Towards better characterization of restricted and unusual interests in Youth with Autism

Abstract

Despite their high prevalence and clinical importance in autism, unusual and restricted interests remain under-researched and poorly understood. This study aimed to characterize the frequency and type of interests in autism by coding caregivers’ open-ended responses in a sample of 237 autistic children and adolescents (M_age= 8.27 years, SD_age= 4.07; range: 2.08–18.25 years). It further aimed to explore the effects of age, sex, cognitive functioning and social and communication deficits on the number and type of interests. We found that 75% of autistic youth had at least one interest and that 50% of those children showed two or more different interests. The most frequent interests were sensory-based (43%), with a majority of these interests relating to the visual modality. Interest within vehicles/transportation, fictional characters, TV/DVD/movies, computers and video games, constructive, mechanical objects, animals and plants, and attachment to specific objects were also prevalent. Logistic regression showed that being male, having a co-occurring intellectual disability and having more severe social and communication impairments were associated with a higher probability of having one or more restricted interests. Sex was significantly associated with the type (χ²= 37.52, Phi= .37, p= .021) of restricted interests, with females showing a significantly higher percentage of creative interests and males significantly higher percentage of interest in characters, vehicles/transportation, computers/video games, mechanical objects and constructive interests. Theoretical and measurement implications are discussed.

Background

A variety of interests and fascinations with specific objects, topics, and actions unusual in terms of their focus, intensity and/or object/theme constitute an important group of symptoms in autism. Early anecdotal reports and small-scale studies have highlighted potentially beneficial effects of certain interests for autistic individuals (Attwood, 2003; Grandin & Duffy, 2004; Mercier et al., 2000). Subsequent studies have confirmed these observations and provided further evidence about the inherent reward value and potential adaptive value of some of the restricted interests, for instance, as facilitators for successful educational and vocational outcomes due to the acquired expertize (Grove, Hoekstra, Wierda, & Begeer, 2018; Patten Koenig & Hough Williams, 2017; also see Harrop et al., 2019). However, robust evidence suggests that certain interests may negatively affect both autistic individuals and their families (Anthony et al., 2013; Boyd et al., 2007; Klin et al., 2007; Mercier et al., 2000; Pierce & Courchesne, 2001; South et al., 2005; Spiker et al., 2012; Turner-Brown et al., 2011). Despite the high prevalence, unusual and restricted interests (RI) remain under-researched (Cho et al., 2017; Carter et al., 2020).

One of the main obstacles to advancing our understanding of RI has been the lack of consensus on their taxonomy and measurement. Previous literature has relied on a range of attentional paradigms, suggesting that autistic individuals show a higher degree of exploration and longer fixations to stimuli RI than to social stimuli (Sasson et al., 2011; Sasson & Touchstone, 2014; Unruh et al., 2016; but see Parsons et al., 2017). Despite affording ecological validity and scalability, the majority of currently available autism severity measures and dedicated repetitive behaviors (RRB) measures contain only a few items that directly relate to RI. Although interview instruments such as the Repetitive Behavior Interview (Turner, 1997) provide an in-depth RI assessment, these assessments are labor-intensive, which significantly limits their utility within research settings.

