Atypical Sensory Behaviors in Young Children with Autism Spectrum Disorders. Rosalind Schaefer Oti

Transcription

1 Atypical Sensory Behaviors in Young Children with Autism Spectrum Disorders by Rosalind Schaefer Oti A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Psychology) in The University of Michigan 2009 Doctoral Committee: Professor Catherine Lord, Chair Professor Sandra A. Graham-Bermann Assistant Professor Julie C. Lumeng Assistant Professor Christopher S. Monk

2 Rosalind Schaefer Oti 2009 iii

3 Acknowledgements I gratefully acknowledge the assistance of those who helped me with this project includi ng the faculty and staff at the University of Michigan Autism and Communication Disorders Center who collected these data, the consultants at the Center for Statistical Consultation and Research, including Joe Kazemi, Brady West, and Kathy Welch, who provided invaluable statistical guidance, and the members of my dissertation committee who were key in guiding the formulation of this work. I am especially grateful to Cathy Lord, who provided the feedback, guidance, and mentorship that made this dissertation possible. I would like to thank the many individuals with autism and their families who participated in the First Words and Toddlers Projects. I would also like to acknowledge that this research was supported by a Student Award Program grant from the Blue Cross and Blue Shield of Michigan Foundation. I would also like to acknowledge the encouragement and support of my family and friends. To my Mom, whose guidance and lessons in my childhood started me on the path to being a psychologist. To Papi, whose never-ending belief in me kept me going. To my friends, Rhiannon Luyster, Jen Richler, Somer Bishop, Sarah Lopez-Duran, and Emily Piltch, whose comments and suggestions added greatly to this dissertation and whose laughter and stories kept me balanced through it all. And finally, to Josie and Coco. ii

4 Table of Contents Acknowledgements. ii List of Tables... iv List of Figures. vi Abstract... vii Chapter I. Introduction.. 1 II. Validation of the Infant/Toddler Sensory Profile Questionnaire for Use with Toddlers with Autism Spectrum Disorders III. Unusual Sensory Behaviors in Children with ASD: Differences from Typical and Nonspectrum Populations and Related Characteristics. 45 IV. Effect of Atypical Sensory Behaviors on Socialization in Autism 82 V. Conclusion References iii

5 List of Tables Table 2.1 Participant Demographics Means, Standard Deviations, and Modes, for the TSP Items Identified as Autism-Specific Items Odds Ratios for NVIQ, Age, and Low Registration for Each Diagnosis Odds Ratios for NVIQ, Age, and Auditory for Each Diagnosis Percent Correct Classification of Diagnosis by Model Internal Consistency Coefficients for Each Sensory Style by Diagnosis Internal Consistency Coefficients for Each Sensory Modality by Diagnosis Percent of Sample Receiving Each Sensory Item Code Effect of ADI Sensory Severity Score on Each TSP Modality and Style Participant Demographics Sample Size at Each Time Point Chi-square Analyses Comparing Distribution of Children with Autism and PDD-NOS to the Standard Distribution Type III Test of Fixed Effects for Models Predicting Variability in TSP Modalities Coefficient Estimates for Age by Diagnosis Interaction Models with the Best Fit for Predicting TSP Sensory Styles Demographics of the First Assessment (FA) Group Demographics of the ADI Assessment Group iv

6 4.3 Demographics of the All Cases Sample Correlations between TSP Sensory Scores and ADOS Algorithm Scores Correlations of the TSP Sensory Modalities with ADOS Total Algorithm Score Comparison between the Mild to Moderate Group and the Severe Group on TSP Sensory Style Effect of TSP Sensory Styles and Modalities on Vineland Socialization Domain Standard Scores. 109 v

7 List of Figures Figure 2.1 Frequency of Derived ADI Sensory Severity Score by Diagnosis Change in Effect Size Over Time for Each Sensory Style Change in Effect Size Over Time for Each Sensory Modality Comparison of the Effect Sizes for Auditory and Low Registration Scores on Auditory Modality by Risk Status Over Time Change in ADOS Sensory Interest Item Over Time. 81 vi

8 Abstract Previous research has established that individuals with autism spectrum disorders (ASD) have higher rates of unusual responses to sensory stimuli than the typical population. Differences in these behaviors between children with autism compared to children with nonspectrum developmental delays or other clinical populations have not been as consistent. Research in this area is limited by inconsistencies in the defini tion of sensory behavior, differences in how the behaviors are assessed, what modalities are highlighted, and how styles of responding are defined (e.g. hyper-responsivity versus sensory sensitivity). The three studies included in this dissertation examined unusual sensory responses in children between the ages of 12 and 36 months who were at higher risk for having ASD. The first study examined the validity of using the Toddler Sensory Profile Questionnaire (TSP) to measure sensory behaviors in children with ASD by identifying items in the TSP that overlap with autism-specific behaviors, examining internal consistency of the TSP domains, and comparing performance on the TSP to other measures of sensory behavior. The goal of the second study was to compare children with ASD to typically developing children and children with nonspectrum developmental disorders in terms of different sensory modalities and processing styles. In addition, the effect of individual characteristics on sensory behavior was examined. Individual characteristics included NVIQ, age, risk status and diagnosis. The third study examined the relationship between atypical sensory behavior and impairments in socialization by vii

9 examining the relationship between scores on the TSP and socialization as measured by the Autism Diagnostic Observation Schedule, The Autism Diagnostic Interview-Revised, and the Vineland Adaptive Behavior Interview. Results from the current studies support previous research that children with autism have higher rates of unusual sensory responses than typically developing peers. However, caution should be used when interpreting findings from studies that use the TSP as several TSP items were identified as autism-specific behaviors. There was some overlap between scores on the TSP and sensory behaviors that parents report on the Autism Diagnostic Interview-Revised. NVIQ, age, and diagnosis were found to have an effect on sensory behavior, but the effect differed based on the sensory modality or style being addressed. Results of Study 3 indicated that there is a relationship between sensory behaviors and socialization, specifically when socialization is measured by the Autism Diagnostic Interview-Revised and the Vineland Adaptive Behavior Interview. Implications of these findings and limitations of the studies and other research regarding sensory behaviors in children with ASD are discussed. viii

10 Chapter I Introduction Autism Spectrum Disorders (ASD) include a group of pervasive developmental disorders that are characterized by impairments in socialization and communication and the presence of restricted and repetitive behaviors (American Psychiatric Association, 2000). In addition, the criteria for autism indicate that impairment in one of these areas must be present by the age of three. Throughout the remainder of this proposal, the term autism spectrum disorders (ASD) will be used to refer to autism as well as the other autism spectrum disorders (Pervasive Developmental Disorder Not Otherwise Specified and Asperger syndrome). Unusual responses to sensory stimuli are more common among individuals with ASD than typically developing individuals (Rogers, Hepburn, & Wehner, 2003; Kern, et al., 2006; Baranek, David, Poe, Stone, & Watson, 2006). Although unusual sensory responses are not an independent category in the core diagnostic criteria for autism, preoccupations with parts of objects, which can be considered a sensory behavior, is included under the category of restricted and repetitive behaviors. Unusual sensory response refers to a range of behaviors that are elicited from the presence of sensory stimuli. These behaviors may be present in any sensory modality: auditory, visual, tactile, gustatory, oral, olfactory, vestibular, and pain and comprise behaviors that include both sensory seeking behaviors such as peering at an object, 1

11 spinning toys/wheels, smelling objects, licking metal, and touching rough surfaces and sensory avoiding behaviors, such as getting upset at the sound of someone coughing, refusing to eat foods of a certain texture, or intolerability for wearing certain fabrics. Although differences in processing sensory stimuli have been suggested from studies examining sensory integration (Smith & Bennetto, 2007; Iarocci & McDonald, 2005), arousal levels (Dawson & Lewy, 1989; Tinbergen & Tinbergen, 1972), tests of higher order cortical sensory perception (Minshew & Hobson, 2008; Goldstein, Johnson, & Minshew, 2001) and startle responses (Perry, Minassian, Lopez, Maron, & Lincoln, 2007) the terms sensory responses and sensory behaviors used in this dissertation, refer to a child s behavioral reaction to sensory stimuli and not his or her sensation or perception of that stimuli. A History of the Role of Sensory Behaviors and ASD The role of unusual sensory response in the conceptualization of ASD has varied over the past several decades, with some researchers viewing atypical sensory responses as one of the underlying features that contribute to the impairments in social and communicative behavior more commonly associated with ASD (Schopler, 1965; Ornitz & Ritvo, 1968; Ornitz, 1988). Others view the abnormal sensory responses seen in children with ASD as a secondary symptom, and not unlike the unusual sensory behaviors seen in children with mental retardation, ADHD, and fragile X syndrome (Rutter, 1978; Wing & Gould, 1979). Autism was first characterized in a seminal paper by Leo Kanner in 1943, in which he described the behaviors of 11 children, whom he reported had a unique syndrome, not heretofore reported, which seems to be rare enough, yet is probably more 2

12 frequent than is indicated by the paucity of observed cases (p. 242). Many of the behaviors that he described including difficulties with social relationships, delays in speech acquisition, echolalia, and repetitive and nonfunctional use of objects are consistent with the core symptoms now included in the diagnostic criteria for autism (American Psychiatric Association, 1994). In addition to what is now considered the core symptoms, Kanner (1943) also described atypical sensory responses among these children. These descriptions included examples of both sensory avoiding behaviors, such as fear of running water, vacuum cleaners, and swings, as well as an increase in sensory seeking behaviors, such as spinning objects and rhythmic movements. While the role of socio-emotional deficits in ASD has remained remarkably consistent over the past 60 years, the degree to which the relationship between atypical sensory responses and ASD has been emphasized has varied. In the years following Kanner s article, atypical reactions to sensory stimuli became a focal point in ASD. In his in-depth description of a young girl with autism, Eveloff (1960) described how the girl liked the feel of water on her tongue (p. 96). He also noted that children with ASD often mouth or smell toys, although he stated that these behaviors were poorly understood. In his conceptualization of the symptoms of ASD, Schopler (1965) explained the occurrence of these behaviors by proposing that individuals with ASD have a differential reliance on receptors. He considered taste and smell to be near receptors (later termed proximal receptors), which were seen as the primary way, along with the other near recepto r of touch, that an individual with ASD interacts with and understands his/her environment. The distance receptors (later termed distal receptors) of vision and audition were thought to be secondary in individuals with ASD. According to Schopler, typical 3

13 development includes a shift from near receptor preference during the first 6 months of life, to a greater reliance on distance receptor usage. Along with this shift comes more of an emphasis on interpersonal interactions, as the typical infant begins to preferentially attend to the visual and auditory cues associated with his or her caregiver. It was proposed that repetitive use of objects and the insistence on sameness as emphasized in Kanner s original description of autism occurs when the infant with autism does not progress beyond near receptor dominance. Other conceptualizations of the symptoms of ASD have also focused on the role of sensory responses. Based on their observations of over 150 individuals with autism, Ornitz and Ritvo (1968) presented a characterization of infantile autism that included five subclusters of symptoms. One of these subclusters was disturbances of perception, which included behaviors related to a heightened awareness of sensory stimuli, heightened sensitivity and irritability, and nonresponsiveness to sensory stimuli. Ornitz and Ritvo proposed that this subcluster may underlie other subclusters, such as disturbances of relating, disturbances of language, and disturbances of motor behavior. Ornitz and Ritvo also went beyond describing the behaviors of children with ASD, to using these behaviors to propose an underlying mechanism that causes ASD. Like Schopler (1965), Ornitz and Ritvo proposed a biological model; one in which the central nervous system of the individual with ASD is unable to regulate the perceptual information that is received. In effect, environmental stimuli are either not adequately modulated or are unevenly amplified (p. 88) and the perceptual system is either overwhelmed or understimulated. Thus, the individual with ASD either engages in sensory seeking behaviors in order to 4

14 stimulate an understimulated sensory system or avoids sensory stimuli as a way to cope with an already overwhelmed and overstimulated sensory system. However, not all conceptualizations put such an emphasis on the role of sensory behaviors in ASD (Bender, 1959; Wing & Gould, 1979). In the years prior to the introduction of autism in the Diagnostic and Statistical Manual, there were two views of the key features of ASD. Rutter (1978) based his definition on impaired social relationships, impaired language skills and communication, and an insistence on samene ss, which were considered universal to ASD, while sensory behaviors and level of cognitive functioning varied by individual. The National Society for Autistic Children (NSAC; 1978), however, defined ASD as: disturbances of developmental rates and sequences, disturbances of speech, language-cognition, and nonverbal communication, disturbances of the capacity to appropriately related to people, objects, and events, and disturbances of responses to sensory stimuli. This debate over the role of unusual sensory response and ASD continued with publication of the Diagnostic and Statistical Manual of Mental Disorders 3 rd Edition (DSM III; American Psychiatric Association, 1980) and the revised version of the DSM III (DSM III R; American Psychiatric Association, 1987). When autism first appeared in the DSM III, unusual sensory response had a primary role in the diagnosis. The criteria for autism in the DSM-III included lack of social responsiveness, impairments in communication, and bizarre responses to the environment, which included atypical sensory responses. With the publication of the DSM III R, criteria for autism changed to a focus on a restricted repertoire of activities and interests, rather than a bizarre response to the environment. This change in focus meant that unusual 5

