)

8.07 (5.55)

7.52 (5.96)

216.735

213.918

213.544

211.276

3

6

7

9

Likelihood ratio testing
Δdf
3
1
2
χ Test
2.81
.37
Statistic
Note: Parameter estimate standard errors listed in parentheses.
*p < .05, **p < .01

2
2.27

Results of the null model were used to compute an intraclass correlation, which showed
that 61% (i.e., ICC = τ00/ (τ00 + σ2) = 12.57 / [12.57 + 7.89] = 0.61) of variation in depression
scores occurred between rather than within participants. Thus, the need for multilevel modeling
was supported. The use of GLM methods to examine this dependent variable would have led to
substantial Type I error and threats to inferential validity.

92

Final model. While the Level 1 model, which added Time as a main effect, yielded
insignificant improvement in model fit, χ2 (3) = 2.81, p = .42, final results of LMM were nonsignificant for depression. In order to help explain some of the significant variance between
individuals at baseline and around the grand mean of the Level 1 model, SES was added as a
Level 2 covariate, but fit was not significantly improved, χ2 (1) = 0.37, p = .35., F (1, 18.45) =
.378, p = .55. However, given the high ICC, it was hypothesized that some other covariate, other
than Time or SES, may account for changes in depression over time. Therefore, psychiatric
comorbidity was built into a Level 2 model, but also did not improve fit, χ2 (1) = 0.286, p = .59,
F (1, 15.63) = .001, p = .98. Therefore, no further modeling was warranted.
However, some information can be gleaned from the null model. Results of the null
model showed a significant grand-mean depression score, Îł00 = 15.58, p < .01, indicating
moderately severe depression (Cassin et al., 2013), on average, across ASSET participants at
baseline. The variance, τ = 12.57, p < .05, of individual mean depression scores (averaged
00

across all three observations) varied significantly around the grand mean. Significant differences
were also observed between each participant’s observed and predicted depression scores over
time, σ2 = 7.89, p < .01. The significance of these random effects indicates that ASSET
participants showed wide individual variation in depressive symptoms at baseline and highly
variable rates of change after the intervention. Next, the level-2 variance estimate was converted
to a standard deviation to aid in interpretation. Converting the intercepts variance estimate to a
standard deviation (i.e., √12.57 =3.55) and assuming approximate normality of distribution of
the residuals, we can conclude that 95% of participants had mean depression scores between
8.62 and 22.54 (i.e., 1.96 deviations from the mean intercepts estimate; 15.58 Âą1.96[3.01]),
ranging widely from mild to severe.

93

Research question 4: Are self-efficacy and psychological wellness correlated? The
relationship between self-efficacy (empathy and social), anxiety, and depression was tested using
Pearson product-moment correlation analysis. Results are summarized in Table 4.14.
Table 4.14
Pearson product-moment correlations among self-efficacy and psychological wellness variables
Empathic SE
Social SE
Anxiety
Depression
Empathic SE
.80**
-.09
.11
Social SE
-.00
-.01
Anxiety
.52**
Depression
*p < .05, **p < .01. Empathic SE = empathic self-efficacy, Social SE = social self-efficacy.
While the correlation between anxiety (as measured by the GAD-7) and depression (as
measured by the PHQ-9) was positive and large, r = .52, n = 39, p < .001, the relationships
between self-efficacy and self-reported anxiety and depression were negligible to small, and not
statistically significant.
Research question 5: Is ASSET associated with improvements in participants’ selfperceived psychosocial functioning (social functioning, self-efficacy, and psychological
wellness) and outcome expectations over time? It was hypothesized that ASSET would
improve self-perceived psychosocial functioning as evidenced by alignment of the four outcome
measures with qualitative statements. Further, it is hypothesized that participants will report
development in career and independent living goals as captured by interview and focus group
data.
To test hypothesis five, and expand, clarify, and deepen understanding of the quantitative
results, two focus groups with graduates of the program were conducted: one with high school
participants (n = 2) and one with college participants (n = 3). Focus group transcripts were
analyzed using layered, content analysis with theory-based coding derived from SCCT
constructs. Cross-checking was used to help ensure representativeness. Using a mixed methods

94

approach, the qualitative analysis was then integrated with the quantitative findings to see if
results were aligned, discordant, or novel.
Social functioning. Content analysis revealed that social functioning, including both
communication and cognition, was the most frequently discussed topic in both focus groups.
Social cognitive flexibility, i.e. shifting thoughts and actions based on context rather than rote
adherence to rules (Geurts, Corbett, & Solomon, 2009), emerged as a prominent subtheme.
Cognitive inflexibility and associated repetitive behaviors have long been considered one of the
more troubling features of autism (Van Eylen et al., 2011). Notably, cognitive inflexibility has
been most consistently observed and measured in the population using natural observation and
behavior rating scales, such as the Behavior Rating Inventory of Executive Function (BRIEF),
while neurocognitive assessment results have been less consistent (Van Eylen et al., 2011).
Thus, it was noteworthy to hear several examples from ASSET focus group participants that
illustrated flexibility in social cognition within functional contexts, such as job interviews and
work. Furthermore, participants were able to link their flexible behaviors to training experiences
in the ASSET program. For example, one college participant, pseudonym Gayle, first discussed
how the training prepared her for job interviews by stating, “I already knew what I was going to
say beforehand and the whole making eye contact thing, and remembering that really helped.”
However, Gayle did more than remember what was taught and practiced; she exhibited shifting
of social cognition in situ when she was able to “feel more comfortable like asking for an
alternative when [the interviewer] asked a question that I couldn’t answer.” Another college
participant and recent college graduate, pseudonym Chris, described shifting his social behaviors
situationally with coworkers when he shared, “I don’t drink that much but that’s what we do with
coworkers.” He went on to demonstrate awareness of some of the social subtleties of socializing

95

with coworkers outside of the office at “bar events” where they “hang out and drink and stuff.”
Chris seemed to be able to capitalize on the social benefits of such activity while understanding
the need to exhibit restraint: “They’re pretty cool. I make sure not to drink too much. I don’t
wanna get too bad but I make sure everyone’s hanging out, having a good time with everyone.”
When probed, he shared that this socialization outside of the office improved working
relationships in the office: “We talked a lot better, hanging out, a lot better, get work done more
effectively as a result. Cuz we actually know each other somewhat.” Finally, one participant,
pseudonym Ed, who had recently graduated high school and taken a 32 hour/week job in retail,
exemplified his social flexibility when he described meeting “a lot of good people” at work but
understanding that socialization needed to be brief and that work tasks take precedence: “We all
have some good short conversations while we’re working. You know, we always, we joke
around a bit but we always focus on getting the work done first.”
In contrast, other stories illustrated struggles with social cognitive flexibility. Gayle
recounted a recent job loss that resulted from her inability to shift the pace of her customer
interactions from the time pressured style of working in fast food, which she was accustomed to,
to a more individualized customer service approach in a new sales job. She observed, “So like
I’m used to working in the fast-paced environment where...rushing is okay and I, like [the
manager] basically said, I was going too fast or like I contributed kind of a weird atmosphere.”
She recounted being aware of incidents in which she knew the supervisor wanted her to slow
down (she recognized the need for flexibility), yet she was unable to inhibit her impulse to
“zoom into the back to get it before like my boss...actually talked to me about like getting it.”
She concluded that “he might’ve been prepared to go back there and get it but I just kinda like
went back there.” Unfortunately, Gayle’s difficulty in shifting her style of customer interaction

96

led to her dismissal. In response to this story, a second college student, pseudonym Harry, stated
that he is “fine with generally whatever speed” his work is at, yet qualified “just as long as it’s
constant and doesn’t take like sudden jumps.” Thus, Harry too seemed most comfortable pacing
his work and workplace social expectations at the rate suited to him rather than the situation. In
addition, Harry also shared a realization that eye contact serves a social communicative purpose
when he shared that he learned that “constant eye contact just makes it feel like you are engaged,
like always engaged in the conversation at hand.”
While all participants shared successes and struggles regarding workplace relationships,
only one focused on difficulties in managing feelings of attraction to coworkers or customers in
the workplace. This 23-year-old high school participant, pseudonym Carl, was enrolled in a
school-to-work transition program designed as a bridge year between completing high school
and entering the workforce. As his internships took place on a college campus where he was in
the presence of his typically developing peers, it is perhaps not surprising that he expressed
frustration regarding wanting to not just date, but find a wife. In several instances, Carl linked
the concept of a romantic relationship to establishing an independent life as an adult. He noted:
“Yes. In fact, I met someone at, I met...an employee that I, that I started taking a liking to but she
already had a guy after I told her I had a crush on her.” He illustrated his struggle with social
cognitive flexibility when he went on to conclude: “In fact, said to myself, yeah. I can’t find a
wife, I guess I’ll just have to stay a bachelor. Just have to stay a bachelor.” It is notable that
when asked about how he felt about sharing his feelings with this employee he said he felt good
but then went on to conclude that he would never find companionship. Moreover, when the
conversation transitioned to a discussion about independent living, Carl expressed additional
frustration regarding living with family, and not being able to find suitable roommates. Yet he

97

also described the idea of living alone as unattainable, and remaining at home as intolerable due
to rigid family routines and being “a lot annoyed by [aunt]...cuz of her overprotection.” He
mentioned three possible roommates, and demonstrated awareness of why the thought them to be
unsuitable when he explained that it might not work out because one friend was “into a lot of
fantasies,” another was a “guy like into girlie stuff,” and a third is not interested and when asked
“did not give a specific answer.” This participant’s conclusion, again, was that he needed a wife
in order to live safely and comfortably as an adult. He reflected, “Yes, in fact, if I was to find a
girl, know what I would not do? Would not go on a date with her. I would just plan marriage
right away. I would just force her into marriage.” These issues could jeopardize Carl’s
employment prospects and are being addressed by the transition staff at his internship site who
will be implementing behavioral programming to address his perseveration on female coworkers.
Carl’s situation illustrates how misreading of social cues by individuals with autism can lead to
social and even legal predicaments. For example, Post and colleagues (2014) observed that
certain ASD-related characteristics, including difficulty accurately interpreting verbal and
nonverbal cues of people uninterested in socially engaging, potentially perseverating on an
individual of interest, and disregard for social or legal repercussions of persistent behavior, can
be perceived as stalking. Thus, while Carl is able to perform the duties and functions of his
internship, his difficulty with social cognition and communication leave him vulnerable to job
loss and the potential for criminal justice involvement.
Yet participants, including Carl, gave examples of prosocial behaviors. For example, Ed
noted that others describe him as having “a really good work ethic,” “always polite,” and “a good
team player.” Carl recounted an experience at work in which he found a cell phone and made
sure that it was returned to its owner. Moreover, improvements in social interaction were noted

98

not just in work settings but also by family. Chris reported that his “parents definitely noticed a
difference,” and saw him as “more social, more active than [he] used to be.” He stated that he
“interact[s] on [his] own a lot better than I used to.” He even ran and won a seat on a computer
programmers’ board and attributed his willingness to engage to the intervention: “I would not
have run for that without EPASS and ASSET helping, improving that.”
Overall, participants shared that they valued being explicitly taught about expectations
for social behavior in the workplace. Chris’s statement was representative: “Knowing what
you’re supposed, expected to do, how to react to certain situations...and even what not to do.”
He went on to describe social scenarios, including “how to hang out with [co-workers] and like
do interviews even, job interviews and stuff like that” as topics he “didn’t really know before the
programs.” Harry felt that his social communication skills improved through participation in
ASSET and its companion program EPASS (Employment Preparation and Assistive Skills
Support). He observed, “The real world is very team based and that requires a lot of good
communication skills and knowing how, knowing like appropriate reactions and expectations
and... while I was familiar with some of the stuff that was covered, ASSET and EPASS in
particular helped me kind of focus on stuff that I was doing that I didn’t quite realize what I was
doing or stuff that people might not have liked very much.”
In conclusion, these qualitative results both confirm and expand the quantitative findings
on social functioning. That is, while participants made gains in both social cognition and
communication, it seems that knowing and doing are two different things. Both the quantitative
and qualitative results point to the need for additional supports in helping some participants take
the social skills they learned in group (social cognition) and more fully apply them to social
interactions (communication) with employers, co-workers, and customers over time. Gayle

99

illustrated this point when stating that concepts like teamwork were more concrete and therefore
easier to implement, whereas “communication was just kind of something I didn’t exactly
understand that well.” Others, such as Harry, shared experiences in which both teamwork and
social communication became more fluid. For example, he described teamwork as “still not [his]
first choice,” but expressed an understanding that some work “realistically can’t be done alone.”
Further, he noted that after getting to know team members “interactions are just made easier,
more bountiful.” Thus, both the quantitative and qualitative results show the tremendous
variation in responses among a group of individuals who were otherwise narrowly screened and
selected for participation.
Self-efficacy. In alignment with Bandura’s (1986) description of self-efficacy as domain
specific, the majority of comments related to self-efficacy were specific to job seeking and
workplace interactions as addressed by the intervention. For example, Chris talked about feeling
“more confident” and socially engaged at work: “I talk to coworkers a lot more, hang out with
them, doing stuff for them. Like even giving birthday gifts and stuff.” In addition to his
confidence in his ability to socially engage at work, Chris poignantly described a new-found
confidence in his ability to have a satisfying career: “I feel a lot more confident after going
through these programs, like doing the work, succeeding in the workplace. I heard a lot of like
bad stories, like I might not be able to get a job after graduation so I feel confident I can
succeed.” He summed this up by saying he felt “more ambitious.” Harry also expressed
increased work-related self-efficacy, but as a college student with limited work experience was
not as specific as the others in his descriptions; he simply stated: “I feel more confident
interacting in a more professional setting.”
Gayle described having bolstered confidence in her ability to work on a team: “I was able

100

to better conclude that I was good at teamwork. So like that was something I was more confident
in saying that I knew.” She also mentioned networking and interviewing as activities she felt
confident and capable engaging in: “I feel I can do interviews very easily...I don’t see myself
struggling at all. I feel more confident in the idea of networking. I haven’t been to a job fair or
anything yet but I feel like I can be easily prepared for that.”
However, as previously noted, while Gayle exited ASSET feeling more confident in her
ability to communicate, her difficulty with customer communication made her realize that
perhaps communication was something she didn’t “exactly under[stand] that well.” Thus, while
her self-efficacy was necessary for her to even give the sales job a try, her confidence was based
on a more concrete, basic form of communication that might be described simply as being polite.
Gayle recalled watching videos on professionalism in ASSET and thinking quite rigidly that
“they were just talking, like people would be talking to each other and asking how their day was
or like just standard communication that is probably more basic and I just took that for like what
communication was meant to be instead of what it was actually.” Despite this setback, Gayle
demonstrated resilience and a willingness to learn from the experience of losing a job when she
stated, “I’m really good at getting the job and I know what I need now to like stay in the job. So I
don’t see myself running into too many obstacles. Or if they are obstacles...I know how to solve
them, if I were to get into the workforce.”
In addition to work-related self-efficacy, one participant demonstrated generalization of
his confidence outside of the workplace. Chris observed: “I feel confident professionally and
personal as well. Like hanging out with people. Hanging out with friends, confident speaking up
more with my friend group. I used to always stay silent all the time. I bring stuff up more, be
involved in the conversations a lot more with friends.”

