Autism Spectrum Disorders (ASDs) are diagnosed in considerably greater numbers of males than females, with estimated ratios in the range of four affected males for every female (CDC, 2014). The mechanisms that give rise to this male bias are not well understood and are the subject of much current debate (e.g., Baron-Cohen et al., 2009; Fombonne, 2009; Werling and Geschwind, 2013). Several biological and non-biological theories have been proposed.

Non-biological models attribute differences in prevalence rate to biases introduced by differences in the presentation of ASD symptoms. Males with ASD have been reported to show more “externalizing behavior” including hyperactivity, aggressive behaviors, and repetitive and stereotyped behaviors and interests (Giarelli et al., 2010; Bölte et al., 2011; Hattier et al., 2011; Mandy et al., 2012; Solomon et al., 2012; Szatmari et al., 2012). On the other hand, females diagnosed with ASD present with more “internalizing behaviors” such as anxiety and depression (Hattier et al., 2011; Solomon et al., 2012). It therefore seems quite likely that the more socially disruptive behaviors in males have a higher likelihood to motivate parents or caretakers to seek clinical evaluations. In females, ASD symptoms are diagnosed when associated with more severe intellectual disabilities. In addition, high functioning ASD in females may be masked by their higher social abilities causing them to remain undiagnosed (Russell et al., 2011; Dworzynski et al., 2012).

Several biological models have been proposed to explain sex differences in ASD prevalence. The Extreme Male Brain (EMB) theory (e.g., Baron-Cohen, 2002) proposes that factors inherent in the male genotype and development that give rise to typically observed sexual dimorphisms in cognition (e.g., empathy and systemizing; Asperger and Frith, 1991) may be exaggerated in people affected with ASD giving rise to disordered social behavior (Baron-Cohen et al., 2003, 2005; Baron-Cohen and Wheelwright, 2004). This proposed “masculinization” can be observed in overt behavior (Ingudomnukul et al., 2007; Knickmeyer et al., 2008) and at the levels of brain structure and function (Lai et al., 2013) and may be linked to the expression of sex-hormones (Auyeung et al., 2006, 2009). Another model proposes that females with ASD carry a higher mutational load than affected males, but that the minimum liability threshold sufficient to cause ASD is higher in females. Evidence in support of this model has been mixed (Goin-Kochel et al., 2007; Hallmayer et al., 2011), but recently, some fairly convincing support emerged from a study by Robinson et al. (2013). The central premise of the Robinson study was that if a female “protective” effect exists, then a simple prediction would be that siblings of females with very high autistic trait scores (i.e., those above the 90th percentile) should show greater autistic trait scores than siblings of males with similarly high trait scores. This is exactly what was found in this large study of almost 10,000 twin pairs. There is also support for this protective notion from studies in females with Turner syndrome, a chromosomal abnormality where one of the X chromosomes is either missing or partially deleted (Bondy et al., 2012). Skuse and colleagues showed that social difficulties in this syndrome were predicated upon the parent of origin of the functioning X chromosome (Skuse et al., 1997). That is, those girls with a paternally derived X showed better social abilities than those with a maternally derived X, pointing to an imprinted genetic locus (or loci) for social cognitive functions expressed only on the paternal X. Of course, all males inherit their X chromosome from the mother, and so they won't express this socially protective gene (or genes), and their threshold for ASD would therefore be lower than it is in females (Skuse, 2000).