15 sensory responses were removed from the diagnostic criteria and were instead included under the heading of associated features. This change is consistent with the current conceptualization of ASD, which views sensory abnormalities as neither universal nor specific to ASD (Rogers, Hepburn, & Wehner, 2003), except to the degree that they contribute to other defining characteristics of autism. Current Views of Abnormal Sensory Responses and ASD Unusual sensory responses are common among children with ASD and estimates have suggested that atypical sensory behaviors may be present in as many as 53% (Szatmari, et al., 2006) to 95% (Ornitz, Guthrie, & Farley, 1977) of children with ASD. Several studies have compared the rates of abnormal sensory responses in children with autism to typically developing children and results indicate that on average children with autism have increased rates of unusual sensory behavior that involve all sensory modalities and can include sensory seeking and sensory avoiding behaviors. While the differences in sensory behaviors between children with autism and typically developing children are generally consistent, the nature of the sensory difficulties, the prevalence of these difficulties, and their relationship to nonverbal IQ and age is difficult to determine as previous studies have utilized a range of measures to assess sensory behaviors, have included very different ages of individuals with autism, and have generally not controlled for NVIQ. In addition, few studies have examined atypical sensory responses in children with autism under the age of three. With epidemiological studies that have suggested an increased prevalence of autism (Fombonne, 2003; Scott, Baron-Cohen, Bolton, & Brayne, 2002), estimates of prevalence rates now reaching 1 in 150 (CDC, 2007), and research 6

16 suggesting that diagnoses of autism made at age two remain stable through age nine (Lord, et al., 2006) there has been an emphasis on early identification of ASD. Until recently, the majority of studies examining indicators of ASD under the age of 3 were based on retrospective parent report and review of early home videotapes. While the majority of these retrospective studies focused on the ability of social behaviors to differentiate children who were later diagnosed with ASD (Werner, Dawson, Osterling, & Dinno, 2000; Charman, et al., 1996; Mars, Mauk, & Dowrick, 1996; Wimpory, Hobson, Williams, & Nash, 2000; Osterling & Dawson, 1994) some retrospective videotape studies have also found differences in sensory behaviors among young children later diagnosed with autism (Baranek, 1999; Osterling & Dawson, 1994). Analyses of videotapes of first birthday parties have found that, in addition to differences in social behavior, children later diagnosed with autism were more likely to cover their ears or engage in self-stimulation (Osterling & Dawson, 1994). Baranek (1999) found that infants between the ages of 9 and 12 months old, who were later diagnosed with autism, differed from typically developing and developmentally delayed infants in their response to both social and nonsocial stimuli. A combination of behaviors, including decreased orienting to visual stimuli, increased mouthing of objects, and greater social touch aversion, best differentiated those who were later diagnosed with ASD from the other two groups. Dawson, Osterling, Meltzoff, & Kuhl (2000) reported a case study of an infant with autism who was seen by medical professionals from birth, due to feeding difficulties, and then evaluated by the authors at 1 and 2 years of age. Notes made in his medical records indicated that at 2.5 months of age the infant was easily over-stimulated if too 7

17 much noise was present and was mildly hypersensitive to tactile stimulation. By 9 months of age, this hypersensitivity was extended to other sensory modalities such as auditory and visual, and by one year of age it was noted that he was insensitive to pain and was over-stimulated by the presence of more than two objects on the table. One of the factors making it difficult to examine the early indicators of ASD prospectively is that the majority of children with ASD are not evaluated until the age of four and on average do not receive a diagnosis of ASD until the age of five (Wiggins, Baio, & Rice, 2006), despite the age of first concern typically being around 18 months (Stone, et al., 1999). Recurrence risk for ASD has been estimated to be somewhere between 5% and 10% (Ritvo, et al., 1989; Cook, 1998; Szatmari, Jones, Zwaigenbaum, & MacLean, 1998; Fombonne, 2003). Because siblings are an identifiable population who are at increased risk for autism, studies of early development of ASD are recruiting younger siblings of children with ASD in order to prospectively characterize the symptoms of ASD at younger ages. Results of sibling studies are just beginning to be published. These studies have prospectively followed children from as young as four months and are finding differences between those who are later diagnosed with ASD and those who are typically developing. These studies have focused on differences in social interactions (Zwaigenbaum, et al., 2005; Yirmiya, et al, 2006), language skills (Klin, Saulnier, Chawarska, & Volkmar, 2008; Landa & Garrett-Mayer, 2006) and visual fixation patterns of social and ambiguous events (Klin & Jones, 2008). However, one study did examine differences in responses to sensory stimuli. Zwaigenbaum and his colleagues (2005) found that at 12 months of age, sensory-oriented behaviors, in addition to differences in eye contact, 8

18 imitation, social smiling, temperament, and social interest and affect significantly predicted diagnosis of autism at 24 months. These behaviors were coded during administration of the Autism Observation Scale for Infants (AOSI: Bryson, Zwaigenbaum, McDermott, Rombough & Brian, 2008), a standardized semi-structured play session. The sensory-oriented behaviors were generally related to stereotyped play with the AOSI materials. With the push to identify early indicators of autism in children under the age of three and results of initial studies indicating that differences in response to sensory stimuli in children who are later diagnosed with autism can be seen by 12 months of age, it is important to continue to examine differences in sensory behaviors among very young children with autism. The current dissertation examined unusual sensory responses in children between the ages of 12 and 36 months who were at higher risk for having ASD, defined here as either being a younger sibling of a child with ASD or as having medical/behavioral risk factors associated with autism, such as cognitive or language delays. The following three chapters examined the validity of using a standardized measure of sensory behaviors with children with ASD, compared children with ASD to typically developing children and children with nonspectrum developmental disorders in terms of different sensory modalities and processing styles, examined which characteristics, including NVIQ, age, and diagnosis account for variability in unusual sensory responses, and examined the relationship between atypical sensory behavior and impairments in socialization. 9

19 Chapter II Validation of the Infant/Toddler Sensory Profile Questionnaire for Use with Toddlers with Autism Spectrum Disorders Sensory behaviors in young children with Autism Spectrum Disorders (ASD) have been assessed in a variety of ways. These have included direct observation of the child (DeGangi, Berk, & Greenspan, 1988; Baranek, Boyd, Poe, David, & Watson, 2007), parent-report questionnaires that combine sensory scores across modalities (Talay-Ongen & Wood, 2000; Dunn, 1997), questionnaires that include items related to sensory behaviors in addition to questions that address a range of other behaviors (Dahlgren & Gillberg, 1989), and parent interviews that include items related to sensory behavior among a group of items that address social, communicative, and restricted/repetitive behaviors (Lord, Rutter, & Le Couteur, 1994). The Toddler Sensory Profile (TSP; Dunn, 2002) is a popular tool for assessing sensory responses among toddlers with ASD, as it is easy to administer and has been standardized for use between 7-36 months of age. While the TSP was not developed as a measure of sensory behaviors among children with ASD specifically, it has become the most widely used measure of sensory processing among autism research sites, including the Collaborative Programs of Excellence in Autism (CPEA) and the Studies to Advance Autism Research and Treatment (STAART) Centers. Inform ation that is generated from the use of the TSP among children with ASD may be used by autism researchers to further clarify the core diagnostic characteristics of autism, 10

20 to develop appropriate interventions, and to understand early indicators of prognosis and outcome. Therefore, it is important that the TSP be examined for its appropriateness for use with this population. The TSP is a 48-item caregiver questionnaire that uses a five-point scale to assess the frequ ency of behaviors that occur in response to sensory stimuli in five sensory modalities: Auditory Processing, Visual Processing, Tactile Processing, Vestibular Process ing, and Oral Sensory Processing in children between the ages of seven months and three years. Items in each of these sensory modalities are then combined to obtain totals in four different general sensory styles (referred to as quadrants by the authors of the TSP): Low Registration, Sensation Seeking, Sensory Sensitivity, and Sensation Avoiding. These four sensory styles are based on a conceptual model that the author of the TSP developed after doing research into the sensory styles of older children (Dunn, 1997). According to Dunn, how a child responds to stimuli in the environment is based on an interaction between the child s neurological threshold, as defined by habituation and sensitization, and behavioral response/self-regulation. The interaction between neurological threshold and behavior can result in four general behavioral styles: 1) Low Registration, which results from a high threshold and a passive response (i.e. individuals do not seek to reach their threshold), 2) Sensation Seeking, which is the combination of a high threshold and an active behavioral response (i.e. an individual seeks out stimulation to reach his/her threshold, 3) Sensory Sensitivity, a style of behavior that results from a low threshold and passive behavioral response, and 4) Sensation Avoiding, which includes a low threshold and active self-regulation. 11

21 The TSP was standardized using a group of 1,094 toddlers under the age of three, with (n = 285) and without (n = 809) disabilities. The standardization sample included 24 children with autism among the group of children with disabilities. Distribution of the sensory processing style scores into the categories of less responsive, typical, and more responsive, was based on the creation of a normal distribution from the scores of 489 children without disabilities in the standardization sample. The internal consistency coefficients for the sensory modalities ranged from α =.42 (vestibular processing) to α =.71 (tactile processing). Internal consistency coefficients for the sensory response styles ranged from α =.7 for sensation avoiding to α =.86 for sensation seeking. Testretest reliability was based on a retest of 32 participants from the standardization sample 2-3 weeks after initial administration of the SPQ. The test-retest correlation coefficient was.86 for the sensory modality scores and.74 for the sensory processing style scores. Initial standardization of the TSP included 24 children with ASD (Dunn, 2002). When these 24 children were compared to a group of 24 age- and gender-matched controls, it was found that the children with ASD had significantly lower scores (indicating that they exhibited the behaviors more frequently) on all of the sensory modalities and three of the four sensory styles. The only sensory style that did not differentiate the children with ASD from the typical sample was Sensory Seeking. The effect sizes of the significant differences ranged from.29 (visual processing) to.51 (oral sensory processing) among the modalities and.51 (sensory sensitivity) to.63 (sensation avoiding) for the sensory styles. The Toddler Sensory Profile is a follow-up instrument to two previous versions of the Sensory Profile Questionnaire (SPQ: Dunn, 1999; Dunn, 1994) that assess behavioral 12

22 responses to sensory stimuli in children over the age of three. The 1999 version includes 125 items and 10 factors while the 1994 version includes 99 items, which combine to form six modalities and two behavioral categories. Both versions of the SPQ have been used to examine differences in sensory responses between children with autism and typically developing children. The author of the SPQ has conducted several of these studies and found consistent differences between children with autism and typically developing children. Using the earlier version of the SPQ, Kientz and Dunn (1997) found that a group of 32 children with ASD, ages 3 to 13 differed significantly from a group of 64 typically developing children on each of the six modalities and two behavioral categories. Effect sizes ranged from 0.50 for body position to 0.81 for touch. Using the more recent version of the SPQ, Dunn, Myles, and Orr (2002) found that children with ASD ages 8 to 14 (n = 42) had greater rates of atypical sensory responses in each of the sensory modalities and response styles compared to 42 typically developing children. Effect sizes ranges from 0.39 for the visual modality to.67 for the tactile modality, with the children with autism displaying greater frequencies of atypical sensory behaviors. The typically developing children were randomly drawn from the standardization sample of the SPQ and although they were also 8 to 14 years of age, their average age was three years less than the group with ASD. Other researchers have found similar results. Kern et. al., (2006) used the SPQ in a cross-sectional study of 104 individuals with ASD, ages 3-56 years, and found that the group with ASD s scores for the four modalities that they included from the SPQ (auditory, visual, touch, and oral) were significantly lower than the scores of the agematched typical control group. Results indicated that tactile processing had the largest 13

23 effect size (1.9) while visual processing had the smallest effect size (0.88). The same sample of individuals with ASD (minus a 56 year old participant) was also found to have greater rates of behaviors than the typical gender and age-matched controls in each of the sensory styles: Low Registration, Sensation Seeking, Sensory Sensitivity, Sensation Avoiding (Kern, Garver, Carmody, Andrews, Trivedi, & Mehta, 2007). Watling, Dietz, & White (2001) found that children with autism, ages 3 to 6, had significantly different scores than an age and gender-matched typical control group in the areas of sensory seeking, emotional reactivity, oral sensitivity, and poor registration. Only a few studies have compared the performance of children with ASD to another non-typical population using the SPQ. Rogers, Hepburn, and Wehner (2003) used the Short Sensory Profile, an abbreviated version of the Sensory Profile, to compare young children with autism to children with fragile X syndrome, other developmental delays, and typical development. They found that the children with autism had higher rates of sensory symptoms compared to the developmentally delayed and typically developing groups, but not compared to the children with fragile X syndrome. Significant differences were found in the areas of tactile sensitivity, taste/smell sensitivity, underreactive/seeking stimulation, and auditory filtering. In a study comparing the sensory behaviors of children with autism and children with ADHD Ermer & Dunn (1998) found that the children with ADHD had greater frequencies of Sensory Seeking behavior than the children with autism, while oral sensitivities and fine motor/perceptual difficulties were more common among the children with autism than the children with ADHD. 14