101

In contrast to these responses from the college focus group, the two high school
respondents shared much more concrete, task-specific examples of augmented self-efficacy. For
example, when asked about changes in self-confidence, Ed shared that he was “very confident”
in his ability to sweep and stock the store “very well.” In the course of the interview, it became
apparent that Carl needed clarification on the definition of confidence. After clarification, he too
listed specific work tasks (cleaning sinks and toilets) as skills he had confidence in performing
well. It is noteworthy that neither high school focus group participant mentioned social
functioning as an area of increased confidence.
In sum, focus group responses regarding self-efficacy were much more circumscribed
and task specific than the questions on the perceived empathy and social self-efficacy measures
administered at baseline, post, and follow-up. In fact, Chris was the only focus group participant
who described gains in social self-efficacy, and Gayle was the only one to specifically describe
continued feelings of confidence in her ability to understand the feelings of others and to be both
sensitive and responsive to the emotional states of others, i.e. empathic self-efficacy. For
example, Gayle shared being “able to assess the situation” and “more clearly understand”
problems experienced by coworkers and a supervisor. Given that the quantitatively measured
improvements in both social and empathy self-efficacy were highly significant at post but no
longer statistically significant at follow-up, the qualitative findings appear to confirm that the
apparent influence of ASSET on social and empathic self-efficacy, while large immediately after
the intervention, may have continued to diminish over time. However, a more circumscribed
self-efficacy related to confidence in being a “worthy” part of the workforce and performing jobrelated tasks remained prominent 1-2 years post.
Psychological wellness. Most reports of anxiety seemed to be situational rather than

102

pathological in nature, describing understandable feelings of being “nervous” in job interviews,
or “a little stressed out” by semester deadlines or work demands. Gayle was the only focus
group participant to describe more significant symptoms of anxiety, including somatic
complaints. When describing her first day in the sales position she eventually lost, she said “I
was just very sick that day and I think it was like because I was anxious, like just, like just
constantly like feeling like I was gonna puke or something.” Gayle could not identify why she
felt this way but, when an interviewer asked if it might be due to not knowing what to expect,
Gayle concluded “that’s the only way I can like make sense out of it.” She also recalled that her
anxiety was noticed by her supervisor and mentioned as one of the reasons she was let go. She
remembered, “Sometimes I’d be nervous with typing in someone’s order... I was really nervous
so I would just double tap everything, so like it caused me to like stay on the computer typing
longer than I should’ve.” This contrasts with her interview experiences, in which she was overly
prepared, in a scripted manner: “I had the confidence to speak and like I had questions
memorized so like I didn’t feel nervous at all.”
In sum, with the exception of Gayle, very little was mentioned about significant anxiety
and nothing indicative of depressive symptoms. In fact, both Chris and Harry reported that their
family members observed improvements in affect. For instance, Harry mentioned overhearing
“some conversations, like [Harry] just looks like he’s more lively, upbeat, more talkative…”
Interestingly, both high school participants seemed to interpret questions about feeling sad, tired,
or lonely in a completely literal fashion. For example, Ed answered, “Well, I can get a little bit
tired at work.” Soon after he explained, “I don’t really work by myself, just a few other people,”
when asked about feelings of loneliness. These qualitative results align with the quantitative
results, which indicated steady declines in both anxiety and depression, on average, over time.

103

Outcome expectations. Participants across both groups expressed expectations to
achieve fulfilling work and independent living. Furthermore, most noted that their career
expectations shifted to become “higher,” yet more “specific,” “realistic” and achievable after the
interventions. Harry expressed this sentiment when he discussed a recent job advancement: “I
am still kinda like a cog in the grand machine of making the game, but I feel like I’m a little less
replaceable.” Ed gave an example that illustrates this shift well when he spoke about being able
to compartmentalize his restricted interests in the arts to leisure rather than career: “Well, I’ve
always wanted to be an actor and artist but turns out, everything I ever wanted to be was right
here in front of me.” He went on to talk about his involvement in community theater and
continuing to draw and share his art with family and friends. Carl also shared a dramatic shift in
career goals. He previously wanted to be an entrepreneur but after the intervention and with the
help of his vocational rehabilitation counselor he became open to considering jobs that were
available in his community rather than building his own business. Finally, while expectations
seemed realistic, participants also felt optimistic and even “ambitious” about work they
previously might not have felt “really worthy of.”
Finally, all participants expressed a desire to live independently but also had questions
about how to achieve this. Carl talked about “learning to live my own” by doing more chores
around the house, yet he also stated that he is “afraid of living in a house alone unless I have a
wife.” Both Gayle and Harry, who are still college students, had expectations to live
independently but had not yet developed plans. While the other participants stated that they
could depend on their families for assistance with this transition, both Gayle and Carl, who selfreported lower SES, struggled to identify an appropriate support network. Gayle said she was
unsure if her mother could help her and Carl identified teachers as individuals that he could ask

104

for help.
As transportation provides access to both employment and independent living,
participants were also asked to discuss community mobility. Chris stated, that he would “have a
car and a license probably...within a year and a half,” as he has chosen to live in a city where he
can access public transit until that time. Similarly, Carl is using the public bus system
independently and has no plans to drive. Interestingly, Ed explained, “I was about 18” when I
went to driver’s ed and… huh. Not really sure why I didn’t get a permit.” Although Ed knows
how to use the public bus system, his parents drive him to and from work. Both Gayle and Harry
expressed fears about driving, including something happening to the vehicle or not knowing how
to communicate effectively with other drivers on the road. Thus, Gayle plans to live in a large
city where she will have access to public transit. Harry is the only participant who currently
holds a driver's license but he stated that he rarely needs to drive himself anywhere. Moreover,
he is “paranoid” that “If I do need to use a car and I go to the lot and it, oh, no, just so happens to
be gone, how long? I wouldn’t know.” He also expressed trepidation about other drivers and
poor road conditions.
Fit of qualitative results with quantitative findings. In an expansion of the quantitative
results, focus group data from participants of higher socioeconomic status described greater
skills generalization, more social cognitive flexibility, and greater job-related success including
more hours worked per week, job retention, and securing employment within the field of interest.
In addition, this subgroup expressed higher work-related social self-efficacy and overall
confidence in their ability to succeed in their chosen career fields. In contrast both Gayle and
Carl reported lower SES and multiple job experiences with limited tenure. One had recently
been let go from a position related to her field of study and the other required behavioral

105

intervention for appropriate workplace interaction. Therefore, when interpreting the results of
this study, it should be noted that privilege, status, and developmental opportunities, not just
disability and the ASSET intervention, have shaped the work-related social, self-efficacy, and
psychological wellness status of each participant in the study. Finally, these qualitative results
align with SCCT and augment the quantitative findings by illustrating how each focus group
participant reconsidered their interests, reset their goals, and altered their career actions after
participating in the ASSET program.
Results summary
In sum, ASSET was associated with improved self-perceived psychosocial functioning as
evidenced by alignment of the four outcome measures with qualitative statements. However,
this conclusion is conditional because higher SES participants exhibited relatively greater gains
in social function, as measured by the SRS, and were able to share examples of successful
application of learned skills to work settings, whereas lower SES participants reported, not just
their successes, but also some social difficulties in their workplaces. Lastly, all ASSET
participants reported bolstered career expectations and various degrees of development in career
and independent living goals at the 1-2-year follow-up focus group.

106

CHAPTER FIVE
Discussion
This chapter uses the study data, theoretical framework, and the literature to: (1)
summarize and explain the major findings, (2) integrate these findings with the existing
literature, (3) consider limitations, (4) draw implications to practice and research, and (5) state
conclusions.
Major Findings and Interpretations
The primary objective of this study was to determine if the ASSET social skills
intervention is associated with significant improvements in social function, self-efficacy, and
psychological wellness. Quantitative results confirmed that, on average, ASSET participants
made statistically significant and clinically meaningful improvements in social cognition,
empathic and social self-efficacy, and anxiety after completing the eight-week intervention.
Qualitative results confirmed and expanded our understanding of these results by giving a longterm (1-2 years post intervention) perspective on participants’ ongoing psychosocial
development. The qualitative data aligned well with SCCT constructs and illustrated how each
focus group participant reconsidered their career outcome expectations, broadened their interests,
reset their career and independent living goals, and altered their career actions using a growth
orientation after participating in the ASSET program. Although our sample may have been too
small to show significant correlations between self-efficacy and psychological wellness, these
constructs did show a positive relationship. Specifically, participants showed very large,
immediate improvements in self-efficacy, and steady gradual improvements in psychological
wellness. Notably, reduction in anxiety and depression appeared to have continued 1-2 years
post, as focus group participants verbally described no depressive symptoms and only normal,

107

situational anxiety when interviewed. This was a striking contrast to baseline, in which at least
75% of the sample reported clinically significant anxiety and 100% of the sample reported
clinically significant depression. However, as discussed below, self-selection bias may have
influenced focus group results. A detailed discussion of these results, including time-related
trends and subgroup analyses for the higher and lower socioeconomic groups, is provided below
and organized by outcome.
Social cognition. On average, ASSET was associated with a small to medium (d = -0.3)
sized and statistically significant improvements in social cognition between baseline and post.
However, these gains were no longer statistically significant at follow-up. Moreover, subgroup
analysis revealed that SES had a moderating effect on social cognitive outcomes. Results
revealed that the significant improvements at post may be further explained by the medium-sized
(d = -0.5) gains made specifically by the higher SES group immediately after participating in
ASSET. It should be noted that the results of the subgroup analysis are hypothesis generating
rather than confirming. Future research is needed to validate the observed differences in social
cognitive change that the lower versus higher SES participants experienced after participating in
ASSET (Wang & Ware, 2013).
It was further observed that participants with lower SES started ASSET with clinically
significant social cognitive dysfunction, while participants with higher SES, on average, showed
subclinical symptoms at baseline as measured by the SRS. In addition, while the higher SES
group exhibited nonsignificant regression of social cognitive gains between post and follow-up,
the lower SES group showed a continued trend of improvement. Finally, both higher and lower
SES focus group participants shared successes and struggles with social cognition, particularly
social cognitive flexibility, at the 1-2-year post-ASSET focus group. In sum, these results

108

suggest that while the higher SES group made immediate and statistically significant gains in
social cognition after participating in ASSET, both groups were able to articulate ongoing
improvement long after completing ASSET.
While generating an additional moderation hypothesis, these results also partially support
hypothesis one. Compared to baseline (T1), students with HFASD who completed the ASSET
program, on average, showed small to medium, statistically significant improvements in social
cognition (a) immediately post intervention (T2), and (b) small to medium-sized, but statistically
non-significant, improvements at two-month follow-up (T3).
Social communication. While, on average, the ASSET intervention was associated with
clinically small improvements in social communication at post and follow-up, these changes
were not statistically significant. However, SES proved to be a significant moderator of the
relationship between ASSET and social communication. When subgroup analysis was
performed, a small to medium-sized (d = -0.4) and statistically significant improvement in social
communication was measured among the higher SES group at follow-up but not at post. While
similar at baseline, the SES group trends diverged over time, with the higher SES group showing
a trend of small to medium improvement between baseline and post, and baseline and follow-up,
and the lower SES group showing a small decline in social communication at post, and no
appreciable change between baseline and follow-up. Results of the focus group largely align
with this trend. While lower SES participants were able to articulate social communication
lessons taught in ASSET, some almost verbatim, their anecdotes of social engagement detailed
struggles with social interaction that, for one, led to job loss and, for another, led to the need for
behavioral intervention. These results suggest that improvements in social communication occur
after growth in social cognition. Thus, it seems fitting that social communication did not

109

significantly improve until the eight-week period after the intervention, when participants had
opportunities to apply social cognitive gains within natural social contexts beyond the training
setting.
In sum, these results partially support hypothesis one and generate new hypotheses
regarding the influence of SES on social communication in this population. Compared to
baseline (T1), higher SES students with HFASD who completed the ASSET program, on
average, showed small to medium-sized, but statistically nonsignificant, improvements in social
communication (a) immediately post intervention (T2), and (b) small to medium-sized,
statistically significant, improvements at two-month follow-up (T3).
Social implications. The disparate outcomes in both social cognition and social
communication between the lower and higher SES groups, point to the need for examination of
additional strategies to address the social skills needs of youth with HFASD and lower
socioeconomic status. This group did make small gains in social cognition, which given more
time and training, may have continued to improve. Further, the slight decline in perceived social
communication skills at post may point to an increased self-awareness of social communication
deficits among the lower SES group as a result of the intervention.
While one interpretation of these findings could be that the intervention was associated
with decreases in social communicative ability, a previous feasibility study of ASSET (Sung et
al., 2017) revealed high levels of satisfaction among participants. Thus, it seems more likely
that, like the participants in Hillier and colleagues’ (2011) study, ASSET group members simply
became more self-aware of their social communication needs and abilities after learning about
some of the nuances in workplace communication. In addition, based on the trend of gradual and
steady gains in social cognition among lower SES participants in conjunction with their

110

relatively low, clinically significant, social cognition scores at baseline, it can also be speculated
that the lower SES group may simply have not had enough time and/or treatment to significantly
improve social cognitive skills. In turn, without sufficiently developed social cognition, it would
not be expected that this group would have made gains in social communication.
Again, it should be noted that knowing and doing are two different things. Similar to the
association between self-awareness of social deficits and heightened anxiety and depression in
the HFASD population (Vickerstaff et al., 2007), there may be an inverse association between
social cognition and social performance (i.e., communication). In effect, the more that
individuals with HFASD learn about social skills, the more likely they may be to recognize their
own deficits. Thus, while self-awareness is a necessary first step in reshaping behavior, this
group would likely benefit from additional interventions that build on their increased selfawareness and, ultimately, improve their communicative performance. Our results indicate that
this conclusion may be particularly salient to young adults with HFASD from lower SES
backgrounds, who may need more extensive interventions, including longer duration and more
individualized services, in order to practice and apply newfound social skills knowledge to
performance contexts. Hillier and colleagues (2011) drew similar conclusions when they found a
lack of significant improvement in their participants’ self-reported peer relationships.
Specifically, they noted that the training may have impacted participants’ self-awareness of
deficits in social behavior and, therefore, the lack of improvement could actually be suggestive
of increased knowledge and understanding. Our results, augment this conclusion by affirming
that additional social skills training and/or more targeted training, such as social communication
training, may be necessary specifically within the lower SES subgroup of the population. Finally,
as the higher SES group was comprised of individuals who were enrolled in college and those

111

who were transitioning from school to work, the robust response of the higher SES group is
likely due to not just educational attainment, but also to a multitude of potential social
advantages and opportunities afforded by higher SES. For instance, greater exposure to a variety
of social contexts, including professional settings, as well as increased parental investment in
educational opportunities may contribute to differing trajectories between individuals from
higher and lower SES groups. Therefore, individuals with HFASD and lower SES may benefit
from additional family supports, such as psychoeducational and career development
interventions, to augment opportunities for expanded social opportunities and increase outcome
expectations.
In contrast, the higher SES focus group participants described how social communication
became more fluid, comfortable, and “bountiful” in both work, home, and community settings.
As the higher SES group also showed earlier gains in social cognition at post, followed by
significantly improved social communication at follow-up, it seems that these participants were
able apply gains in social knowledge to performance contexts. Knowing that difficulty
generalizing learning is problematic for this population (McKenzie, Evans, & Handley, 2010),
this is an important finding. Again, it should be noted that, although the lower SES group seems
to have made gains in social awareness, they had greater difficulty with social performance.
Therefore, individuals with HFASD and lower SES may require greater intensity, duration of
intervention, and more individualized treatment approaches to realize comparable gains to their
higher SES peers.
Empathic self-efficacy. Even having started ASSET with relatively high empathic selfefficacy scores (“moderately well” at baseline), on average, ASSET participants demonstrated
very large (d = 0.9) improvements in empathic self-efficacy at post. However, some decline