Studies that have utilized dedicated instruments such as the Yale Survey of Special Interests (South et al., 1999), the Interests Scale (Bodfish, 2004), and the Special Interests Motivation Scale (Grove, Roth, & Hoekstra, 2016) have confirmed a high prevalence of RI in autism, ranging from 75% to 88%, and demonstrated that majority of autistic individuals show more than one type of RI (Anthony et al., 2013; Grove et al., 2018; South et al., 2007). Although noted studies differ in terms of RI taxonomy, interests related to specific objects and non-social domains, such as mechanics, were frequent (Klin et al., 2007; South et al., 2007; Turner-Brown et al., 2011). These studies have also emphasized that RI is a complex and heterogeneous domain that encompasses at least (i) interests that while typical in terms of the topic (e.g., interests in mathematics or arts, TV shows, video games) tend to be overly intense in terms of focus, level of engagement and/or inflexibility, and (ii) interests that are atypical in terms of the topic (e.g., road signs, numbers, dates or train tables). Further, these studies have highlighted the prominence of sensory-related interests and fascinations (e.g., bright or vivid colors and watching spinning objects) and behaviors related to seeking sensory stimulation (e.g., sensing textures) (Anthony et al., 2013; Baron-Cohen & Wheelwright, 1999; Klin et al., 2007). Although fascination with specific objects, parts of objects and sensory stimuli are traditionally considered as either part of repetitive sensory-motor behaviors RRB subdomain (Leekam et al., 2011) or aspect of the sensory features RRB subdomain (American Psychiatric Association, 2013), the empirical classification of these behaviors and fascinations is currently unclear. More specifically, while several factor analytic studies found that items capturing these behaviors load onto the repetitive (sensory) motor behaviors factor (Bishop et al., 2013; Grove et al., 2021; Lidstone et al., 2014; Richler et al., 2010), other studies have reported that these behaviors load together with restricted interests-related items (Frazier & Hardan, 2017; Hooker et al., 2019; Mirenda et al., 2010; Uljarević et al., 2021) or form a separate factor, with somewhat distinct correlates when compared to the sensory-motor RRB domain. Therefore, given somewhat unclear conceptual boundaries and content of the RI domain, it is crucial to comprehensively capture the full RI spectrum and understand their correlates with demographic, cognitive and clinical variables.

While reviewed studies have made important contributions to the field, these studies were limited by sample size and/or almost exclusive focus on individuals with IQ in the normative range. Therefore, our study aimed to further increase our understanding of the unusual and restricted interests domain by exploring their specific presentation in a relatively large, well-characterized sample of autistic youth that spanned a broad spectrum of cognitive functioning. Previous studies did not comprehensively explore the association between specific interests with demographic characteristics and clinical traits. Therefore, in addition to describing the full spectrum of unusual and restricted interests, we aimed to understand potential continuities and discontinuities between distinct aspects of RI domain by exploring their association with age, sex, cognitive functioning and social and communication deficits.

Methods

Participants

The sample included 237 autistic youth (M_age= 8.27 years, SD_age= 4.07; range: 2.08–18.25 years; 44 females) recruited as a part of the Western Australian Autism Biological Registry (WAABR) study. All children received a clinical diagnosis of ASD based on the assessment by a multidisciplinary team comprising a paediatrician, clinical psychologist, and speech pathologist. The diagnosis was further verified using the Autism Diagnostic Observation Schedule (ADOS; Lord et al., 2000) administered by research-reliable assessors. Fifty-seven participants had a comorbid diagnosis of intellectual disability (ID). Please see Table 1 provides detailed demographic and descriptive information.

Table 1.

Demographic and descriptive information

Characteristics	Mean (SD) (Except where indicated otherwise)
Age	8.27 years (4.07)
Sex (Number [%])
Female	44 (18.56)
Male	193 (81.44)
Intellectual Disability (Number [%])
No	180 (76)
Yes	57 (24)
ADOS Module (Number [%])
Module 1	61 (25.74)
Module 2	85 (35.86)
Module 3	78 (32.91)
Module 4	13 (5.49)
ADOS Total CSS Score	6.20 (2.31)
ADOS RRB CSS Score	5.77 (1.95)
ADOS SA CSS Score	6.67 (2.34)
SRS-2 T Total Score	89.54 (2.01)

Note: ADOS= Autism Diagnostic Observation Schedule; CSS= Calibrated Severity Score; RRB= Restricted and Repetitive Behaviors; SA= Social Affect; SRS-2= Social Responsiveness Scale.