24 Although findings from research using the SPQ to examine differences between children with ASD and typical children have been consistent, it is unclear whether these differences hold for children under the age of three. As reported above, during the standardization of the Toddler Sensory Profile (TSP), 24 children with ASD (ages months) were compared to 24 age- and gender-matched typical controls. Results indicated that the children with ASD had significantly lower scores on each of the sensory modalities and styles, except Sensation Seeking. However, no additional studies have compared the performance of a large group of children with ASD to children with typical development or nonspectrum developmental delays. Therefore, these findings need to be replicated with a larger and independent sample. It is also important to consider the influence of individual items on the TSP, as some of the behavioral responses included on the TSP may be consistent with behaviors characteristic of autism, such as poor eye contact and lack of a response to name. It is possible that the elevated scores of children with autism may be due to these autism-specific behaviors rather than to unique sensory responses. Furthermore, no previous studies have compared the performance of children on the TSP to other parent-report or observational measures of sensory responses. The aim of the current study was to examine the validity of the TSP with toddlers with autism between the ages of 12 and 36 months in three ways. First, experts identified items on the TSP that overlapped with the diagnostic criteria for autism. Second, the internal consistency of the modalities and the processing style quadrants was determined for each diagnostic group. Finally, concurrent validity was examined by comparing the TSP to the sensory-related items included on the Autism Diagnostic 15

25 Observation Schedule (ADOS: Lord, et al., 2002) and the Autism Diagnostic Interview- Revised (ADI-R: Lord, Rutter, Le Couteur, 1994). Following these aims, three hypotheses were made: 1) it was hypothesized that items identified by the experts as autism-specific items, would be more common among the autism group and the group means for these items would be significantly lower for the autism group, indicating that these behaviors occurred more frequently, 2) internal consistency of the modalities and sensory styles would be consistent with what was reported in the TSP manual, and 3) the concurrent validity would be greater for the TSP and ADI-R than the TSP and ADOS. Method Participants Participants included 89 children, ages 12 months to 3 years. The participants were included in the First Word/Toddlers Projects at the University of Michigan Autism and Communication Disorders Center (UMACC). These projects addressed early identification of ASD in infants and toddlers who were considered to be at higher risk for having autism. Higher risk status was defined by either being the younger sibling of a child already diagnosed with ASD or by having a medical or behavioral risk factor associated with autism (e.g. language delay, seizures). Participants in the First Words Project included children with speech and language impairments. Participants had to be less than 3 years of age to enroll in the study. Participation in this project included assessments every six months. A subset of First Words participants were also enrolled in the Toddlers Project. All participants in the Toddlers project were suspected of having autism. Participation in the Toddlers 16

26 Project included monthly assessments and children must have been under the age of 16 months at the time of enrollment. Based on most recent consensus diagnoses, which were determined at the mean age of 27.5 months (s.d. = 7.8), 17 (19.1%) participants were considered typically developing, 25 (28.1%) had a current diagnosis of autism, 25 (28.1%) had a current diagnosis of PDD-NOS, and 22 (24.7%) were diagnosed with nonspectrum developmental delays (see Table 2.1). The nonspectrum developmental delay group was a heterogeneous group that included language disorders, intellectual disability of unknown etiology, Down syndrome, William syndrome, fetal alcohol syndrome, and behavioral disorders. Across all diagnoses, 35 (39.3%) participants were siblings of an older child with autism and 54 (60.7%) were considered high risk due to the presence of medical and/or behavioral risk factors. There were 71 males and 18 females. As shown in Table 2.1, the typically developing participants were significantly younger at their first assessment than the other three groups. This is because a greater proportion of the typical children who were recruited were siblings. The average age of the first visit for siblings was 16.6 months (s.d. = 6.6), while the average age for those with medical/behavior risk factors was significantly higher (x = 24.2, s.d. = 8.0; t = -4.7, p <.001). Measures As part of the First Words/Toddlers Project, participants completed several measures that were used in the current study. These measures are described below. (1) The Toddler Sensory Profile Questionnaire (TSP; Dunn, 2002) is a 48-item caregiver questionnaire that assesses typical behaviors in 5 sensory modalities: 17

27 Auditory Processing, Visual Processing, Tactile Processing, Vestibular Processing, and Oral Sensory Processing in children between the ages of 7 months and three years. Parents rate their child s behavior on a five-point scale from almost never to almost always. Items from each domain are added to yield a raw score for each sensory modality. These raw scores are then plotted on a grid, which indicates whether an individual is less responsive, typical, or more responsive in each sensory modality. For purposes of data analysis, these groups were converted into a categorical variable. Both the raw scores and the categorical variable were used in the analyses. Items in each sensory modality are also combined to obtain totals in four different general sensory styles: low registration, sensation seeking, sensory sensitivity, and sensation avoiding. The four sensory styles are based on Dunn s (1999) model of sensory processing that takes into consideration an individual s neurologic threshold and coping strategy. The scores in each sensory processing style are then divided into three categories: less sensory responsiveness, typical performance, and more sensory responsiveness. The TSP was standardized using a group of 1,100 toddlers under the age of three, with and without disabilities. The standardization sample included 18 children with autism and 18 children with Down syndrome among the group of children with disabilities. Distribution of the sensory processing style scores into the categories of less responsive, typical, and more responsive, was based on the normal distribution of 489 children without disabilities in the standardization sample. The internal consistency coefficients for the sensory modalities ranged from α =.42 (vestibular processing) to α =.71 (tactile processing). Internal 18

28 consistency coefficients for the sensory response styles ranged from α =.7 (sensation avoiding) to α =.86 for sensation seeking. Test-rest reliability was based on a retest of 32 participants from the standardization sample 2-3 weeks after initial administration of the SPQ. Test-retest reliability was.86 for the sensory modality scores and.74 for the sensory processing style scores. (2) Mullen Scales of Early Learning (MSEL: Mullen, 1995) is a standardized assessment of cognitive functioning for children between birth and 5 years, 8 months of age that is administered directly to the child. Standard scores on visual reception (nonverbal problem solving), fine and gross motor skills, and expressive and receptive language abilities yield an overall IQ score. As part of the First Words/Toddlers Project, the MSEL is administered at 6 month intervals as close to 12, 18, 24, 30, and 36 months as possible, by an examiner blind to history and previous diagnosis. The MSEL has been used in many studies of children with autism (McConachie, Le Couteur, & Honey, 2005; Lord, Shulman, & DiLavore, 2004). NVIQ s were generated by taking into consideration the child s standard scores in the visual reception and fine motor domains. In cases where the raw scores were too low to yield a standard score, a ratio IQ was used. A ratio IQ was generated by dividing the child s age equivalent by the chronological age. (3) The Autism Diagnostic Observation Schedule (ADOS; Lord, Rutter, DiLavore, & Risi, 1999) is a semi-structured, standardized assessment, that examines communication, social interaction and play behavior. One of five modules is selected based on the child s language level. Children in the First Words and Toddler studies were administered Module 1 (preverbal or single words), the 19

29 Toddler Module, which is a modified version of Module 1 intended for use with children between the ages of 12 months and 30 months, or Module 2, which is used with children who have phrase speech. The assessment involves a variety of social presses designed to elicit behaviors relevant to a diagnosis of autism. A standardized diagnostic algorithm is calculated and scores are compared to cut-off scores to indicate autism, autism spectrum, and non-autism spectrum classifications. For the Toddler Module, sensitivity of the ASD algorithm cutoff score ranges from 83% to 91%, depending on the age and verbal ability of the child. Specificity is also high, with a range from 91% to 94% depending on age and verbal ability (Luyster, et al., submitted). Sensitivity and specificity is high for both Modules 1 and 2; 97% and 94% for module 1, respectively, and 95% and 87% for module 2 (Lord, et al., 2000). Internal consistency coefficients of the socialization domain ranged from α = (depending on module), the communication domain ranged from α = , and the restricted and repetitive behavior domain ranged from α = for modules 1 and 2. Cronbach s alpha was high for the social-communication total scores α = across modules (Lord, et al., 2000). (4) The Autism Diagnostic Interview Revised (ADI-R; LeCouteur, Lord, & Rutter, 2003) is a standardized 90-minute caregiver interview that generates separate scores for socialization, communication and restricted and repetitive behaviors in children referred for possible autism. Algorithm items have been shown to have high interrater reliability, with multi-rater kappas for 12 out of 15 items in the reciprocal social interaction domain exceeding.70 and kappas for all 7 items in 20

30 the restricted and repetitive behaviors domain exceeding.63 (Lord, et al 1994). Internal consistency for each of the algorithm domains was also high: α =.95 for the socialization domain,.84 for the communication domain, and.69 for the restricted and repetitive behaviors domain. The algorithm adequately discriminates individuals with autism from a chronologically age-matched nonautistic mentally impaired comparison group (Lord et al., 1994). Caregivers of children in the First Words/Toddlers Projects complete the Toddler ADI, which is a revised version of the ADI-R that is currently being tested for use in children under the age of three. Results Identification of Individual Autism-Specific Items In order to examine which behaviors included on the TSP were descriptions of autism-specific behaviors, 7 autism experts were given a list of 65 behaviors (the 48 items from the TSP, and 17 items from the Social-Communication Questionnaire (Rutter, Bailey, & Lord, 2003) and the Social Responsiveness Scale (Constantino & Gruber, 2005) and asked to rate whether or not each behavior was characteristic/highly indicative of autism. Expert status was defined by being reliable on either the Autism Diagnostic Observation Schedule (ADOS), a semi-structured play-based observational assessment for children suspected of having ASD or the Autism Diagnostic Interview-Revised (ADI- R), a autism diagnostic parent interview, or both. Seven TSP items were identified by all 7 experts as diagnostic of autism. These items include: My child avoids playing with others I have to speak loudly to get my child s attention 21

31 I have to touch my child to gain attention My child takes a long time to respond, even to familiar voices It takes a long time for my child to respond to his/her name when it is called My child enjoys looking at moving or spinning objects My child avoids eye contact with me Next, the distributions of scores by diagnostic groups were examined for each of these items. Table 2.2, shows the mean, standard deviation, and modal score for each item by diagnosis. Differences between diagnostic group means were compared using one-way analysis of variance and pairwise comparisons were examined using Tukey s HSD. Significance levels were set at.002 to account for multiple comparisons. The modal scores shown in Table 2.2 indicate that the behaviors identified by the autism experts, were in fact seen more often in the group diagnosed with autism than in the other three diagnostic groups, except for the item my child avoids eye contact with me, which parents most frequently endorsed as seldom occurring. Mean scores were significantly lower for the children with autism as compared to the typical group for all items except my child enjoys looking at moving or spinning items and my child avoids eye contact with me. Four of the seven items that were identified are included in the Auditory Processing domain and five of the seven items are part of the Low Registration sensory style. Given that the majority of autism-specific items were included in these two domains, a Multinomial Logistic Regression was used to examine whether Low Registration and/or Auditory Processing significantly added to a model that predicted diagnosis. Model 1 included only the covariates NVIQ and age. NVIQ and age alone 22

32 accounted for approximately 45% of the variance in diagnosis (Nagelkerke r 2 =.447) and classified approximately 47% of the cases accurately. In Model 2, NVIQ, age, and Low Registration were entered into a Multinomial Logistic Regression using a forward stepwise approach. The results indicated that Low Registration significantly contributed to the final model (X 2 (3) = 11.4, p =.01). Model 2 demonstrated acceptable model fit (X 2 (9) = 55.4, p <.001) and accounted for approximately 53% of the variance in diagnosis (Nagelkerke r 2 =.529). There was 59.3% correct classification of diagnosis based on Model 2. As shown in Table 2.3, when autism was set as the reference group, the typical group was differentiated from the autism group by NVIQ and age, whereas only Low Registration significantly differentiated PDD-NOS and NSDD from the group with autism. In Model 3 NVIQ, age, and Auditory scores were entered into a Multinomial Logistic Regression using a forward stepwise approach. The results indicated that the Auditory modality significantly contributed to the final model (X 2 (3) = 12.1, p =.007). Model 3 demonstrated acceptable model fit (X 2 (9) = 54.8, p <.001) and accounted for approximately 53% of the variance in diagnosis (Nagelkerke r 2 =.525). There was 55.6% correct classification of diagnosis based on the complete model. As shown in Table 2.4, when autism was set as the reference group, the typical group was differentiated from the autism group by NVIQ and auditory scores. Auditory also significantly differentiated PDD-NOS and NSDD from the group with autism. Although the addition of either Low Registration or Auditory only increased the amount of explained variance in diagnosis by 9%, examination of the increase in percent correct classification for each diagnosis showed that, although the addition of Low 23

