Alexithymia in autism spectrum disorder

ABSTRACT

Objective

Alexithymia is a trait characterised by difficulty identifying and describing one’s own emotions and externally orientated thinking. Alexithymia is of clinical interest in people with autism spectrum disorder (ASD) given research that has highlighted elevated levels of overall alexithymia in people with ASD. Presently, little is known about what specific facets of alexithymia might be impaired in ASD, or whether deficits are present for both negative and positive emotions. This study therefore aimed to fill this gap, establishing a facet-level profile of alexithymia in people with ASD.

Method

Using the Perth Alexithymia Questionnaire, levels of alexithymia were assessed in sample of 55 people with a diagnosis of ASD and compared with 246 people in a community control sample.

Results

We found that all facets of alexithymia (across both valence domains) were substantially elevated in ASD (N = 55) compared to a community control sample (N = 246).

Conclusions

Assessing all facets of alexithymia, across both valence domains, may help identify subgroups with particular social and communication difficulties, and in turn, support the development of personalised interventions.

Key Points

What is already known on this topic:

1. Alexithymia is a multidimensional construct, comprised of at least three interrelated components: difficulty identifying one’s own feelings (DIF), difficulty describing feelings (DDF), and an externally orientated thinking style (EOT).

2. The Perth Alexithymia Questionnaire (PAQ) is a psychometrically sound measure of alexithymia.

3. Alexithymia commonly occurs in people with Autism Spectrum Disorder (ASD).

What this paper adds:

1. In the ASD group, alexithymia was higher across all facets (i.e., DIF, DDF, EOT) compared to the community group, supporting the elevated levels of alexithymia in people with ASD.

2. The PAQ may be important for recognising variability and individual differences among people with ASD, as whilst one third of participants with ASD scored in the “high alexithymia” range, two thirds did not.

3. Identifying unique profiles of alexithymia may have key implications for personalised treatment and interventions.

KEYWORDS:

• Alexithymia
• autism spectrum disorder (ASD)
• Perth alexithymia questionnaire
• PAQ
• emotion processing

Introduction

The term alexithymia, meaning “without words for emotions” in Greek, was coined by psychoanalysts, Nemiah and Sifneos (1970), to describe a cluster of emotion processing deficits commonly seen in psychiatric patients. Empirical research has since strongly supported the status of alexithymia as a multidimensional personality trait, comprising at least three core facets: difficulties identifying one’s own feelings (DIF); difficulties describing feelings (DDF); and externally orientated thinking (EOT) whereby one tends to not focus their attention on their emotions (D. A. Preece et al., 2020; D. Preece et al., 2017; Watters et al., 2016). Some authors also consider difficulties daydreaming and reduced emotional reactivity (difficulties emotionalising) to be additional core components of alexithymia (e.g., Vorst & Bermond, 2001). However, over the past few decades, most empirical data has not supported the inclusion of these components within the alexithymia construct. In factor analyses, difficulties daydreaming and difficulties emotionalising do not load on the same general alexithymia factor as DIF, DDF, and EOT; alexithymia is typically uncorrelated with daydreaming frequency, and is associated with higher (not lower) levels of reactivity for negative emotions (for a review, see Preece et al., 2020). Consequently, most alexithymia research has operationalised alexithymia with measures that assess only DIF, DDF, and EOT (e.g., Bagby et al., 1994). In this study, we define alexithymia according to the attention-appraisal model of alexithymia (D. Preece et al., 2017), which conceptualises the construct as being comprised of only DIF, DDF and EOT. This model has a high level of empirical support (for a review, see Luminet et al., 2021).

The alexithymia trait is normally distributed in the general population, with around 10% having high levels of alexithymia to a clinically significant level (Taylor et al., 1999). Rates of high alexithymia are more prevalent in samples with mental health conditions (e.g., depression, anxiety, psychosomatic, eating, and personality disorders), thus underscoring the status of alexithymia as an important transdiagnostic risk factor (Poquérusse et al., 2018). Considering the overlap between alexithymia traits, and the communication and social skills deficits in people with autism spectrum disorder (ASD), significant interest has recently emerged in relation to the role of alexithymia and ASD symptomology (Kinnaird et al., 2019; Poquérusse et al., 2018). While the relationship is not yet well understood, there is typically a high level of alexithymia in people with ASD (i.e., around 40–65% of people with ASD having high levels of alexithymia; Bird & Cook, 2013), leading to the proposed notion that some deficits in ASD may be driven by co-occurring alexithymia (Kinnaird et al., 2019). For instance, some research has reported associations between higher levels of alexithymia and reduced social interactions (Gerber et al., 2019), poor facial recognition (Cook et al., 2013), impaired interoception (Shah et al., 2016), impaired affective theory of mind (Pisani et al., 2021), and impaired recognition of the emotional valence of verbal and nonverbal vocal cues (Lindner & Rosén, 2006; Philip et al., 2010). Alexithymia is, by definition, specific to difficulties processing one’s own emotions rather than difficulties recognising other people’s emotions (i.e., affective theory of mind), but many models of emotional intelligence highlight that the capacity to recognise one’s own emotions provides an important foundation for understanding emotions in others (e.g., Bird & Cook, 2013; Israelashvili et al., 2019). Such findings accord well with the well supported “alexithymia hypothesis” of ASD, suggesting that the emotional deficits in ASD may be (at least in part) an effect of the greater proportion of elevated alexithymia levels in populations with ASD (Bird & Cook, 2013).

