The main purpose of this study was to advance previous work on the “alexithymia hypothesis” which demonstrated that alexithymia accounts for deficits in emotion recognition abilities, empathy, and interoceptive accuracy in adult samples of participants with ASD. To extend their findings to other components of emotion processing, we examined the effects of alexithymic and autistic traits on the production of spontaneous facial expressions in children with and without ASD, as they watched emotionally salient video stimuli. Consistent with Bird and Cook’s “alexithymia hypothesis” , two hierarchical multiple regression analyses confirmed that alexithymic traits, but not autistic traits, predicted variance in participants’ facial expression production.
As described in the introduction, children with ASD have been shown to produce diminished facial expressions relative to neurotypical children and are less likely to reciprocate other people’s expressions in real-world or experimental settings [18, 20, 22–24]. Our study provides support for Brewer et al.’s  prediction that alexithymic traits may be contributing to reduced facial expression production characteristic of ASD. In Brewer et al.’s study, participants with and without ASD attempted to recognize facial expressions posed by neurotypical individuals and individuals with ASD. They found that both neurotypical participants and participants with ASD were less able to infer the emotions expressed by posers with ASD compared to the expressions of the NT posers, providing support for the notion that individuals with ASD produce atypical, less recognizable facial expressions. While they found that alexithymic and ASD traits were highly correlated in their sample (r = .667), they did not report an association between alexithymia and facial expression production.
Why is alexithymia associated with diminished facial expression production? As discussed in the introduction, the conflict hypothesis proposes that highly alexithymic individuals may suppress their own facial displays to defend against negative affect and to avoid conflict [29, 32, 33]. This explanation would suggest that alexithymic individuals might unconsciously suppress their unfavorable negative emotions such as anger or sadness in an effort to distance themselves from distressing inner emotions and to mitigate potential external conflict. Such a possibility would also predict the expression of non-distressing emotions (e.g., joy) to be unaffected by alexithymia. Our data support this notion, as we observed in our sample that higher alexithymic traits were associated with less negative expressions, but they were not significantly associated with the amount of positive expression. This result is consistent with Rasting et al.’s  findings that the expression of negative emotion, but not positive emotion, was correlated with alexithymia in clinical patients during therapeutic interventions.
It is important to consider that a relative lack of facial expressions does not necessarily indicate a lack of physiological emotional arousal; that is, bodily signals of emotion may be present even in the absence of facial displays of emotion. It has been suggested that alexithymia is characterized by a deficit in interoceptive accuracy despite the fact that behavioral and autonomic reactivity are present [15, 57]. These considerations may help to explain the previously reported relationship between alexithymia and emotion regulation . Not being consciously aware of one’s emotions as they arise, actively suppressing one’s emotions or refusing to acknowledge that one is experiencing negative emotions would inhibit one’s ability to regulate emotions as they increase in intensity—shedding light on the seeming paradox by which alexithymic individuals are prone to displaying minimal nonverbal emotional expression most of the time but are also prone to intense emotional outbursts [4, 27, 46]. As emotion regulation issues—and ensuing emotional outbursts—are common in ASD [59–61], future examinations of the “alexithymia hypothesis” may find that alexithymia is a major contributor to emotion regulation difficulties in people with ASD. Support for this idea comes from research that shows that compared to neurotypicals, participants with ASD are more likely to use suppression (e.g., “I keep my emotions to myself” and “When I am feeling negative emotions, I make sure not to express them”) as an emotion regulation strategy  as measured by the Emotion Regulation Questionnaire . Thus, it appears that a relative disposition towards ignoring or suppressing one’s emotions compounded by poor emotional insight (e.g., alexithymia) may contribute to emotion regulation problems in ASD .
Revisiting the alexithymia hypothesis
When we examined the alexithymia hypothesis using correlational and regression analyses, we observed a clear pattern of results in support of the hypothesis such that alexithymia, but not ASD traits (measured by the AQ), were associated with facial expression production. When we examined the alexithymia hypothesis at the group level of diagnosis, we found that that the ASD group expressed significantly more neutral expression than the neurotypical group, which would be expected considering the ASD group had higher levels of alexithymia. However, no group differences emerged on levels of negative expression. Similarly, no group differences were observed in level of positive expression, but this is consistent with our correlational analyses and the “conflict hypothesis” that would predict positive expression production to be unassociated with alexithymia. In sum, the groupwise comparisons reveal a murkier pattern of results than what was found in the regression analyses. However, these results should be interpreted with caution as the analyses were underpowered as indicated in Table 3. In addition, it is quite possible that there are other factors besides alexithymia that contribute to differences in facial expression production in the ASD population that were unaccounted for in the present study, and future research is needed to uncover these additional mechanisms.
