Sensory sensitivity is characterised by over- (hyper) or under- (hypo) responding to sensory stimuli, and can occur within a number of different sense domains (e.g., visual, auditory, olfactory, gustatory, tactile, vestibular, proprioceptive). Hyper-sensitivity typically incorporates sensory overload and avoidance behaviours, while hypo-sensitivity incorporates sensory under-responsivity and seeking behaviours. For example, a person with hyper-sensitivity might find strong smells overwhelming (i.e., sensory overload) and avoid them (i.e., sensory avoidance), while a person with hypo-sensitivity might fail to notice strong smells at all (i.e., sensory under-responsivity or ‘dampening’) and actively seek them out (i.e., sensory seeking). Children can find sensory information more disruptive than adults (Ahn et al.,
2004), but their sensory sensitivities are poorly understood. Limited research has concentrated to a large extent on clinical groups (see below), but beyond clinical groups, our knowledge is vague. One barrier has been a lack of instruments that can measure childhood sensory sensitivities across both clinical and non-clinical populations, while also adequately capturing the complexity of sensory experiences and their multifaceted nature. Any useful measure would need to be able to test multiple sense domains (vision, audition, olfaction etc.), for both hyper and hypo sensitivity, with both sensory and behavioural components (i.e., overload/dampening vs. avoidance/seeking). The current paper presents an instrument which aims to fulfil all these requirements.
We noted above that research in sensory sensitivities has tended to focus on clinical populations, and in particular, on children with an Autism Spectrum Condition (ASC), also known as Autism Spectrum Disorder (ASD; henceforth autism) where sensitivities are particularly high (Baranek et al.,
2007; Billstedt et al.,
2007; McCormick et al.,
2016; Robertson and Simmons,
2013; Watling et al.,
2001). Indeed sensory sensitivities now form part of the diagnostic criteria for autism within the Diagnostic and Statistical Manual of Mental Disorders (DSM-V: American Psychiatric Association
2013). Other conditions, too, show sensory sensitivities, including Attention Deficit Hyperactivity Disorder (Cheung & Siu,
2009; Ghanizadeh,
2011; Panagiotidi et al.,
2018; Yochman et al.,
2004), childhood developmental delays (Baranek et al.,
2007; McCormick et al.,
2016), Tourette’s syndrome (Isaacs & Riordan,
2020), and premature births (Rahkonen et al.,
2015). Moreover, sensory sensitivities are found not only within clinical populations, but are also in approximately 1 in 6 typically developing children (Ahn et al.,
2004; with a notable degree of within-group variation; Brockevelt et al.,
2013; Cheung & Siu,
2009) meaning as many as five children with sensitivities arise within the average English primary school classroom. The exact sensory profile of typically and non-typically developing groups requires more research (Tomchek et al.,
2014), but in autistic children for example, sensory behaviours are known to cluster together, such as tactile/movement sensitivity, taste/smell sensitivity, auditory/smell sensitivity. And environmental interactions are also important; hence a child may find loud noise in a supermarket unbearable but play music loudly at home.
It is important to have a clear understanding of (and validated tests for) sensory sensitivities in children, for a number of reasons. Sensory sensitivities can have a significant impact on children’s lives, and where sensitivities arise, they present early in development (Ben-Sasson et al.,
2010). Dunn, and colleagues found that children with higher levels of sensory processing issues were more likely to have difficulties in everyday routines such as eating, playing, family interactions, and other social/activity participation (Dunn et al.,
2016). Hyper-sensitivities in particular (described by Dunn et al. as ‘Low threshold’) were associated with higher levels of anxiety, shyness and more challenging behaviours. Indeed, sensory sensitivities have been linked to elevated anxiety in both typically and non-typically developing children (Green et al.,
2012; Lane et al.,
2012; Neil et al.,
2016; Reynolds & Lane,
2009). Conversely, hypo–sensitivities (described by Dunn et al. as ‘high threshold’) are associated with under-responsivity and potentially repetitive self-harming behaviours. In otherwise typically developing populations, sensory sensitivities are also linked to reduced play behaviours (Bundy et al.,
2007), compulsive-like behaviour (Dar et al.,
2012), modulated movement (Buitendag & Aronstam,
2010; Gal et al.,
2010), and feeding problems (Davis et al.,
2013; Smith et al.,
2020). They can also impede social participation (Engel-Yeger and Dunn,
2011; Hochhauser & Engel-Yeger,
2010), sleep behaviours (Reynolds et al.,
2011), and classroom learning (Ashburner et al.,
2008; Baranek et al.,
2013; Dunn,
2007; Pfeiffer et al.,
2015). For example, a child with hypo-sensitivities is likely to miss information in the classroom from sensory dampening (e.g., not hearing the teacher) while a child with
hyper-sensitivities (e.g., aversion to loud noise) may find the classroom a stressful environment, potentially leading to disruptive behaviour (Cheng & Boggett-Carsjens,
2005). Knowing the triggers and behavioural patterns of children with sensory sensitivities, as well as how to test for them, can enable schools to adapt (Dynia et al.,
2022; Goodman-Scott & Lambert,
2015).