Having identified a relationship between alexithymia and ASD, we argue that it is important to gain a better understanding of the specific characteristics of alexithymia (i.e., facet-level profile) in this population. To date, most studies of alexithymia in relation to ASD have used the 20-item Toronto Alexithymia Scale (TAS-20) to assess alexithymia. However, the TAS-20 was originally designed to just provide a total scale score (i.e., as an overall marker of alexithymia), and the scale developers advise against deriving or using any facet-level scores (i.e., subscale scores) (Bagby et al., 2008). For instance, on occasions when facet-level scores have been derived, the TAS-20 EOT facet has not met standard reliability thresholds (e.g., Kooiman et al., 2002). That is, whilst it is well established that overall levels of alexithymia are typically elevated in ASD, it remains unclear whether this reflects elevations in particular facets of alexithymia, or whether all facets might be impaired. Indeed, elsewhere in the alexithymia field, clinicians and researchers are becoming increasingly interested in examining alexithymia at the facet level (e.g., Goerlich, 2018; Greene et al., 2019), as this aligns with the multidimensional nature of the construct and may better inform precision medicine approaches to treatment (Collins & Varmus, 2015).

An additional gap concerns whether emotion processing deficits in ASD might be valence-specific (i.e., difficulties in the processing of negative emotions, positive emotions, or both). The TAS-20 does not specify an emotional valence in most of its items, and recent statistical work has highlighted that it assesses alexithymia only with respect to negative emotions (see D. Preece et al., 2020). As such, the extent to which ASD alexithymic difficulties may extend to the positive valence domain remains unclear. Outside the TAS-20 there are several other well-validated self-report measures (e.g., Perth Alexithymia Questionnaire [PAQ; D. Preece et al., 2018]; Bermond-Vorst Alexithymia Questionnaire (Vorst & Bermond, 2001)) or observer-rated measures of alexithymia (e.g., Toronto Structured Interview for Alexithymia); however, to date only the recently introduced PAQ provides valence-specific assessments of the construct, and thus it may have promising utility in this area.

The present study

Our aim was to address the abovementioned gaps by comprehensively examining the facet-level profile of alexithymia in ASD. To do this, we utilised the Perth Alexithymia Questionnaire (PAQ; D. Preece et al., 2018), which was designed to assess all facets of alexithymia and do so across both negative and positive emotions. Psychometric studies conducted across a range of sample types have consistently supported the validity and reliability of the PAQ, at both the overall and subscale levels (e.g., Becerra et al., 2021; Fynn et al., 2022; Greene et al., 2020; D. Preece et al., 2020). As such, we compared PAQ scores across an ASD sample and a community control sample, to establish the extent to which deficits were present for specific alexithymia facets and/or specific emotional valences.

Method

Participants and procedure

ASD sample

Participants with ASD were recruited from a local Australian NGO which provides social skills coaching programs to young adults over 18 years and have received a diagnosis of ASD. The sample comprised 55 people (74.5% male, 25.5% female) aged between 18 and 25 years (M = 21.33, SD = 2.43). Approximately 81.8% percent were Caucasian, with more than half (60%) having completed secondary school as their highest level of education, and 32.7% a tertiary degree or diploma.

Community control sample

Participants in the control sample were adults recruited by an online survey company (Qualtrics Panels) to be representative of the Australian population in terms of gender, age, and geographic region. The sample comprised 246 people (51.6% male, 48.4% female) with an average age of 50.82 (range = 18–84). Most of the sample were Caucasian (84.6%). In terms of highest level of education, for 19.1% it was year 12 high school, and for 44.3% it was a tertiary degree or diploma. One third of participants (33.3%) reported that they had a history of a mental health disorder (mostly depression and/or anxiety). One additional participant (not included in the final sample number cited above), listed that they had an ASD diagnosis, so they were excluded from the sample to maximise differentiation between the community control sample and the ASD sample. Participants completed the PAQ as part of an online survey.