The “alexithymia hypothesis” is intuitively appealing due to its simplicity and robust implications. Bird and Cook  argue that, given the wide heterogeneity in emotion processing abnormalities observed in the ASD population, it may be useful to create diagnostic subtypes of ASD reflecting the presence or absence of alexithymia. This idea is certainly worthy of continued exploration. However, one of the limitations of the alexithymia hypothesis is that it fails to explain why individuals on the autism spectrum are much more likely than the neurotypical population to possess a strong disposition for alexithymia. Alexithymia and ASD are independent constructs, but research is needed to uncover the reasons for why ASD and alexithymia often co-occur. Investigating potential differences between alexithymia in ASD and non-ASD populations may help shed light on this issue, in addition to conducting rigorous developmental and longitudinal research [2, 3]. Indeed, the present study is only the second to examine alexithymia in children with ASD, and no studies to date have examined alexithymia longitudinally in this population. Further, while the present research, consistent with past studies [13–15], shows that alexithymia “predicts” other emotional symptoms in the statistical sense, future research with more sophisticated research designs including longitudinal and structural equation modeling techniques are needed to untangle the relative causal associations among these variables.
Alexithymia is thought to represent an impaired affective representation system, in which the alexithymic individual may be aware they are upset but not able to identify which emotion they are experiencing , perhaps in part due to impaired interoceptive accuracy. Lacking the ability to differentiate internal affective states could inhibit one’s ability to associate affective internal states with perceptual cues that represent the emotional states of others . For example, an fMRI study showed that in individuals with ASD, alexithymia is associated with impairments in the ability to report one’s own emotions, as well as with self-reported empathy, and that this relationship may be due to reduced activity in the anterior insula . This complex interplay between emotion processing in self and other may explain why alexithymia is accounting for atypical emotion recognition, empathy, emotional awareness, and here—expression production—in ASD. By this line of reasoning, alexithymia may contribute to deficits in paired associate learning in which internal representations of affective states are not conditioned to be paired with external representations. Thus, in the context of the present study and past research, alexithymia may contribute to deficits in external representations of emotions in others (recognizing others’ facial expressions) or in the self (representing one’s own emotions with accurate and salient facial expressions).
Deficits in emotion recognition and empathy have been associated with ASD, but recent evidence suggests alexithymia accounts for these effects [13, 14]. As Bird and Viding  describe, empathy involves an “affective coding” process by which an observer’s internal affective state matches that of an object through specialized perceptual systems that are sometimes beneath conscious awareness. Alternatively, emotion recognition requires a “cognitive coding” process by which individuals are consciously aware of and can assign an appropriate label (e.g., anger) to another’s emotional state. This cognitive encoding process could be aided by the affective coding process , by which individuals arrive at emotion attributions by “simulating” others’ emotions in oneself. However, in many social situations, an observer’s emotional state may be very different than that of another. In such cases, individuals must selectively attenuate one’s interoceptive signals in order to prevent interference with the decoding process. . Quatrocki and Fristron review evidence that the oxytocin system is responsible for selective attenuation of interoceptive signals , which may help explain why oxytocin administration improves emotion recognition abilities to a greater extent for individuals who are more highly alexithymic and less emotionally expressive .
Utility of the CAM for ASD research
Although not a main aim of this study, we examined group differences in levels of alexithymic traits between children with and without an ASD diagnosis. We were initially skeptical of the CAM’s utility in the ASD population given the strong verbal component of the items. For example, items on the CAM include “Has trouble finding words or getting words out when talking about his/her own feelings,” “Verbal expressions of feelings do not match non-verbal expressions of feelings,” and “Says “I don’t know” when asked why he/she is upset.” In other words, if participants with ASD scored higher than the neurotypical group on the CAM, would these differences be better explained by true differences in alexithymic traits or by deficits in verbal ability? We found that the ASD group in our sample scored more than one standard deviation higher on the parent-report Children’s Alexithymia Measure (CAM) than the neurotypical control group. Importantly, the two groups were matched on measures of verbal (and nonverbal) IQ, and verbal IQ was not significantly associated with alexithymia, indicating the higher CAM scores in the ASD group were not explained by lower verbal ability. This finding replicates the large group differences in CAM scores between children with and without ASD reported by Griffin et al.  and complements studies showing higher rates of alexithymia in adults with ASD [8, 67].