Measuring sensory sensitivity in children
There are a number of existing instruments to identify sensory sensitivities in children (Jorquera-Cabrera et al.,
2017), each with their own strengths and limitations. The Sensory Profile 2 is one of the most commonly used questionnaires, available in a full 86-item version (Dunn,
1999,
2014) as well as a 38 item short form (Short Sensory Profile; SSP2) and is designed for parents or teachers of children from birth to 14 years (SSP2). The Sensory Profile provides scores for six sense domains (auditory, visual, touch, movement, body position, oral) and for four outcome domains (seeking, avoiding, sensitivity, registration). It has excellent psychometric properties (i.e., Cronbach’s alphas ranging 0.74 to 0.92 for the SP2, and 0.79 to 0.86 for the SSP2; Dunn et al.,
2016) but also has certain limitations. As well as excluding olfactory sensitivities, the latest versions (SP2, SSP2) as with the original, do not specifically split sensitivities by hyper versus hypo domains. Another measure of childhood sensory sensitivities is the Sensory Experiences Questionnaire (Ausderau & Baranek,
2013; Baranek et al.,
2006; Baranek,
2009; Little et al.,
2011). Its latest edition (version 3.0) currently comprises 105 items for 2–12 year olds (Baranek,
2009) across five sense domains (auditory, visual, tactile, gustatory, and vestibular) and four outcome domains (described by the authors as Hyper, Hypo, Sensory seeking, and Enhanced perception; i.e., they use the terms hyper/hypo for over/under-sensing rather than related behaviours e.g., seeking). Whilst showing adequate psychometric properties for most scales (Cronbach’s alphas ranging 0.68 to 0.72 for subscales, 0.80 full scale: Baranek,
2009), this measure was developed specifically for sensory symptoms in children with autism or developmental disabilities so it is unclear whether it has suitability for normative populations, where internal consistency is poor (Cronbachs alpha = 0.63 for the full scale; Baranek,
2009). Finally, the Sensory Processing Measure is a questionnaire for caregivers of 5–12-year-olds (Parham et al.,
2007), and comprises 62 items across five sense modalities (visual, auditory, tactile, proprioception, vestibular) and has excellent psychometric qualities (Cronbach’s alphas ranging 0.77 to 0.95 for home form subscales, 75 to 0.95 for class form subscales; Parham et al.,
2007). However, the key focus of this questionnaire is tangential to sensory sensitivities per se, measuring instead sensory processing, praxis, and social participation in school. Other behavioural assessments are available, such as the Sensory Integration and Praxis Tests (Ayres,
1989), and the Sensory Processing Scale Assessments (Schoen et al.,
2017). But these assessments might be considered too labour-intensive for many research situations (where sensory sensitivity can be just one aspect of assessment among many). For example, a complete Sensory Integration and Praxis Test comprises a two-hour battery, incorporating a number of different elements beyond sensory sensitivities themselves.