33 Registration (Model 2) or Auditory (Model 3) did not increase the correct classification of typical children, either domain increased the correct classification of children with autism by 26% (See Table 2.5). Model 2 and 3 were rerun without the autism-specific items. Results indicated that when either Low Registration or auditory were included without the autism-specific items both were dropped from the forward stepwise approach, as neither significantly contributed to the model s ability to predict diagnosis. Internal Consistency of the Sensory Style Domains In order to examine how well the items included in each of the sensory style domains measure the same construct, a Cronbach s alpha was generated for each style. In order to compare groups on a measure for research, alpha levels of 0.7 and greater are acceptable (Bland & Altman, 1997). For the purposes of this study, if an alpha level was less than 0.7, pairwise correlations between each of the items included in the relevant sensory domain were examined. Items that did not correlate significantly with the other items or that were negatively correlated were dropped and Cronbach s alpha was generated again. Table 2.6 shows the Cronbach s alpha for each of the sensory styles, in addition to the alphas once individual items were removed. Table 2.7 shows the Cronbach s alpha for each of the sensory modalities. As shown in Table 2.6, the internal consistency coefficients for the sensory styles were good for Low Registration and Sensation Avoiding. For Sensory Sensitivity, the coefficients were generally high, except for the typical group, for which the alpha level was slightly below the acceptable level. Except for PDD-NOS, the coefficients for 24

34 Sensation Seeking were all below 0.7. Even after removing items that had the weakest correlations, the alpha levels did not exceed 0.7. In general, the internal consistency coefficients for the sensory modalities (Table 2.7) were not as high as the coefficients for the sensory styles. The auditory modality is the only modality that had coefficients well above 0.7 for each of the diagnostic groups. The tactile modality had the next highest coefficients, which increased with the exclusion of two of the tactile items: my child enjoys splashing during bathtime and my child uses hands to explore food and other textures. The other three modalities were generally poor, with individual diagnoses having alpha levels in the acceptable range. Concurrent Validity Between the TSP, ADOS, and ADI Performance on ADOS sensory item The Autism Diagnostic Observation Schedule (ADOS; Lord, Rutter, DiLavore, & Risi, 1999) is a play-based semi-structured diagnostic observational assessment that includes one sensory-related item that is coded on the basis of spontaneous sensory behaviors. The item, Abnormal or idiosyncratic response to sensory stimuli is coded either, 0 for no behavior present, 1 for possible sensory behavior, 2 for definite sensory behavior, and 3 for sensory behavior present to an extent that it interferes with other tasks. Performance on the TSP was examined by these ADOS sensory scores. First, distribution of the ADOS sensory item scores by diagnosis was examined. Results indicated that there was a significantly different distribution of scores across diagnostic groups (X 2 (9) = 53.9, p <.001). Table 2.8 shows the number of children with each ADOS sensory code, by diagnosis. Pairwise comparison of these groups showed that distribution of the scores for the children with autism was significantly different than 25

35 the group with PDD-NOS (X 2 (3) = 14.7, p <.01), the group with NSDD (X 2 (3) = 9.5, p <.05), and typical children (X 2 (3) = 52.6, p <.001), with a greater proportion of toddlers with PDD-NOS, NSDD, and typical children having no sensory behaviors during the ADOS, compared to the children with autism. There was no significant difference in the distribution across scores when comparing children with PDD-NOS and children with NSDD (X 2 (3) =.32, p = n.s.). However, there was a significant difference between PDD-NOS and typical children (X 2 (3) = 17.5, p =.001) and NSDD and typical children (X 2 (3) = 20.1, p <.001). Performance on the TSP was then examined by three ADOS sensory groups across all diagnoses: no evidence of sensory interests (score of 0), sensory interest possible (score of 1), and definite sensory interest (score of either 2 or 3). First, a Chisquare analyses was used to compare the distribution of these three sensory groups on each of the sensory modality and sensory style categorical groups (less responsive than others, typical performance, and more responsive than others). When all diagnostic groups were combined in the analyses there were significant differences in distribution for the auditory modality (X 2 (4) = 18.8, p =.001), visual modality (X 2 (4) = 14.0, p <.01), and oral modality (X 2 (4) = 11.1, p <.01). Results were not significant for the vestibular modality (X 2 (4) = 2.5, p = n.s.) or tactile modality (X 2 (4) = 9.4, p = n.s.). There were also significant differences by sensory style: low registration (X 2 (4) = 13.7, p <.01) and sensation avoiding (X 2 (4) = 14.6, p <.01). Results approached significance for sensory seeking (X 2 (4) = 11.4, p <.05) and were not significant for sensory sensitivity (X 2 (4) = 5.5, p = n.s.). However, when looking only within each diagnostic group, therefore controlling for diagnosis, these results were no longer significant. 26

36 Next we examined whether there were significant differences in means on the modalities and styles, between those who showed no sensory behavior during the ADOS and those who showed definite sensory behavior. When comparing the no sensory behavior (code of 0) group to the group with definite sensory behavior (collapsing codes of 2 or 3), with all diagnostic groups combined, there were significant differences in the mean differences in three of the styles: Low Registration (t(213) = 3.7, p <.001), Sensory Seeking (t(213) = 2.3, p <.05), and Sensation Avoiding (t(213) = 2.8, p <.01). To account for multiple comparisons, a Bonferonni correction was used and significance level was set at.01. Therefore, the difference in Sensory Seeking is no longer considered significant. For Low Registration and Sensation Avoiding the definite sensory behavior group had significantly lower scores, indicating greater frequency of the unusual sensory behaviors included on the TSP. The definite sensory behavior group also had significantly lower scores on two of the sensory modalities: auditory (t(213) = 3.5, p <.01), and visual processing, (t(213) = 3.3, p <.01), and a trend toward significance for the oral modality (t(213) = 2.1, p <.05). In order to account for repeated measures and to control for diagnosis, age at testing, and NVIQ, a Linear Mixed Model was used. Age and NVIQ were entered as covariates and diagnosis and ADOS sensory group were entered as fixed factors. Results indicated that ADOS sensory group did not have a significant effect on mean score of the sensory modalities or on the mean scores of the sensory styles on the TSP. Performance on the ADI-R sensory items Three items on the ADI-R address unusual sensory responses. These items are: 1) abnormal/idiosyncratic response to sensory stimuli, 2) undue general sensitivity to noise, 27

37 and 3) unusual sensory interests. Each of these items is scored from 0, indicating no presence of the behavior to 3, indicating presence of the behavior to the degree that it interferes with functioning. First, the individual effect of each ADI item on the TSP modalities and styles was examined using Linear Mixed Modeling. Diagnosis, age, and NVIQ were controlled for in the model and significance level was set at.004. The item Unusual Sensory Interests significantly predicted Low Registration (F(4, 87.0) = 5.3, p =.001) and the relationship with the Auditory modality approached the adjusted significance level (F(4, 101.5) = 4.0, p =.005). In order to examine if the effects of the Unusual Sensory Interest item on Low Registration and the auditory modality were being driven by the autism-specific items identified above, the analyses were repeated for Low Registration and auditory without these items. Results indicated that Unusual Sensory Interest still significantly predicted Low Registration (F(4, 96.1) = 4.6, p =.002) and there was a trend toward significance for the Auditory modality (F(4, 16.6) = 5.1, p =.008). The other two ADI-R sensory items did not significantly predict any of the TSP styles or modalities. The relationship between the ADI sensory items and the TSP was also examined by creating an ADI sensory severity score. This score was created by summing the scores from the three sensory item codes, yielding a score from 0 to 9. Figure 2.1 shows the frequency of each severity score by diagnosis. As can be seen in figure 2.1, the most common severity score for children with autism was 3, whereas the most common severity score for the typical children was 0, for PDD-NOS it was 0 or 1, and for children with nonspectrum disorders it was 1. Only one child in the sample had a score of 8, 28

38 indicating that the child received a score of 3 on two of the items and a score of 2 on the third item. This child was in the nonspectrum DD group. In order to test the hypothesis that individuals with greater ADI-R sensory severity scores would have lower scores on the TSP modalities and styles, a Linear Mixed Model was used. NVIQ and age did not predict any of the sensory styles or sensory modalities, so these variables were removed and the analyses were repeated without them. Results are shown in Table 2.9. After controlling for diagnosis and removing NVIQ and age from the model, it was found that ADI-R sensory severity significantly predicted three styles: Low Registration, Sensory Sensitivity and Sensation Avoiding. ADI-R sensory severity score also predicted three of the sensory modalities: Auditory, Tactile, and Oral. As was done with the analyses with the ADOS sensory item, the effect of the ADI-R sensory severity score was run on the Low Registration and Auditory modalities without including the identified autism-specific items. The results indicated that after removing the autismspecific items, the ADI-R sensory severity score still significantly predicted both of these domains. Conclusions The main aim of this study was to determine whether the Toddler Sensory Profile Questionnaire (TSP) is an appropriate instrument to use to measure atypical sensory behavior in young children with autism spectrum disorders. Several different approaches were used to address this aim. First, because the items on the TSP address behavioral responses to everyday situations, it was important to individually examine each TSP item to determine whether the behavior could be explained by autism characteristics in general 29

39 rather than being specifically in response to sensory input. Second, the internal consistency coefficients for each sensory modality and processing style were examined to determine how well the items were correlated with each other and how well they measured the same construct. Third, the concurrent validity was examined between the TSP, the sensory item on the ADOS, and the three sensory items on the ADI-R. Autism experts identified seven of the 48 items on the TSP as being autismspecific items. Six of the seven items (shown in Table 2.2) are behaviors that are in response to social situations. Therefore, if children with autism have lower scores on these items, it would be difficult to attribute their behavior to anything besides their diagnosis. These items form a minority of the total items on the TSP but the majority of these items are included on the same modality (auditory) and the same style (Low Registration). Examples of auditory items that were not identified as autism-specific items included: my child tries to escape from noisy environments, my child finds ways to make noise with toys, and my child startles easily at sound, compared to other children the same age. Examples of Low Registration items that were not identified as autism-specific items included: my child bumps into things, seeming to not notice objects in the way, my child is unaware of food or liquid left on lips, and my child does not recognize self in the mirror. Although analyses in the current study suggested that differences in Low Registration and auditory behaviors remained after the autism items were removed, examination of the changes in explained variability suggested that removing the autism items from these domains reduced the increase in explained variability. These findings suggest that differences in auditory sensory behaviors and Low Registration on the TSP may be affected by the inclusion of the autism-specific 30

40 items. Therefore, future analyses that indicate higher rates of low registration or atypical responses to auditory stimuli among children with autism, as assessed by the TSP, must be interpreted with caution and comparisons between children with autism and other diagnostic or typical groups should examine auditory and Low Registration without the autism-specific items. The appropriateness of including individual items on the TSP was also examined by checking internal reliability within each modality and style. The internal consistency coefficients were better for the sensory styles than the sensory modalities. Among the styles, all of the styles had acceptable alpha levels except Sensation Seeking. This is in contrast to the standardization of the TSP, which found that Sensation Seeking had the highest internal consistency of any of the styles. Among the modalities, only the auditory modality had acceptable alpha levels across diagnoses. The typical group had poor internal consistency coefficients for the visual modality and the vestibular modality. The internal consistency coefficient for vestibular was similar to the vestibular internal consistency coefficient in the standardization of the TSP (.39 in the current study versus.42 in the standardization study). On the other hand, the internal consistency for the visual domain was much higher in the standardization study than that seen in the current study. Despite the low internal consistency for the vestibular modality for the typical group, the consistency among the group with autism reached an acceptable level. Internal consistency across diagnoses and domains improved when some of the items were dropped. Overall, the face validity for the dropped items was not good. For example, the items my child enjoys splashing during bath time or my child enjoys looking at self in the mirror do not seem to reflect unusual sensory responses. However, 31

41 it is important to keep in mind that a midrange score (my child sometimes does the behavior) on the TSP is typical. Interestingly, the internal consistency was generally poor for the typical group (visual: α =.29; oral: α =.48, vestibular: α =.39). This is surprising for two reasons. First, these alpha levels are much lower than either of the ASD groups, yet the TSP was des igned to assess sensory responses in typical populations, not autism. Second, the TSP was standardized on a group of typical children and the alpha levels of the current study s typical group are much lower than the alpha levels for the standardization group. This may be due to the differences in sample size (17 versus 809) or may be due to the large number of at-risk siblings in the current typical sample; 58.8% of the typical sample were siblings, which was the largest percentage of any of the groups. Although these participants were believed to be typical on the basis of performance on standardized testing, no history of developmental delay or current concerns, and clinical judgment, subtle differences in pointing, rate of communicating, and daily living skills have been found between typical siblings of children with autism and children with no family history of autism (Toth, Dawson, Meltzoff, Greenson, & Fein, 2007). This leaves the possibility that, although judged to be typically developing, the typical sample in the current study was not consistent with the typical sample in the standardization sample. There is some consistency between the TSP and sensory behavior on the ADI and ADOS, however most of these associations can be accounted for by diagnosis. Initial examination of scores on the TSP by ADOS sensory score indicated that those with scores of 2 or 3 on the Unusual Sensory Interest item on the ADOS had greater frequency of unusual sensory behavior in the areas of Low Registration and Sensation 32