112

occurred between post and follow-up, and the original gains, though still positive, were no longer
significant at follow-up, having effectively returned to baseline levels. Further, no significant
differences were noted between the higher and lower SES groups. The qualitative results of the
1-2-year post-intervention focus group partially supported the quantitative trend of loss in
empathic self-efficacy gains over time. Notably, only one focus group participant specifically
described continued feelings of confidence in the ability to understand the feelings of others and
to be both sensitive and responsive to the emotional states of others by being “able to assess the
situation” and “more clearly understand” problems experienced by coworkers and a supervisor.
While empathic self-efficacy did show a long-term return to baseline, this baseline status was a
rating of “moderately well.” Thus, while the gains in self-efficacy were not durable, participants
reported maintenance of moderate empathic self-efficacy over time. In essence, on average,
participants seem to have experienced a temporary boost in perceived empathic self-efficacy
from “moderately well” to “well” levels immediately after completing ASSET. The study’s
theoretical framework may shed light on this trend, as SCCT posits that learning directly impacts
self-efficacy (Lent et al., 2002). Therefore, the boost in self-efficacy measured immediately after
the ASSET intervention may be indicative of the learning that took place.
In sum, these results partially support hypothesis two. Compared to baseline (T1),
students with HFASD who completed the ASSET program showed statistically significant and
clinically very large improvements in empathic self-efficacy (a) immediately post intervention
(T2), but (b) not at two-month follow-up (T3).
Social self-efficacy. As indicated by the PSSE, on average, participants started ASSET
already feeling “moderately well” that they were able to express opinions, share personal
experiences, work cooperatively, and manage interpersonal conflict. While this baseline average

113

seems relatively high for the population, average improvement immediately post-intervention
was clinically very large (d = 0.9) and statistically significant, but like empathic self-efficacy,
declined to near baseline levels at follow-up. There was no significant difference between lower
and higher SES participants on the PSSE.
In sum, these results partially support hypothesis two. Compared to baseline (T1),
students with HFASD who completed the ASSET program showed significant and clinically
very large improvements in social self-efficacy (a) immediately post intervention (T2), but (b)
not at two-month follow-up (T3). Once again, the qualitative data appeared to back this trend of
declining social self-efficacy months to years after completing ASSET. Specifically, only one
focus group comment was analyzed as being specific to social self-efficacy in the 1-2 year postintervention focus group interview.
Self-efficacy implications. On average, ASSET participants demonstrated dramatic
gains in both empathic and social self-efficacy immediately after participating in ASSET
followed by regression almost back to baseline. These self-efficacy trends suggest that, while
the semantics of feeling “moderately well” versus “well” are arguable, it is clear that
participation in ASSET was associated with increased feelings of both empathic and social selfefficacy in our sample. This finding aligns with the study’s theoretical framework, i.e. learning
experiences, like ASSET, directly impact self-efficacy beliefs (Lent, Brown, & Hackett, 1994,
2000, 2002.) Yet, while the large, these gains in social and empathic self-efficacy were fleeting,
leaving us to speculate whether or not the boost in self-confidence had any appreciable impact on
functional outcomes. The qualitative results shed some light and pointing to gains in careerrelated self-efficacy that seem likely to impact participants’ work participation. Specifically, at
the 1-2-year follow-up focus group, participants described lasting feelings of work-related

114

confidence, optimism, and even ambition. For example, despite reporting “hearing bad stories”
about employment outcomes among their peers with autism, ASSET participants expressed longterm confidence not just in their ability to better assess social situations but, most notably, in
their ability to “succeed in the real world.”
Anxiety. On average, ASSET participants began the program with self-reported,
clinically moderate levels of anxiety as measured by the GAD-7, and experienced steady
declines in self-reported symptoms at both post and follow-up. These effects were medium to
large (d = -0.6) at follow-up, while small (d = -0.3) and statistically non-significant at post. The
lack of statistical significance at post may have been due to high variance. Furthermore, the final
model indicated that the average decrease in anxiety from baseline to follow-up was not only
statistically significant, but also clinically significant, as scores fell from the clinically moderate
to mild range.
In sum, these results partially support hypothesis three. Compared to baseline (T1), on
average, students with HFASD who completed the ASSET program showed clinically medium
to large, but statistically nonsignificant, improvements in anxiety (a) immediately post
intervention (T2), and (b) statistically significant yet clinically small improvements at two-month
follow-up (T3). In addition, this trend of decreasing anxiety over time is supported by the
qualitative, 1-2-year follow-up focus group results, in which all but one anecdote related to
anxiety was categorized as normal, situational anxiety that did not interfere with work or social
functioning (Simpson, 2010).
Depression. Estimated marginal means at baseline, post, and follow-up revealed a trend
of steady and growing declines in mean depression scores over time. While these changes were
not statistically significant, they are descriptive of group mean change over time. Specifically, as

115

measured by multiple dependent t-tests, group mean depression scores declined, on average, 0.05
points between baseline and post, and 0.09 points between baseline and follow-up. While not
statistically significant, these changes represent a small to medium effect size at post (d = -0.4),
and a medium to large effect size at follow-up (d = -0.7).
In sum, these results, while promising, do not support hypothesis three. Compared to
baseline (T1), students with HFASD who completed the ASSET program showed medium-sized
but statistically nonsignificant improvements in depression (a) immediately post intervention
(T2), and (b) and large but statistically nonsignificant gains at two-month follow-up (T3).
Psychological wellness implications. Baseline levels of both anxiety and depression
were striking in our sample of young adults with HFASD. Yet, improvements, particularly in
anxiety, were also notable. While at least 75% of all anxiety scores were clinically significant at
baseline, at least 75% dropped below the level of clinical significance by follow-up. Although
these steady and meaningful declines in anxiety among our participants show the promise of
group based social skills interventions in improving psychological wellness, the high baseline
anxiety scores also mirror and highlight the prevalence and intensity of anxiety in this
population. The literature estimates that nearly half of all young people with ASD struggle with
anxiety and depression (White & Roberson-Nay, 2009), and our sample exceeds these findings.
Across three observations between baseline and follow-up, on average, 54% of participants selfreported clinically significant anxiety on the GAD-7, and 100% self-reported clinically
significant depression as measured by the PHQ-9. Strikingly, on average, participants’
depressive symptoms were reported to be moderately severe before the start of the intervention.
However, there was also large variation in scores, ranging from mild to severe, among this group
of young adults. Moreover, improvements in depression, though statistically nonsignificant

116

showed a large (d = -0.7) effect size at follow-up. Given this large effective size, the multiple
testing correction imposed on the t tests, and the fact that multilevel modeling was not possible,
it is speculated that additional within-participant data (reassessment at time four, five, etc.) may
have provided additional power to detect significant change, particularly in light of the fact that
the qualitative analysis did not identify any statements indicative of depressive symptoms.
While the quantitative improvements cannot be extrapolated to be long-term trends, the
qualitative data does suggest continued improvement in psychological wellness, at least among
the focus group participants. Notably, only one instance of debilitating anxiety was recounted at
the focus group, and none of the participants described feelings of depression, even when
probed. However, these results should be interpreted with caution. In addition to a myriad of
potential confounders, the possibility of self-selection attrition (Shadish, Cook, & Campbell,
2002) should not be overlooked in the 1-2-year post-intervention focus group. Not only was the
focus group months to years after the ASSET intervention had ended, it was also a research-only
activity with no intervention component. Therefore, the individuals who did choose to take part
in the focus group may have been internally motivated individuals. Moreover, as both anxiety
and depression have a negative impact on social and vocational engagement in this population,
particularly during adolescent transition (Anderson et al., 2014; Shattuck et al., 2012), it seems
unlikely that individuals experiencing significant symptoms of depression and/or anxiety would
self-select to participate in this optional activity. Thus, transferability of these qualitative
findings to the larger HFASD population should be done cautiously, particularly when
considering the alarmingly high rates of suicidal ideation and intentionality in this population
(Cassidy et al., 2014; Raja, Beidel, & Murray, 2008).
Relationship between self-efficacy and psychological wellness. While the inverse

117

relationship between anxiety/depression and self-efficacy has been well documented in the
literature (Tahmassian & Jalali Moghadam, 2011), no significant correlations were identified
within our sample. The only statistically significant correlation identified was that between
anxiety and depression, which, not surprisingly, showed a large, positive relationship. It was
speculated that this anomaly may be due to the fact that both anxiety and depression steadily
declined, while both social and empathic self-efficacy increased sharply at post and then declined
approximately to baseline at follow-up. In conclusion, a larger sample or more withinparticipant observations over time may be needed to detect significant correlations between selfefficacy and psychological wellness.
In sum, hypothesis four was not supported. While self-efficacy did exhibit a negative
relationship with anxiety/depression, this relationship was weak and statistically nonsignificant.
Self-perceived, long-term psychosocial functioning and outcome expectations.
ASSET was associated with improvements in self-perceived psychosocial functioning as
evidenced by alignment of the trends in social and psychological outcome measures with
qualitative statements. However, hypothesis five can only be described as conditionally
supported based on convergence of the quantitative and qualitative results. This conclusion is
conditional because higher SES participants exhibited relatively greater gains in social function,
as measured by the SRS, and were able to share examples of successful application of learned
skills to work settings, whereas lower SES participants reported, not just their successes, but also
some social difficulties in their workplaces. Lastly, all focus group participants reported
bolstered career expectations and various degrees of development in career and independent
living goals one-two years after completing ASSET. However, it should be assumed that not
only ASSET, but also other learning experiences and social supports, while not social skills

118

training per se, likely had an influence on focus group participants’ self-perceived, long-term
psychosocial functioning and outcome expectations.
Psychosocial implications. According to SCCT, efficacy can influence one’s career
thinking, behaving, and feeling, and even shape outcome expectations (Lent et al., 1996). While
quantitative measures of career outcome expectations were not used in this study, focus group
questions were used to target this construct. In response, 1-2-year follow-up focus group
participants from both higher and lower SES groups expressed expectations to achieve fulfilling
work and independent living. Furthermore, most noted that their career expectations shifted to
become “higher,” yet more “specific,” “realistic” and achievable after the interventions.
Although their expectations were high, some participants also described potential barriers to
successful career engagement. These barriers included poorly developed independent living
skills (e.g., learning to perform self-care tasks without prompting or monitoring), fear of driving,
and dependence on family for transportation despite having working knowledge of the public
transit system. Overall, participants expressed optimism about their future prospects, and pride
in their accomplishments. However, these results, particularly the focus group findings, must be
interpreted with great caution. While focus group participants indicated that they had not
participated in any additional social skills training since completing ASSET, their reported
changes could be the result of any number of confounding factors, including but not limited to
maturation, mentoring, or other treatments.
Integration of Findings with the Literature
In comparison to other social skills training programs designed for adolescents and young
adults with HFASD, ASSET has both parallel and distinguishing features. Four comparable
studies, based on manualized social skills interventions of varying emphases, approaches, and

119

durations are discussed here in order to illustrate ways in which ASSET converges, diverges, and
adds to the extent literature.
Three of these programs had similar sample sizes consisting of 13-17 participants, 2-5 of
which were female (Gantman et al, 2011; Herbrecht et al., 2009; Hillier et al., 2007). Only one
program, studied by Hillier and colleagues in 2011, had a substantially larger sample (n = 42)
with multiple interventions occurring across sites. While comparison of this manualized
intervention across sites may have provided useful results, such comparisons are rarely
performed in autism research (White et al., 2007). Based on the scarcity of such examinations of
manualized interventions, this seems to be a logical next step for the ASSET program. With the
exception of Gantman and colleagues (2011) who used a randomized controlled trial, the other
studies, like ASSET, were quasi-experimental in design. In comparing methodology, ASSET is
the only study that reports use of linear mixed-effects modeling to account for unbalanced data,
although missing data is specifically noted as a problem by Hillier et al. (2011), and Herbrecht et
al. (2009).
A decade ago, White and colleagues (2007) observed that many autism intervention
studies did not report effect sizes or data, such as standard deviations and change scores, that
could be used to calculate effect sizes. Without such metrics, meaningful comparison of the
intervention literature is unlikely. Thankfully, two of the four studies offered as comparisons to
ASSET did include sample size calculations and are detailed below. As calculated using
Cohen’s d, ASSET is associated with: (1) small to medium effects on social cognition at post, (2)
small to medium effects on social communication at follow-up among higher SES participants,
(3) very large effects on self-efficacy at post, and (4) medium to large effects on anxiety at
follow-up. In comparison, Herbrecht and colleagues found large effects (𝜂 2) in the areas of

120

social and adaptive competence, social behavior, and peer interaction. However, it should be
noted that seven of this study’s 17 participants received the same training two years prior. In
considering the implications of these findings, it would be interesting to know if there was any
significant dosage effect from repeating the intervention. Again, this may point to an area for
future study with ASSET. Such dosage effects may be particularly relevant to the lower SES
subset of the ASSET group, as their improvements in social functioning only approached
significance.
Hillier and colleagues (2011) also reported effect sizes (d), and found small, positive
effects on both anxiety and depression. Given that their intervention was delivered in one-hour
sessions over eight weeks, comparable to ASSET’s eight-week/1.5-hour format, it is noteworthy
that ASSET resulted in larger effect sizes for both anxiety. Therefore, it is speculated that the
increased duration of the individual sessions potentially contributed to the increased effect sizes.
This comparison is particularly relevant to Hillier et al. (2011) as their study is the only other
intervention besides ASSET to target vocational social skills, in part, and measure both anxiety
and depression. It is also noteworthy that they delivered content using a format described by
McMahon and colleagues (2013) as a social skills support group. In contrast, ASSET, as well as
Herbrecht and colleagues (2009) and Gantman and colleagues (2011), followed a structured
learning approach, which incorporates didactic learning in conjunction with peer and facilitator
feedback (McMahon, Lerner, & Britton, 2013).
Given the relatively small samples and reliance on general linear modeling in most
comparable studies, the use of covariates and control variables is also limited. In fact, McMahon
and colleagues (2013) observed that the current social skills literature has only just begun to
assess moderation and mediation of outcomes. Herbrecht and colleagues (2009) conducted the

121

only other comparable study, besides ASSET, to consider demographic covariates as potential
confounders. Specifically, they examined IQ and medication. Noting their observation that
more complex social skills development, as is the case in vocational contexts, is dependent upon
higher cognitive and verbal capacities, future examination of the ASSET datasets should control
for IQ. While our initial assumption was that limiting eligibility to high functioning individuals
would control the influence of cognitive variability, our quantitative findings and qualitative
observations indicate that even among this circumscribe group, there is tremendous variability in
social, adaptive, and emotional function. Furthermore, while psychiatric comorbidity was ruled
out as a potential covariate in ASSET, Herbrecht and colleagues (2009, p. 334) found “some
hints” that participants who were medicated had better results. Again, this may indicate another
area for future consideration.
Additionally, some autism researchers, including Herbrecht et al. (2009) have used alpha
levels of .10 to compensate for issues like attrition, data loss, and difficulty recruiting adequate
sample sizes with this population. It should be noted that had we set alpha at .10, social
cognition, social communication, and social self-efficacy would have reached levels of statistical
significance at follow-up for both lower and higher SES groups; similarly, anxiety, would have
yielded statistically significant results at post, not just at follow-up. However, introducing the
possibility that our conclusions could be wrong in up to one out of ten occasions was not
considered to be conservative enough to assess ASSET as a potential evidence-based practice.
Further, results of our previous pilot study (Sung, Connor, Chen, Lin, Kuo, & Chun, 2017),
suggested that the use of additional within-participant observations (T3) could be used to explore
changes in not only self-efficacy and social functioning over time, but also psychological
constructs, such as anxiety and depression, which are highly comorbid with autism. Thus, the