Materials

The Perth alexithymia questionnaire

The PAQ (D. Preece et al., 2018) is a 24-item self-report measure of alexithymia designed to measure DIF, DDF, and EOT. Each item is answered on a 7-point Likert Scale, with higher scores indicating higher alexithymia. The PAQ is comprised of five subscales, as valence-specific subscales are available for the DIF and DDF facets: Negative-Difficulty Identifying Feelings (N-DIF; “When I’m feeling bad, I can’t tell whether I’m sad, angry or scared”), Positive-Difficulty Identifying Feelings (P-DIF; “When I’m feeling good, I get confused about what emotion it is”), Negative-Difficulty Describing Feelings (N-DDF; “When I’m feeling bad, I can’t talk about those feelings in much depth or detail”), Positive-Difficulty Describing Feelings (P-DDF; “When something good happens, it’s hard for me to put into words how I’m feeling”), and General-Externally Orientated Thinking (G-EOT; “I prefer to just let my feelings happen in the background, rather than focus on them”). These subscales can also be summed into a total scale score, as an overall marker of alexithymia. Previous Australian adult norms have indicated an average PAQ total score of 81.97 (SD = 30.91) (N = 748; D. Preece et al., 2018); alexithymia scores in our present study within 1SD of this normative mean can thus be considered “average range”, scores 1SD above as “high alexithymia”, and scores 1SD below as “low alexithymia”. All PAQ subscales and the total scale score had good reliability in both our samples (Cronbach’s α > .80).

Analytic strategy

To examine the differences in PAQ scores between the two groups, we conducted a Multivariate Analysis of Covariance (MANCOVA). The five PAQ subscale scores (N-DIF, P-DIF, N-DDF, P-DDF, G-EOT) were included as the dependent variables. Participant age and gender were included as covariates to control for possible demographic effects; controlling for demographic effects can be important in this area, as some previous work has highlighted alexithymia levels can, on average, be higher in males and those of older age (e.g., Mattila et al., 2006; Salminen et al., 1999).

In the event of a significant overall MANCOVA, to isolate the source of the effect, we conducted five follow-up Univariate Analyses of Covariance (ANCOVAs), each using one of the five PAQ subscales as the dependent variable. In the interest of completeness, we also conducted a further ANCOVA using the PAQ total scale score as the dependent variable (i.e., to examine differences in overall levels of alexithymia). Effect sizes were judged based on partial η² values with the following criteria: .01 = small, .06 = medium, .14 large (Cohen, 1988).

Results

Our MANCOVA highlighted that there was a significant difference between the ASD and community groups on an overall linear composite of the five PAQ subscale scores (F(5, 293) = 5.187, p < .001, partial η² = .081; medium effect size).¹ Our follow-up ANCOVAs highlighted that the ASD group had significantly higher levels of alexithymia on all five subscales of the PAQ: N-DIF (F(1, 297) = 15.427, p < .001, partial η² =. 049; small effect size), P-DIF (F(1, 297) = 24.200, p < .001, partial η² = .075; medium effect size), N-DDF (F(1, 297) = 12.824, p < .001, partial η² = .041; small effect size), P-DDF (F(1, 297) = 22.443, p < .001, partial η² = .070; medium effect size), and G-EOT (F(1, 297) = 6.762, p < .010, partial η² = .022; small effect size). Our ANCOVA of the PAQ total scale score also confirmed that overall levels of alexithymia were significantly higher in the ASD group (F(1, 297) = 20.235, p < .001, partial η² = .064; medium effect size). Indeed, around one-third (32.7%) of the ASD group scored in the “high alexithymia” range on the PAQ total scale score (60% in average range, 7.3% low range) compared to only 10.2% of the community group (71.1% in average range, 18.7% low range). Estimated marginal means for both groups (adjusted to control for demographic covariates) are provided in Table 1.

Table 1. Estimated marginal means.

	ASD Group		Community Group
Scale	Estimated Marginal Mean	Standard Error	Estimated Marginal Mean	Standard Error
N-DIF	17.31	1.00	12.78	.40
P-DIF	15.42	.88	10.70	.35
N-DDF	19.41	1.07	14.92	.43
P-DDF	17.67	.97	12.63	.39
G-EOT	32.86	.73	27.74	.73
Total	102.66	4.75	78.09	1.91

Covariates in the model were evaluated at the following values: Age = 45.43, Gender = 1.44 (1 = male, 2 = female).