42 Avoiding, as well as the modalities of auditory and visual, compared to children who did not display sensory interests during the ADOS. When diagnosis was controlled for, howeve r, these differences were no longer significant, indicating that differences on the TSP could be explained by diagnosis. As shown in Table 2.4, ADOS sensory score was highly related to diagnosis, with a much greater percentage of toddlers with autism receiving a score of 2 or 3 than any other diagnosis. There are several explanations for the lack of consistency between the ADOS and TSP. First, the ADOS is a relatively brief observational measure of behavior. It is possible that behaviors which the caregiver reports as occurring frequently on the TSP, do not occur so frequently that they would be observed in a 45-minute observational period. In addition, the TSP asks about responses to situations that are not part of the ADOS, such as eating in a noisy environment and being tipped backward while bathing. Similarly, the sensory behaviors that are most often observed and coded as a 2 or 3 by the examiner during the ADOS, are not included in the TSP. The majority of children who scored a 2 or 3 on the ADOS were given this code due to prolonged visual examination/inspection of an object, often accompanied by squinting at the object. This behavior is not included among the visual behaviors assessed on the TSP. Additionally, several children were coded with a 2 on the ADOS due to repeatedly sniffing the testing materials. Olfactory sensory behaviors are also not addressed on the TSP. Another explanation for the lack of convergence between the ADOS and TSP is that the sensory behavior that is coded on the ADOS primarily addresses atypical sensory seeking behavior and not behaviors that would be considered under low registration or sensation avoiding on the TSP. In fact, covering one s ears during the ADOS, which could be 33

43 considered sensory sensitivity to noise or sensation avoiding, is not coded under the ADOS sensory item, but rather is coded as a repetitive/stereotyped behavior. There was greater convergence between the TSP and ADI-R. After controlling for diagnosis, scores on the ADI-R item Unusual Sensory Interest had a significant effect on Low Registration and auditory. This effect remained after removing the autism- Abnormal/Idiosyncratic specific items. Undue General Sensitivity to Noise and Response to Sensory Stimuli did not have a significant effect on any TSP sensory styles or modalities. An ADI-R sensory severity score, which was created by combining the scores from all three ADI-R sensory items, had a significant effect on auditory and Sensation Avoiding behaviors, even after controlling for diagnosis. The ADI sensory severity score significantly predicted Low Registration and Auditory Processing after removing the autism-specific items. In fact the effect was stronger for Low Registration without these items. The finding that the ADI-R sensory items were more related to the TSP scores than the ADOS sensory item after controlling for diagnosis may have been due to the fact that both the TSP and the ADI are based on parent report and can therefore reflect behavior that occurs on a daily basis and in response to a variety of situations. In addition, the ADI-R sensory severity score had the greatest convergence with the TSP, likely because the severity score encompassed a wider range of behaviors, which is more similar to the TSP. 34

44 Table 2.1 Participant Demographics Current Toddlers/First Words Participants Most Recent Diagnosis Typical Autism PDD Nonspectrum DD N Age 14.5 (5.4) 22.7 (7.2) 23.1 (8.8) 22.5 (8.8) Gender Male 11 (64.7%) 22 (88%) 20 (80.0%) 18 (81.8%) Female 6 (35.3%) 3 (12%) 5 (20.0%) 4 (18.2%) Risk Status Sibling 10 (58.8%) 9 (36.0%) 9 (36.0%) 7 (31.8%) Not Sibling 7 (41.2%) 16 (64.0%) 16 (64.0%) 15 (68.2%) Number of Visits 1 visit only 4 (23.5%) 7 (28%) 6 (24.0%) 10 (45.5%) 2 visits only 2 (11.8%) 8 (32%) 6 (24.0%) 5 (22.7%) More than 2 visits 11 (64.7%) 10 (40%) 13 (52.0%) 7 (31.8%) Cognitive abilities (initial visit) Verbal IQ* 99.5 (14.5) 51.0 (23.0) 87.5 (19.5) 65.8 (26.6) Nonverbal IQ* (15.5) 86.6 (18.6) 97.6 (17.2) 80.9 (24.4) Full Scale IQ* (10.0) 70.2 (16.4) 92.0 (13.7) 74.4 (24.5) *Average =

45 Table 2.2 Means, Standard Deviations (top), and Modes (bottom) for the TSP Items Identified as Autism-Specific Items Typical N=17 Autism N=25 PDD-NOS n=25 NSDD n=22 My child avoids playing with others 4.4 (0.8) 2.8 (1.3) 3.6 (1.2) 4.2 (1.1) I have to speak loudly to get my child s 3.7 (1.0) 2.4 (1.3) 3.3 (1.0) 3.3 (0.9) b 3 I have to touch my child to gain his/her (1.6) 3.75 (1.1) 3.4 ( 1.2) attention c (1.1) My child takes a long time to respond, 4. 2 (0.8) 2.5 (1.3) 3.9 (.93) 3.6 ( 1.2) even to familiar voices d It takes a long time for my child to 3.9 (0.9) 2.2 (1.1) 3.4 (1.1) 3.5 (1.1) respond to his name when it is called My child enjoys looking at moving or 2. 8 (1.3) 2.2 (1.5) 2.8 ( 1.4) 2.9 ( 1.2) spinning objects My child avoids eye contact with me 3.9 (1.2) 3.2 (1.1) 3.9 (1.2) 4.0 (0.9) Meaning of modal scores: 1 = Almost Always 2 = Frequently 3 = Occasionally 4 = Seldom 5 = Almost Never a Autism significantly lower than Typical and NSDD (p <.001) b Autism significantly lower than Typical (p <.001) c Autism significantly lower than Typical (p <.01) d Autism significantly lower than Typical and PDD-NOS (p <.001) and NSDD e Autism significantly lower than Typical (p <.001) and PDD-NOS and NSDD (p <.01) 36

46 Table 2.3: Odds Ratio s for NVIQ, Age, and Low Registration for Each Diagnosis in Relation to Autism Odds Ratio 95% CI Significance Typical NVIQ Age Low Registration PDD-NO S NVIQ Age Low Registration NSDD NVIQ Age Low Registration Autism Referent 37

47 Table 2.4: Odds Ratio s for NVIQ, Age, and Auditory for Each Diagnosis in Relation to Autism Odds Ratio 95% CI Significance Typical NVIQ Age Auditory PDD-NOS NVIQ Age Auditory NSDD NVIQ Age Auditory Autism Referent 38

48 Table 2.5: Percent Correct Classification of Diagnosis by Model Model 1 Model 2 Model 3 Typical 76% 76% 77% PDD-NOS 38% 52% 38% NSDD 40% 45% 45% Autism 39% 65% 65% Model 1 = NVIQ and age Model 2 = NVIQ, age, Low Registration Model 3 = NVIQ, age, Auditory 39

49 Table 2.6: Internal Consistency Coefficients for Each Sensory Style by Diagnosis Cronbach s Alpha Quad 1 Low Quad 2 Sensation Quad 3 Sensory Quad 4 Sensation Registration Seeking Sensitivity Avoiding Typical (w/o items 6, 12, 38).69 (w/o item 9) Autism (w/o items 6, 12, 38) PDD-NOS (w/o items 6, 12, 38) NSDD (w/o items 6, 12, 38) Item 6: My child enjoys making sounds with his/her mouth Item 9: My child is distracted and/or has difficulty eating in noisy environments Item 12: My child finds ways to make noise with toys Item 38: My child enjoys rhythmical activities (for example, swinging, rocking, car rides) 40

50 Table 2.7: Internal Consistency Coefficients for Each Sensory Modali ty by Diagnosis Cronbach s Alpha Auditory Processing Visual Processing Tactile Processing Vestibular Processing Oral Processing Typical (w/o item 19) Autism (w/o item 19) PDD-NOS (w/o item 19) NSDD (w/o item 19) (w/o items 34 & 35) (w/o item 36) (w/o i tem 46) (w/o items 34 & 35).59 (w/o item 36).54 (w/o item 46) (w/o items 34 & 35).50 (w/o item 36).51 (w/o item 46) (w/o items 34 & 35).21 ( w/o item 36).68 (w/o item 46) 41 Item 19: My child enjoys looking at own reflection in the mirror Item 34: My child enjoys splashing during bath time Item 35: My child uses hands to explore food and other textures Item 36: My child requires more support for sitting than other children the same age (for example, infant seat, pillows, towel roll) Item 46: My child resists having teeth brushed 41

51 Table 2.8: Number of participants and percent of sample receiving each ADOS sensory item code by diagnostic group AD OS Code = 0 ADOS Code = 1 ADOS Code = 2 ADOS Code = 3 No sensory Possible sensory Definite sensory Sensory interest interest observed interest interest interferes Typical 59 (73.8%) 9 (11.3%) 2 (2.1%) 0 Autism 18 (22.8%) 27 (34.2%) 18 (22.8%) 9 (11.4%) PDD-NO S 47 (47.4%) 22 (23.2%) 12 (12.6%) 5 (5.3%) Nonspectrum D D 32 (46.4%) 18 (26.1%) 10 (14.5%) 4 (5.8%) 42 42

52 Table 2.9: Effect of ADI Sensory Severity Score on Each TSP Modality and Style Modality Style Effect of ADI-R Sensory Severity Score Auditory F(1, 76.3) = 8.5, p <.01 Visual F(1, 74.4) = 2.4, p = n.s. Tactile F(1, 75.1) = 6.3, p <.05 Vestibular F(1, 75.4) = 1.7, p = n.s. Oral F(1, 75.9) = 6.6, p <.05 Low Registration F(1, 74.5) = 6.6, p <.05 Sensation Seeking F(1, 73.8) = 0.2, p = n.s. Sensory Sensitivit y F(1, 74.6) = 6.3, p <.05 Sensation Avoidin g F(1, 75.3) = 11.2, p =.001 Low Registration (w/o autism items) F(1, 73.5) = 7.3, p <.01 Auditory (w/o autism items) F(1, 74.6) = 8.6, p <.01 43

53 Figure 2.1: Frequency of Derived ADI Sensory Severity Scores by Diagnosis 44

54 Chapter III Unusual Sensory Behaviors in Children with ASD: Differences from Typical and Nonspectrum Populations and Related Characteristics Studies comparing sensory responses of children with ASD to typically developing children have consistently found a greater incidence of unusual responses among children with autism. Talay-Ongan and Wood (2000) used the Sensory Sensitivity Questionnaire Revised, a 54-item questionnaire that assesses sensory sensitivity in the auditory, tactile, visual, gustatory, olfactory, and vestibular domains, to compare children with autism between the ages of 4 and 14 years (n = 27) to a group of age-matched typically developing peers (n = 30). Results indicated that the group with autism had significantly greater rates of hyper- and hypo-sensitivities across all modalities, with effect sizes ranging from.56 for the olfactory domain to 2.53 for the auditory domain. These findings are consistent with earlier studies that examined differences between children with autism and typical controls using a different parentreport measure, the Sensory Profile Questionnaire (Dunn, 1997). In a study of 32 children with autism, between the ages of 3 and 10 years of age, Kientz and Dunn (1996) found that individuals with ASD had significantly more atypical behaviors than typically developing individuals in all sensory modalities. In addition, behaviors consistent with both hypersensitivities (difficulty functioning in noisy environments and expressing discomfort during grooming) and hyposensitivity (continually seeking out activities that 45

55 involve movement) differentiated the two groups. Using an earlier version of the Sensory Profile Questionnaire, Watling, Deitz, and White (2001) also found differences between 40 children with autism and 40 age- and gender-matched controls without autism (between the ages of 3 and 6 years) in the areas of sensory seeking, oral sensitivity, and low registration. While the majority of the studies comparing children with autism to typically developing peers, such as those reported above, have examined differences after the age of three years a few studies have looked for differences in younger populations. Dahlgren & Gillberg (1989) used a questionnaire to assess differences in behavior in the first two years of life in a group of children with autism (n = 26) compared to a group of children with mental retardation (n = 20) and a typically developing comparison group (n = 25). The results indicated that, in addition to differences in social relationships, communication, and play-behavior, the children with autism were also more likely to have abnormal reactions to sound stimulation and to engage in bizarre looking at objects, patterns, and movement. These items were found to differentiate the children with autism from the other two comparison groups. One limitation of this study, however, which the authors acknowledge, is the retrospective design. At the time that the caregivers completed the questionnaire, the participants ranged in age from 7 to 13 years, thus, one could question the accuracy of parent s report of the child s behavior prior to 2 years of age. However, another study did find significant differences on sensory items of a standardized diagnostic parent interview that was administered when the children were 2 years of age. By comparing rates of behaviors of individual items on the ADI-R (Lord, Rutter, & LeCouteur, 1994), Richler, Bishops, Kleinke, & Lord (2007) found that 46

56 at age 2 the prevalence of unusual sensory interests and abnormal/idiosyncratic response to sensory stimuli was higher among the group with ASD than the typically developing children or children with developmental delays. Studies comparing children with autism to children with other developmental delays or disorders have not been as consistent. Baranek, Boyd, Poe, David, and Watson (2007) developed a play-based observational assessment tool, called the Sensory Processing Assessment for Young Children (SPA). This measure provides opportunities to observe a child s reaction to a range of multisensory toys. Using the SPA, Baranek and her colleagues examined the behavior of 56 children with autism, 30 with nonspectrum developmental delays (NSDD), and 53 with typical development. Results indicated that both the ASD and NSDD groups scored significantly higher than the group with typical development, but did not score significantly differently from each other. The results of another study suggest that differences between ASD and NSDD may be modified by the type of response. Baranek, et al., (2006) found that a pattern of hyporesponsiveness to both social and nonsocial stimuli differentiated the children with ASD from NSDD and typically developing control groups. Hyperresponsiveness did not differentiate the children with ASD from the developmentally delayed group, but did distinguish those with ASD from the typically developing group. In addition, hyporesponsiveness specifically to social stimuli differentiated the children with autism from PDD-NOS, suggesting that there may be a relationship between the increased social impairment in autism and atypical sensory responses. The atypical sensory responses among individuals with ASD are also present in other disorders, such as Attention Deficit Hyperactivity Disorder (ADHD; Dunn & 47