122

current study expanded the sample size, collected follow-up quantitative data at T3, and
incorporated a 1-2-year follow-up focus group to better understand long-term impacts of the
program. In sum, the pilot study not only established preliminary efficacy, but also verified that
the program is acceptable to participants, and feasible for an array of professionals who can
potentially facilitate and deliver the ASSET intervention within community contexts.
In terms of data collection, both the ASSET pilot and the current ASSET study are the
only comparable studies that used mixed methods designs to better incorporate participant
perspectives. In the pilot study, qualitative data was elicited to inform the iterative,
developmental design of the curriculum, assess participant satisfaction, and probe facilitator’s
experiences in order to assess feasibility of the program. In the current ASSET study, qualitative
data were instrumental in confirming and expanding the quantitative results, and providing a
longer-term (1-2 years post program completion) perspective on changes in psychosocial status,
career outcome expectations, and functional independence among focus group participants.
Given some of the measurement challenges with this population, as noted in the limitations
section, other studies have used trained observers to document participant change (e.g., Hillier et
al., 2007). As video data was collected throughout the cohort two and three ASSET
interventions, such observational coding and measurement of participant behavior is a future
consideration.
Other features that make the ASSET intervention and this study unique include the
examination of individual level factors, such as socioeconomic status and psychiatric
comorbidity, which provide another level of specification in matching the intervention to the
needs of individuals. McMahon and colleagues (2013) offer an extensive list of other potential
mediators or moderators, including symptom severity and group cohesiveness, that could be

123

examined to further specify: ‘‘What treatment, by whom, is most effective for this individual
with that specific problem, and under which set of circumstances?” (Paul, 1976).
In addition to being theoretically derived and tailored to the population, ASSET is one of
few studies in ASD social skills research to take a long view on impacts. Similarly, Herbrecht
and colleagues (2009) measured quantitative changes in their participants across four time
points. Yet, beginning with the pilot study, ASSET has been both quantitatively and
qualitatively examined for appropriateness to the population, acceptability, and feasibility, not
just through solicitation of participant feedback, but also through focus group and survey data
collected from parents and group facilitators (Sung et al., 2017). Finally, the use of 2-month
follow-up measures and interviews, and a 1-2-year participant follow-up focus group helped in
identifying lasting, transferrable effects. As noted by McMahon and colleagues (2013), effects
that do not last might not be interpretable as meaningful.
Finally, and most notably, ASSET and the intervention examined by Hillier et al., (2011)
are the only two social skills programs, to date, to address both vocational and psychological
functioning among adults with ASD. As noted by Hillier et al. (2011), the established
relationships between anxiety, depression, and social functioning makes social skills
interventions a logical choice for not only providing skills training, but also addressing
psychological wellness. It is noteworthy, that the intervention designed by Hillier and colleagues
(2011) specifically targeted anxiety and depression. In contrast, the ASSET intervention targets
soft skill development exclusively. In light of the fact that ASSET is associated with significant
improvements in anxiety and promising trends in depression, while primarily targeting social
skills development, our results support previous findings that show an association between social
functioning and psychological wellness.

124

Previous research has described an inverse relationship between cognition and selfassessed social competence. Furthermore, individuals with relatively high IQs, but potentially
low self-perceived social competence may, subsequently, experience elevated levels of
depression (Vickerstaff et al., 2007). This trend did not appear to play out in our data. While, on
average, our participants made gains in social cognition, and potentially the self-awareness that
comes with knowledge, they also concurrently made small and steady, while statistically
nonsignificant, improvements in depression. While Vickerstaff and colleagues (2007) note a
general association between low social appraisal/high IQ and depression among individuals with
HFASD, it is speculated that ASSET may have offered some protection from negative selfappraisal as participants were able to examine their social strengths and needs within the safe and
supportive milieu of the group.
Theoretical alignment. The results of this study largely support the tenets of SCCT
(Lent & Brown, 1996) and mostly align with the theoretical framework laid out in chapter two.
Our results indicate that autism-related characteristics (SCCT personal inputs), including
impaired social cognition and social communication, in conjunction with contextual affordances
(socioeconomic status), contributed to both the learning needs and responses to training among
ASSET participants. Further, this learning experience was associated with large and positive
changes in social and empathic self-efficacy which, in turn, were reflected in participants’
statements regarding augmented career ambition, optimism, and realism (outcome expectations).
In alignment with the framework, three constructs (the ASSET intervention, bolstered selfefficacy, and outcome expectations) can be described as primary drivers of the measurable
improvements made in social functioning and psychological wellness among ASSET

125

participants. In fact, both social and empathic self-efficacy showed very large and positive 1
correlations with social functioning. Future study using mediation analysis to confirm this
relationship in the HFASD population is warranted.
Limitations
While the results of this study contribute to our understanding about the vocational and
psychosocial needs of young adults with HFASD, generalization to the larger population should
be applied cautiously. This section details some of the limitations of this study.
Threats to external validity. Given the quasi-experimental nature of the study and
limited sample size, several limitations should be considered before making any interpretations.
First, the study design lacks random assignment and, therefore, the potential effects of treatment
expectancy cannot be controlled. However, the community versus clinic-based delivery of the
intervention may lessen treatment expectancy, as non-clinical settings are traditionally not
associated with treatment. Second, change was assessed in one group over time and, therefore,
causality cannot be determined between independent and dependent variables. Third, the
generalizability of the findings is limited by the use of a non-random convenience sample.
Finally, as noted earlier, the possibility of self-selection attrition cannot be overlooked when
interpreting the qualitative data, as not only did these individuals complete the eight-week
observation, but they were also motivated to return 1-2 years post. Thankfully, overall attrition
was very low (1/18). However, the potential for self-selection bias is inherent to the voluntary
nature of participation. Furthermore, our study was limited to three observations over time;
additional observations would have allowed for growth curve analysis, giving a more robust
description of any possible trends. Moreover, sample size, use of statistically conservative
1

Table 4.3 lists negative numbers in the correlations between social cognition, social communication, and empathic
and social self-efficacy. This is due to the scaling of the SRS, for which decreases in scoring reflect improved social
function.

126

modeling, and instrument sensitivity may have limited our ability to detect a statistically
significant improvement. Finally, the qualitative data collected through focus group
methodology did not reach saturation, as recommended by Carlsen and Glenton (2011) but,
rather, was practically limited by time and recruitment constraints to one group of two highschool-enrolled participants, and one group of three college-enrolled participants. Thus, the
application of these results to populations other than young adults with HFASD or use of the
intervention outside of group-based, community contexts should be considered cautiously.
Measurement biases. The use of self-report measures introduces a threat to internal
validity, as the data is susceptible to self-report error and bias (Remler & Van Ryzin, 2011). All
measures used in this study were based on perception; these perceptions may have been
influenced by a variety of factors including but not limited to mood, fatigue, and motivation to
report accurately. Furthermore, independent behavioral observations were not used to confirm or
refute results of these measures, and evaluators were not blinded, i.e., they were part of the team
conducting the intervention. Perhaps, most importantly, several authors have expressed concerns
regarding the accuracy with which individuals with autism are able to introspect their mental
states and emotions (Baron-Cohen, 1997; Kleinhans et al., 2011) in completing self-assessments.
Much like Hillier and colleagues (2011), and as visually depicted in the chapter four boxplots,
our sample showed tremendous variability in responses to the SRS, PESE, PSSE, GAD-7, and
PHQ-9, which resulted in high standard deviations, and intraclass correlation coefficients. The
use of LMM in our study was one way to help account for some of this individual level variance
and, thereby, minimize Type II error. For example, the use of LMM enabled us to detect
significant changes in social communication that were not detected using dependent samples t
tests.

127

While using data from multiple raters could have offered corroboration of the reliability
of participant report measures, this was not possible in the current study. Unfortunately,
independent rater reliability could not be established as there were no significant correlations
between teacher, parent, and self-report on the SRS. Due to this predicament, and the fact that
for some observations, data was not collected by all three stakeholders, there was no effective
way to determine which rater had the highest reliability. Thus, an average of all available ratings
for each participant at each observation was used.
It was further noted that, despite having confirmed ASD diagnoses, at least 75% of the
low SES group and 100% of the high SES group started ASSET with SRS social communication
scores already within normal limits. This discrepancy may indicate an initial lack of insight into
social communication abilities before the intervention. In addition, this trend may suggest that
the SRS, while sensitive to change in young children with ASD, may not be as suited to
capturing the more nuanced developmental changes that occur in adults. Moreover, the SRS
may lack sensitivity in detecting change in the high functioning population, in which individuals,
especially adults, may be more likely to test within normal limits.
In addition, this study was limited by lack of parsimonious measures of social
functioning. While both the social cognitive and social communication subscales of the SRS
showed utility in highlighting change over time, the SRS was designed for diagnostic
confirmation rather than assessment of social skills growth. Moreover, the use of two subscales
as opposed to one comprehensive assessment of social functioning, introduced the potential for
increased Type I error in this study. Therefore, future studies would benefit from a dedicated
measure not just of social function, but of work-related social skills in adult populations.
Threats to internal validity. In addition to the threats to external validity inherent to the

128

study design and sampling, the following concerns were noted regarding internal validity. First,
the researcher and author of this thesis was also a group facilitator, curriculum developer, and
funded program manager for the intervention and, consequently, may have inadvertently
exhibited researcher bias. This threat was addressed through team collaboration, faculty
supervision, and cross checking of the data collection, structuring, and analyses. In addition,
consultation from the Michigan State University, Center for Statistical Training and Consulting
was sought in order to ensure validity and appropriateness of all quantitative analyses.
Furthermore, there have been a total of four facilitators involved in the intervention, thus creating
a threat to internal validity and fidelity. However, the use of manualized lesson plans, in-session
observation by faculty and research assistants, and weekly team debriefing meetings minimized
these threats. Fidelity of treatment can be further enhanced and more systematically addressed in
future studies through use of fidelity checklists completed by independent raters. Lastly, each
cohort consisted of a college and high school group that each met the eligibility criteria but did
have some difficult to measure, qualitative differences, such as level of career aspiration and
family support for postsecondary education and employment. However, this threat appears to
have been mitigated through examination of SES, which significantly correlated with the study’s
social constructs and showed commonalities across the college and high school groups.
Implications for Practice
Group-based interventions show promise in reducing service costs and caseload
demands, while capitalizing on the power of peer interaction, group process, and the creation of
safe therapeutic spaces in which skills generalization can begin (Gantman et al., 2011).
Moreover, community-based interventions can lessen practical constraints, such as lack of
transportation or lack of awareness of services, which can be barriers to seeking vocational

129

rehabilitation. Given the interdisciplinary, iterative, developmental design of the ASSET
curriculum, in conjunction with the manual and associated training materials, ASSET is
amenable to delivery by numerous professions, including counseling, occupational therapy,
social work, and special education. Furthermore, ASSET requires minimal infrastructure (group
seating, computer, projector, binders, printed materials, and video recording from mobile phones
or stand-alone cameras), making it suited to any geographic location.
With training tailored to the HFASD population through use of visual supports, video
modelling and feedback, and reduction of extraneous sensory stimuli (Wong et al., 2014),
ASSET mitigates the lack of access to quality community-based rehabilitation by bringing
vocational rehabilitation strategies to community settings. As ASSET addresses vocational
readiness and is associated with gains in social functioning, self-efficacy, and psychological
wellness, it can potentially help to fulfill federal mandates, such as WIOA (2014). Furthermore,
the group interaction provides opportunities for vicarious learning (Bandura, 1986) through
modeling (both live and video modeling) of positive social behaviors by peers and guest experts,
and participation in group activities that reinforce learning and promote increased self-efficacy.
Moreover, the group experience offers individuals with HFASD not only the opportunity
to practice social skills, but to experience the sense of belongingness that comes from being part
of a group with shared experiences. Hiller and colleagues (2011) speculated that this opportunity
to develop relationships with other young adults with HFASD partially contributed to their
samples’ significantly improved anxiety and depression. Our results further suggest that groupbased social skills training is associated with more than just skills acquisition. For instance,
although ASSET was not designed to address psychological wellness, on average, ASSET
participants experienced significant reduction in anxiety and an overall trend of lessening

130

depression. Therefore, practitioners working with this population should not overlook the
secondary psychological wellness benefits associated with the ASSET social skills training
program. Furthermore, as demonstrated by our sample, practitioners should be cautious not to
screen out potential participants from skills training due to clinical levels of anxiety and/or
depression. Thus, in addition to more targeted mental health approaches, social skills training
should be considered an integral part of promoting, maintaining, and improving psychological
wellness for young adults with HFASD.
Furthermore, the symptoms of both anxiety and depression may not present as they do in
the general population, as they can be masked by ASD characteristics including flat affect and
social withdrawal (Hillier et al., 2011). Not only might this partially explain our surprise in
seeing how high our participants scored on measures of anxiety and depression, but it also
implies that practitioners must be careful to not overlook clinical signs and symptoms by
assuming them to be typical ASD behaviors or mannerisms. In conclusion, there is still much to
learn about depression and psychological wellness among individuals with ASD (Hedley &
Young, 2006). Our results underscore the prevalence and magnitude of these issues, as well as
the potential to utilize group interventions to partially address psychological wellness through
community-based, group interventions in addition to the more traditional, individual, clinical
and/or pharmaceutical approaches.
In addition to being developed based on the unique needs of young adults with HFASD,
ASSET is framed by theory. The development of the ASSET curriculum was guided by SCCT,
thereby creating a common theoretical language through which practitioners can examine why,
how, with whom, and when to implement the program. Furthermore, SCCT constructs have
been previously validated as effective approaches to bolstering career-related self-efficacy

131

among other young adult populations, including college students with disabilities (Corrigan et
al., 2001). In addition to the SCCT framework, the results of the current study, as well as the
pilot efficacy and feasibility study (Sung et al., 2017), can be used to guide practitioners in
assessing the suitability of ASSET in the service of their clients with HFASD. In sum, ASSET
provides practitioners with a level-three (Chan et al., 2009) evidence-based, and manualized
approach to fostering vocational social skills development and psychological wellness among
young adult clients with HFASD. Recognizing that social dysfunction is a core feature of autism
(APA, 2013), and the fact that, on average, one out of every two individuals with HFASD has
comorbid anxiety and/or depression (White & Roberson-Nay., 2009), ethical practice demands
that we use empirically validated interventions with this vulnerable population.
In addition to being HFASD specific and theoretically guided, ASSET is manualized.
The ASSET curriculum was developed through a 3-year, iterative, developmental process that
included an interdisciplinary team of researchers, educators, and research assistants. This
process helped to ensure rigor and the development of an end product (including manual,
handouts, and PowerPoints) that is practical and user friendly. Manualization is a step rarely
achieved in the development of social skills interventions (White et al., 2007), understandably so
given the tremendous time, staff commitment, and expertise needed to effectively build a useful
manual. However, the provision of a validated manual provides reassurance to practitioners that
the program can be delivered with ease and consistency. Furthermore, manualization not only
aids the user in delivery of the intervention, but also aids researchers by providing additional
assurance that the program is delivered with fidelity. These steps all contribute to ensuring that
the program is safe and effective. Notably, our results show that, despite having multiple group
facilitators, ASSET was still associated with significant improvements across two cohorts and

132

four groups of participants.
In summary, this study and the previous feasibility study (Sung et al., 2017) offer
practitioners evidence that the ASSET intervention is: (1) suited to young adults with HFASD,
(2) acceptable to this population, (3) designed for ease of delivery, and (4) associated with
significant, consistent, and positive effects on social cognition, social and empathic self-efficacy,
and anxiety.
Implications for Research
While many of our findings aligned with previous research, some results were
unexpected. Although the literature has long described self-efficacy and anxiety as negatively
correlated (Bandura, 1988), our results failed to show a statistically significant relationship
between self-efficacy and psychological wellness (Table 4.3). The direction of the relationships
was mostly negative (PESE and depression showed a very small positive, nonsignificant
correlation), but the correlations failed to reach levels of significance. It is speculated that this
may be due to lack of adequate power to detect change. However, future work will be required
to confirm the influence of self-efficacy on psychological wellness among young adults with
HFASD, and to explore whether anxiety and depression differ in their relationships to selfefficacy.
Another striking finding was the unexpectedly high social functioning scores indicated by
the SRS. Based on clinical and qualitative observations, it seems unlikely that a majority of our
sample, despite confirmed diagnosis of autistic disorder, had SRS social cognitive and social
communication scores within normal limits. This conclusion aligns with those drawn by
McMahon and colleagues (2013) in a systematic review of the group social skills literature.
Specifically, these authors noted that this instrument’s lack of design to evaluate efficacy of