The study of sex differences in ASD traits is challenged by the relatively lower prevalence of females diagnosed with ASD and complicated by developmental changes and biases inherent in the diagnostic process due to differences in the presentation of symptoms. These circumstances may have led to inconsistencies in reports on sex differences in core ASD traits with some studies reporting evidence for sex differences in core diagnostic features (McLennan et al., 1993; Carter et al., 2007; Hartley and Sikora, 2009; Lai et al., 2011; Mandy et al., 2012) and other studies providing evidence to the contrary (Tsai and Beisler, 1983; Pilowsky et al., 1998; Holtmann et al., 2007; Solomon et al., 2012). Biological and non-biological models of sex differences in ASD are clearly not mutually exclusive, and both have important implications for the investigation of genetics, brain function and their relationship to the overt symptoms of ASD. These implications extend to research on ASD traits in the neurotypical population as there is mounting evidence for the heritability of ASD traits in unaffected individuals (Robinson et al., 2011; Lundström et al., 2012) and it has been suggested that “at least some of the genetic and environmental factors associated with ASD are the same as those that cause individual differences in autism-like behavior below the clinical threshold.” (Robinson et al., 2013). If indeed autistic traits are represented on a continuous spectrum that extends into the unaffected population then the sexual dimorphism in ASD characteristics should as well. In other words, if there is a sexual dimorphism of core ASD symptoms in individuals with an ASD diagnosis, then these sex differences would likely also manifest as different distributions of ASD traits in the unaffected population. For example, Matsuyoshi et al. (2014) examined sensitivity to eye gaze direction in 128 unaffected adults (64 females), a task in which individuals with ASD display robust deficits (see Senju and Johnson, 2009 for a review). The investigators found that individual differences in autistic traits predicted performance in this task in male but not female participants suggesting that direct-gaze perception may not constitute an autistic endophenotype in both sexes. Lai et al. (2012) studied four key cognitive domains including mentalizing and emotion perception, executive function, perceptual attention to detail and motor function in 128 male and female adults with and without ASD (32 per group). They found that deficits in mentalizing and facial emotion perception in individuals with ASD compared to controls were similar in both sexes. However, attention to detail and dexterity involving executive function were found to be impaired only in male ASD participants. The authors suggested that performance in the social cognitive domain is equally impaired in male and female individuals with ASD, whereas sex differences are observed in non-social cognitive domains. These findings lend support to the notion of sex differences in the disease phenotype and associated traits in the “normal” population and represent compelling reasons to consider sex differences when studying ASD traits in affected and unaffected samples.

In a previous cross-sectional study of 84 children with ASD and 142 neurotypical children between the ages of 5 and 17, we provided strong support for severe multisensory deficits in audiovisual speech perception during childhood and a subsequent recovery around 12 years of age in children with an ASD (Foxe et al., 2015). These large and consistent deficits in audiovisual gain between 5 and 12 years of age could not be explained by unisensory speech perception deficits alone or differences in eye-gaze as assessed with eye-tracking. These findings raise important questions about the neural mechanisms underlying these observed deficits, their possible heritability and link to biological sex. An advanced understanding of sex differences in ASD may benefit our understanding of the genetic, neurobiological and environmental factors involved in the development of ASD. Basic research on sex differences in ASD has practical implications for therapeutic intervention and may inform clinicians to delineate more personalized treatments for this diverse disorder.

In the current study we assessed sex differences in the perception of auditory, visual and audiovisual speech presented in varying levels of noise in typically developing children between the ages of 5 and 17 years of age. We also explored possible sex differences in speech perception in neurotypical adults. We finally examined sex differences in speech perception performance in a sample of ASD children between 8 and 17 years of age.

In this study, we assessed whether male and female participants differed in their ability to benefit from visual speech when target words were presented under varying background noise conditions. We tested sex differences in a sample of TD children and adults, and children diagnosed with ASD. In the TD child sample, females were significantly superior in recognizing speech in noise under auditory alone conditions than were males. Even larger performance differences were found under multisensory conditions, with the females benefitting significantly more from the addition of visual speech than the males, particularly under low intelligibility conditions (i.e., higher background noise). The females also performed better under pure speechreading conditions. These sex differences in children were fully absent in the sample of adult participants with the exception of the speechreading condition, in which case the males were slightly but significantly better at speechreading than the females. We then tested whether male/female performance differences were present in a sample of ASD children and found that ASD females performed significantly better under audiovisual conditions than ASD males, a difference that was not apparent for the auditory-alone condition in which no visual articulatory information was provided. Similarly, we found no evidence for sex differences in the ASD sample in speechreading, thus ruling out a purely unisensory account of differences in multisensory gain. Further, eye-tracking data made it clear that these sex differences were not due to different gaze patterns.