57 Bennett, 2002; Mangeot, et al., 2001) and fragile X syndrome (Baranek, et al., 2005; Rogers, Hepburn, & Wehner, 2003). Studies comparing individuals with ASD and fragile X syndrome have suggested that rates of atypical sensory behaviors are comparable between the two groups (Rogers, Hepburn, & Wehner, 2003), although the high rate of autism among individuals with fragile X syndrome makes it difficult to draw conclusions regarding differences between these two groups. For example, Rogers, Hepburn, & Wehner (2003) reported that 35% of their sample with fragile X syndrome also met criteria for autism. Without consideration of this overlap between their two samples, they concluded that there was no difference in rates of atypical sensory behaviors between the children with autism and children with fragile X syndrome. Another study compared sensory behaviors of children with ASD to those of children with ADHD and typical controls (Ermer & Dunn, 1998). Results indicated that while the children with ADHD had higher rates of sensory seeking behavior than children with ASD, children with ASD had higher rates of oral sensory sensitivity and fine motor/perceptual difficulties. Several of the studies reported above did not take IQ or mental age into consideration when examining group differences (Watling, Deitz, & White, 2006; Talay- whether or Ongan & Wood, 2000; Ermer & Dunn, 1998), making it difficult to determine not cognitive differences between the groups were driving the differences in sensory behavior. The studies that have examined the association between cognitive impairment and the presence of unusual sensory behaviors have generally found that IQ is not associated with sensory behaviors. Lord, Rutter, & Le Couteur (1994) compared a group of children with ASD (mean nonverbal IQ = 62) to a group of same-aged intellectually 48

58 impaired children without ASD (mean nonverbal IQ = 63). They found that parents more frequently reported unusual sensory responses in the children with ASD, suggesting that differences in unusual sensory behaviors are not associated with impairments in NVIQ. In addition, using an abbreviated version of the Sensory Profile, Rogers, Hepburn, and Wehner, (2003) also found that children with autism, between the ages of 26 and 41 months had significantly more sensory symptoms in the areas of taste/smell sensitivity, tactile sensitivity, and auditory filtering compared to the children with developmental delay of mixed or unknown etiology and that sensory scores were not associated with developmental level or ratio IQ. Likewise, in both her study of parent report of sensory behaviors (Baranek, et al, 2006) and her direct observation of sensory behaviors (Baranek, et al, 2007) Baranek and her colleagues found that IQ did not have a significant effect on responses to sensory stimuli. Results of research examining the occurrence of atypical sensory behaviors across development have been inconsistent, with some studies suggesting that at very young ages sensory symptoms decrease as mental age increases (Baranek, et al., 2007; Baranek, et al., 2006). Others have concluded that there is no relationship between developmental level and sensory abnormalities (Rogers, Hepburn, & Wehner, 2003). According to one cross-sectional study of ASD, atypical sensory responses decrease with age and become more similar to typical controls by middle adulthood (Kern, et al. 2006). However, another cross-sectional study of children ages 4 to 14 found that sensory symptoms were more common among the 6 to 9 years olds than 4 to 5 year olds and were most common among 10 to 14 year olds (Talay-Ongan & Wood, 2000), suggesting that atypical sensory responses may actually increase in frequency with age. Qualitative studies of sensory 49

59 difficulties in adults with ASD, have found that difficulties with sensory stimulation can make learning and social interaction more difficult (Jones, Quigney, & Huws, 2003). Autobiographical accounts of these experiences in adults indicate that responses to sensory stimuli in the environment can cause distress (Grandin & Scariano, 1996), although at other times, or in other forms, sensory stimuli can lead to enjoyment and comfort (Jones, Quigney, & Huws, 2003; O Neill & Jones, 1997). While some studies have examined the effect of age and NVIQ on sensory behaviors, no studies have yet examined how risk status is associated with sensory symptoms. Siblings of children with autism are at greater risk for both having autism themselves and for having other developmental difficulties, such as language delay. Recurrence risk for ASD has been estimated to be between 5% and 10% (Ritvo, et al., 1989; Cook, 1998; Szatmari, Jones, Zwaigenbaum, & MacLean, 1998; Fombonne, 2003). In addition, they may be more likely to exhibit unusual sensory responses if they have a sibling who has high rates of unusual sensory behaviors; one study found moderate heritability estimates for sensory behaviors in the auditory and tactile modalities among a large sample of twins without autism (Goldsmith, Van Hulle, Arneson, Schreiber, & Gernsbacher, 2006). If atypical sensory behaviors are more common among children with autism and sensory behaviors have a genetic basis, then it could by hypothesized that atypical sensory behaviors would be seen more often among siblings of children with autism (even if they themselves do not have autism), then nonsiblings. To date, the relationship between sibling status and the presence of sensory behaviors has not been examined. 50

60 The current study examined the presence of sensory symptoms within ASD by comparing the groups with autism and PDD-NOS to the typical standardization sample as well as by examining effect sizes of the comparison between autism and PDD-NOS. It was hypothesized that both the groups with autism and PDD-NOS would have significantly greater rates of unusual sensory behaviors compared to the standardization sample. However, it was also hypothesized that frequency of unusual sensory behavior would be more common among the group with autism than the group with PDD-NOS, as demonstrated by the effect sizes. The association between atypical sensory behaviors and characteristics including age, diagnosis, NVIQ, and risk status (sibling versus behavioral/medical risk factor) was also examined. On the basis of previous literature, it was hypothesized that diagnosis would have a significant effect on Low Registration and Auditory sensory behaviors and that age would have a significant effect on Sensation Seeking, Tactile, and Oral behaviors. In order to compare differences in observed sensory behaviors, the effect of diagnosis, NVIQ, and age on the sensory item included in the ADOS was also examined. It was hypothesized that diagnosis and age would have a significant effect on ADOS sensory item codes, with those with the autism having significantly higher scores on this item than the other three diagnostic groups. Method Participants Participants included 89 children, ages 11 months to 3 years. These were the same participants included in the analyses described in Chapter 2. Each of the participants were included in the First Word/Toddlers Projects at the University of Michigan Autism and Communication Disorders Center (UMACC). These projects 51

61 address ed early identification of ASD in infants and toddlers who were considered to be at higher risk for having autism. Higher risk status was defined by either being the younger sibling of a child already diagnosed with ASD or by having a medical or behavioral risk factor associated with autism (e.g. language delay, seizures). Participants in the First Words Project included children with speech and language impairments and participants had to be less than 3 years of age to enroll in the study. Participation in this project included assessments every six months. A subset of First Words participants were also enrolled in the Toddlers Project. All participants in the Toddlers project were suspected of having autism. Participation in the Toddlers Project included monthly assessments and children must have been under the age of 16 months at the time of enrollment. Based on most recent consensus diagnoses, which were determined at the mean age of 27.5 months (s.d. = 7.8), 17 (19.1%) participants were considered typically developing, 25 (28.1%) had a current diagnosis of autism, 25 (28.1%) had a current diagnosis of PDD-NOS, and 22 (24.7%) were diagnosed with nonspectrum developmental delays (see Table 3.1). The nonspectrum developmental delay group was a heterogeneous group that included language disorders, intellectual disability, Down syndrome, William syndrome, fetal alcohol syndrome, and behavioral disorders. With all diagnoses combined, 35 (39.3%) participants were siblings of an older child with autism and 54 (60.7%) were considered high risk due to the presence of medical and/or behavioral risk factors. There were 71 males and 18 females. Sixty one of these participants were seen more than once. Age at initial visit ranged from 7 to 36 months (x 52

62 = 21.2, s.d. = 8.4). The average number of visits was 3.7 (s.d. = 3.5) with a range from As shown in Table 3.1, the typically developing higher risk participants were significantly younger at their first assessment than the other three groups. This is because a greater proportion of the typical children were recruited as siblings. The average age of the first visit for siblings was 16.6 months (s.d. = 6.6), while the average age for those with medical/behavior risk factors was significantly higher (x = 24.2, s.d. = 8.0; t = -4.7, p <.001). The diagnostic groups also varied in terms of their cognitive abilities. While the typically developing children had average verbal and nonverbal IQ scores, those with autism and nonspectrum developmental delay had verbal IQ scores in the range of mild intellectual disability and nonverbal IQ scores in the low average range. Those with PDD-NOS had verbal IQ scores that were low average and nonverbal IQ scores that were in the average range. Because the First Words/Toddler s projects were longitudinal, it is helpful to look at the sample sizes by diagnosis at different age groups. Five age groups were established, which correspond to the age groups used in the scoring of the outcome measure (Toddler Sensory Profile Questionnaire), which is described in detail below. Table 3.2 below shows the total number of assessments during each Sensory Profile age group by diagnosis. Some participants had more than one assessment during the same age group. Thus, some of the numbers are larger than the total sample size in each group reported in Table 3.1. The number in parentheses indicates the number of initial assessments for individual children during that age range. Because of the small sample 53

63 sizes in the 7 to 12 month group, of all but the typical children, assessments in this age range are not included in the analyses described below. Measures The measures used in the current study include the Toddler Sensory Profile (TSP; Dunn, 2002), Mullen Scales of Early Learning (MSEL; Mullen, 1995), Autism Diagnostic Observation Schedule (ADOS; Lord, Rutter, DiLavore, & Risi, 1999), and The Autism Diagnostic Interview Revised (ADI-R; LeCouteur, Lord, & Rutter, 2003). Please refer to Study 1 for a description of these measures. Procedures Children in the First Words project completed a comprehensive diagnostic evaluation approximately every six months. Prior to the evaluation, the caregiver was sent a packet of questionnaires, which included the Toddler Sensory Profile Questionnaire (TSP). The ADI-R and Vineland were completed at the initial appointment with only the caregiver(s) present. A second appointment with the child was held within a week of this initial appointment. During this second session an examiner worked directly with the child and administered the ADOS and the Mullen. At least one caregiver was in the room during this testing. Children in the Toddler study completed monthly testing sessions at UMACC during which they completed the ADOS. The TSP was also completed monthly. Like the First Words Project, the ADI-R and Mullen were completed in six-month intervals, by clinicians who were blind to the child s history and previous diagnosis. 54

64 A best estimate diagnosis was completed after each assessment. The best estimate diagnosis was based on all information available from the ADOS, various measures, and clinical judgment. Results Preliminary analyses indicated that the children with autism and the children with PDD-NOS had significantly different distributions of scores on each of the TSP sensory modalities and sensory styles (results reported below). Therefore, all of the following analyses were conducted separately for autism and PDD-NOS, unless otherwise indicated. Atypical Sensory Behaviors within ASD Autism and PDD-NOS Compared to the Typical Standardization Sample As noted above, the Toddler Sensory Profile yields several different sensory scores. First, continuous scores are obtained for each of the sensory modalities and each of the sensory processing styles. Second, each of these continuous scores is broken down into three categories: less responsive than others, typical performance, and more responsive than others. The range of scores included in each of these categories is based on a normalized distribution of scores found among typical toddlers in the standardization sample (Dunn, 1997). If the autism group had a normal distribution, 16% of the sample would be included in the less responsive category, 68% would fall in the typical performance category, and 16% would be in the more responsive category. Therefore, performance of the toddlers with autism and PDD-NOS can be examined by examining the distribution of the sample across these three categories. Thus, first, performance of the children with autism and PDD-NOS was compared to the standardization sample by examining the distribution of these groups across the 55

65 three categories using Chi-square analyses. Table 3.3 shows the Chi-square values and significance levels for the children with autism and PDD-NOS compared to the normal distribution on each of the sensory modalities and sensory styles. Significance level was set at.005, to account for the number of a nalyses. Results indicated that the distribution of children with autism across the three categories (less responsive, typical, and more responsive) was significantly different than a normal distribution for each of the sensory modalities. This was due to a greater than expected proportion of the sample scoring in the more responsive than others category for each modality. There was also a significantly greater proportion of children with PDD-NOS in the more responsive than others category in the auditory domain than expected compared to a normal distribution. As opposed to the children with autism, the children with PDD-NOS had a normal distribution for all other modalities. For the sensory styles, the children with autism had a significantly different distribution for Low Registration, Sensory Sensitivity, and Sensation Avoiding, whereas the children with PDD-NOS had a significantly different distribution only for Low Registration. These significant differences were due to a greater than expected proportion of the sample falling in the more responsive range than other categories for both autism and PDD-NOS. Comparison of Atypical Sensory Behaviors across Diagnoses Effect Sizes Diagnostic groups were compared using both the continuous and categorical sensory scores from the TSP. First, effect sizes were generated using the continuous raw scores for each sensory style and sensory modality at each TSP age group. Autism was 56

