Three ‘driving forces’ behind the silence: Social anxiety, autism spectrum symptoms, and language difficulties as correlates of selective mutism in young children

Selective mutism (SM) is a childhood disorder characterized by an inability to speak in specific challenging social situations, while speech is normal in other less demanding settings. For instance, in a school context, a child with SM may not respond verbally to a teacher’s questions or requests, nor engage in verbal interactions with classmates, whereas at home the child speaks freely and confidently with family members (American Psychiatric Association, 2022; World Health Organization, 2022). However, this prototypical description should be interpreted in the light of the current understanding that psychiatric conditions, including SM, are dimensional rather than strictly categorical. Thus, some children with SM may occasionally exchange a few words with their teacher or whisper softly to a nearby friend in class, although their overall speech output remains clearly reduced compared to situations in which they feel more comfortable (Muris and Ollendick 2021a).

SM is a relatively rare disorder, with reported prevalence rates in non-clinical populations ranging between 0.03% and 2%, depending on the assessment method and specific participant characteristics (e.g., age range, immigrant status; Bergman et al., 2002; Elizur & Perednik, 2003; Karakaya et al., 2008; Kopp & Gillberg, 1997; Kumpulainen et al.,1998; Sharkey & McNicolas, 2012). Findings on gender differences are inconsistent: some studies indicate that SM is somewhat more prevalent in girls (Sharkey & McNicolas, 2012), whereas others report an equal gender distribution (Bergman et al., 2002) or even a higher prevalence in boys (Karakaya et al., 2008). SM typically has an early onset, most often emerging during the preschool years, although referrals frequently occur only after school entry (Viana et al., 2009). While the most severe manifestations often diminish over time, longitudinal studies have found that a substantial proportion of children diagnosed with SM continue to display shyness, reduced verbal output, and psychiatric problems – particularly anxiety disorders – during adolescence and adulthood (Koskela et al., 2023; Remschmidt et al., 2001; Steinhausen et al., 2006).

Research on the ‘driving forces’ behind the selective non-speaking of children with SM has focused on three main factors. The first factor, ‘anxiety’, is grounded in the premise that children refrain from speaking in specific situations because these settings elicit substantial fear and apprehension. Evidence consistently indicates that anxiety – and social anxiety in particular – plays a central role in SM (Muris & Ollendick, 2015). For example, Vogel et al. (2019) asked 65 children and adolescents with SM (aged 8–18 years) to describe their main fear in situations where they could not speak. The majority reported social-evaluative fears (i.e., apprehension about the negative evaluation or reactions by others). A follow-up questionnaire revealed that they endorsed social anxiety-related cognitions (e.g., “I don’t speak, because others might think I’m stupid” or “I don’t speak, because I think that others could laugh at me”) at levels comparable to peers diagnosed with social anxiety disorder. In a subsequent study, parent-based surveys indicated that children with SM not only experience fear and anxious cognitions, but also frequently engage in avoidance behaviors, actively attempting to escape or evade social situations that demand verbal responses (Vogel et al., 2024). Other studies have reported strong similarities in the clinical presentation of SM and social anxiety disorder (Manassis et al., 2003; Milic et al., 2020; Poole et al., 2021; Schwenck et al., 2021; Sharp et al., 2007; Yeganeh et al., 2006), and a meta-analysis by Driessen et al. (2020) found that the two conditions are frequently comorbid. In sum, this body of evidence highlights social anxiety as a prominent driving force in SM and supports its classification as an anxiety disorder in current psychiatric nosologies (American Psychiatric Association, 2022; World Health Organization, 2022).

The second factor implicated in SM is ‘autistic traits’, which refer to features of autism spectrum disorder (ASD), a neurodevelopmental condition characterized by aberrant brain development and functioning (Hirota & King, 2023). Muris and Ollendick (2021b) have argued that the core characteristics of ASD may contribute to the selective non-speaking observed in SM. Specifically, the persistent social communication and interaction difficulties experienced by individuals high on the autism spectrum can make it challenging for them to engage in verbal exchanges. In addition, restricted and repetitive behavior patterns, such as rigidity, sensory overstimulation, and demand avoidance, may further limit verbal expressiveness in certain social situations. Over the past decade, evidence supporting a role for ASD in SM has steadily grown.

For example, Steffenburg et al. (2018) retrospectively examined medical records of 97 clinically referred children with SM (aged 4–18 years) and found that 63% met DSM-IV (American Psychiatric Association, 1994) criteria for ASD, with an additional 20% displaying subclinical ASD symptoms. Similarly, Suzuki et al. (2020) compared children aged 4–11 years with SM, children with ASD, and typically developing peers using teacher ratings of autistic traits on the Autism Spectrum Quotient (Baron-Cohen et al., 2001). Children with SM received ratings comparable to those of the ASD group, and both groups scored significantly higher than the typically developing group. Finally, in a mixed clinical and non-clinical sample of 6- to 12-year-olds, Muris et al. (2024), found that children with SM scored as high on a parent-report autism screening tool as a clinical control group, many of whom had an official ASD diagnosis, and marked higher than typically developing peers. Additional analyses showed that 48% of the SM group scored in the highest 10% on the autism screen, with 28% exceeding the clinical cut-off (highest 2%). Altogether, these findings indicate that ASD and autistic traits represent another driving mechanism in SM, potentially contributing to its onset and maintenance (see also Iimura et al., 2024; Keville et al., 2023; Klein et al., 2019; Muris et al., 2021).