133

intervention, along with the tendency of self and other reporters to inflate scores at baseline, can
lead to spurious results. Moreover, they cautioned that achieving accurate baseline scores is
difficult in general due to a lack of clinician/researcher familiarity of normal baseline behavior
for each individual and the fact that participants may exhibit inhibited/atypical behavior patterns
as they “acclimate to the intervention setting” (p. 33). Furthermore, we failed to find any
significant correlation between self-reported social functioning and teacher and/or parent report.
This finding aligns with other studies identified through systematic review of the literature
(McMahon et al., 2013). Thus, both our experience and that of others in this field indicate that
age-normed, population specific, social skills instruments designed to capture therapeutic
change, as opposed to developmental patterns or diagnostic criteria, are needed.
Not only was it difficult to measure social functioning in the current study, but we also
struggled to find an appropriate means of assessing work-related social self-efficacy. As a result,
we combined social and empathic measures to approximate a measure of participant confidence
in the ability to exercise theory of mind in understanding the feelings of coworkers and to be
both sensitive and responsive to their emotional states (empathic), while also being able to
express opinions, share personal experiences, work cooperatively, and manage interpersonal
conflict in the workplace (social). However, neither measure is specifically work-related.
Moreover, SCCT posits that self-efficacy is domain specific. Therefore, a more targeted
measure of work-related social self-efficacy needed. This is not only an area for future research
but also a limitation in our study.
Finally, the note of caution mentioned above regarding the atypical expression of mood
and anxiety disorders in this population is relevant not just to practitioners, but to families,
educators, and researchers. Based on the results of the current ASSET study, it is recommended

134

that future intervention studies outline procedures for addressing clinically significant assessment
results even when behavioral observations do not seem aligned.
Future directions. While the group process effect cannot be controlled in this ex post
facto analysis of the existing database, this study opens a future line of inquiry into the potential
effects of augmenting traditional learning interventions with group methods derived from the
counseling, psychology, and therapy literature. For example, closer comparisons of social skills
support groups, such as those examined by Hillier and colleagues (2007; 2011) with structured
learning interventions, such as ASSET and the PEERS program (Laugeson et al., 2012), could
help tease out how much of the treatment effect is due to group process versus skills training.
In addition to better understand the clinical aspects of social skills interventions,
continued research on manualized social skills curricula for schools is needed (White, Scahill et
al., 2007), particularly for students with ASD who benefit from high frequency and intensity of
intervention, along with opportunities for practice within natural contexts in order to generalize
social skill performance (Lopata et al., 2013; Toomey, Lopata, Volker, & Thomeer, 2009). Such
“dosage” effects and opportunities for generalization are not possible in clinical settings. Based
on our results, particularly in the area of psychological wellness, the small but steady
improvements observed in anxiety and depression over time are hypothesized to show larger
effects if participants are exposed to greater frequency and duration of the group intervention,
and they have increased opportunities to practice emerging social skills in a supported
environment. Increasing the frequency and duration of the intervention is not sustainable in
clinical environments, therefore future studies guided by implementation science (Locke et al.,
2016) are needed to determine the most efficient and effective ways of inserting vocational social
skills training and psychosocial supports into environments where young adults with HFASD

135

spend much of their time. Clearly, this includes high schools, community colleges, and four-year
institutions.
Research will also be necessary in order to develop measures of soft skills functioning
and self-efficacy that stand alone and capture changes in work-related social function, as
opposed to using multiple subscales of primarily diagnostic instruments. Development of such
tools will not only make the assessment process less laborious for participants and researchers,
but will reduce the additional Type I error introduced when testing multiple subscales. While it
must be acknowledged that the use of blinded observers is superior to rating instruments both in
terms of specificity and bias reduction, the costs, in terms of time, training, and supervision, are
great. Therefore, our use of video data collection should be further explored as a potentially
more efficient approach to observational rating.
In addition to more sensitive and targeted measures, future studies are needed to further
validate SCCT as a theory with utility in guiding vocational social skills interventions and
research. While our study adds to the SCCT literature on the impacts of learning (ASSET) on
self-efficacy, career outcome expectations, and social functioning and psychological wellness,
future work using larger sample sizes and mediation analysis can help to confirm if self-efficacy,
as opposed to outcome expectations, accounts for most of the influence on psychological
wellness and social functioning. Understanding the primary drivers of these outcomes can help
to guide refinement of existing interventions and development of new ones.
Additionally, further examination of moderators and other potential mediators can shed
more light on subgroups for which ASSET is most efficacious. Specifically, future research with
stratified random assignment to treatment (Wang & Ware, 2013) is needed to ascertain if SES is
a true moderator of social functioning among ASSET participants and within the HFASD

136

population. Moreover, Wang & Ware (2013) suggest that follow-up studies should include
“stratified randomization of treatment assignments to ensure sufficient representation in the
subgroups of interest so that the study has adequate power to detect the moderation effect” (p.
112). Also, larger samples would allow for the examination of self-efficacy as a potential
mediator of the relationship between social skills training and psychological wellness.
Finally, some broad thoughts regarding the research horizon are offered. As the ASSET
pilot phase and manual development is complete, next steps should include multi-site testing of
the manualized intervention to ensure that it can be delivered with fidelity by a variety of
professionals, such as occupational therapists, vocational rehabilitation counselors,
psychologists, social workers, and school counselors in a variety of academic, community, and
clinical settings. White and colleagues (2007) lament that this step is often overlooked in autism
research, and they go on to stress that multi-site testing can greatly increase confidence in
reporting efficacy as opposed to mere associations. Ideally, multi-site testing should be
implemented using random assignment and control if causal inferences are to be made (White et
al, 2009).
Finally, there is much known about barriers to employment for this population, yet there
is little known about factors that facilitate not just job acquisition, but long-term and meaningful
career fulfillment (Lee & Carter, 2012). Accordingly, expansion of the ASSET database, not
necessarily by increasing sample size, but perhaps by increasing within-participant observations
over time, could both increase statistical power to detect significant change and facilitate a
longer view of what psychosocial challenges and successes occur over the long-term amongst
these young adults.

137

Conclusion
The quantitative results of this mixed methods, quasi-experimental, single group,
staggered enrollment, repeated measures study support an empirical association between the
ASSET intervention and statistically significant, clinically relevant improvements in social
cognition, social communication, empathic self-efficacy, social self-efficacy, and anxiety among
young adults, ages 18-24 years, with autism and average to above-average intelligence.
Moreover, although statistically significant changes in depression were not measured in this
small sample (n = 17), ASSET was associated with steady and growing improvements in selfreported depression up to two months after the intervention was completed.
Furthermore, results of the qualitative analysis suggest varying degrees of social skills
generalization in employment, social, and family contexts up to two years after completing the
program. Specifically, greater transferability of social skills to successful employment
experiences was reported from participants with higher socioeconomic backgrounds; however,
all participants described optimistic, realistic, and positive career expectations and bolstered
confidence in workplace interactions. In addition, few reports of work-related anxiety were
described by participants, and those that were reported could be characterized as normal,
situational reactions to stressful situations. Finally, while no depressive symptoms were shared
in the 1-2-year follow-up focus group, this result should be interpreted with caution due to the
high incidence of depression in the population, and the potential masking of symptoms by
characteristics associated with autism.
In sum, this study provides empirical support that ASSET can: (1) partially meet the need
for manualized, evidence-based strategies to address work-related social skill development
among adults with HFASD, (2) help young adults with HFASD capitalize on the iterative power

138

of bolstered self-efficacy and optimistic career expectations when facing the social challenges
inherent to HFASD, (3) mitigate barriers to delivering such services by making them
community-based and interdisciplinary, and (4) provide an innovative and efficient mechanism
for addressing psychological wellness while targeting transferable skills.
Knowing that comprehensive, theory-based interventions are needed in post-secondary
transition service delivery (Murray & Doren, 2016), ASSET has been designed to help address
the federal mandate for improved transition outcomes for youth with disabilities (WIOA, 2014)
more seamlessly than traditional VR approaches by bringing the intervention to the community
rather than bringing the clients to the office. ASSET achieves this by using an innovative,
interdisciplinary, group-format service delivery model in traditional educational or community
settings. Building on a three-year, iterative, developmental curriculum design, and the results of
a pilot and feasibility study, this in-depth examination of the psychosocial implications of
ASSET provides empirical evidence that carefully designed and administered, group-based
vocational rehabilitation interventions go beyond teaching “skills to pay the bills,” and offer
participants opportunities to increase psychological wellness, refine career outcome expectations,
and engage with their families and communities during the critical period surrounding postsecondary transition.

139

APPENDICES

140

APPENDIX A
Consent Form
Michigan State University
Participant Information and Consent Form
You are cordially invited to participate in a research-based skill training project. Researchers
are required to provide a consent form to inform you about this project, to convey that your
participation is voluntary, to explain risks and benefits of participation, and to empower you to
make an informed decision. You should feel free to ask the researchers any questions you may
have.
PROJECT TITLE: Job-related Social Skills Training Group for Improving Employment
Outcome of Youth with Autism
RESEARCHER: Connie Sung, Ph.D., CRC, Assistant Professor of Rehabilitation Counseling
PROJECT FACILITATOR: June Chen, Ph.D., Research Associate
Annemarie Connor, M.S., Research Assistant
DEPARTMENT AND INSTITUTION: Office of Rehabilitation and Disability Studies, MSU
1. PURPOSE OF PROJECT
The proposed project aims to adapt, further develop, and validate the effectiveness of an
eight-week job-related social skills training group for transition-aged individuals with autism
spectrum disorder (ASD), namely ASSET (Assistive Social Skills and Employment Training).
In this project, researcher will examine the feasibility and applicability of the ASSET program
among youth/young adults with ASD; evaluate the effectiveness of the ASSET program on
individual’s social and empathic skills, self-efficacy, as well as work-related social behaviors.
2. WHAT YOU WILL DO
If you decide to participate in this project, you will be invited to attend 11 individual
and/or group sessions. Each session will be around 1.5 hours in length.
First (1st session), you will attend an initial screening interview session, during which,
you will be asked to complete an assessment and individual interview about your job-related
ability, experiences and career goals. You will then be asked to fill out a set of questionnaires at
home prior to the beginning of the ASSET program.
Second (2nd – 9th session), if you are determined to be eligible for the ASSET program,
you will be invited to attend eight weekly 1.5-hour intervention sessions, start from October 3rd,
2014 to November 21st, 2014. Each session will cover a topic of job-related social skills and
corresponded activities.
Third (10th session), you will attend an exit interview session, during which, you will be
interviewed about your intervention experience with the ASSET program and you will be asked

141

to fill out another set of questionnaires.
Finally (11th session), you will attend a follow-up interview session two months after
completing the ASSET program and you will be asked to fill out the last set of questionnaires.
3. POTENTIAL BENEFITS
As a participant, you will receive the ASSET training program for free, which is a
specifically designed intervention program for youth/young adults to improve their job-related
social skills. Upon participation, we anticipate you will benefit in following areas: interpersonal
social skills in workplace, job-related management skills, teamwork skills, employment-related
communication skills, and problem solving skills.
Further, your participation may also contribute to the further development of the ASSET
program. Results and products of this project can potentially be used for job-related social skills
training for other youth/young adults with ASD to improve their employment outcomes.
4. POTENTIAL RISKS
There are no foreseeable risks associated with participation in this study except that you
may experience anxious feelings about being in a new environment, meeting new people and
doing activities with them. You are encouraged to speak with the researcher and discuss any
issues related to study participation. Trained researchers will take multiple precautions to prevent
unnecessary anxiety and/or work with you throughout the process to reduce your anxiety level.
5. PRIVACY AND CONFIDENTIALITY
All information gathered in this study will be used only for research purposes and be
accessible only by the research personnel affiliated with this project. All data for this project will
be kept strictly confidential. No names or identifiers will be revealed for non-research purposes.
The results of this study may be published or presented at professional meetings, but the
identities of all participants will remain confidential. All research materials will be treated
confidentially and will be stored in a locked file cabinet and/or password protected computer.
The project involves group activities and discussion. The privacy of your personal
information will be kept strictly confidential during these sessions. All the participants in the
group will be asked to keep everything to be discussed in the group private, but this cannot be
guaranteed. In any event, the researchers will take multiple precautions to protect the identity and
confidentiality of all participants in the study. If the researchers find out during the process that
there is any suspected life threatening or illegal abuse or neglect, either to yourself or others, the
researchers will have to report the issues to state officials as required by the law. (Initials
:__________)
The three individual interview sessions (1st, 10th, & 11th session) will be audiotaped
and/or note-taken with your permission for research purposes. The audio recording will then be
transcribed by the researcher and the tape will be erased once the transcriptions are checked for
accuracy. Transcripts of your interview may be reproduced in whole or in part for use in
presentations or written products that result from this project. Neither your name nor any other
identifying information (e.g., your voice) will be used in presentation or in any written products
resulting from the study. (Initials :__________)
6. YOUR RIGHTS TO PARTICIPATE, SAY NO, OR WITHDRAW
We very much hope that you will participate in this project and attend all scheduled

142

sessions. However, your participation is completely voluntary. Refusal to participate will involve
no penalty or loss of benefits to which you are otherwise entitled. You have the right to say no.
You may choose not to answer specific questions or to stop participating at any time. There is no
consequence of withdrawal or incomplete participation.
7. COSTS AND COMPENSATION FOR BEING IN THE PROJECT
There is no cost participating in this project other than the value you place on your time.
And, you will receive three gift cards for compensation at three different time points upon
completion of the project: pre-intervention interview ($15), post-intervention interview ($15) and
2-month follow-up interview ($15).
8. CONTACT INFORMATION
You may ask any questions about the research at any time. If you have concerns or
questions about this study, such as scientific issues or how to do any part of it, please contact the
researchers:
- Dr. Connie Sung at 517-353-1638, or email: csung@msu.edu , or regular mail at:
Michigan State University, 460 Erickson Hall, East Lansing, MI 48824;
- Dr. June Chen at 517-353-8909, or email: junechen@msu.edu, or regular mail at:
Michigan State University, 176 Erickson Hall, East Lansing, MI 48824;
If you have any questions about your role and rights as a research participant, would like
to obtain information or offer input, or would like to register a complaint about this study, you
may contact, anonymously if you wish, the Michigan State University’s Human Research
Protection Programs at 517-355-2180, fax: 517-432-4503, or email: irb@msu.edu, or regular
mail at: 207 Olds Hall, East Lansing, MI 48824.
9. AUDIOTAPE PERMISSION
You are voluntarily to give researcher the permission of audiotaping the three interview
sessions (session 1, session 10 and session 11). The interviews will not be audiotaped if you do
not want it to be. Also, you can ask the audiotaping to be turned off at any time. Please check the
box to indicate your permission to the researcher:
□ I voluntarily agree to be audiotaped during the interviews, have the tape transcribed and
have the transcript used in presentations and written products. (Initials :__________)
□ I do not want my interview to be audiotaped. I voluntarily agree researcher take notes
during the interview and used it in presentations and written products. (Initials :__________)
10. DOCUMENTATION OF INFORMED CONSENT
Again, your participation is completely voluntary and confidential. Your signature below
means that you voluntarily agree to participate in this project. You will be given a copy of this
signed consent form to keep.

________________________________________
Signature of Participants
________________________________________
Signature of Researcher

143

_____________________________
Date
_____________________________
Date

APPENDIX B
Demographics
Assistive Social Skills & Employment Training (ASSET) Program
Please fill out the following form:

1.

Name

2.

What is your ethnic/racial background?
Caucasian
African American
Hispanic/Latino
Asian & Pacific Islanders
Native/Indian American
Other (Please specify) ______________________

3.