Procedures and Measures

The study was approved by the Princess Margaret Hospital Human Research Ethics Committee. Parents/guardians provided written informed consent. As a part of the WAABR study, all of the participating families received a questionnaire pack that contained (i) a case-history questionnaire relating to the mother’s pregnancy and the early development of their autistic child, (ii) several questionnaire measures of children’s ASD symptoms (Social Responsiveness Scale-2 [SRS-2; Constantino & Gruber, 2012]), communication (Children’s Communication Checklist-2 [CCC-2; Bishop, 2003]) and sleep problems (Children’s Sleep Habits Questionnaire [CSHQ; Owens et al., 2000]), and (iii) the Autism Spectrum Quotient (AQ; Baron-Cohen et al., 2001) as a measure of parental autistic traits. Children’s restricted interests were coded based on an open-ended question which asked: “Has your child ever shown a more intense interest than other children?”. If parents answered “Yes”, they were asked to provide an additional description. All responses were provided in the written form. Descriptions were coded following the procedure described below. In order to characterize the severity of social and communication impairments, we used the social communication/interaction (SCI) subscale T score of the SRS-2. The SRS-2 is a 65-item measure designed to index severity in social impairments and the presence of RRB. Each item is rated on a 4-point scale (from 1 = Not True to 4 = Almost Always True). Higher scores indicate higher trait severity/atypicality.

Coding of Restricted Interests

The coding approach utilized in this study followed the approach used in our previous work (Lau et al., 2020) which coded common and autism-related anxiety experiences in ASD from optional open-ended Spence Children’s Anxiety Scale-Parent item (Spence, 1999). We firstly categorized parental descriptions into (a) insufficient/unclear/incomplete responses and (b) interpretable responses. Responses where parents provided single words (e.g., “toys”), only a few words (e.g., “dog (stuffed toy)”‘; “anything plastic”) or more detailed descriptions that nevertheless did not contain sufficient information (e.g., “_can spend hours playing with a single toy for hours on end eg. tennis ball”; “Used to get focused on one item and obsess with it for a couple of weeks” or “Used to be obsessed with objects, oblivious when someone entered the room”) were coded as insufficient/unclear. No formally established or universally agreed-upon RI categories currently exist. Therefore, the systematic search of the literature was conducted and the first author derived categories that have consistently surfaced across (a) the existing general RRB measures such as the Repetitive Behavior Scale-Revised (RBS-R; Bodfish et al., 2000), the Repetitive Behavior Interview (Turner, 1997), the Repetitive Behavior Questionnaire-2 (RBQ-2; Leekam et al., 2007), (b) dedicated restricted interests instruments including the Interests Scale (Bodfish, 2004), the Yale Survey and Interview of Special Interests (Klin & Volkmar, 1996; South, Klin, & Ozonoff, 1999), the Survey of Favorite Interests and Activities (Smerbeck, 2019), and the Special Interests Motivation Scale (Grove et al., 2016, and (c) studies that have coded narrative descriptions of RI (e.g., Baron-Cohen & Wheelwright, 1999; Cho et al., 2017; DeLoache et al., 2007). Given the noted RI heterogeneity and complexity, the coding scheme aimed to capture a wide range of interests. Derived categories were piloted on 15% of the sample to explore their utility for organizing caregivers’ descriptions and explore whether any additional categories would emerge. Given that no additional RI categories emerged, we proceeded to categorize the remaining responses. RI categories/themes are listed in Table 2. Where more than one RI was listed, each was coded into a separate theme/category. Given that parental descriptions were generally brief, a conservative approach was adopted and responses were coded based on available information only, without inferring the mechanisms or nature of RI. The third and fourth authors independently coded all available responses. The first author then reviewed all codes. Where discrepancy in coding occurred, the disagreement was reviewed and a consensus was reached. Once all responses were categorized, we conducted quantitative counting of the number and type of restricted interests across the sample and for each individual participant.

Table 2.

