The Adult Repetitive Behaviours Questionnaire-2 (RBQ-2A): A Self-Report Measure of Restricted and Repetitive Behaviours

Restricted and repetitive behaviours (RRBs) form one of the core diagnostic criteria for autism spectrum disorder (ASD; American Psychiatric Association 2013; World Health Organization 1993). This class of behaviours, driven by a desire for sameness and dislike of change (Kanner 1943), includes a wide range of motor and sensory behaviours and restricted activities that are highly frequent in their repetition and invariant in their manifestation. These behaviours are also found in neurotypical (NT) individuals and those with other developmental disorders and neuropsychological conditions (for reviews see Langen et al. 2011; Leekam et al. 2011).

Caregiver interviews and questionnaires are the most frequently used measures of RRBs. Observation measures, while effective for measuring motor and sensory behaviours, may be less sensitive for measuring less frequent restricted behaviours (e.g., Harrop et al. 2014; Honey et al. 2012). Factor analytic studies of RRBs using caregiver interviews and questionnaires have identified two sub-groups; one comprising repetitive sensory and motor behaviours such as hand flapping and rocking (RSMB), and the other comprising more abstract behaviours such as routines and circumscribed interests, which are collectively referred to as insistence on sameness (IS). This binary grouping has been found in as many as eleven previous studies of individuals with ASD (e.g., Bishop et al. 2006; Bishop et al. 2013; Cuccaro et al. 2003; Georgiades et al. 2010; Lidstone et al. 2014; Mooney et al. 2009; Papageorgiou et al. 2008; Richler et al. 2007; Richler et al. 2010; Shao et al. 2003; Szatmari et al. 2006) and in studies of NT children (e.g., Evans et al. 1997; Leekam et al. 2007b).

However, other studies have identified alternative solutions ranging from three to five different factors (e.g., Bishop et al. 2013; Honey et al. 2008; Lam and Aman 2007; Lam et al. 2008; Mirenda et al. 2010). Such differences may be due to the use of RRB measures that are different in terms of their scope and format, such as the Repetitive Behaviour Scale-Revised (RBS-R; Bodfish et al. 1999) and the Autism Diagnostic Interview-Revised (ADI-R; Lord et al. 1994). For example, the RBS-R is a questionnaire comprising questions about self-injurious behaviours, which may form a separate factor (e.g., Bishop et al. 2013; Mirenda et al. 2010). On the other hand, the ADI-R is an interview that has been reported to undersample RRBs (Lam et al. 2008).

In contrast to an extensive literature on RRBs in children with ASD and in children with neurotypical development, there is limited research on RRBs in adulthood. Some factor analysis studies of RRBs in ASD have included adults in their samples (e.g., Cuccaro et al. 2003; Georgiades et al. 2010; Lam et al. 2008; Papageorgiou et al. 2008; Shao et al. 2003). However, conclusions from these studies about RRBs in adults are limited; either because the samples span a limited age range, or because the adult samples were not separated from the child samples in the analysis. A minority of studies of ASD that have directly compared RRB symptoms in adults with those in children have found lower levels of RRBs in adults than in children (Esbensen et al. 2009; Fecteau et al. 2003; Piven et al. 1995). This pattern remains the same across the subtypes of RRBs and is consistent across gender and intellectual disability (ID), with the exception that motor stereotypies do not reduce as much over time in adults with comorbid ID (Esbensen et al. 2009). These findings indicate that RRBs in adulthood may present differently than in childhood, which has implications for clinical practice and research.

Caregiver-report methods such as the RBS-R are suitable for use with adults. However, certain items may not be applicable, such as items related to play behaviours and toys. Furthermore, once an adult leaves home caregivers may not be able to report as accurately on their behaviours. Currently there are few self-report measures of RRBs available that are suitable for adults. While there is a self-report interview of obsessive-compulsive symptoms, the Yale-Brown Obsessive-Compulsive Scale (YBOCS; Goodman et al. 1989), to our knowledge there are no published self-report measures for the full range of RRBs relevant to the diagnosis of ASD. RRBs and obsessive-compulsive behaviours overlap but they do not capture the same construct. For example, the YBOCS includes questions about intrusive imagery, which is not a feature of RRBs. The Autism-Spectrum Quotient (AQ; Baron-Cohen et al. 2001) is a measure of autistic traits that includes items related to RRBs (e.g., It does not upset me if my daily routine is disturbed). However, factor analyses suggest that the AQ does not provide an adequate or reliable assessment of RRBs in NT adults (Kloosterman et al. 2011; Lau et al. 2013).