66 used as the reference group therefore the effect sizes shown in Figures 3.1 and 3.2 reflect the difference between autism and the other diagnostic groups. Effect sizes were not generated for the 7 to 12 month age range due to the small number of participants in each diagnostic group (except typical) for that age group. Likewise, an effect size was not generated for the difference between the autism and typical group after 24 months because of the small sample size in the typical group. As shown in Figure 3.1, the effect sizes for the differences between autism and the other three groups became larger for each sensory style, except sensation seeking, as the participants moved from months of age to months of age. For comparisons of the children with autism to those with typical development, the effect size increased from below 1 to above 1 for each of the sensory styles except sensation seeking. This change in effect sizes was due to a decrease in scores among the children with a utism (indicating greater frequency of atypical behaviors) and an increase in scores (indicating a reduction in atypical behaviors) for the other three groups. As the groups moved into the 25 to 30 month age group, the difference between PDD-NOS, NSDD, and autism continued to increase for Low Registration and Sensation Avoiding, with the effect sizes being larger for Low Registration. The largest effect size of 2.4 was for the comparison between autism and NSDD at 25 to 30 months of age. By 31 to 36 months of age, all of the effect sizes had become smaller. A similar pattern was found for the effect sizes for the modalities. As shown in Figure 3.2, the effect size for the comparison of children with autism to typically developing children increased from below 1 to above 1 for each of the sensory modalities, with the exception of oral processing. The effect sizes for the comparison of autism to 57

67 the other two diagnostic groups also increased from the first age group into the 25 to 30 month age group, with the greatest increase being in the auditory modality. For children with autism versus PDD-NOS, the effect size remained consistent from the 25 to 30 month age group into the 31 to 36 month age group for the visual and oral domains, while the effect sizes comparing autism and NSDD decreased into the 31 to 36 month age group in all modalities. Like the effect sizes for the sensory styles, the increases in effect size were due to both a decrease in scores among the children with autism and an increase in scores among the other three groups. The decrease in effect sizes in the 31 to 36 month groups was due to both an increase in scores for the group with autism and a decrease in scores for the other two diagnostic groups. Due to the similarities between auditory and Low Registration, both in terms of the changes in effect sizes over time and the significant difference in distribution of both autism and PDD-NOS compared to the standardization sample (as reported above), effect sizes were examined for the auditory modality without the overlapping Low Registration items and vice-versa. For the auditory modality, excluding Low Registration items also resulted in removing all of the autism-specific items reported in Chapter 2. For Low Registration, one additional item that was considered an autism-specific item, but was not included in the auditory domain was also excluded. This item was my child avoids eye contact with me. Figure 3.3 compares the effect sizes of auditory and low registration with all items included to auditory and low registration without the overlapping and autism-specific items. As can be seen in the graphs, without the overlapping and autism- specific items there is still a slight increase from the first age group to the third age group but the amount of increase is greatly reduced, suggesting that the large increase in the 58

68 effect sizes in the full auditory and low registration domains is being driven mostly by the autism-specific items. Predictors of Sensory Behaviors on the TSP (All Diagnoses Included) Diagnosis, risk status (sibling versus non sibling), NVIQ, and age were examined as predictors of sensory behavior using Linear Mixed Modeling. To allow for the intercept to be different from zero, age was centered at the mean for each diagnosis: for typical, for autism, for PDD-NOS and for nonspectrum DD. The initial model included all main effects, all two-way interactions, and two three-way interactions (age by diagnosis by nonverbal IQ and age by diagnosis by risk). In all models, age was entered as both a linear variable and a quadratic variable (age by age). Correlations between each observation within a subject were modeled by an autoregressive (AR1) covariance structure, allowing for assessments that occurred closer together to be more correlated than assessments that occurred farther apart. Nonsignificant interactions were removed from the model one at a time and the model was rerun each time. Changes in Restricted Log Likelihood and Bayesian Criterion were compared to examine goodness of fit. Table 3.4 shows the Type III test of fixed effects for the final models predicting the TSP sensory modalities that were determined to have the best fit, based on the process described above. Significance levels were set at.01. Significant results are bolded. None of the three-way interactions are included in Table 3.4, as they were not significant for any of the modalities. The vestibular modality is not included in Table 3.4, due to the low Cronbach s alpha coefficients for the vestibular domain. As reported in Chapter 2, Cronbach s alpha was low for each diagnostic group 59

69 except autism; for the NSDD group, 0.39 for the typical group, 0.45 for PDD-NOS, and 0.70 for autism. Auditory modality: There was a significant main effect of diagnosis, a quadratic age effect, and a significant interaction between age and risk. The relationship between age and risk is shown in Figure 3.4. At younger ages there is greater divergence between the siblings and nonsiblings, with siblings having higher scores on the auditory processing modality (indicating lower frequency of atypical auditory sensory behaviors). Over time, the scores for the siblings become lower and the scores for the non siblings increase. By the time they are 36 months of age, there was virtually no difference between the siblings and non siblings. The data included in this interaction reflects both longitudinal (within subjects) and cross-sectional (between subjects) data. The Linear Mixed Model analyses used to test the interaction takes this into consideration. The main effect of diagnosis can be further explored by examining the coefficient estimates. These estimates indicate that the group with autism had significantly lower scores than the three other diagnostic groups (typical: t(80.4) = 4.2, p <.001; PDD-NOS t(79.9) = 2.9, p <.01; NSDD: t(79.9) = 2.6, p <.05). The typical group had the highest scores, reflective of more typical behaviors. There was no significant difference between the NSDD group and the PDD-NOS group. In order to ensure that the main effect of diagnosis was not being driven by the autism-specific items (as reported in Chapter 2) the model was rerun on the auditory modality without these items. The results indicated that even after removing the autism specific items, diagnosis still had a main effect on auditory behaviors (F(3, 73.9) = 4.6, p <.01), although there was no longer a significant difference between the autism group and the NSDD group (t(79.6) = 0.9, p = n.s.). 60

70 Visual Modality: Similar to the findings for the auditory modality the relationship between risk and visual scores approached significance and this relationship varied by age. Also consistent with the auditory scores, as the nonsiblings got older, their visual scores increased (improved) and became more consistent with the siblings. The sibling scores, however, remained stable over time. Unlike the auditory modality, there was no main effect of diagnosis. Also unlike the auditory modality, there was a trend toward an interaction between diagnosis and age, with the difference in trajectories for each diagnostic group approaching significance. The typical group had the greatest increase (improvement) in the visual modality scores over time and the group with autism had the smallest increase. Table 3.5 shows the amount of change in the visual modality score over time by diagnosis. For the autism group, for every 1 month increase in age, there was a 0.09 point increase in the visual score (a 2.16 point increase in score over 24 months). This increase was not significant. The typical group showed the greatest change over time; for every one month increase in age, the visual score changed by 0.53 points (12.72 points over 24 months), indicating that as the typical group got older, on average they had less frequent unusual visual sensory behaviors. While this change over time was not significantly different from zero for the autism group, it was significant for the typical group (t(73.2) = 4.0, p <.001). Tactile Modality: For the tactile modality, no interaction terms were significant, nor did they add to the fit of the model, therefore they were removed. The final model included age, risk, diagnosis, and NVIQ. There was a main effect of diagnosis, with the coefficient estimate being significantly lower for autism than PDD-NOS (t(80.2) = 3.3, p 61

71 <.01) and typical (t(91.5) = 2.8, p <.01). There were no significant differences between the other diagnostic groups. There was a non-significant trend for age; as age increased scores on the tactile modality also increased, indicating that as the participants got older they engaged in tactile sensory behaviors less frequently. Oral Modality: Results indicated that NVIQ had a main effect on oral sensory scores. As NVIQ scores increased, oral sensory scores increased, indicating that across diagnosis, individuals with higher NVIQ had lower frequency of atypical oral sensory behaviors. Predictors of TSP Style Scores As was done above for the TSP modalities, age, risk status, NVIQ, and diagnosis were examined as predictors of TSP style scores. Table 3.6 shows the models that provided the best fit using Linear Mixed Modeling. Significance level was set at Low Registration: There was a significant effect of diagnosis on Low Registration style. The estimates of fixed effects showed that the autism group had significantly lower Low Registration scores than all three other diagnostic groups (typical: t(88.6) = 3.9, p <.001; PDD-NOS: t(79.6) = 4.4, p <.001; NSDD: t(84.7) = 3.6, p =.001). There were no other significant differences between any other diagnostic groups. The change in low registration behaviors, over time, followed a quadratic trend with the frequency of low registration behaviors becoming less frequent and then more frequent over time. In order to ensure that the autism-specific items were not driving the significant main effect of diagnosis, the model was run excluding the autism-specific items. Results indicated that there still was a main effect of diagnosis on Low Registration without the autism items (F(3, 81.6) = 5.6, p <.01). The autism group continued to be significantly 62

72 different from the other three diagnostic groups (typical: t(80.0) = 2.9, p <.01; PDD- NOS: t(75.5) = 3.8, p <.001; NSDD: t(83.4) = 2.6, p =.01). The rest of the model also remained unchanged, with a significant quadratic age effect. Sensation Seeking: Diagnosis did not have a significant effect on sensation seeking. This was expected, as earlier chi-square analyses had indicated that neither the autism nor PDD-NOS groups had distributions that were significantly different than a normal distribution. Risk and NVIQ were also not significant. Age had a significant effect on Sensation Seeking; as the children got older, they engaged in less sensory seeking behaviors. Sensation Avoiding: There was a significant main effect of diagnosis on the Sensation Avoiding style. The estimates of fixed effects showed that the autism group had significantly lower Sensation Avoiding scores than the group with PDD-NOS and the group with NSDD (PDD-NOS: t(84.7) = 2.7, p =.01; NSDD: t(83.6) = 3.0, p <.01). The autism group was not significantly different from the typically developing group. Although the diagnosis by risk interaction was not significant, it was included in the final model, as including it greatly improved the fit statistics. Predictors of Sensory Behaviors Seen During the ADOS The effect of age, NVIQ, and diagnosis on sensory behaviors was also examined using the ADOS sensory interest item. Figure 3.5 shows the trajectories of change in score by diagnostic group. The program SAS proc mixed was used to examine these trajectories. Results indicated that there was a significant effect of diagnosis (F(3, 81) = 9.27, p <.0001). According to the Test of Effect Slices, there was a significant difference between the diagnoses at ages 13 to 18 months (F(3, 73) = 10.89, p <.0001), 19 to 24 63

73 months (F(3, 73) = 5.68, p <.01), and 31 to 36 months (F(3, 73) = 6.39, p <.001). There was a non-significant trend toward an interaction between age and diagnosis, with different trajectories for the different groups. While the group with autism appeared to be showing a general increase in observed sensory behavior over time, the typically developing participants showed a steady decrease. Conclusions Results of the current study indicate that unusual sensory behaviors are more common among toddlers with autism than their typically developing peers, children with PDD-NOS or children with NSDD. However, the results of this study highlight the importance of examining individual modalities and styles of sensory responding separately, as the differences between the diagnoses change depending on which modality or style is being considered. For example, when children with autism and PDD- NOS were compared to the typical standardization sample, which had a normal distribution, it was found that the distribution for those with autism was significantly different, with more children with autism being categorized as more responsive than others for each of the sensory modalities and styles except Sensation Seeking. These findings are consistent with the results of the standardization of the TSP, as reported in the TSP manual (Dunn, 2002). One difference between the results of the current study and those of the standardization study, however, is that the distribution for the children with PDD-NOS was only different from the normal distribution on the auditory modality and Low Registration, whereas the standardization sample included all ASD s together. These findings, as well as the large effect sizes between autism and PDD-NOS, discussed 64

74 later, suggest that future studies examining sensory behaviors in ASD should examine behaviors separately for autism and PDD-NOS. It is important, also, to consider the lack of difference between autism and the typical population on Sensation Seeking. This lack of difference is surprising given that sensory seeking behaviors are often what is coded for the Unusual Sensory Interest item on the ADOS. For example, behaviors such as visual inspection, sniffing or licking nonfood items, and repetitive rubbing of certain textures would all be coded on the ADOS as sensory behaviors. While these behaviors appear to have a sensation seeking quality, they are not included on the TSP. Therefore, it is possible that the Sensation Seeking style on the TSP may not be tapping into the sensation seeking behaviors that individuals with autism engage in. Results also indicated that the differences between groups changed over time. The group means for each modality and style were examined at each of the age groups that are used in the TSP. It was found that generally as the children got older, the effect sizes (differences between the groups, taking into consideration standard deviation) became larger until the 25 to 30 month age group, after which time the effect sizes became smaller. The change in effect size followed this pattern for each of the sensory modalities and styles except, Sensation Seeking, which showed little change from the month group to the to the month group but then peaked at months and Sensory Sensitivity, which peaked at months and then declined. Across all modalities and styles, increases in effect size were due to both lower scores among the children with autism (indicating greater frequency of unusual sensory behaviors) and 65