Restricted interests coded from parental narratives

Main Restricted Interests Theme	Subthemes if applicable	Responses
		n	%
Sensory	Total Sensory	72/165	43.7
	Spinning objects wheels, fans, revolving doors	39/72	54.17
	Visual shiny objects, lights, light switches	16/72	22.22
	Materials/texture sand, water, soft objects	14/72	19.44
	Manipulating objects ball bouncing, door opening/closing	11/72	15.28
	Olfactory	1/72	1.39
Vehicles/Transportation	Total Vehicles/Transportation	31/165	18.79
	Cars and trucks	19/31	61.29
	Trains	16/31	51.61
	Planes and rockets	2/31	6.45
	Any mode of transportation	1/31	3.22
Characters (from movies, books, cartoons, etc.)	Total Characters	24/165	14.54
	Thomas the Tank Engine	12/24	50.0
	Marvel Universe Characters	3/24	12.5
	Ben 10	2/24	8.33
	Dora the Explorer	2/24	8.33
	Other 1 each of Wiggles, Teletubbies, Dr. Who, Angry Birds, Star Wars	5/24	20.83
TV/DVD/Movies	Total	22/165	13.33
Idiosyncratic	Total Idiosyncratic	19/165	11.51
	Cleaning related	7/19	36.84
	Rocks	2/19	10.53
	Other e.g. keys, toilets, shop signs, license plates	10/19	52.63
Computers/Tablets/Video Games	Total	16/165	9.70
Constructive (Building/putting things together)	Total Constructive	14/165	9.09
	Lego	9/14	64.28
	Puzzles	5/14	35.72
Mechanical Objects	Total Mechanical Objects	14/165	8.48
	Vacuum Cleaners	7/14	50
	Electric Toothbrushes	2/14	14.28
	Other 1 each of windmills, dryer, escalators, unspecified/a range of mechanical objects	5/14	35.72
Animals and Plants	Total Animals and Plants	12/165	7.27
	Dinosaurs	5/12	41.67
	Cats and dogs	3/12	25
	Insects	2/12	16.67
	Horses	2/12	16.67
	Sharks	2/12	16.67
	Plants	2/12	16.67
	Other 1 each of birds and crocodiles	2/12	16.67
Restricted, Stereotypical Play*	Total Play	11/165	6.67
	Toys	7/11	63.64
	A wide range of objects	4/11	36.36
Attachment	Total Attachment	10/165	6.06
	Specific Toy	7/10	70
	Blanket	2/10	20
	Blocks	1/10	10
Creative	Total Creative	5/165	3.03
	Drawing and painting	4/5	80
	Acting	1/5	20
Books and reading	Total	4/165	2.42
Time/date	Total	4/165	2.42
Science and Technology	Total	4/165	2.42
Other	Total Other	15/165	9.09
	Miscellaneous flags, guns, playing cards	4/15	26.67
	Music	3/15	20
	Letters and numbers	2/15	13.33
	Collecting	2/15	13.33
	Food	1/15	6.67
	Sports	1/15	6.67
	Classifying/Organizing Information	1/15	6.67
	Facts and learning	1/15	6.67

Note:

^*Stereotypical play was not a part of the original coding scheme, however, eleven parents reported that their children engaged in repetitive play by arranging objects in stereotypical manner (e.g. straight line).

Statistical Analyses

Logistic regression was conducted to explore the role of chronological age, sex, cognitive level (presence/absence of ID) and severity of social and communication impairments (indexed by the SRS-2 SCI T score) in predicting the presence/absence of RI.

Community involvement

There is no community involvement in this study.

Results

Of the 237 parents who responded to the question about whether their children exhibited restricted interests, 58 (24.47%) answered “No” and 179 (75.53%) “Yes”. Fourteen parents were removed due to insufficient/unclear responses. Eighty-eight parents described that their child engaged in one, 53 in two, 16 in three, 6 in four and two in five different interest categories/themes. As a comparison, over half (52.6%) of children and adolescents were rated as showing no unusual and restricted interest on the relevant ADOS item. Sensory-related interests (43.66% of all responses), interests in vehicles/transportation (18.79%), characters (14.54%), TV/DVD/movies (13.33%), and idiosyncratic (11.51%) categories emerged as most frequent. Although less frequent, interests within computers/tablets/video games (9.70%), constructive (e.g., puzzles, legos) (9.09%), mechanical objects (8.48%), animals and plants (7.27%) and attachment to specific objects (6.06%) were endorsed by at least 5% of the sample. Table 2 provides detailed information in terms of coded themes/categories.

Logistic regression was utilized to characterize the contribution of age, sex, the presence of co-occurring intellectual disability and social and communication impairments in predicting the presence of RI. Logistic regression was conducted to explore the role of chronological age, sex, cognitive level (presence/absence of ID) and severity of social and communication impairments (indexed by the SRS-2 SCI T score) in predicting the presence/absence of RI. The model was significant (χ²= 22.01, p< .001), explaining 19.3% of variance (Nagelkerke R²), correctly classifying 75.3% of responses. Male sex, ID and higher SRS-2 SCI T scores were associated with a higher likelihood of the presence of interests (Table 3).