Therefore, in the current study we adapted and tested a parent-report questionnaire to provide the first self-report RRB questionnaire suitable for adults with ASD. Following the pattern of previous research on RRBs in both NT and ASD children, we assessed the questionnaire initially in NT adults and then applied it to an ASD sample. In contrast to research on RRBs in NT children, research on the full range of RRBs in NT adults is sparse and limited to particular behaviours such as pre-sleep rituals and transition objects (Markt and Johnson 1993). Therefore, new evidence on self-reported RRBs in NT individuals will enable comparison with evidence from adults with ASD, providing further insight into the presentation of these behaviours in adults both with and without ASD. Beyond comparison purposes, it would be useful to understand the pattern of RRBs in an adult NT population. Furthermore, given the increasing need by clinicians for briefer and more streamlined methods for diagnosis, a self-report format for eliciting information on RRBs in able adults has application as a supplement to add information to other diagnostic methods.

The Repetitive Behaviour Questionnaire-2 (RBQ-2; Leekam et al. 2007b) is a twenty item questionnaire, with items directly derived from a standardised clinical interview tool, the Diagnostic Interview for Social and Communication Disorders (DISCO; Wing et al. 2002). The DISCO has good inter-rater reliability and discriminant validity (Leekam et al. 2002; Maljaars et al. 2012; Nygren et al. 2009) and shows strong agreement with outputs from the ADI-R (Nygren et al. 2009) and Autism Diagnostic Observation Schedule (ADOS; Maljaars et al. 2012). Items from the DISCO and converging items from a semi-structured interview, the Repetitive Behaviours Interview (RBI; Turner 1996, unpublished doctoral thesis), were adapted into a questionnaire measure, the RBQ-2. The RBQ-2 includes 20 RRB items; 13 identical items taken from both interviews, five items unique to the DISCO and two unique to the RBI (Leekam et al. 2007b).

The RBQ-2 was originally tested in a large sample (N = 679) of NT two-year-olds (Leekam et al. 2007b). There was satisfactory endorsement of all RRBs, and exploratory factor analysis supported both a four- and two-factor solution. The four-factor solution comprised: repetitive motor movements, adherence to routine, restricted interests, and unusual sensory interests. The two-factor solution comprised RSMB, which corresponded to repetitive motor movements and unusual sensory interests, and IS, which corresponded to adherence to routine and restricted interests. The reliability and validity of the RBQ-2 has since been further supported in NT 15-month-olds (Arnott et al. 2010). Finally, the RBQ-2 has also been assessed in children and adolescents with ASD (N = 120; Lidstone et al. 2014). Reflecting Leekam et al.’s (2007b) findings, principal components analysis (PCA) for this ASD sample also resulted in two components: RSMB and IS, with good internal consistency across the whole scale (α = .86) and for both RSMB and IS (α = .79, α = .83, respectively). Overall, the similarity of results across studies, satisfactory endorsement of items and good internal consistency support the construct validity of the RBQ-2 in children.

For the current research, the RBQ-2 was adapted into an adult self-report questionnaire, the Adult RBQ-2 (RBQ-2A). As the RBQ-2A has been adapted into a self-report measure, it is only accessible to participants with sufficient cognitive resources and verbal ability to complete the questionnaire. Approximately half of children with ASD are reported to have an IQ in the average range (e.g., Charman et al. 2011), suggesting that the RBQ-2A will be accessible to a similar proportion of adults with ASD. However, given the hetereogeneity of ASD, this represents a potential limitation of the RBQ-2A. Nevertheless, there is still need for a self-report measure of RRB for adults with ASD for the reasons discussed earlier.

Two studies are reported here. In Study 1, to build upon the findings with NT children, the RBQ-2A was administered to a group of NT young adults. Consistent with the majority of previous factor analytic research, PCA analysis was used to identify factors and the internal consistency of the measure was also assessed. In Study 2, the RBQ-2A was administered to a sample of NT adults with a broader age range and a sample with an ASD diagnosis to establish whether the RRB scores in those with a diagnosis of ASD was higher than those in the NT group. A secondary aim of Study 2 was to assess whether the subscales derived from Study 1 were reliable in a more representative sample.