The present study is just the second to examine alexithymia in children with ASD using the CAM. In addition to using the CAM, Griffin et al.  assessed alexithymia using the Children’s Alexithymia Questionnaire-Self Report (CAQ-SR; ). While both measures were sensitive enough to demonstrate higher levels of alexithymia in the ASD participants, correlational analyses revealed no relationship between the two measures indicating that children and their parents may be using different sources of information or that the content of the respective measures is highly dissimilar. Self-report measures of alexithymia have been criticized because, by definition, highly alexithymic individuals may lack the awareness to accurately assess their own levels of emotional awareness . On the other hand, parent-report measures have the disadvantage of relying on parents to accurately assess their own children’s emotional awareness by indicators such as how often they express and talk about their emotions verbally, which may not accurately reflect how emotionally aware their children actually are. Increased research is needed to evaluate the relative strengths and weaknesses of self and other-report measures of alexithymia in children with ASD.
Given that our groups were matched on levels of verbal ability, we can tentatively conclude that the large group differences in alexithymia were not explained by differences in verbal ability. However, future research may benefit from developing measures that specifically assess emotion-word vocabulary (rather than domain-general vocabulary) as a control variable. It is possible that children with high levels of alexithymia have normal verbal intelligence and vocabulary knowledge but lack sufficient word labels to describe emotional states due to deficient emotional understanding. While continued research is needed to verify the appropriateness of the CAM for research in the ASD population, our initial findings seem promising. We do, however, caution that the ASD participants in our sample had average to above average IQ. It remains a likely possibility that the CAM would not be an appropriate tool for use in lower functioning children with ASD as it would be difficult to distinguish whether high scores on the CAM reflect true alexithymic tendencies or possible language or communication delays.
Our sample size was small, limiting the power to detect group difference in expression production and potentially producing unreliable effects in the between groups analysis. As a result, clear conclusions about whether alexithymia contributes to abnormal facial expression in the ASD population remain lacking. Future research on this topic will benefit from matching ASD and neurotypical groups on alexithymia (e.g., ) to shed more light on this issue. It is also possible that reactions to videos do not authentically capture how expressive participants are in real-world settings. Indeed, many of the studies on facial expressions in ASD reviewed in this article examined facial expression use during naturalistic social interactions. In a similar vein, it is possible that individuals with ASD, in comparison to neurotypicals, display markedly reduced expression only when regulating social behavior (i.e., during social interactions) but are not less expressive when passively viewing stimuli. This possibility offers an interesting possibility for future research.
In addition, our small sample size precluded the possibility of examining within group correlations. In fact, to our knowledge, virtually all studies that have examined alexithymia in ASD have utilized small samples (N < 100). Large-scale studies are urgently needed to more accurately examine how much of the ASD population displays high levels of alexithymia and to further our understanding of the full range of emotional processing differences that may be accounted for by alexithymia within this population. In addition, while the video stimuli we used were effective in that they yielded wide variance in facial expression production in our participants, the ways in which participants express emotions in response to videos may not be reflective of how they express emotions in real-world settings. Future research should replicate the present study in more ecologically valid contexts to explore how facial expressions may be used for empathizing and understanding others’ emotions.
Mentioned earlier, in addition to diminished facial expressions, individuals with ASD also tend to display abnormal facial expressions, which could not be directly assessed from the facial expression analysis software we used. Future research would benefit from finding ways to objectively measure abnormal facial expressions. For example, Ekman and Friesen  and Ekman et al.  have identified 46 specific action units controlled by facial muscles (e.g., “cheek raise” and “nostril dilation”) that combine in various ways to represent each of the basic emotions. An ambitious but worthwhile aim of future research will be to identify combinations of action units that are common in ASD, but atypical in the general population, to identify the precise combinations of facial movements that contribute to abnormal facial expressions in the ASD population. In addition, Brewer et al.  recently introduced a clever paradigm whereby atypical expression production in ASD was measured by how well others could interpret the intended emotion conveyed by facial expressions
A final limitation is that we only assessed alexithymia using a parent-report questionnaire as time limitations of the camp did not allow us to collect self-report data from the participants. Future research should continue to utilize self- and other-report measures.