Table 3.

Logistic regression predicting the presence/absence of restricted interests

Variable	B	SE	Wald Test	p	Exp(B)	95% CI
CA	−.012	.05	.056	.81	.988	.89, 1.09
Sex	−1.22	.458	7.15	.007	.29	.12, .72
Cognitive Level	.89	.59	2.31	.13	2.45	.77, 7.79
SRS-2 SCI T score	.09	.03	9.45	.002	1.09	1.03, 1.16

Note: CA: Chronological age; SRS-2: Social Responsiveness Scale; SCI: Social and Communicative Impairments

Follow-up models within a sample of 165 participants with at least one RI showed that (i) ID was not associated with either the number (χ²= 1.25, Phi= .09, p= .87) or the type (χ²= 20.60, Phi= .27, p= .55) of interests, (ii) sex was not associated with a number (χ²= 2.72, Phi= .13, p= .60) but it was with a type (χ²= 37.52, Phi= .37, p= .021) of interests, with females showing a significantly higher percentage of creative interests and males significantly higher percentage of interest in characters, vehicles/transportation, computers/video games, mechanical objects and constructive interests. Age and SRS-2 SCI T scores were not associated with either the number or the type of interests. Given that previous studies exploring the relationship between sensory-based interests with age, IQ and the severity of social and communication difficulties have reported inconsistent findings (South et al., 2005; Uljarević et al., 2021) a supplemental analysis was conducted to address this question. Supplemental analysis dividing participants based on the presence/absence of sensory-based interest showed that age (F= .49, p= .49, ηp²= .003), ID (χ²= .33, Phi= .045, p= .56) or SRS-2 SCI T score (F= 1.55, p= .22, ηp²= .013) were not associated with the presence of sensory-based interests.

Discussion

In this study, we sought to better characterize the frequency and type of unusual and restricted interests (RI) by coding caregivers’ open-ended responses in a relatively large sample of autistic youth. Consistent with previous studies (Anthony et al., 2013; Grove et al., 2016; Turner-Brown et al., 2011), we found that 75% of our sample had at least one RI, with half of those children showing two or more different interests. The most commonly identified interests in this study were broadly consistent with previous investigations reporting a high prevalence of sensory-related interests (Baron-Cohen & Wheelwright, 1999; Klin et al., 2007), interests in fictional characters and TV/movies (Klin et al., 2007; South et al., 2005), computers and video games (Anthony et al., 2013), biology (South et al., 2005; Turner-Brown et al., 2011) and object attachment (Baron-Cohen & Wheelwright, 1999; Turner-Brown et al., 2011). Although previous investigations that have relied on checklist format observed a high frequency of interests related to facts/verbal memory and learning (Anthony et al., 2013; Klin et al., 2007; South et al., 2005), this was not the case in our study. Specific coding of open-ended parental responses in our study could have contributed to inconsistent findings given that parents might have focused on sensory- and object-related interests as they tend to be perceived as context inappropriate and/or most severe and limiting rather than to learning and amassing information, which might be perceived as more adaptive.

Being male, having a co-occurring ID and more severe social and communication impairments were associated with a higher probability of having at least one interest. Evidence of a positive relationship with more severe social and communication impairments reported here provides support to previous empirical findings (Boyd et al., 2007; Mercier et al., 2000; Pierce & Courchesne, 2001). Although the nature of the relationship between social and communication impairments and restricted interests remains poorly understood, it has been suggested that early emerging atypical reward processing prioritizing non-social over social stimuli might explain this link (Carter et al., 2020). In addition to associations reported here and in other studies, this hypothesis is also supported by studies demonstrating longer fixations to stimuli related to restricted interests than to social stimuli (Sasson et al., 2011; Sasson & Touchstone, 2014; Unruh et al., 2016; but see Parsons et al., 2017) and neuroimaging studies reporting structural and functional alterations in brain regions involved in social reward processing (Clements et al., 2018). However, despite noted evidence, it is important to highlight the fact that it is currently unclear whether the association between social deficits and restricted interests might be due to atypicality in social (Carter et al., 2020) or in general (Clements et al., 2019) reward processing. In addition to the reward processing theory of restricted interests, it is also possible that these behaviors disrupt social development due to pervasiveness and specific focus (Atwood, 1998; Koegel, et al., 1974; Klin et al., 2007). It is also important to note that several studies have reported that autistic children and adults experience certain RI as highly rewarding and that they might play an adaptive role, including facilitating interactions, educational and vocational outcomes (Clark & Adams, 2020; Grove et al., 2018; Patten et al., 2017). Longitudinal studies are needed to clarify the direction and nature of the relationship between social functioning and restricted interests.