Given the limitations of existing sensory processing measures (primarily in the breadth of sense domains covered, the populations they are limited to, or the time requirements) we sought to produce a single comprehensive instrument that would be relatively fast and easy to administer to the parents of children from a relatively young age, for both clinical and non-clinical populations. We chose a parent report in particular (as opposed to children’s self-report) for several reasons. Firstly, although children below 8 years can reliably self-report in domains such health and well-being (Riley,
2004; Smees et al.,
2019), more complex domains such as problem-behaviour or personality typically require more labour-intensive approaches for younger children (Measelle et al.,
1998; Norwood,
2007; Rebok et al.,
2001; Rinaldi et al.,
2022; Ringoot et al.,
2017). Secondly, for our measure to be valuable to children who are both typically and non-typically developing, a parent-report would be required given that conditions where sensory sensitivities arise (e.g., autism) can often present with poorer reading comprehension and/or introspection skills (Frith & Happe,
1999; Kinnaird et al.,
2019; Robinson et al.,
2017).
We have named our novel child measure the Parent-completed Glasgow Sensory Questionnaire (GSQ-P) because it is based on an equivalent pre-existing adult measure (the GSQ), developed previously by Robertson and Simmons (Robertson & Simmons,
2013). The original GSQ had been shown to have good validation in adults, in both neuro-typical and autistic populations, and in the UK and cross-culturally (Kuiper et al.,
2018; Sapey-Triomphe et al.,
2018; Ujiie & Wakabayashi,
2015), displaying excellent internal reliability for the global scale (Cronbach’s Alpha ranging 0.93 - 0.95), as well as correlating strongly with both autistic and sub-autistic traits (Horder et al.,
2014; Kuiper et al.,
2018; Ujiie & Wakabayashi,
2015). Here we create a version applicable to children, to be completed by their parents. Exactly like the original adult measure, our own test for parents was designed to assess hyper and hypo-sensitivity (21 items each) split equally across seven sense domains (e.g., visual, auditory, olfactory, gustatory, tactile, vestibular, proprioceptive), incorporating both sensory and behavioural dimensions (i.e., sensory overload/avoidance behaviour, as well as sensory under-responsivity/seeking behaviour). Importantly, although items from the existing adult GSQ were originally chosen to reflect knowledge within the autism literature on sensory sensitivities (Baranek et al.,
2006; Robertson & Simmons,
2013) it has been successfully validated and utilised in normative adult populations. Furthermore, adult instruments have been successfully adapted to a parent-report perspective (Rinaldi et al.,
2020). Our tool, therefore, took the form of an appropriate re-wording of the GSQ, to create a parent-report for describing children from both typically-developing and clinical populations.
Our new child measure, the GSQ-P, has already received a degree of validation from an earlier study, which showed our tool has concurrent validity against the Sussex Misophonia Scale for Adolescents (Simner et al.,
2022). Misophonia is a type of sensory hyper-sensitivity to certain of sound (e.g., eating sounds; Jastreboff & Jastreboff,
2001). High scores on the Sussex Misophonia Scale for Adolescents significantly mapped onto high scores within our GSQ-Parent, but only for its domain of sensory hyper-sensitivity. This is exactly as we might expect since misophonia is a type of hyper- (but not hypo-) sensitivity. Here we seek additional validation for our tool by investigating its psychometric properties and convergent validity, testing the parents of a large sample of children aged 6–11 years, looking at both the general population as well as children within our sample who have learning vulnerabilities. For this latter, we used the UK schooling classification of Special Educational Needs and Disabilities (SEND). The SEND system in England and Wales is designed to provide educational support to children and young people (aged 0–25) who have additional needs, as laid out in the SEND Code of Practice 2014 (Department for Education and Department of Health,
2015) and the Children and Families Act 2014 (Department_of_Education,
2014). SEND status signifies that a child has a learning difficulty and/or a disability that requires additional support in school, and includes multiple conditions including autism, sensory impairment and mental health problems. Needs cover aspects of cognition and learning; social, emotional and mental health; communication and interaction; and sensory and/or physical needs (e.g., vision impairment). Hence this population is a heterogeneous group but is known to show meaningfully group-wise characteristics (Evangelou et al.,
2008; Gaspar et al.,
2016; Schwab,
2019). This classification allows us to retrieve a usable sample size and, importantly, is considered to represent a unified group not only psychologically (Wigelsworth et al.,
2015) but also for educational purposes (e.g., overseen by a single co-ordinator in schools).