What was your official diagnosis?
Autistic disorder
Asperger’s syndrome
Pervasive developmental disorder

4.

At what age were you first diagnosed?

5.

What type of social security benefits are
you receiving (Please check all that apply)
Not receiving any
SSDI (Social Security Disability Insurance)
SSI (Supplemental Social Income)
Medicare
Medicaid
Other (Please specify)
______________________

6.

What is your current employment status?
Employed (work as __________________________,
work _______ hours/week, weekly earnings: $______)
Unemployed, and actively looking for a job
Unemployed, and not looking for a job
Volunteering
Other (Please specify)_________________________

144

APPENDIX C
Questionnaire (Pre-Intervention)
Assistive Social Skills & Employment Training (ASSET) Program
Dear Participant,
Congratulations! You are being selected to participate in the 8-week ASSET Program. To help us to get to know you
better and customized our training program, please complete and submit the following questionnaire prior to
September XXth, 2014. Should you have any questions, please contact Dr. June Chen at junechen@msu.edu.

SECTION A (Demographic Information)
Instructions: Please check the options or fill in the blanks as best descriptions as your situations.
Name: ____________________________________
Date: ____________________________________
7.

What is your ethnic/racial background?
a. Caucasian
b. African American
c. Hispanic/Latino
d. Asian & Pacific Islanders
e. Native/Indian American
f. Other (Please specify) ________

2. What type of social security benefits are you receiving (Please check all that apply)
a. Not receiving any
b. SSDI (Social Security Disability Insurance)
c. SSI (Supplemental Social Income)
d. Other (Please specify) _______________
3. What is your current employment status?
a. Employed (work ____ hours/week)
b. Unemployed, and actively looking for a job
c. Unemployed, and not looking for a job
d. Volunteering
e. Other (Please specify)_________________________
4. Since you are employed now,
● What is your current job? _______________
● Approximately how much is your weekly earning? _______________
● Approximately how many hours do you work per week? _______________
5. Since you are unemployed now, have you ever been employed?
a. Yes
b. No
6. Since you have been employed before,
● What was your last job? _______________
● Approximately how much was your weekly salary for the last job? _______________
● Approximately how many hours did you work per week for the last job? _______________

145

SECTION B (General Self-Efficacy)
Instructions: Please read each of the statements below and indicate how true you think it is for you. There is no
right or wrong answer, please be open and honest in your response.
Not at all
Hardly Moderatel Exactly
true
y true
true
1. I can always manage to solve difficult problems if I try hard
1
2
3
4
enough.
2. If someone opposes me, I can find the means and ways to
1
2
3
4
get what I want.
3. It is easy for me to stick to my aims and accomplish my
1
2
3
4
goals.
4. I am confident that I could deal efficiently with unexpected
1
2
3
4
events.
5. Thanks to my resourcefulness, I know how to handle
1
2
3
4
unforeseen situations.
6. I can solve most problems if I invest the necessary effort.
1
2
3
4
7. I can remain calm when facing difficulties because I can
1
2
3
4
rely on my coping abilities.
8. When I am confronted with a problem, I can usually find
1
2
3
4
several solutions.
9. If I am in trouble, I can usually think of a solution.
1
2
3
4
10. I can usually handle whatever comes my way.
1
2
3
4

SECTION C (Perceived Social Self-Efficacy)
Instructions: Below is a list of statements about yourself. Please rate how well you can do the following.
How well can you…
Not well at Slightly Moderately Well Very
all
well
well
well
1.
express your opinion to people who are talking
1
2
3
4
5
about something of interest to you?
2.
work or study well with others?
1
2
3
4
5
3.
help someone new become part of a group to
1
2
3
4
5
which you belong?
4.
share an interesting experience you had with other
1
2
3
4
5
people?
5.
actively participate in group activities?
1
2
3
4
5

SECTION D (Perceived Empathy Self-Efficacy)
Instructions: Below is a list of statements about yourself. Please rate how well you can do the following.
How well can you…
Not well at Slightly Moderately Well
Very
all
well
well
well
1.
read your friends’ needs?
1
2
3
4
5
2.
recognize when someone wants comfort and
1
2
3
4
5
emotional support, even if (s)he does not overtly
exhibit it?
3.
recognize whether a person is annoyed with you?
1
2
3
4
5
4.
recognize when a person is inhibited by fear?
1
2
3
4
5
5.
recognize when a companion needs your help?
1
2
3
4
5
6.
recognize when a person is experiencing
1
2
3
4
5
depression?

146

SECTION E (Social Functioning Scale)
Instructions: Please read each of the statements below and choose the one that comes closest to how you have been
in the past two to six weeks.
1 I complete my tasks at work and home Most of the
Quite often
Sometimes
Not at
satisfactorily.
time
all
2 I find my tasks at work and at home
Most of the
Quite often
Sometimes
Not at
very stressful.
time
all
3 I have no money problems.
No problems at Slight worries
Definite
Very
all
only
problems
severe
problem
s
4 I have difficulties in getting and
Severe
Some problems Occasional
No
keeping close relationships.
difficulties
problems
problem
s at all
5 I have problems in my romantic
Severe
Some problems Occasional
No
relationship.
difficulties
problems
problem
s at all
6 I get on well with my family and other Yes, definitely
Yes, usually
No, some
No,
relatives.
problems
severe
problem
s
7 I feel lonely and isolated from other
Almost all the
Much of the
Not usually
Not at
people
time
time
all
8 I enjoy my spare time.
Very much
Sometimes
Not often
Not at
all

SECTION F (Autism Spectrum Quotient-10)
Instructions: Read each of the following statements. Please rate to what extent you agree or disagree.
Definitely
Slightly
Slightly Definitely
disagree
disagree
agree
agree
1.
I often notice small sounds when others do not.
0
1
2
3
2.
I usually concentrate more on the whole picture,
0
1
2
3
rather than the small details.
3.
I find it easy to do more than one thing at once.
0
1
2
3
4.
If there is an interruption, I can switch back to
0
1
2
3
what I was doing very quickly.
5.
I find it easy to ‘read between the lines’ when
0
1
2
3
someone is talking to me.
6.
I know how to tell if someone listening to me is
0
1
2
3
getting bored.
7.
When I’m reading a story I find it difficult to work
0
1
2
3
out the characters’ intentions.
8.
I like to collect information about categories of
0
1
2
3
things (e.g. types of car, bird, train, plant etc).
9.
I find it easy to work out what someone is thinking
0
1
2
3
or feeling just by looking at their face.
10. I find it difficult to work out people’s intentions.
0
1
2
3

147

SECTION G (General Anxiety Disorder-7)
Instructions: How often have you been bothered by any of the following problems over the PAST TWO WEEKS?
Please select the answer that best describes you.
Over the PAST TWO WEEKS,
Not at all Several days More than half Nearly
the days
every day
1.
2.
3.
4.
5.
6.
7.

Feeling nervous, anxious or on edge
Not being able to stop or control worrying
Worrying too much about different things
Trouble relaxing
Being so restless that it is hard to sit still
Becoming easily annoyed or irritable
Feeling afraid as if something awful might happen

0
0
0
0
0
0
0

1
1
1
1
1
1
1

2
2
2
2
2
2
2

3
3
3
3
3
3
3

SECTION H (Patient Health Questionnaire-9)
Instructions: Over the PAST TWO WEEKS, how often have you been bothered by any of the following
problems?
Over the PAST TWO WEEKS,
Not at all
Several days More than Nearly
half the every day
days
1.
Little interest or pleasure in doing things
0
1
2
3
2.
Feeling down, depressed, or hopeless
0
1
2
3
3.
Trouble falling or staying asleep, or sleeping too
0
1
2
3
much
4.
Feeling tired or having little energy
0
1
2
3
5.
Poor appetite or overeating
0
1
2
3
6.
Feeling bad about yourself — or that you are a
0
1
2
3
failure or have let yourself or your family down
7.
Trouble concentrating on things, such as reading
0
1
2
3
the newspaper or watching television
8.
Moving or speaking so slowly that other people
0
1
2
3
could have noticed. Or the opposite — being so
fidgety or restless that you have been moving
around a lot more than usual
9.
Thoughts that you would be better off dead, or of
0
1
2
3
hurting yourself in some way
10.

If you checked off any problems, how difficult
have these problems made it for you to do your
work, take care of things at home, or get along with
other people?

Not difficult
at all

148

Somewhat
difficult

Very
difficult

Extremely
difficult

APPENDIX D
Psychosocial Questionnaire (Post-Intervention)
Assistive Social Skills & Employment Training (ASSET) Program
Dear Participant,
Congratulations! You have finished our 8-week ASSET Program. To help us to better explore how our program
works for you, please complete the following survey and bring it back to your exit interview. Should you have any
questions, please contact us at assetmsu@gmail.com or 517-353-1638.
Name: ____________________________________

SECTION A (Usage Rating Profile-Intervention-Revised)
Instructions: Below is a list of statements about your experience in the ASSET program. Please rate to what extent
you agree or disagree.
Strongly
Strongly
Disagree Not sure Agree
disagree
disagree
1. I am motivated to participate in this program.
1
2
3
4
5
2. Each session of the program was implemented within the
1
2
3
4
5
duration of time as stated.
3. The information and materials covered in this program
1
2
3
4
5
was appropriate and relevant to employment.
4. The quality of instruction was good.
1
2
3
4
5
5. The amount of time required to participate in this
1
2
3
4
5
program was reasonable.
6. The amount of weekly assignment was reasonable and
helpful for my learning.
7. I would need consultative support to participate this
1
2
3
4
5
program.
8. The
T expectation of participating in this program was
1
2
3
4
5
clear.
9. The facilitators made the information in this program as
1
2
3
4
5
interesting as possible.
10. I have positive attitudes about participating in this
1
2
3
4
5
program.
11. I could easily understand the content covered in this
1
2
3
4
5
program.
12. I participated in this program with a good deal of
1
2
3
4
5
enthusiasm.
13. The
T strategies used for addressing the topics of this
program were effective.
14. IIhave learned new skills through participation in this
program.
15. I would be interested in participating in a program similar
1
2
3
4
5
to this one in the future.
16. Overall, the program is beneficial for me.
1
2
3
4
5
17. I would recommend this program to others.
1
2
3
4
5
18. I would be interested in participating in the EPASS
1
2
3
4
5
program (extension of ASSET program) in spring
semester.

149

SECTION B (Perceived Improvement)
Instructions: Below is a list of statements about your perceived improvement after participating in the ASSET
program. Please rate to what extent you feel you have improved in following aspects.
After participating in the ASSET
Not improved
Slightly
Moderately
Largely
program, my ability in…
at all
improved
improved
improved
1.
communicating with others has
1
2
3
4
2.
showing positive attitudes and enthusiasm
1
2
3
4
has
3.
working in a team has
1
2
3
4
4.
networking with others has
1
2
3
4
5.
critical thinking has
1
2
3
4
6.
problem solving has
1
2
3
4
7.
being professional has
1
2
3
4

SECTION C (Perceived Confidence)
Instructions: Below is a list of statements about your self-efficacy after participating in the ASSET program. Please
rate to what extent you feel confident about yourself in following aspects.
After participating in the ASSET program, my Much less Less
Equally
More Much more
self-efficacy in…
confident confident confident confident confident
1. soft skills (job-related social skills) is
1
2
3
4
5
2. transition from school to work is
1
2
3
4
5
3. finding and maintaining a job in the future is
1
2
3
4
5
4. becoming a better employee is
1
2
3
4
5
5. working with others in the workplace is
1
2
3
4
5
6. succeeding in the workplace is
1
2
3
4
5

SECTION D (General Self-Efficacy)
Instructions: Please read each of the statements below and indicate how true you think it is for you. There is no
right or wrong answer, please be open and honest in your response.
Not at all Hardly true Moderately Exactly
true
true
1. I can always manage to solve difficult problems if I try
1
2
3
4
hard enough.
2. If someone opposes me, I can find the means and ways to
1
2
3
4
get what I want.
3. It is easy for me to stick to my aims and accomplish my
1
2
3
4
goals.
4. I am confident that I could deal efficiently with
1
2
3
4
unexpected events.
5. Thanks to my resourcefulness, I know how to handle
1
2
3
4
unforeseen situations.
6. I can solve most problems if I invest the necessary effort.
1
2
3
4
7. I can remain calm when facing difficulties because I can
1
2
3
4
rely on my coping abilities.
8. When I am confronted with a problem, I can usually find
1
2
3
4
several solutions.
9. If I am in trouble, I can usually think of a solution.
1
2
3
4
10. I can usually handle whatever comes my way.
1
2
3
4

150

SECTION E (Perceived Social Self-Efficacy)
Instructions: Below is a list of statements about yourself. Please rate how well you can do the following.
How well can you…
Not well Slightly Moderately Well
Very well
at all
well
well
1.
express your opinion to people who are talking
1
2
3
4
5
about something of interest to you?
2.
work or study well with others?
1
2
3
4
5
3.
help someone new become part of a group to
1
2
3
4
5
which you belong?
4.
share an interesting experience you had with
1
2
3
4
5
other people?
5.
actively participate in group activities?
1
2
3
4
5

SECTION F (Perceived Empathy Self-Efficacy)
Instructions: Below is a list of statements about yourself. Please rate how well you can do the following.
How well can you…
Not well Slightly Moderately Well
Very well
at all
well
well
1. 1.
read your friends’ needs?
1
2
3
4
5
2.
recognize when someone wants comfort and
1
2
3
4
5
emotional support, even if (s)he does not overtly
exhibit it?
3.
recognize whether a person is annoyed with
1
2
3
4
5
you?
4.
recognize when a person is inhibited by fear?
1
2
3
4
5
5.
recognize when a companion needs your help?
1
2
3
4
5
6.
recognize when a person is experiencing
1
2
3
4
5
depression?

SECTION G (Social Functioning Scale)
Instructions: Please read each of the statements below and choose the one that comes closest to how you have been
in the past two to six weeks.
1 I complete my tasks at work and
Most of the
Quite often
Sometimes
Not at all
home satisfactorily.
time
2 I find my tasks at work and at home
Most of the
Quite often
Sometimes
Not at all
very stressful.
time
3 I have no money problems.
No problems
Slight worries Definite
Very severe
at all
only
problems
problems
4 I have difficulties in getting and
Severe
Some
Occasional
No problems at
keeping close relationships.
difficulties
problems
problems
all
5 I have problems in my romantic
Severe
Some
Occasional
No problems at
relationship.
difficulties
problems
problems
all
6
7
8

I get on well with my family and
other relatives.
I feel lonely and isolated from other
people
I enjoy my spare time.

Yes,
definitely
Almost all the
time
Very much

151

Yes, usually
Much of the
time
Sometimes

No, some
problems
Not usually

No, severe
problems
Not at all

Not often

Not at all

SECTION F (Autism Spectrum Quotient-10)
Instructions: Read each of the following statements. Please rate to what extent you agree or disagree.
Definitely
Slightly
Slightly
Definitely
disagree
disagree
agree
agree
I often notice1.small sounds when others do not.
0
1
2
3
2.
I usually concentrate more on the whole
0
1
2
3
picture, rather than the small details.
3.
I find it easy to do more than one thing at once.
0
1
2
3
4.
If there is an interruption, I can switch back to
0
1
2
3
what I was doing very quickly.
5.
I find it easy to ‘read between the lines’ when
0
1
2
3
someone is talking to me.
6.
I know how to tell if someone listening to me
0
1
2
3
is getting bored.
7.
When I’m reading a story I find it difficult to
0
1
2
3
work out the characters’ intentions.
8.
I like to collect information about categories of
0
1
2
3
things (e.g. types of car, bird, train, plant etc).
9.
I find it easy to work out what someone is
0
1
2
3
thinking or feeling just by looking at their face.
10. I find it difficult to work out people’s
0
1
2
3
intentions.