Clearly, multisensory speech perception is an important aspect of social communication. Therefore, possible answers to the observed sex differences might be found in sex differences in the development of social communication skills in general. Indeed, there is an extensive literature on the development of social communication in males and females which most frequently shows that females display greater, or at least earlier, development of skills in this domain. On average, females start to talk earlier than males (Fenson et al., 1994) and score higher on tests of verbal fluency (Hyde and Linn, 1988). Girls and women exhibit more eye contact than males (Hall, 1985), show greater ability to detect and understand emotional facial expressions (Rosenthal et al., 1979; Happe, 1995; Baron-Cohen et al., 1997, 1999) and there is accumulating evidence that preadolescent girls show relatively higher abilities in tasks assessing social understanding such as inferring other people's mental states (Theory of Mind; Hatcher et al., 1990; Bosacki and Astington, 1999; Calero et al., 2013).

It has been suggested that differences in social communication may have their origins at the earliest stages of development during intrauterine exposure to sex hormones (Auyeung et al., 2006, 2009; Chapman et al., 2006) thereby affecting brain structure and function relevant to social communication. Female newborns look longer at animated faces than mobile mechanical objects whereas newborn males showed the opposite pattern (Connellan et al., 2000). These genetic/epigenetic/hormonal origins of sex differences may be further enhanced by differential socialization, especially by parents (Stern and Karraker, 1989). Mothers have more verbal communication with their daughters than with their sons (Leaper et al., 1998) and parents show preferential acknowledgement of their infant daughter's emotional displays than their son's (Malatesta and Haviland, 1982). These factors may explain why female toddlers and infants show greater nonverbal communication skills (Clarke-Stewart, 1973; Fenson et al., 1994), vocabulary acquisition (Huttenlocher et al., 1991) and frequency of social initiations (Klein and Durfee, 1978). The evidence for differences in integration abilities between males and females remains far from conclusive, although women have been shown to be better at lip reading when target words were presented in a sentence context (Johnson et al., 1988; Watson et al., 1996). Similarly, some investigators have reported increased sensitivity to the so-called McGurk effect in women (Aloufy et al., 1996; Öhrström and Traunmüller, 2004). The McGurk illusion refers to a rather dramatic multisensory perceptual phenomenon whereby presentation of incongruent visual articulatory inputs can greatly modify the speech sound that is heard (McGurk and MacDonald, 1976; Saint-Amour et al., 2007). Irwin et al. (2006) showed that women displayed greater influence of visual speech on heard speech than men, but did not find evidence for sex differences in speechreading abilities.

Apart from greater abilities in AV-speech perception in ASD girls than ASD boys, our findings suggest that AV-benefit is not as affected by ASD in females as in males when stratifying performance by sex. Small relative decrements in AV-benefit in ASD females compared to TD males may exist, but did not reach statistical significance. ASD females do show lower performance than their unaffected counterparts, but they appear to be affected to a lesser extent than their male peers. Given our evidence for a sexual dimorphism in audiovisual speech perception in the TD population, and the apparent role of audiovisual speech perception for social communication (e.g., reading speech cues from faces) one may be inclined to interpret these findings within the framework of the EMB- theory. However, a closer look reveals that our pattern of results does not entirely conform to the EMB- theory. Although, to our knowledge, this prediction has not been explicitly stated by EMB proponents, it is implicit in the theory that if ASD brings about an increase or exaggeration of a masculine phenotype then one would expect that the effect of the disorder should have a greater impact on females than on males since males with ASD approach the extreme end of the proposed “maleness” spectrum and the difference between affected and unaffected males might therefore be subject to a ceiling effect. In contrast to this prediction, our data show larger differences in males than in females. However, the interpretation of our findings in light of the EMB theory should be exercised with caution, since it is not entirely clear whether the ability to read visual speech cues can be interpreted as an act of “empathizing” as has been proposed for the ability to read emotional expressions on people's faces (Baron-Cohen et al., 2005).