The third and final factor, referred to as ‘language difficulties’, relates to the idea that some children remain silent in certain situations due to discomfort stemming from language impairments. Research examining the link between SM and language difficulties is limited. Kristensen (2000) conducted an early investigation involving clinically referred children with SM aged 3–16 years (n = 54) and found that they were more likely to show delays in language development compared to age-, gender-, and socioeconomic status-matched non-clinical controls (n = 108; 51.9% vs. 11.1%). From a diagnostic perspective, half of the children with SM (50%) met DSM-IV criteria for a communication disorder (i.e., expressive language disorder, mixed receptive-expressive language disorder, and phonological disorder) compared with only 11.5% of controls. Subsequent studies using standardized language assessments (such as the Peabody Picture Vocabulary Test) have similarly identified language deficits among children with SM (Klein et al., 2012; Manassis et al., 2007; Nowakowski et al., 2009), particularly when evaluating more advanced and complex linguistic abilities. On the whole, this evidence suggests that language difficulties or delays may contribute to children’s muteness in specific contexts and thus play a role in the development of SM.

Overall, at least three factors have been identified as contributing to the selective non-speaking behavior of children with SM: social anxiety, autistic traits, and language difficulties (Muris and Ollendick 2015, 2021a, b). Most studies, however, have examined the relationship between each factor and SM in isolation, overlooking the fact.

Overlookig the fact that these factors are often interrelated. For instance, research shows that social anxiety frequently co-occurs in children with ASD (Spain et al., 2018) and is also common among young people with language impairments (Brownlie et al., 2016; Voci et al., 2006). Similarly, ASD is associated with a delayed language development and deficits in both language comprehension and expression (Georgiou & Spanoudis, 2021).

Few studies have investigated the relative contributions of social anxiety, autistic traits, and language deficits to SM symptoms. One exception is a study by Manassis et al. (2007) which examined the additive effects of social anxiety and language deficits in 44 children with SM aged 6–10 years, finding that both factors explained a significant portion of the variance in SM symptoms. Starke (2018) conducted a longitudinal study of 30 preschool children at risk for SM over a period of nine months and found that anxiety was the strongest predictor of non-speaking behavior, whereas the effect of language difficulties did not reach statistical significance. More recently, Muris et al. (2021, 2024) explored the unique contributions of social anxiety and autistic traits in two samples of preschool (N = 172) and primary school children (N = 169), including some clinically referred participants. Their findings indicated that social anxiety accounted for most of the variance in SM symptoms, but autistic traits also made a distinct, independent contribution.

From a developmental psychopathology perspective, it is important to recognize that SM may arise through equifinality – that is, different pathways can lead to the same behavioral outcome (Cicchetti, 2016). Identifying the relevant contributing factors has direct clinical relevance, as it can refine diagnostic assessments and support the tailoring of interventions to individual profiles. As noted above, previous studies have examined the distinct roles of social anxiety and language difficulties (Manassis et al., 2007; Starke, 2018), as well as social anxiety and ASD-related traits (Muris et al., 2021, 2024). However, to date, no study has investigated the relative contributions of all three potential ‘driving’ factors – social anxiety, autistic traits, and language difficulties – to SM symptoms. The present investigation aimed to fill this research gap. An online survey was administered to parents of preschool children aged 3 to 6 years (N = 81), recruited either from the general population (n = 52) or a clinical setting/special online forum for parents of children with SM (n = 29). The survey included questionnaires measuring SM symptoms as well as the three hypothesized contributing factors: social anxiety, autistic traits, and language difficulties/delay. It was hypothesized that each factor would be positively associated with SM symptomatology. In addition, regression analyses were conducted to examine the unique contributions of the three factors. It was predicted that social anxiety would account for a significant and substantial proportion of the variance in SM symptoms, whereas autistic traits and language difficulties were expected to make smaller, but still significant unique contributions.

The present study investigated three potential underlying mechanisms – social anxiety, autistic traits, and language difficulties – thought to drive the selective non-speaking behavior of children with SM. Parents of preschool children, recruited from both the general population and a clinical setting/specialized online forum for parents of children with SM, completed a survey comprising standardized questionnaires assessing the relevant constructs. The findings provided clear evidence that social anxiety is a prominent feature of SM (Muris & Ollendick, 2015). Specifically, PAS social anxiety scores showed a strong positive correlation with SM symptom levels in this sample of 3- to 6-year-olds. The corresponding scatterplot revealed a clear linear trend, with children from the (sub)clinical group clustering in the upper-right quadrant, indicating the co-occurrence of high SM symptoms and high social anxiety (see also Muris et al., 2024, for similar results in children aged 6–12 years). Moreover, 83% of children with SMQ scores in the clinical range also displayed elevated social anxiety scores.