Investigations that have measured RI using more general autism severity or RRB measures have reported a positive (Bishop et al., 2013), negative (Uljarević et al., 2020) and the lack of a significant relationship with IQ (Lam et al., 2008), however, as noted, some of these studies utilized subscales that combined interests atypical in terms of focus, theme and/or intensity, which could have contributed to the inconsistent findings. There are several possible interpretations of the association between the presence of restricted interests and IQ status. Several studies have shown a strong association between IQ and cognitive control at both behavioral and genetic levels (Engelhardt et al., 2016; Friedman et al., 2008). Given that previous studies have reported an association between restricted interests and deficits in cognitive control (Anthony et al., 2013; Faja & Darling, 2019), IQ status may serve as a proxy of (lower) cognitive functioning which would explain the association with restricted interests reported here. Further, as noted, due to their consuming nature and overly narrow focus, restricted interests can negatively impact the general development and learning opportunities (Atwood, 1998; Koegel, et al., 1974; Klin et al., 2007). However, given the complexity of the restricted interests domain, it is very likely that distinct mechanisms might underpin associations reported in our investigation. Therefore, further research in larger samples utilizing a multi-method approach is needed to provide a better understanding of the relationship between cognitive functioning and RI and contextualize findings reported in our study.

While the sample size utilized in this study was larger than in previous studies, the number of female participants was still limited. Further, it was possible to only code what parents have reported without the ability to further probe and characterize reported interests. Relatedly, it is not possible to conclude that children whose parents did not provide an answer did not have RI. However, the percentage of children that did not show RI in this study is in line with several previous studies (Anthony et al., 2013; Grove et al., 2016; Turner-Brown et al., 2011). Further, 52.6% of children and adolescents were rated as showing no RI on the relevant ADOS item. The nature of the question could have primed parents to report only interests and related behaviors that were either most easily observable and/or created the most difficulties, therefore undersampling interests perceived as less impactful, more typical and/or adaptive. Further, this study relied on parental responses. It is possible that children would interpret the questions differently and focus on distinct aspects of the RI domain, most likely emphasizing more conceptual RI perceived as particularly rewarding and something that children are particularly good at (Clark & Adams, 2020). Therefore, it will be crucial for future studies to incorporate multi-informant design to provide a more comprehensive assessment and identify aspects of this complex RRB subdomain that might be differently perceived and reported by different informants.

Despite noted limitations, this study was able to extend current knowledge about RI and characterize the relationship between the presence, number and type of RI with sex, cognitive functioning and social and communication symptoms—an aspect that was not addressed in detail in previous investigations. Crucially, this study offers important information for future measurement development. Although several existing instruments provide a detailed RI assessment, these measures are lengthy and not feasible to include in more general investigations that do not specifically focus on this particular domain. Although the current study was limited by relying on coding of parental responses to a single open-ended question, it demonstrated that this approach can yield rich information about a wide spectrum of RI in a time-efficient manner. It will be important to further develop the approach used in our study by designing additional questions to capture the impact and function of different interests and developing Natural Language Processing algorithms to allow automatic extraction of this information from parental responses. These questions could be embedded within a more general RRB instrument that would also incorporate a checklist to quantify key subtypes of restricted interest. This new combined assessment method would allow scalable and time-efficient RI capture and enable further understanding of this important yet poorly characterized symptom domain.