The pattern of our findings suggests a more parsimonious model of the observed sex differences. AV-speech perception can be regarded as an instance of language and social communication for which, as discussed above, previous evidence for a sexual phenotypic difference has been provided. The ability to integrate visual speech might therefore have genetic as well as environmental origins such as differential parental socialization resulting in an earlier development of these abilities in females and a sustained advantage into the late teenage years (Ross et al., 2011) relative to age-matched males. This developmental trajectory flattens toward early adulthood when females reach their performance maximum which allows males to “catch up” explaining the absence of performance differences in adults. ASD imposes a developmental delay in both males and females that resolves in the early teenage years rather than an irreversible and sustained impairment (Foxe et al., 2015). The fact that females with ASD are affected to a lesser extent might be afforded by “protective factors” inherent in the female genotype (Skuse et al., 1997; Skuse, 2000; Robinson et al., 2013). Unfortunately, the exploration of the developmental course of AV speech processing stratified by sex was not possible within the framework of this study due to the low subject numbers in our female ASD sample and should be subject to future studies.

High-density recordings of electrophysiological brain activity have revealed that the neural integration of multisensory inputs is impaired in children with ASD (Russo et al., 2010; Brandwein et al., 2013, 2015), and it is also the case that ASD children are not as effective at deploying attention to a relevant unisensory stream when there are competing multisensory inputs (Murphy et al., 2014). These studies included only a small proportion of female participants with ASD, precluding consideration of the role of sex in these deficits. An important question then is whether these deficits are equivalently seen in both male and female participants with ASD, or if they are sexually dimorphic as we see in the present study. These electrophysiological studies presented simple stimuli of no obvious higher-order communicative or social significance, and as such the findings were interpreted as representing the breakdown of basic sensory and attentional processes, although these could well have cascading consequences for higher-order functions such as multisensory speech perception. It is instructive that even fundamental deficits in multisensory integration processes and in the basic sensory processing of auditory tonal stimuli were found to be related to the severity of clinical symptoms in ASD children (Brandwein et al., 2015). This would suggest to us the possibility that while impaired communication among sensory cortices is part of the broader autism phenotype, protective factors may serve to “rescue” multisensory speech processing functions in females with ASD. Alternatively, it is also possible that even basic multisensory integrative processing is spared in females with ASD. Future work will be required to determine the extent to which this sparing is observed for other types of multisensory integrative processes, and whether it extends to non-social processing.

A limitation of this study is that the generalization of our findings from our ASD sample is only possible to the population of high functioning individuals with ASD. Using tasks adapted to individuals with low functioning ASD, future research may determine whether sex differences can also be observed in this population.

In conclusion this study provides evidence for sex differences in the ability to integrate heard and seen speech under noisy environmental conditions in a large sample of typically developing children and teenagers between the ages of 5 and 17 years. These differences were absent in a sample of healthy adults. We further show that multisensory speech processing is less affected in ASD females than males. Our findings underline the importance of considering sex differences in the search for autism endophenotypes. An advanced understanding of sex differences in ASD may benefit our understanding of the genetic, neurobiological and environmental factors involved in the development of ASD. Basic research on sex differences in ASD has practical implications for therapeutic intervention and may inform clinicians to delineate more personalized treatments for this diverse disorder.

JF, LR, and SM conceived the project, analyzed the data and wrote the paper. VDB and LR collected the data. VDB provided support with editing the manuscript. LR constructed the stimulus set, and HPF conducted and analyzed the eye-tracking component of the experiments. All authors discussed the results, commented on the manuscript and approved the final version.