SECTION G (General Anxiety Disorder-7)
Instructions: How often have you been bothered by any of the following problems over the PAST TWO WEEKS?
Please select the answer that best describes you.
Over the PAST TWO WEEKS,
Not at all
Several
More than Nearly every
days
half the days
day
2.
3.
4.
5.
6.
7.

Feeling nervous,
1. anxious or on edge
Not being able to stop or control worrying
Worrying too much about different things
Trouble relaxing
Being so restless that it is hard to sit still
Becoming easily annoyed or irritable
Feeling afraid as if something awful might
happen

0
0
0
0
0
0
0

152

1
1
1
1
1
1
1

2
2
2
2
2
2
2

3
3
3
3
3
3
3

SECTION H (Patient Health Questionnaire-9)
Instructions: Over the PAST TWO WEEKS, how often have you been bothered by any of the following
problems?
Over the PAST TWO WEEKS,
Not at all Several days More than
Nearly
half the days every day
Little interest1.or pleasure in doing things
0
1
2
3
2.
Feeling down, depressed, or hopeless
0
1
2
3
3.
Trouble falling or staying asleep, or sleeping
0
1
2
3
too much
4.
Feeling tired or having little energy
0
1
2
3
5.
Poor appetite or overeating
0
1
2
3
6.
Feeling bad about yourself — or that you are a
0
1
2
3
failure or have let yourself or your family down
7.
Trouble concentrating on things, such as
0
1
2
3
reading the newspaper or watching television
8.
Moving or speaking so slowly that other people
0
1
2
3
could have noticed. Or the opposite — being so
fidgety or restless that you have been moving
around a lot more than usual
9.
Thoughts that you would be better off dead, or
0
1
2
3
of hurting yourself in some way
10.

If you checked off any problems, how difficult Not difficult
have these problems made it for you to do your
at all
work, take care of things at home, or get along
with other people?

153

Somewhat
difficult

Very
difficult

Extremely
difficult

APPENDIX E
Focus Group Protocol (1-2 years Post-Intervention)

Interview Questions
This focus group is designed to explore how ASSET impacts participants’ self-perceived
psychosocial functioning (social functioning, self-efficacy, and psychological wellness) and
outcome expectations over time.
1.

2.

3.
4.
5.
6.
7.

8.

Discuss any changes you have noticed in yourself since participating in ASSET
a. Probe for changes in:
i. Work-related self-efficacy
ii. Social functioning
iii. Anxiety
iv. Depression
Discuss any changes others have noticed in you since participating in ASSET
a. Probe for changes in:
i. Work-related self-efficacy
ii. Social functioning
iii. Anxiety
iv. Depression
How have these changes impacted your ability to work with others in school, home,
leisure, or work settings?
Discuss anything that has been easier for you to do since participating in ASSET.
Discuss anything that you feel you do better since participating in ASSET.
Discuss anything that you feel more confident doing since participating in ASSET.
What are your current work and independent living expectations/goals
a. Have they changed since participating in ASSET?
i. If yes, please explain.
What else would you like to share with us about your experiences in ASSET and any
changes since then?

154

APPENDIX F
Human Subjects Institutional Review Board Approval

155

REFERENCES

156

REFERENCES

Albright, J. J., & Marinova, D. M. (2015). Estimating multilevel models using SPSS, Stata, SAS
and R. Retrieved from https://scholarworks.iu.edu/dspace/handle/2022/19737.
American Psychiatric Association [APA]. (2013). Diagnostic and statistical manual of mental
disorders (5th ed.). Arlington, VA: Author.
Anderson, K. A., P. T. Shattuck, B. P. Cooper, A. M. Roux, and M. Wagner (2014). Prevalence
and correlates of postsecondary residential status among young adults with an autism
spectrum disorder. Autism 18(5): 562–570.
Bandura, A. (1997). Self-efficacy: the exercise of control. New York: Freeman.
Bandura, A. (1986). Social foundations of thought and action: a social cognitive theory.
Englewood Cliffs, NC: Prentice-Hall.
Baron-Cohen, S. (1997). Mindblindness: an essay on autism and theory of mind. Cambridge:
The MIT Press.
Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Clubley, E. (2001). The autismspectrum quotient (AQ): Evidence from Asperger syndrome/high-functioning autism,
males and females, scientists and mathematicians. Journal of Autism and Developmental
Disorders, 31(1), 5.
Betz, N. E. (2007). Career self-efficacy: Exemplary recent research and emerging directions.
Journal of Career Assessment, 15(4), 403-422.
Bliese, P. D., & Hanges, P. J. (2004). Being both too liberal and too conservative: The perils of
treating grouped data as though they are independent. Organizational Research Methods,
7, 400–417.
Bolte, S., Poustka, F., & Constantino, J. N. (2008). Assessing autistic traits: Cross-cultural
validation of the Social Responsiveness Scale (SRS). Autism Research, 1, 354-363. DOI:
10.1002/aut/49
Brereton, A. V., Tonge, B. J., & Einfeld, S. L. (2006). Psychopathology in children and
adolescents with autism compared to young people with intellectual disability. Journal of
Autism and Developmental Disorders, 36, 863–870.
Brown, S. D., Lent, R. W., & Gore, P. A., Jr. (2000). Self-rated abilities and self-efficacy beliefs:
Are they empirically distinct? Journal of Career Assessment, 8, 223-235.
Brown, L. A., Wiley, J. F., Wolitzky-Taylor, K., Roy-Byrne, P., Sherbourne, C., Stein, M. B., …

157

Craske, M. G. (2014). Changes in self-efficacy and outcomes expectancy as predictors of
anxiety outcomes from the calm study. Depression and Anxiety, 31(8), 678–689.
http://doi.org/10.1002/da.22256http://doi.org/10.1002/da.22256
Cappadocia, M. C., & Weiss, J. A. (2011). Review of social skills training groups for youth with
Asperger Syndrome and high functioning autism. Research in Autism Spectrum
Disorders, 5, 70-78.
Carlsen, B., & Glenton, C. (2011). What about N? A methodological study of sample-size
reporting in focus group studies. BMC Medical Research Methodology, 11(1), 26-26.
Carter, A., Davis, N., Kiln, A., & Volkmar, F. (2005). Social development in autism. In FR.
Volkmar, R. Paul, A. Klin, & D. Cohen (Eds.). Handbook of autism and pervasive
developmental disorders (3rd ed., pp 312-334). Hoboken, NJ: Wiley.
Cassidy, S., Bradley, P., Robinson, J., Allison, C., McHugh, M., & Baron-Cohen, S. (2014).
Suicidal ideation and suicide plans or attempts in adults with Asperger's syndrome
attending a specialist diagnostic clinic: A clinical cohort study. The Lancet Psychiatry,
1(2), 142-147.
Chadsey-Rusch, J. (1990). Social interactions of secondary-aged students with severe handicaps:
Implications for facilitating the transition from school to work. JASH, 15(2), 69-78.
Chan, F., Bezyak, J., Ramirez, M., Chiu, C., & Sung, C. (2010). Concepts, challenges, barriers,
and opportunities related to evidence-based practice in rehabilitation counseling.
Rehabilitation Education, 24(3,4), 179-190.
Chen, J. L., Leader, G., Sung, C., & Leahy, M. (2014). Trends in employment for individuals
with autism spectrum disorder: a review of the research literature. Review Journal of
Autism and Developmental Disorders, 2, 1–13.
Cloutier, H., Kurtz, A., May, J., & Hagner, D. (2014). Person--centered planning for transitionaged youth with autism spectrum disorders. The Journal of Rehabilitation, 80(1), 4.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation
analysis for the behavioral sciences, Third Edition. Mahwah, NJ: Earlbaum.
Cone, J. D., & Foster, S. L. (2006). Dissertations and theses from start to finish: Psychology and
related fields (2nd ed.). Washington, DC: American Psychological Association.
Connor, A. (2017). The comprehensive disability counseling needs of transition-age youth: An
Interagency Case Study. Manuscript in preparation.
Constantino, J. N. (2012). Social Responsiveness Scale. Los Angeles: Western Psychological
Services.

158

Corrigan, M. J., Jones, C. A., & McWhirter, J. J. (2001). College students with disabilities: An
access employment group. Journal for Specialists in Group Work, 26, 339-349.
Constantino, J. N., Lavesser, P. D., Zhang, Y., Abbacchi, A. M., Gray, T., & Todd, R. D. (2007).
Rapid quantitative assessment of autistic social impairment by classroom teachers.
Journal of the American Academy of Child and Adolescent Psychiatry, 46, 1668-1676.
Croen, L. A., Zerbo, O., Qian, Y., Massolo, M. L., Rich, S., Sidney, S., & Kripke, C. (2015). The
health status of adults on the autism spectrum. Autism, 19(7), 814-823.
Creswell, J. W. (2009). Research design: qualitative, quantitative, and mixed methods
approaches. Los Angeles: Sage.
Davis, C. S. (2002). Statistical methods for the analysis of repeated measurements. New York,
NY: Springer.
Dooley, D. (2003). Unemployment, underemployment, and mental health: Conceptualizing
employment status as a continuum. American Journal of Community Psychology, 32
(1/2), 9-20.
Elksnin, N., & Elksnin, L. K. (2001). Adolescents with disabilities: The need for occupational
social skills training. Exceptionality, 9, 91–105.
Facts About ASDs. (n.d.). Retrieved February 29, 2016, from
http://www.cdc.gov/ncbddd/autism/facts.html
Federal Autism Research: Updated Information on Funding from Fiscal Years 2008 through
2012. (n.d.). Retrieved February 29, 2016, from http://www.gao.gov/products/GAO-15583R
Fetters, M. D., Curry, L. A., & Creswell, J. W. (2013). Achieving integration in mixed methods
Designs—Principles and practices. Health Services Research, 48(6pt2), 2134-2156.
http://www.gao.gov/products/GAO-15-583R
Field, A. P. (2013). Discovering statistics using IBM SPSS statistics: And sex and drugs and rock
'n' roll (4th ed.). Los Angeles: Sage.
Fouad, N. A., & Guillen, A. (2006). Outcome expectations: Looking to the past and potential
future. Journal of Career Assessment, 14(1), 130-142.
Fremstad, S., Vallas, R. (2013). The facts on social security disability insurance and
supplemental security income for workers with disabilities. Retrieved from
https://www.americanprogress.org/issues/poverty/reports/2013/05/30/64681/the-facts-onsocial-security-disability-insurance-and-supplemental-security-income-for-workers-withdisabilities/)

159

Gerhardt, P., & Holmes, D. (1997). Employment: Options and issues for adolescents and adults
with autism. In D. Cohen & F. Volkmar (Eds.), Handbook of autism and pervasive
developmental disorders (2nd ed., pp. 650–664). New York: Wiley & Sons.
Geurts, H. M., Corbett, B., & Solomon, M. (2009). The paradox of cognitive flexibility in
autism. Trends in Cognitive Sciences, 13, 74–82.
Greenhaus, J. H. & Callanan, G. A. (Eds.) (2006). Encyclopedia of career development
Thousand Oaks, CA: SAGE Publications Ltd. doi: 10.4135/978141295267
Hayes, A. F. (2006). A primer on multilevel modeling. Human Communication Research, 32(4),
385-410.
Heck, R. H., Thomas, S. L., & Tabata, L. N. (2014). Multilevel and longitudinal modeling with
IBM SPSS (Second ed.). New York: London.
Hedges, L. V. (2007). Effect sizes in cluster-randomized designs. Journal of Educational and
Behavioral Statistics, 32, 41-370.
Hedley, D., & Young, R. (2006). Social comparison processes and depressive symptoms in
children and adolescents with Asperger syndrome. Autism, 10(2), 139–153.
Howlin, P., Alcock, J., & Burkin, C. (2005). An 8-year follow-up of a specialist supported
employment service for high-ability adults with austim or Aspberger’s syndrome.
Autism, 9(5), 533-549.
Howlin, P., & Moss, P. (2012). Adults with autism spectrum disorders. The Canadian Journal of
Psychiatry, 57(5), 275–283.
Hillier, A. J., Fish, T., Siegel, J. H., & Beversdorf, D. Q. (2011). Social and vocational skills
training reduces self-reported anxiety and depression among young adults on the autism
spectrum. Journal of Developmental and Physical Disabilities, 23(3), 267–276.
http://doi.org/10.1007/s10882-011-9226-4
Hua, C. B. (2002). Career self-efficacy of the student who is gifted/learning disabled: A case
study. Journal for the Education of the Gifted, 25, 375-405.
Hurlbutt, K., & Chalmers, L. (2004). Employment and adults with Asperger syndrome.
Focus on Autism and Other Developmental Disabilities, 19, 215-222
Individuals with Disabilities Education Act Public Data and Resources. (2013). Fall 2013
Individuals with Disabilities Education Act Data. Available from
http://www2.ed.gov/programs/osepidea/618-data/index.html

Interagency Autism Coordinating Committee (IACC) (2011). IACC Strategic Plan for Autism
Spectrum Disorder Research. Washington, DC: US Department of Health and Human
Services.
160

Jahoda, M. (1982). Employment and unemployment. Cambridge; Cambridge University Press.
Jang, J., Matson, J. L., Adams, H. L., Konst, M. J., Cervantes, P. E., & Goldin, R. L. (2014).
What are the ages of persons studied in autism research: A 20-year review. Research in
Autism Spectrum Disorders, 8(12), 1756–1760. doi:10.1016/j.rasd.2014.08.008
Joseph, R. M., Ehrman, K., McNally, R., & Keehn, B. (2008). Affective response to eye contact
and face recognition ability in children with ASD. Journal of the International
Neuropsychological Society: JINS, 14(6), 947–955.
http://doi.org/10.1017/S1355617708081344

Kim, J. A., Szatmari, P., Bryson, S. E., Streiner, D. L., & Wilson, F. J. (2000). The prevalence of
anxiety and mood problems among children with autism and Asperger syndrome.
Autism, 4, 117–132.
Kleinhans, N. M., Richards, T., Johnson, L. C., Weaver, K. E., Greenson, J., Dawson, G., et al.
(2011). fMRI evidence of neural abnormalities in the subcortical face processing system
in ASD. Neuroimage, 54, 697–704.
Kohler, P. D. (1994). On-the-job training: A curricular approach to employment. Career
Development for Exceptional Individuals, 17(1), 29-40.
Laugeson, E. A., Frankel, F., Gantman, A., Dillon, A. R., & Mogil, C. (2012). Evidence-Based
Social Skills Training for Adolescents with Autism Spectrum Disorders: The UCLA
PEERS Program. Journal of Autism and Developmental Disorders, 42(6), 1025–1036.
http://doi.org/10.1007/s10803-011-1339-1
Laugeson, E. A., Gantman, A., Kapp, S. K., Orenski, K., & Ellingsen, R. (2015). A randomized
controlled trial to improve social skills in young adults with autism spectrum disorder:
The UCLA PEERSÂŽ program. Journal of Autism and Developmental Disorders, 45(12),
3978-3989.
Laugeson, E. A., & Park, M. N. (2014). Using a CBT Approach to Teach Social Skills to
Adolescents with Autism Spectrum Disorder and Other Social Challenges: The PEERS
Method. Journal of Rational - Emotive & Cognitive - Behavior Therapy, 32(1), 84–97.
http://doi.org/http://dx.doi.org.proxy2.cl.msu.edu/10.1007/s10942-014-0181-8

Lawler, E. E., & Finegold, D. (2000). Individualizing the organization: Past, present, and future.
Organizational Dynamics, 29, 1-15.
Leahy, M. J., Chan, F., Lui, J., Rosenthal, D., Tansey, T., Wehman, P., … others. (2014). An
analysis of evidence-based best practices in the public vocational rehabilitation program:
Gaps, future directions, and recommended steps to move forward. Journal of Vocational
Rehabilitation, 41(2), 147–163.
Leahy, M.J., Tarvydas, V.M., Connor, A., Landon, T. (in press). Rehabilitation counseling. In
Nassar-McMillion (Ed.), Professional Counseling: Orientation to Past, Present, and
161