For our study, parents completed our novel GSQ-P, as well as measures of their a well-being, anxiety, and empathy, all known to be associated with sensory sensitivities in both adults and children; i.e., lower level of empathy, poorer well-being, and higher levels of higher anxiety (Ashburner et al.,
2013; Horder et al.,
2014; Robertson & Simmons,
2013). In sum, we will explore the psychometric properties of the GSQ-P, examining its factor structure, and demographic predictions. We hypothesise that our GSQ-P will be psychometrically robust, and importantly, will distinguish between typically and non-typically developing children. Secondly, we aim to examine the links between sensory sensitivities and a range of behavioural and emotional measures (specifically anxiety, behavioural difficulties, and ability to empathise). We hypothesise strong associations between all the three measures and sensory sensitivities, providing additional convergent validity for our tool. Finally, we will use our measure to explore the underlying relationship within and between hyper- and hypo-sensitivities (e.g., examining how different traits cluster). Below we describe our empirical investigation, including details of how we created the GSQ-P by adapting the corresponding adult measure to make it fitting for children.
Discussion
This paper evaluated the psychometric properties of our novel adaptation of the adult version of the GSQ (Robertson & Simmons,
2013) into both full and reduced short–form scales (GSQ-P, rGSQ-P). Re-designed to measure sensory sensitivities in children via parent report, the long form GSQ-P had 42 items (mirroring the adult version; Robertson & Simmons,
2013), crossing hyper- and hypo-sensitivity within each of seven sense domains (visual, auditory, gustatory, olfactory, tactile, vestibular, proprioception) giving three questions per cell. The short form (rGSQ-P) had 24 items, again crossing hyper- and hypo-sensitivity but this time with two questions only in each of six sense domains (proprioception was excluded given cross-loading in the long-form; see below). Both versions are freely available to use (presented in entirety in the Appendix), as comprehensive yet brief measures of sensory sensitivity suitable for parents reporting on younger children, and validated here on 6–11-year-olds.
The adult version has been used extensively elsewhere for research (Horder et al.,
2014; Ward et al.,
2018) and we here present a comparable version for children. The original adult version of the GSQ has been found elsewhere to have excellent psychometric properties and this was replicated here for the overall scale in both long and reduced forms. The suitability of both the GSQ-P and rGSQ-P for general sensory sensitivity is also evidenced by their relatively similar overall Cronbach’s alphas as well as from their similarity in convergent validity with other measures. Both showed expected empathy and well-being associations known to be co-morbid with sensory sensitivities in children (Boterberg & Warreyn,
2016; Dean et al.,
2017; Green et al.,
2012). Both the long and short form (GSQ-P and rGSQ-P) were unrelated to age and gender, making it a useful measure for all children across the 6–11 age group. In addition, both successfully distinguished between SEND and typically-developing populations. It may be useful in future studies where numbers of SEND children were available to investigate profiles of sensory sensitivity within specific SEND categories. The sensitivity sub-domains of both GSQ-P and rGSQ-P (total Sensitivity, hypo-sensitivity, hyper-sensitivity) also showed very good (rGSQ-P) or excellent (GSQ-P) internal consistency (i.e., collapsing across senses). Exploratory and cluster analyses of the full scale (GSQ-P) supported two main factors, which largely reflected hyper and hypo-sensitivity. As found previously in adults (Sapey-Triomphe et al.,
2018), we found in children that hyper and hypo-sensitivities co-exist in the same person, but more so for some senses (vision, audition, gustation, and proprioception) than others (i.e., tactile). However, there were notable issues with individual questionnaire items due to cross-loading (and hence increasing the correlation between hyper and hypo domains) or poor factor loadings, and these ultimately led to our generating the reduced scale (rGSQ-P). Items in the proprioception modality were found to be particularly problematic and it is possible that proprioceptive items are more difficult to assess accurately from a parental perspective than sensitivities in other modalities. It should be noted that the correlation between total proprioception score (combining hyper and hypo proprioceptive sensitivities) and GSQ-P total sensitivity score (excluding proprioception) was strong (
rs = 0.66) suggesting, parents are likely to notice proprioceptive sensitivities but may be misidentifying hyper and hypo responses. The removal of poorer items made hyper and hypo domains in our reduced scale more distinct (
r = 0.58 compared to
r = 0.78 for the full scale). By removing problematic items all hyper and hypo-sensitivity items to now load correctly. Correlation with the full 42-item scale was still extremely high (
r = 0.94–0.97 for hyper-, hypo-, and total sensitivity scales). It is hoped the rGSQ-P may therefore be a useful addition to the psychometrically evaluated instruments available, especially when time is at a premium, and for future research wishing to specifically tease apart hyper- and hypo- experiences. In contrast, the long form offers the opportunity to investigate sensitivity if comparison with adult populations is required (i.e., the adult GSQ).