Regression analyses further confirmed the central role of social anxiety: across all models examining the relative contributions of the three constructs, social anxiety consistently emerged as a significant predictor, with the largest β or odd ratio. This indicates that, independent of autistic traits and language difficulties, social anxiety accounted for a substantial proportion of the variance in SM symptoms (Muris et al., 2021, 2024). These findings underscore the importance of social anxiety in the selective non-speaking behavior of children (Viana et al., 2009; Muris & Ollendick, 2015; Sharp et al., 2007) and lend further support to the classification of SM as an anxiety disorder (American Psychiatric Association, 2022; World Health Organization, 2022).

Consistent with previous studies (Steffenburg et al., 2018; Suzuki et al., 2020; Iimura et al., 2024; Keville et al., 2023; Klein et al., 2019), the present findings indicate that autistic traits are implicated in at least a subset of children with SM. SMQ scores were positively associated with ASQ scores, and this was not only true for the total autistic traits score but also for the subscales of interactive-communicative difficulties and restricted-repetitive behaviors. Notably, 34% of children with SMQ scores in the clinical range (34%) also scored in the top 10% of ASQ total scores. Furthermore, in regression models controlling for the other constructs, autistic traits – particularly interactive-communicative difficulties and to a lesser extent restricted-repetitive behaviors – remained a statistically significant predictor of SM symptoms.

This finding aligns with results from our similar study in a mixed clinical/non-clinical sample of older children aged 6–12 years (Muris et al., 2024), but contrasts somewhat with Muris et al. (2021), who reported that the unique association between ASD an SM was primarily driven by the interactive-communicative difficulties. It is worth noting, however, that the latter study included only non-clinical children, who likely exhibited low levels of – and therefore limited variability in – restricted-repetitive behaviors. Overall, the present findings add to the evidence that, in addition to social anxiety, the social deficits and behavioral characteristics associated with ASD may also contribute to the selective non-speaking of children with SM (Muris and Ollendick 2021b).

Support for our hypothesis that language difficulties would be positively associated with SM symptoms was limited. The correlation between SMQ and ELS scores was not statistically significant, although an association did emerge when using the categorical measure of language delay (which accounted for the child’s age). Specifically, 24% of children with SMQ scores in the clinical range displayed a language delay, compared to only 5% of those in the normative range. However, a logistic regression analysis controlling for social anxiety and autistic traits revealed no unique contribution of language delay, suggesting that the observed association was largely carried by overlap with these other variables. This aligns with previous studies showing that the link between language difficulties and SM is not always straightforward (Manassis et al., 2007; Nowakowski et al., 2009).

Meanwhile, several factors may explain why the present study was not optimally suited to demonstrate this link. The instruments used to assess language deficits and delay (including the ELS) often have limited evidence for validity (So & To, 2022). Furthermore, characteristics of the present sample may have reduced our ability to detect an effect: ELS scores were generally low, with a maximum observed score of 6 on a scale ranging from 0 to 26, and – unexpectedly – (sub)clinical children were reported by parents to have fewer than one language difficulty on average. Including more children diagnosed with a developmental language disorder might have increased variability in ELS scores, improving the ability to test the hypothesized relationship with SM symptoms. Encouragingly, when (sub)clinical cases were excluded from the present sample, the correlation between ELS and SMQ scores became positive and statistically significant.

Several additional findings emerged in this study that merit brief discussion. First, ASD symptoms were positively and significantly associated with both social anxiety symptoms and language difficulties, consistent with previous research linking these clinical phenomena (Georgiou & Spanoudis, 2021; Spain et al., 2018). No significant correlation was observed between language difficulties and social anxiety, which contrasts with prior literature (Brownlie et al., 2016; Voci et al., 2006). This discrepancy may be due to the relatively young age of the children in the current sample; social anxiety may emerge later when children with language difficulties become more self-conscious and fear negative evaluation (Higa et al., 2008). Second, while children in the (sub)clinical group generally scored higher on the SMQ and most other measures compared to those in the non-clinical group, not all of them displayed clinically elevated SM levels. This may be due to the fact that it is unknown whether some had already received treatment for the condition. Furthermore, a notable proportion of the non-clinical group exhibited elevated levels of psychopathology. Specifically, 30.8% of non-clinical children had SMQ scores in the clinical range and/or elevated scores on other indices (PAS social anxiety: 28.8%, ASQ total symptoms: 11.5%, and ELS language delay: 17.3%). This may reflect the study’s recruitment framing (‘children who sometimes do not speak’), which could have attracted parents of children with some behavioral concerns. Third, most scales demonstrated good internal consistency. The exception was the ELS, which yielded a relatively low Cronbach’s alpha of 0.54. Because alpha depends on inter-item correlations, this likely reflects the few positive endorsements on this measure and the resulting limited variability in scores (Tavakol & Dennick, 2011).