Future Trends. ACA Press.
Leary, M. R., & Atherton, S.C. (1986). Self-efficacy, social anxiety, and inhibition in
interpersonal encounters. Journal of Social and Clinical Psychology, 4(3), 256-267.
Lee, G. K., & Carter, E. W. (2012). Preparing transition-age students with high functioning
autism spectrum disorders for meaningful work. Psychology in the Schools, 49, 9881000.
Lent, R. W., & Brown, S. D. (1996). Social cognitive approach to career development: An
overview. The Career Development Quarterly, 44(4), 310-321.
Lent, R. W., Brown, S. D., & Hackett, G. (1994). Toward a unifying social cognitive theory of
career and academic interest, choice, and performance. Journal of Vocational Behavior,
45(1), 79-122. doi:10.1006/jvbe.1994.1027
Lent, R. W., Brown, S. D., & Hackett, G. (2000). Contextual supports and barriers to career
choice: A social cognitive analysis. Journal of Counseling Psychology, 47, 36-49.
doi:10.1037/0022-0167.47.1.36
Lent, R. W., Brown, S. D., Nota, L., & Soresi, S. (2003). Testing social cognitive interest and
choice hypotheses across Holland types in Italian high school students. Journal of
Vocational Behavior, 62, 101-118.
Lent, R. W., Brown, S. D., Schmidt, J., Brenner, B., Lyons, H., & Treistman, D. (2003).
Relation of contextual supports and barriers to choice behavior in engineering majors:
Test of alternative social cognitive models. Journal of Counseling Psychology, 50, 458–
465. doi:10.1037/0022-0167.50.4.458
Lent, R. W., Sheu, H. B., Singley, D., Schmidt, J. A., Schmidt, L. C., & Gloster, C. S.
(2008). Longitudinal relations of self-efficacy to outcome expectations, interests, and
major choice goals in engineering students. Journal of Vocational Behavior, 73, 328335. doi:10.1016/j.jvb.2008.07.005
Lent, R.W., Brown, S. D., Talleyrand, R., McPartland, E. B., Davis, T., Chopra, S. B., et al.
(2002). Career choice barriers, supports, and coping strategies: college students’
experiences. Journal of Vocational Behavior, 60, 61-72.
Levy, S.E., Mandell, D.S., & Schultz, R.T. (2009). Autism. Lancet, 374, 1627–1638
Leyfer, O. T., Folstein, S. E., Bacalman, S., Davis, N. O., Dinh, E., Morgan, J., et al. (2006).
Comorbid psychiatric disorders in children with autism: interview development and rates
of disorders. Journal of Autism and Developmental Disorders, 36, 849–861.
McMahon, C., Lerner, M., & Britton, N. (2013). Group-based social skills interventions for
adolescents with higher-functioning autism spectrum disorder: A review and looking to

162

the future. Adolescent Health, Medicine and Therapeutics,4 , 23-38.
Lorenz, T., Frischling, C., Cuadros, R., & Heinitz, K. (2016). Autism and overcoming job
barriers: Comparing job-related barriers and possible solutions in and outside of autismspecific employment. PLOS ONE, 11(1), e0147040.
http://doi.org/10.1371/journal.pone.0147040
Locke, J., Beidas, R. S., Marcus, S., Stahmer, A., Aarons, G. A., Lyon, A. R., . . . Mandell, D. S.
(2016). A mixed methods study of individual and organizational factors that affect
implementation of interventions for children with autism in public schools.
Implementation Science, 11(1)
McKenzie, R., Evans, Jonathan St. B. T, & Handley, S. J. (2010). Conditional reasoning in
autism: Activation and integration of knowledge and belief. Developmental Psychology,
46(2), 391-403. doi:10.1037/a0017412
Müller, E., Schuler, A., Burton, B., & Yates, G. (2003). Meeting the vocational support needs of
individuals with Asperger Syndrome and other autism spectrum disabilities. Journal of
Vocational Rehabilitation, 18, 163-175.
Muris, P. (2001). A brief questionnaire for measuring self-efficacy in children with
affective problems. Journal of Psychopathology and Behavioral Assessment, 23, 145149.
MuthĂŠn, B. O., & Satorra, A. (1989). Multilevel aspects of varying parameters in structural
models. In R. D. Bock (Ed.), Multilevel analysis of educational data (pp. 87−99). San
Diego: Academic Press.
MuthĂŠn, B. O., & Satorra, A. (1995). Complex sample data in structural equation modeling. In P.
V. Marsden (Ed.), Sociological methodology, 1995 (pp. 267−316). Oxford: Blackwell.
Moe, K. O., & Zeiss, A. M. (1982). Measuring self-efficacy expectations for social skills: A
methodological inquiry. Cognitive Therapy and Research, 6(2), 191–205.
National Longitudinal Transition Study-2 [NLTS2]. (2009). Data tables
[Data file]. Available from
http://www.nlts2.org/data_tables/index.htmlhttp://www.nlts2.org/data_tables/index.html
Newman, L., Wagner, M., Knokey, A. M., Marder, C, Nagle, K., Shaver, D., et al. (2011). The
post-high school outcomes of young adults with disabilities up to 8 years after high
school: A report from the National Longitudinal Transition Study-2 (NCSER No. 20113005). Menlo Park, CA: SRI International.
Newman, L. (2007, April). Facts from NLTS2: Secondary school experiences of students
with autism. Menlo Park, CA: SRI International. Retrieved from
www.nlts2.org/fact_sheets/nlts2_fact_sheet_2007_04.pdfhttp://www.nlts2.org/fact_sheet

163

s/nlts2_fact_sheet_2007_04.pdf
Paul, G.L. (1967). Strategy of outcome research in psychotherapy. Journal of Consulting
Psychology, 31, 109–118.
Peugh, J. L. (2010). A practical guide to multilevel modeling. Journal of School Psychology,
48(1), 85-112.
Peugh, J. L., & Enders, C. K. (2005). Using the SPSS mixed procedure to fit cross-sectional and
longitudinal multilevel models. Educational and Psychological Measurement, 65(5),
717-741.
Pew Research Group (2012). Chapter 2: Young adults in the U.S. labor market. Pew Research
Center, Washington, D.C. Retrieved from
http://www.pewsocialtrends.org/2012/02/09/chapter-2-young-adults-in-the-u-s-labormarket/
Post, M., Haymes, L., Storey, K., Loughrey, T., & Campbell, C. (2014;2012;). Understanding
stalking behaviors by individuals with autism spectrum disorders and recommended
prevention strategies for school settings. Journal of Autism and Developmental
Disorders, 44(11), 2698-2706.
Rao, P., Beidel, D., & Murray, M. (2008). Social skills interventions for children with
Asperger’s syndrome or high-functioning autism: A review and recommendations.
Journal of Autism and Developmental Disorders, 38, 353-361.
Raykov, T., & Marcoulides, G. A. (2010). Group Comparisons in the Presence of Missing Data
Using Latent Variable Modeling Techniques. Structural Equation Modeling, 17, 135149.
Remler, D. K., & Van Ryzin, G. G. (2011). Research methods in practice: Strategies for description and
causation. Thousand Oaks: SAGE Publications.

Riesen, T., Schultz, J., Morgan, R., & Kupferman, S. (2014). School-to-work barriers as
identified by special educators, vocational rehabilitation counselors, and community
rehabilitation professionals. The Journal of Rehabilitation, 80(1), 33.
Segrin, C. (1996). The relationship between social skills deficitsand psychosocial problems: A
test of a vulnerability model. Communication Research, 23, 425–450.
Segrin, C. (2000). Social skills deficits associated with depression. Clinical Psychology Review,
20, 379–403.
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental
designs for generalized causal inference. Boston: Houghton Mifflin.
Shattuck, P. T., Narendorf, S. C., Cooper, B., Sterzing, P. R., Wagner, M., & Taylor, J. L.
(2012). Postsecondary education and employment among youth with an autism spectrum
164

disorder. Pediatrics, 2011–2864.
Shek, D. T. L., & Ma, C. M. S. (2011). Longitudinal data analyses using linear mixed models in
SPSS: Concepts, procedures and illustrations. The Scientific World Journal, 11, 42-76.
doi:10.1100/tsw.2011.2
Simpson, H. B. (2010). Anxiety disorders: Theory, research, and clinical perspectives.
Cambridge: Cambridge University Press.
Singer, J.D., & Willett, J.B. (2003). Applied longitudinal data analysis. New York: Oxford
Press.
Skrondal, A., & Rabe-Hesketh, S. (2008). Multilevel and related models for longitudinal data. In
J. d. L. E. Meijer (Ed.), Handbook of Multilevel Analysis (pp. 275-299). New York, NY:
Spring Science + Business Media.
Smith, T., Scahill, L., Dawson, G., Guthrie, D., Lord, C., Odom, S., . . . Wagner, A. (2007).
Designing research studies on psychosocial interventions in autism. Journal of Autism
and Developmental Disorders, 37(2), 354-366.
Snijders, T., & Bosker, R. (1999). Multilevel Analysis: An introduction to basic and advanced
multilevel modeling. London: Sage Publications.
Spain, D., & Blainey, S. H. (2015). Group social skills interventions for adults with highfunctioning autism spectrum disorders: A systematic review. Autism, 19(7), 874–886.
SPSS (2005). Linear mixed-effects modeling in SPSS: An introduction to the mixed procedure.
Retrieved from
http://www.spss.ch/upload/1126184451_Linear%20Mixed%20Effects%20Modeling%2
0in%20SPSS.pdf
Spybrook, J., Raudenbush, S. W., Liu, X. F., Congdon, R., & Martinez, A. (2008). Optimal
design software for multi-level and longitudinal research: Documentation for the
“Optimal Design” software (Version 1.77) [Computer software]. Available at
http://sitemaker.umich.edu/group-based
Sterling, L., Dawson, G., Estes, A., & Greenson, J. (2008). Characteristics associated with
presence of depressive symptoms in adults with autism spectrum disorder. Journal of
Autism and Developmental Disorders, 38, 1011–1018.
Stewart, M. E., Barnard, L., Pearson, J., Hasan, R., & O’Brien, G. (2006). Presentation of
depression in autism and Asperger syndrome. Autism, 10(1), 103–116.
Sung, C., Connor, A., Chen, J., Lin, C. Kuo, H., Chun, J. (2017). Development, feasibility and
preliminary efficacy of an employment-related social skills intervention for young adults
with high functioning autism. Manuscript in preparation.

165

Tabachnick, B.G. & Fidell, L.S. (2013). Using multivariate statistics for the social sciences (6th
edn). Boston: Pearson Education.
Taylor, J. L., & Seltzer, M. M. (2011). Employment and postsecondary educational activities for
young adults with autism spectrum disorders during the transition to adulthood. Journal
of Autism and Developmental Disorders, 41, 566–574. doi:10.1007/s10803-010-1070-3
Toomey, J. A., Lopata, C., Volker, M. A., & Thomeer, M. L. (2009). Comprehensive
intervention for high-functioning autism spectrum disorders: An in-depth case study. In
M. T. Burton (Ed.), Special education in the 21st century (pp. 95–118). Hauppauge, NY:
Nova Science.
Twisk, J.W.R. (2006) Applied Multilevel Analysis: A Practical Guide for Medical Researchers.
Available at: https://www.cambridge.org/core/books/applied-multilevelanalysis/7755169EDD3B2304B041773D48F2BA6F.
United States Department of Labor (2017). Disability resources. Retrieved from
https://www.dol.gov/odep/topics/disability.htm.
Van Eylen, L., Boets, B., Steyaert, J., Evers, K., Wagemans, J., & Noens, I. (2011). Cognitive
flexibility in autism spectrum disorder: Explaining the inconsistencies? Research in
Autism Spectrum Disorders, 5(4), 1390-1401.
Vickerstaff, S., Heriot, S., Wong, M., Lopes, A., & Dossetor, D. (2007). Intellectual ability, selfperceived social competence, and depressive symptomatology in children with highfunctioning autistic spectrum disorders. Journal of Autism and Developmental Disorders,
37, 1647–1664.
Vocational Rehabilitation Act Amendments of 2014 (Workforce Innovation and Opportunity
Act). § 113-128.
Wang, R., & Ware, J. H. (2013). Detecting moderator effects using subgroup analyses.
Prevention Science : The Official Journal of the Society for Prevention Research, 14(2),
111–120.
Wechsler, D. (2011). Wechsler Abbreviated Scale of Intelligence, second edition. San Antonio,
TX: Pearson.
Wehman, P. H., Schall, C. M., McDonough, J., Kregel, J., Brooke, V., Molinelli, A., . . . Thiss,
W. (2014). Competitive employment for youth with autism spectrum disorders: Early
results from a randomized clinical trial. Journal of Autism and Developmental
Disorders, 44(3), 487-500.
Wehman, P., Smith, M. D., & Schall, C. (2009). Autism and the transition to adulthood: Success
beyond the classroom. Baltimore: Paul H. Brookes Pub. Co.

166

West, B. T., Welch, K. B., & Galecki, A. T. (2014). Linear mixed models: A practical guide
using statistical software (Second edition). Boca Raton: CRC Press.
Westbrook, J.D., Fong, C.J., Nye, C., Williams, A., Oliver, W., Cortopassi, T. (2013). Pregraduation transition services for improving employment outcomes among persons with
autism spectrum disorders: A systematic review. Campbell Systematic Reviews
2013:11. DOI: 10.4073/csr.2013.11
White, S. W., Koenig, K., & Scahill, L. (2007). Social skills development in children with autism
spectrum disorders: A review of the intervention research. Journal of Autism and
Developmental Disorders, 37(10), 1858-1868.
White, S. W., & Roberson-Nay, R. (2009). Anxiety, social deficits, and loneliness in youth with
autism spectrum disorders. Journal of Autism and Developmental Disorders, 39, 1006–
1013.
White, S. W., Scahill, L., Klin, A., Koenig, K., & Volkmar, F. R. (2007). Educational placements
and service use patterns of individuals with autism spectrum disorders. Journal of Autism
and Developmental Disorders, 37, 1403–1412. doi:10.1007/s10803-006-0281-0
Wilkinson, L., & The Task Force on Statistical Inference. (1999). Statistical methods in
psychology journals. American Psychologist, 54, 594-604.
Williams, D. M. (2010). Outcome expectancy and self-efficacy: Theoretical implications of an
unresolved contradiction. Personality and Social Psychology Review, 14(4), 417–425.
http://doi.org/10.1177/1088868310368802http://doi.org/10.1177/1088868310368802
Winer, B. J., Brown, D. R., & Michels, K. M. (1991). Statistical principles in experimental
design. (3rd ed.). New York: McGraw-Hill.
Winter, B. (2013). Linear models and linear mixed effects models in R with linguistic
applications. arXiv:1308.5499. Retrieved from http://arxiv.org/pdf/1308.5499.pd
Wong, C., Odom, S. L., Hume, K. Cox, A. W., Fettig, A., Kucharczyk, S., ... Schultz, T. R.
(2014). Evidence-based practices for children, youth, and young ddults with Autism
Spectrum Disorder. Chapel Hill: The University of North Carolina, Frank Porter Graham
Child Development Institute, Autism Evidence-Based Practice Review Group.
Workforce Innovation and Opportunity Act of 2014. Retrieved from
https://www.congress.gov/113/bills/hr803/BILLS-113hr803enr.pdf
Zeiss, A. M., Lewinsohn, P.M., & Munoz, R. F. (1979). Nonspecific improvement effects in
depression using interpersonal skills training, pleasant activity schedules, or cognitive
training. Journal of Consulting and Clinical Psychology, 1979, 47, 427-439.

167