Cronbach’s Alphas within individual senses were somewhat poorer than for the broader domains. Specifically, when looking at gross sensitivities within any sense (e.g., collapsing across hyper- and hypo-sensitivities within vision) the long form had “acceptable” or “good” Cronbach’s alphas for all but the Tactile sense, while the short form was generally “acceptable” at best. This suggests that if scores for individual sense domains are required (e.g., visual; i.e., when collapsed over hyper- and hypo- items) the long form GSQ-P would provide the most reliable diagnostic scores. A similar pattern of lower reliability for sense domains has been found in an adult-GSQ sample (Sapey-Triomphe et al.,
2018) and could be explained by a relatively small number of items per scale (Heo et al.,
2015) or, alternatively, reflect the reduced utility in measuring sensitivities in this way. A possibility that needs to be explored is that hyper and hypo-sensitivities may not necessarily be constrained to the same sense modality. A child who exhibits visual hyper-sensitivity (e.g., aversion to bright lights) may have hypo-sensitivities in other domains (e.g., gustatory dampening). Hence collapsing items within any given sense may be somewhat fallacious, and could explain our low correlations between hyper/hypo questions within senses.
We also found further evidence that hyper-sensitivity may stand apart conceptually and psychologically from hypo-sensitivity. Our cluster analysis showed a more cohesive structure within hyper-sensitivities versus hypo-sensitivities (in reduced form rGSQ-P). Specifically, hyper-sensitivity questions clustered by sense modality (e.g., tactile questions grouped together; visual questions grouped together), whereas hypo-sensitivity questions did not. Instead, hypo- responses grouped by behavioural outcomes, with clusters of seeking-behaviours (e.g., seeking olfactory, tactile or auditory stimuli), clusters of sensory dampening (e.g., not feeling pain; not feeling cold), and an apparent cluster of repetitive behaviours (e.g., repetitively playing the same piece of music [auditory], spinning round and round [vestibular]). Hypo versus hyper-sensitivity may therefore spring from different mechanisms. Other differences are that hypo- outcomes are often enjoyable (Kapp et al.,
2019) in contrast to hyper-sensitivities, and may help regulate sensory input and reduce stress (Liss et al.,
2005; Steward,
2015). Furthermore, we found that poorer emotional well-being relates more to hyper-sensitivity [SDQ-Internalising/ Anxiety], while externalising behaviours relates more to hypo-sensitivity [SDQ-Externalising: hyperactivity, Conduct problems]. In summary, we have found a number of ways in which hyper-sensitivity stands apart from hypo-sensitivity, despite falling within the single dimension of ‘sensory sensitivities’.
In conducting our study we were aware of certain limitations. First, we acknowledge that this analysis is exploratory. Future validations using confirmatory techniques would therefore be useful. Second, as with any parent report measure, the child is not describing his/her own experiences directly. Although parent and child reports of behaviour and traits are known to converge in many ways (Powers et al.,
2003; Zhou et al.,
2008), their perspectives sometimes differ, for example in measures of wellbeing, behaviour, and personality – where somewhat lower associations between parent and child perspectives have been reported (Barbaranelli et al.,
2003; Brown et al.,
2008; Phares et al.,
1989). Future work is in progress to investigate an alternative, self-report, child version of the GSQ (see Brown et al.,
2021) which could allow triangulation of the data and a greater understanding of underlying sensitivities as they emerge through modes of reporting. It would also provide important insights on how the two perspectives of parent and child might differ. It would also be useful in future studies to seek test-retest reliability, and discriminant validity, for both the GSQ-P and rGSQ-P. Although sample limitations make it not possible to analyse using generalizability theory (G-theory: Brennan,
1992) we hope future studies may use these techniques to explore potential forms of variance across different type of responder (e.g., mother or father), and we might also seek to investigate measurement invariance in respect to age, ethnicity, and gender.