75 high scores among the other three groups (indicating reduced frequency of unusual sensory behaviors). Among the modalities, auditory had the largest effect sizes, with effect sizes greater than 1.0 when comparing autism to each of the other comparison groups. When comparing autism to the typical group, the effect size was largest for the month age group (1.77). When autism was compared to the PDD-NOS group, the effect size reached 2.1 at the month age group, and reach 2.2 for the same age group when compared to NSDD. These results are consistent with the findings of Talay-Ongen & Wood (2000) who also found that the effect sizes for the difference between the autism and typical group were the largest when comparing them on the auditory modality. When examining effect sizes by style, the effect sizes were largest for Low Registration. In Chapter 2, it was reported that the majority of autism-specific items were auditory and Low Registration items, therefore the effect sizes were examined for auditory and Low Registration without these items (Figure 3.3). While these results indicate that the large effect sizes, with all items included, were partly driven by the autism-specific items, the effect sizes were still large for some of the comparisons. Specifically, at months the effect size was 1.13 for the difference between autism and typical on the auditory domain and 1.35 at months for the difference between autism and PDD-NOS. These results suggest that differences remain between the groups even without the inclusions of the autism-specific items. The effect sizes for the autism and typical groups at the first two age groups (13-18 and months) were larger and showed a greater change from month age-group to the month age-group than the effect sizes for autism and the other two 66

76 diagnostic groups. Unfortunately, the effect sizes between the group with autism and the typically developing group could only be examined for the month and month age groups, due to the small size of the typical sample after 24 months. It would be interesting for future research to examine if the effect sizes for the difference between the autism and typical groups followed the pattern seen for the NSDD and PDD-NOS groups or if the effect sizes continue to increase or plateau. The current study also provides information on demographics that are related to unusual sensory behaviors. Contrary to what was hypothesized, age did not have a significant effect on oral and tactile modalities, although it did have a significant effect on the auditory modality, the visual modality, Low Registration, and Sensation Seeking. For each of these domains, frequency of unusual sensory behaviors decreased with increasing age. The findings related to Sensory Seeking behaviors are consistent with the standardization of the TSP, which also found a decrease in sensory seeking responses with increasing age among the typically developing standardization sample (Dunn, 2002). The findings are also consistent with Baranek et al. (2006) who found decreases in unusual sensory responses with increasing mental age in a group of children with autism, PDD-NOS, & NSDD ages 5-80 months. For the auditory modality, the relationship between diagnosis and age was moderated by risk status, such that the siblings of children with autism actually showed increases in unusual sensory behavior over time, while nonsiblings had a decrease in sensory behaviors over time. While it is difficult to explain this interaction between risk status and age, one possibility may be differences in reporting between parents who already have one child with autism and those who do not. Those that are already familiar 67

77 with autism and the unusual sensory behaviors may be underreporting unusual sensory responses at very young ages because they are comparing the more subtle sensory responses of their young children to the more extreme sensory responses of their older children. Regardless, it appears that by 36 months, there is no difference between the two groups, suggesting that risk status, independent of diagnosis, is not related to unusual sensory behaviors as the children get older. After controlling for age, NVIQ, and risk status, diagnosis was found to be a significant predictor for auditory, tactile, Low Registration, and Sensation Avoiding. With all the items included, there was a significant difference between autism and the other three groups on each of the Low Registration and auditory domains. However, when the autism items were removed from auditory and Low Registration, the significant difference between autism and the other three groups remained for Low Registration but was no longer significant for autism and NSDD on the auditory modality. Likewise, autism was significantly different from the typical group and PDD-NOS on the tactile domain, but was not different than NSDD. Diagnosis also had a significant effect on unusual Sensation Avoiding behaviors, with autism being significantly different than PDD-NOS and NSDD, but not typical development. These results may help explain the inconsistent findings regarding the differences between autism and NSDD. Using an observational measure of sensory behaviors, Baranek, et al (2007) did not find differences in sensory aversion to a variety of sensory toys between children with autism and children with NSDD. However, using a parentreport questionnaire, she and her colleagues (2006) found a difference between children with autism and children with NSDD in terms of hyporesponsiveness but not 68

78 hyperresponsiveness. While the findings of the current study do not support the findings of these previous studies, given that the children with autism were found to differ from the children with NSDD on both domains that can be considered hypo- and hyperresponsivenss (Low Registration and Sensation Avoiding, respectively) they do suggest that differences between the two diagnostic groups may vary according to type of domain assessed and measure of assessment (parent-report versus observational). Examining the group differences on the ADOS sensory item and comparing these differences to the results of the TSP, also point to the influence of the method of assessment. While the effect sizes between the groups were generally the largest at the month age group on the TSP, this was precisely the time that the NSDD and PDD- NOS groups had ADOS sensory items scores that were most similar to the autism group. In other words, at months the NSDD and PDD-NOS groups were reported to have the lowest frequency of unusual sensory behaviors while at the same time they were observed to exhibit the most unusual sensory behaviors during the ADOS. These incongruent findings illustrate several important points. First, as mentioned above, the behaviors that are coded on the ADOS may not be behaviors that are addressed by the TSP. In order to increase the validity of parent-report of sensory behaviors, whether through future revisions of the TSP or other parent-report questionnaires, it would be important for future research to examine the behaviors that clinicians are observing and coding as sensory behaviors in order to ensure that these behaviors are also being included in these measures. Second, these results also suggest that it is difficult and perhaps futile to compare and interpret contrasting findings from different measures of sensory behaviors, as the behaviors that are included in the different measures may be 69

79 addressing very different constructs. Finally, the different findings of the ADOS and TSP raise the question of the accuracy of parent report. In fact, one study of sensory behaviors (Watling, Deitz, & White, 2001) reported that clinical observations of a few of the participants were not consistent with what the parents had reported. These authors suggested that parents may be under-reporting occurrences of their child s atypical sensory behavior, which would be consistent with the contrasting findings of the current study. 70

80 Table 3.1 Participant Demographics Current Toddlers/First Words Participants Most Recent Diagnosis Typical Autism PDD Nonspectrum DD N Age at First Assessment 14.5 (5.4) 22.7 (7.2) 23.1 (8.8) 22.5 (8.8) Gender Male 11 (64.7%) 22 (88%) 20 (80.0%) 18 (81.8%) Female 6 (35.3%) 3 (12%) 5 (20.0%) 4 (18.2%) Risk Status Sibling 10 (58.8%) 9 (36.0%) 9 (36.0%) 7 (31.8%) Not Sibling 7 (41.2%) 16 (64.0%) 16 (64.0%) 15 (68.2%) Number of Visits 1 vis it only 4 (23.5%) 7 (28%) 6 (24.0%) 10 (45.5%) 2 visits only 2 (11.8%) 8 (32%) 6 (24.0%) 5 (22.7%) More than 2 visits 11 (64.7%) 10 (40%) 13 (52.0%) 7 (31.8%) Cognitive abilities (initial visit) Verbal IQ* 99.5 (14.5) 51.0 (23.0) 87.5 (19.5) 65.8 (26.6) Nonverbal IQ* (15.5) 86.6 (18.6) 97.6 (17.2) 80.9 (24.4) Full Scale IQ* (10.0) 70.2 (16.4) 92.0 (13.7) 74.4 (24.5) *Average =

81 Table 3.2 Sample size at Each Time Point Typical Autism PDD NSDD 7 to 12 months 12 (7) 4 (2) 0 (0) 4 (2) 13 to 18 months 48 (10) 20 (9) 20 (8) 20 (7) 19 to 24 months 20 ( 0) 24 (4) 26 (6) 15 (3) 25 to 30 months 3 (0) 14 (5) 26 (4) 18 (3) 31 to 36 months 1 (0) 17 (5) 24 (6) 12 (6) 72

82 Table 3.3 Chi-square Analyses Comparing Distribution of Children with Autism and PDD-NOS to the Standard Distribution of Children Without Disabilities on the TSP Modality Autism n = 25 PDD-NOS n = 25 Auditory X 2 (2) = 76.2*** X 2 (2) = 26.4*** Visual X 2 (2) = 24.3*** X 2 (2) = 0.54 Tactile 2 X (2) = 11.8*** 2 X (2) = 4.3 Vestibular X 2 (2) = 14.7*** X 2 (2) = 3.2 Oral X 2 (2) = 13.6*** X 2 (2) = 4.7 Low Registration X 2 (1) = 77.1*** X 2 (2) = 45.9*** Sensation Seeking X 2 (2) = 1.2 X 2 (2) = 2.1 Sensory Sensitivity X 2 (2) = 29.8*** X 2 (2) = 3.3 Sensation Avoiding X 2 (2) = 67.0*** X 2 (2) = 8.5 ***p<.001 Style 73

83 Table 3.4: Type III Test of Fixed Effects for Mode ls Predicting Variability in TSP Modalities 74 Auditory Visual Tactile Oral Diagnosis F(3, 84.8) = 5.7** F(3, 79.0) = 2.5 F(3, 84. 7) = 4.4** F (3, 66.9) = 1.6 NVIQ F(1, 131.4) = 0.6 F(1, 121.4) = 1.3 F(1, 130.4) = 1.2 F(1, 107.6) = 7.9** Risk Status F(1, 80.3) = 5.5 F(1, 79.8) = 4.3 F(1, 83.7) = 3.5 F( 1, 69.6) = 0.6 Age F(1, 48.0) = 3.3 F(1, 50.5) = 15.1** F(1, 72.2) = 3.2 F( 1, 66.0) = 4.9 Age*Age F(1, 94.4) = 9.0** F(1, 79.2) = 8.0** F( 1, 67. 1) = 3.5 Age*Diagnosis F(3, 41.4) = 3.3 Age*NVIQ Age*Risk Status F(1, 53.0) = 8.5** F(1, 52.4) = 6.2 Risk Status*Diagnosis Risk Status*NVIQ NVIQ*Diagnosis **p <.01 74

84 Table 3.5: Coefficient Estimates for Age by Diagnosis Interaction for the Visual Modality Variable Estimate SE Df T Sig. Intercept Linear Age Quadratic Age Age*Dx (Dx = Typical) Age*Dx (Dx = PDD-NOS) Age* Dx (Dx = NSDD) Age*Dx (Dx = Autism) Referent 75

85 Table 3.6: Models with the Best Fit for Predicting TSP Sensory Styles 76 Low Registration Sensation Seeking Sensory Sensitivity Sensation Avoiding Diagnosis F(3, 86.3) = 8.6*** F(3, 84.8) = 1.4 F(3, 85.1) = 2.4 F(3, 82.1) = 4.3** NVIQ F(1, 124.1) = 1.6 F(1, 135.0) = 4.9 F(1, 124.6) = 0.6 F(1, 119.7) = 0.4 Risk F(1, 79.0) = 2.9 F(1, 83.0) = 3.0 F(1, 147.2) = 4.8 F(1, 85.2) = 0.1 Age F(1, 35.6) = 4.6 F(1, 147.1) = 16.5*** F(1, 132.8) = 0.0 F(1, 132.8) = 1.6 Age*Age F(1, 88.1) = 11.8** Age*Dx Age*NVIQ Age*Risk F(1, 40.6) = 6.2 Risk*Dx F(3, 78.6) = 2.1 Risk*NVIQ F(1, 137.5) = 5.5 NVIQ*Dx ** p <.01 ***p <

86 Figure 3.1: Change in effect size over time for each sensory style: Difference between autism and comparison groups Low Registration Sensation Seeking Effect Size Autism:PDD-NOS Autism:NSDD Autism:Typical ize Effect S Autism:PDD-NOS Autism:NSDD Autism:Typical months months months months Age Groups months months months months Age Groups 77 Sensory Sensitivity Sensation Avoiding Effect Size Autism:PDD-NOS Autism:NSDD Autism:Typical Effect Size Autism:PDD-NOS Autism:NSDD Autism:Typical months months months months months months months months Age Groups Age Groups 77

87 Figure 3.2: Change in effect size over time for each sensory modality*: Difference between autism and comparison groups Effect Size Auditory months months months months Age Groups Autism:PDD-NOS Autism:NSDD Autism:Typical Effect Size Visual months months months months Age Groups Autism:PDD-NOS Autism:NSDD Autism:Typical Tactile Oral 78 Effect Size Autism:PDD-NOS Autism:NSDD Autism:Typical Effect Size Autism:PDD-NOS Autism:NSDD Autism:Typical months months months months months months months months Age Groups Age Groups * Vestibular was not included due to the low internal consistency found for this modality (see Chapter 2) 78

88 Figure 3.3: Comparison of the Effect sizes for Auditory and Low Registration: All items included (top) and effect sizes without overlapping or autism-specific items (bottom). Effect Size Auditory Autism:PDD-NOS Autism:NSDD Autism:Typical Effect Size Low Registration Autism:PDD-NOS Autism:NSDD Autism:Typical months months months months months months months months Age Groups Age Groups 79 Effect Size Auditory Without Autism-specific or Overlapping Low Registration Items Autism:PDD-NOS Autism:NSDD Autism:Typical months months months months Age Groups Effect Size Low Registration Without Autism-specific or Auditory Items Autism:PDD-NOS Autism:NSDD Autism:Typical months months months months Age Groups 79

89 Figure 3.4: Scores on Auditory Modality by Risk Status Over Time 80