For Study 1, the RBQ-2A was administered to a NT university student sample and the structure of the RBQ-2A assessed using principal components analysis (PCA). It was expected that two components would emerge, as this structure is the most consistent finding, and that they would be broadly similar to that found in the original RBQ-2 (Leekam et al. 2007b; Lidstone et al. 2014). Its internal consistency was also assessed using Cronbach’s alpha. Finally, we also included the Autism-Spectrum Quotient (AQ; Baron-Cohen, et al. 2001), which has been used widely to assess the presence of a variety of autistic traits in the general population (e.g., Hurst, Mitchell, Kimbrel, Kwapil et al. 2007; Kloosterman et al. 2011; Stewart and Austin 2009). It was expected that scores on the RBQ-2A would be significantly correlated with scores on the AQ.

The aim of Study 1 was to develop and test the RBQ-2A as a self-report measure of RRBs in NT adults. An existing parent report measure of RRBs, the RBQ-2 (Leekam et al. 2007b), was adapted into an adult self-report measure and administered to a university student sample. PCA resulted in a two-component structure, one comprising motor behaviours, RMB, and the other behaviours related to routines and a preference for sameness, IS. As predicted, scores on the RBQ-2A were also correlated with another measure of autistic traits, the AQ (Baron-Cohen et al. 2001).

The first component, RMB, is similar to RSMB found in previous research with the RBQ-2 (e.g., Leekam et al. 2007b; Lidstone et al. 2014). Five of the six RMB items consistently load onto the factor that in previous research included motor and sensory items (RSMB), the exception being item one, arranging objects. The major difference between RMB found here and RSMB in previous research is the lack of sensory items loading onto this component. The second component corresponded to IS. This result was more comparable to previous research using the RBQ-2 in an ASD sample, with five items (13–17) loading in the same way as in Lidstone et al.’s (Lidstone et al. 2014) study.

In summary, the components yielded by the present PCA are similar to previous research with NT children and autistic children using the RBQ-2, with the exception of sensory items. Items two to six load onto RSMB in the child version of the questionnaire (Leekam et al. 2007b; Lidstone et al. 2014) and RMB in the present study, and items 13 to 17 load onto IS across all three studies, supporting the construct validity of the questionnaire.

The most probable reason for the difference between the present PCA solution and previous research is that the present sample comprised NT adults whereas previous research examined NT children (Leekam et al. 2007b) and children and adolescents with ASD (Lidstone et al. 2014). Certain types of behaviours may be associated with younger children or children with ASD rather than NT adults. For example, mean scores on items 3 (spinning) and 11 (carrying around objects) were higher in NT children (Arnott et al. 2010; Leekam et al. 2007b) than in the present study. Moreover, autistic individuals show higher levels of sensory symptoms than NT individuals (e.g., Ben-Sasson et al. 2009; Kern et al. 2006; Leekam et al. 2007a; Rogers and Ozonoff 2005) and these items were not well endorsed by the present sample.

The different loading of certain items may also reflect the fact that certain behaviours do not clearly fall into one particular category. For example, eating a small range of foods (item 19) formed part of IS in the present study but has previously loaded on to RSMB (Lidstone et al. 2014) as well as IS (Leekam et al. 2007b); eating a small range of food may be a result of sensory issues or insistence on sameness and is therefore conceptually related to both subscales.

There are some limitations in terms of the sample. Firstly, the sample comprised only university students and is therefore limited in age and IQ distribution. Second, it might be considered that the size of the sample is relatively small for PCA. However, the literature is equivocal regarding the appropriate sample size for PCA and factor analysis (e.g., Tabachnick and Fidell 2014; Williams et al. 2010) and the assumptions for PCA were met. Therefore, the data were deemed suitable for analysis. Overall, the results of Study 1 support the construct validity of the RBQ-2A and suggest that it is useful as a self-report questionnaire in an adult population.

In Study 1, the sample comprised young NT adults. In Study 2, the RBQ-2A was administered to older NT adults and to adults with ASD. It was hypothesised that the ASD sample would score significantly higher than the NT group on the RBQ-2A. This study also explored the reliability of the subscales found in Study 1 in a more representative NT sample. As all RBQ-2A items were administered, including the sensory items that did not load on to the PCA in Study 1, this study also offered the opportunity to examine the RBQ-2A sensory scores in ASD as compared with NT adults.