Discussion

Our study identified that all facets of alexithymia (i.e., DIF, DDF, EOT), across both the negative and positive emotional valence domains, were impaired in people with ASD compared to a community sample. These findings are consistent with results of existing work using the TAS-20, which highlighted elevated levels of overall alexithymia in people with ASD (e.g., Griffin et al., 2016; Milosavljevic et al., 2016). Crucially, by extending this work to also examine the facet-level and valence-specific profiles of alexithymia with the PAQ, our novel study adds key depth to the understanding of alexithymia in people with ASD. This facet-level detail is important, as previous research outside the ASD field has recognised alexithymia as a multifaceted construct that should consider valence (e.g., Greene et al., 2020). Our data may therefore help to enrich theoretical models of ASD. For example, our findings are consistent with the alexithymia hypothesis of ASD (Bird & Cook, 2013). Within this popular model, given that emotion processing of both negative and positive emotions is core to social interaction and understanding in daily life (e.g., Costa et al., 2017; Poquérusse et al., 2018) and conceptually helps to underpin theory of mind skills (e.g., Demers & Koven, 2015; Pisani et al., 2021), our findings provide empirical support that deficits across all facets of alexithymia may contribute to the poor social functioning characterising ASD. That is, deficits in both the frequency with which one focuses attention on their own emotions (EOT), as well as one’s capacity to accurately identify and describe their own negative and positive feelings (DIF, DDF), each likely contribute to these social difficulties.Our findings thus underscore the importance of comprehensive alexithymia assessments as part of routine clinical practice in working with people with ASD. Such assessment may be important for identifying and accounting for individual-level variability, as whilst many of our ASD group participants (32.7%) had high levels of alexithymia, around two-thirds did not. As more studies take this approach, this may help to identify subgroups of ASD that have particular strengths or weaknesses across their facet-level alexithymia profiles (e.g., informed by using techniques like latent profile analysis with larger ASD samples), thus enabling a more personalised approach to treatment focused on the most prominent domains of difficulty (Fernandes et al., 2017). However, in the interim as a default approach, our data suggest that treatment approaches for alexithymia in ASD should ideally be prepared to address all facets of the construct across both valence domains. That is, treatment approaches aiming to improve both the frequency with which one focuses attention on negative and positive emotions (EOT), as well as the capacity to accurately appraise negative and positive emotions (DIF, DDF). This might take the form of skills-based treatment programs or therapy modalities, including psychoeducation about emotions, therapist guidance around the labelling of different emotional states, and emotion-focused experiential exercises like “mindfulness of emotions” (for a more detailed discussion of alexithymia treatment approaches, see D. Preece et al., 2017). Moreover, there are some data outside the ASD field suggesting that high levels of alexithymia (if not accounted for in treatment) may impair the effectiveness of treatments for other psychiatric disorders (e.g., Leweke et al., 2009). Going forward, it will be important to examine the extent to which this is also the case for ASD, and the extent to which understanding the facet-level profile of ASD clients may help streamline treatment effectiveness.

Limitations and future directions

One of the most salient limitations is the relatively small sample size in the ASD group. The ASD sample were also disproportionately male and were all young adults, thus precluding detailed analysis of demographic differences. As past work in the ASD field has highlighted gender and age effects (Wijngaarden Cremers et al., 2014), future work drawing upon larger sample sizes will be needed to determine the generalisability of our preliminary results regarding alexithymia. As we controlled for age and gender statistically, demographic differences between the ASD and community group were accounted for in some way; however, the use of matched samples in future research may also improve generalisability. In future, larger samples will also enable detailed analyses of individual-level variability in alexithymia patterns (i.e., using techniques like latent profile analysis), as opposed to a focus on overall group differences. Furthermore, rather than using self-report exclusively, future work might benefit from a multi-method approach, incorporating structured interviews for alexithymia (e.g., Bagby et al., 2006) and/or parent/informant reports (e.g., Griffin et al., 2016). It also bears mentioning that our participants with ASD may not be representative of all ASD people, given these participants’ ability to participate in a social skill-based coaching program and successfully complete a self-report questionnaire. As such, our findings may be limited to those on the higher functioning end of the autism spectrum, who are known to have stronger language and communication skills (Kinnaird et al., 2019).

Conclusions

Overall, our facet-level and valence-specific analyses suggest that all facets of alexithymia, across both negative and positive emotions, can be impaired in people with ASD. As the emotion-related and social interaction deficits in ASD may be driven by co-occurring alexithymia, our findings point to the importance of comprehensive assessments of alexithymia as part of routine clinical practice in people with ASD. A breakdown of alexithymia profiles using valence-specific and facet-level analysis in people with ASD may have key implications for personalised treatment and interventions. It is suggested that deriving profiles and subgroups may be an effective treatment strategy in targeting specific social and communication difficulties in people with ASD. Moving forward, larger scale studies will assist in continuing to determine the precise role of alexithymia as a multidimensional construct in neurodiverse populations.

Ethical approval

Ethics approval was granted by the University Human Research Ethics Committee (approval number 2000000340). The guidelines of this committee were followed. All participants provided consent for their data to be used. The authors declare no conflicts of interest.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Gender was not a significant covariate (p = .397). Age was a significant covariate (p = .004).