A strength of our paper is that we investigated sensory sensitivity in a population sample recruited from mainstream schools. It is the first study to date, as far as we are aware, to measure sensory sensitivities in children who are both typically-/non-typically developing (SEND) using six or more sense modalities and two domains of sensitivity (hyper-/ hypo-), rather than a single global measure (Ben-Sasson et al.,
2008; Schoen et al.,
2014). This affords potentially important novel insights. The size of our SEND sample meant we could not explore sensitivity profiles within specific sub-groups of SEND children. Future work would be highly desirable to investigate symptom presentation in a larger sample of children with specific psychological diagnoses, including for example targeted samples with autism, ADHD, anxiety disorders, intellectual disabilities, and specific learning disabilities. This type of finer focus would contribute to knowledge about the role of sensory issues in different diagnoses, and allow further validation of our questionnaire across multiple additional groups.
Finally, future studies might also compare our measure alongside existing sensory sensitivity questionnaires such as the widely used Sensory Profile 2. Given our interests in Open Science, we presented a study that could be replicated by others without cost. However, other sensory sensitivity measures are available, such as the Sensory Experiences Questionnaire or indeed the Sensory Integration and Praxis Tests which we were unable to run on our large sample of >600 children given its labour-intensity (e.g., 2 h complete test). This type of concurrent validity in future studies would be particularly important, although we remind the reader to our recent work showing concurrent validity between the rGSQ-P and the Sussex Misophonia Scale for Adolescents (Simner et al.,
2022). High scores on the Sussex Misophonia Scale for Adolescents mapped onto high scores within the rGSQ-P domain of sensory hyper-sensitivity, but not hypo-sensitivities—exactly as we might expect since misophonia is a type of hyper- (but not hypo-) sensitivity. Hence our measures also shows concurrent validity against another scale of hyper-sensitivity. For convergent validity, beyond that already shown here, our rGSQ-P has now also been compared against measures of creativity (Smees et al.,
2022), where our hypo-sensitivity subscale predicted higher scores on the creativity-linked trait of Openness to Experiences (Kaufman et al.,
2016) and greater orientation towards creative activities involving movement and dance (exactly as we might expect from the sensory seeking aspect of the hypo-sensitivity trait). In summary, we have presented a range of ways in which our measure of childhood sensory sensitivities shows convergent validity, as well as showing convergent and concurrent validity elsewhere.
Our current study has provided a robust instrument to detect sensory sensitivities in research contexts, but it also has other potential applications. The questionnaire is quick and easy to administer so could be a useful additional tool in educational and therapeutic settings. We noted in our Introduction the many ways in which children with sensory sensitivities face challenges, for example, finding school and classroom environments particularly overwhelming (Dunn,
2007). Many typically developing children experience higher levels of sensory sensitivities (Little et al.,
2017) but present with behavioural and emotional responses rather than articulating their sensory issues (Cheng & Boggett-Carsjens,
2005). Children demonstrating such difficulties may therefore benefit from a sensory sensitivity assessment so that their exact sensory profile might be understood. A similar routine assessment might be particularly beneficial for groups known to have higher sensory challenges (e.g., SEND groups, as shown here). Their sensory profile could then be written into their support plan (e.g., into their
Educational, Health and Care Plan [EHCP] for schools in England where we tested). This type of application could be particularly beneficial because significant improvements are known to be found when environments are adapted for children’s sensory needs (Dynia et al.,
2022). Our novel tool, completed quickly and simply by their parents, would therefore be highly valuable.
To conclude, the measures we have presented here provide robust indicators of sensory sensitivity as an overall scale, and (in the short form especially) for the separate dimension of hyper-sensitivity as distinct from hypo-sensitivity. The rGSQ-P short form would be especially useful where time is limited and shows a particularly high correspondence to the full scale, although it no longer has functionality within the proprioceptive domain. The GSQ-P long form would be more useful where comparison with adult populations is required (i.e., the adult GSQ), and allow researchers to track sensory sensitivities longitudinally. We hope this validation might enable future researcher a better understanding of the sensory sensitivities of children, especially as viewed through the eyes of their caregivers.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.