Overall, these results indicate that the RBQ-2A is a useful new self-report measure for assessing RRBs in adults. Study 1 found a two-component structure in a NT university student sample that approximately corresponds to previous research using other measures of RRBs, with the exception of sensory items. Study 2, using a more representative sample of adults, found that participants with ASD score significantly more highly than IQ-matched NT participants on the RBQ-2A total and subscale scores, which would be expected from an accurate measure of RRBs. The internal consistency of the RBQ-2A and its subscales was high for adults with ASD, providing further support to its reliability as a measure of RRBs for adults on the autism spectrum. Both studies showed that RRBs are significantly associated with AQ score and support the use of the RBQ-2A as a measure of RRBs in NT adults. Subsidiary analyses in Study 2 also indicated that although the university sample in Study 1 had higher levels of RRB than the adults in Study 2, this was unlikely to be due to differences in age. Given the potential relationship between RRBs and anxiety, it can be speculated that the higher incidence of anxiety traits in university populations (Andrews and Wilding 2004; Stallman 2010), alongside the screening for significant psychopathology in the Study 2 NT sample, may have biased the Study 1 group to relatively higher scores. Further research is needed to explore both the association between psychopathology symptoms and RRBs, and whether the RBQ-2A can accurately distinguish between ASD and psychological disorders that involve high levels of RRB, such as OCD and other specific anxiety disorders.

Another unexpected finding from Study 1 is that most sensory items from the RBQ-2A did not load onto either component. Furthermore, adding sensory items to the RMB subscale for the NT group in Study 2 reduced the mean score on this subscale. These findings indicate that sensory symptoms are not common in the NT participants across both studies, whereas they are highly prevalent within the autistic population (e.g., Boyd et al. 2010; Kern et al. 2006; Leekam et al. 2007a). This may be partly explained by the fact that some of the sensory items contain references to ‘special’ interests and items, which may not be relevant for NT adults. Alternatively, it may be that the RBQ-2A simply does not capture a wide enough range of sensory behaviours, as it includes just six items from the original set of 25 items in the DISCO (items 7, 8, 9, 10, 18 and 19). A previous study of the general population found evidence of a wider range of sensory behaviours with a more detailed questionnaire, the Glasgow Sensory Profile (Robertson and Simmons 2013). This questionnaire covers seven modalities, including auditory, vestibular and proprioceptive, which are not included in the RBQ-2A. Nevertheless, the RBQ-2A was able to discriminate between the ASD and NT groups both with and without the sensory items.

There are also some important limitations to consider for the studies reported here. Although the samples across both studies include a fairly wide range of ages from 18 to 50 years old, these findings cannot be generalised to adults of an older age. Furthermore, as the RBQ-2A was adapted from a measure for children, it may be missing certain items that are applicable only to adults. As discussed in the Introduction, the RBQ-2A as a self-report measure is currently suitable only for more able adults. Therefore the RBQ-2A, and any associated findings, are only generalizable to this population. In both studies, there was significant positive correlation between the AQ and the RBQ-2A. However this correlation might be partly explained by the fact both are self-report measures. As mentioned, it is therefore important to compare the RBQ-2A with other measures of autistic traits such as interviews or informant-report questionnaires. The purpose of the RBQ-2A is to describe a profile of RRBs; it does not measure social communicative behaviours and therefore is not suitable as a stand-alone diagnostic tool for ASD, which includes both domains as necessary and essential conditions for a diagnosis.

Nevertheless, the studies presented here represent an important new contribution with the development of an adult self-report measure of RRBs, which can be used with both ASD and NT populations. The need for such a measure is indicated by the findings of both studies, which indicate that self-reported RRBs in adulthood may present slightly differently than carer-reported RRBs in childhood. Specifically, although the subtypes of RRBs remain the same, the specific behaviours that are endorsed differ. The potential clinical applications of the RBQ-2A include its use as a signposting questionnaire or as a supplement to diagnostic interviews such as the DISCO. Its utility may be especially helpful given that the AQ does not give an adequate or reliable assessment of RRBs across typical populations (e.g., Kloosterman et al. 2011; Lau et al. 2013). It may also be useful for other clinical conditions that show RRBs, such as obsessive-compulsive disorder, Gilles de la Tourette syndrome and Parkinson’s disease (Langen et al. 2011). From a research perspective, the RBQ-2A allows for the opportunity to accurately and reliably explore RRBs directly in adults both with and without ASD. Overall, these results show that the RBQ-2A is a promising self-report measure of RRBs in adults. However, further research should involve older and more diverse NT participants that include representation of a range of ethnic and SES groups, as well as a larger sample of adults with ASD. Although the RBQ-2A is a descriptive questionnaire that can only identify a profile of behaviours as perceived by self-informants, further research comparing self- and other-informant RBQ-2A questionnaires and its use with clinical interviews may help to assess how well the RBQ-2A complements and streamlines the diagnostic process in clinical practice.