Cross-cultural validation and measurement invariance of the Perth Alexithymia Questionnaire (PAQ): a study in Iran and the USA

ABSTRACT Objective Alexithymia is a trait defined by difficulty in identifying and describing feelings, as well as externally oriented thinking. It is an important transdiagnostic risk factor for a range of psychopathologies, and therefore its assessment is of substantial interest. Recently, the Perth Alexithymia Questionnaire (PAQ) was developed to try to enable more comprehensive assessments of alexithymia. To date, no studies have examined the PAQ’s psychometric properties among adolescents, and few have examined non-Western populations. Method To address these gaps, here we examined the psychometric properties of the PAQ among three samples of Iranian adolescents (N = 557, 53% female, Mage = 14.94, SD = 1.29), Iranian adults (N = 926, 62% female, Mage = 32.52, SD = 9.65) and American adults (N = 242, 40% female, Mage = 40.69, SD = 11.91). Results Confirmatory factor analysis supported the intended five-factor model (that distinguished between different facets of alexithymia across positive and negative emotions) within all three samples. This five-factor model was invariant across genders, ages and cultural groups. Furthermore, the PAQ subscales showed good internal consistency, test–retest reliability and concurrent validity. Conclusions Overall, the present findings indicate that PAQ has strong psychometric properties among both Middle Eastern and Western samples, and functions similarly across adults and adolescents. The PAQ therefore appears to be a useful tool for comprehensively operationalising alexithymia across a diverse range of populations. Key Points What is already known about this topic: Heightened alexithymia contributes to the development and maintenance of numerous forms of psychopathology. The assessment of alexithymia has been limited as it primarily focuses only on negative emotions. The Perth Alexithymia Questionnaire (PAQ) was recently developed to provide an integrated and valence-sensitive assessment of alexithymia. What this topic adds: The PAQ can be used to measure alexithymia in both adults and adolescents. The intended five-factor structure of the PAQ, which distinguishes between different facets of alexithymia across positive and negative emotions, was supported. The intended factor structure of the PAQ was found to be invariant in terms of gender, age, and cultural groups.

Alexithymia is a multidimensional construct that is characterised by difficulty identifying one's own feelings (DIF), difficulty describing one's own feelings (DDF) and an externally oriented thinking style (EOT) whereby one tends to not focus on internal emotional states (Apfel & Sifneos, 1979;Preece et al., 2020). 1  Alexithymia was first documented by psychoanalytic psychiatrists, who coined the term to describe the emotion processing deficits of patients suffering from psychosomatic disorders (Nemiah & Sifneos, 1970;Sifneos, 1973).Subsequent work has demonstrated that the relevance of alexithymia is not limited to psychosomatic patients, and that it is a trait that is normally distributed in the general population with around 10% of people having problematically high levels of alexithymia (Parker et al., 2008).Twin studies have indicated that around one-third of the variance in alexithymia levels is attributable to genetic factors, with the remainder being environmental factors (Jørgensen et al., 2007).Alexithymia appears to impair emotion regulation, with alexithymia characterised by high usage of avoidant and maladaptive emotion regulation strategies (Preece, Mehta, Petrova, Sikka, Bjureberg, Becerra, et al., 2023), and is a vulnerability factor for numerous other forms of psychopathology (Preece et al., 2022;Tesio et al., 2019).For example, high levels of alexithymia are associated with heightened depression (Celikel et al., 2010;Honkalampi et al., 2001), substance abuse and alcohol dependence (Cruise & Becerra, 2018;de Bruin et al., 2019), posttraumatic stress disorder (Passardi et al., 2019), eating disorders (Speranza et al., 2007;Westwood et al., 2017), as well as the severity of symptoms among general psychiatric samples (McGillivray et al., 2017).
The contribution of alexithymia to a wide range of disorders indicates the critical importance of its study for researchers and clinicians, thus requiring precise assessment for its operationalisation.To date, the most commonly used instrument to examine alexithymia over the past two decades has been the 20-item Toronto Alexithymia Scale (TAS-20; Bagby et al., 1994), a self-report measure originally developed to assess overall alexithymia levels by providing a total scale score (Bagby et al., 2007).However, the need for a more detailed assessment of alexithymia (i.e., robust assessment of the DIF, DDF and EOT facets) has been increasingly acknowledged (e.g., Preece et al., 2017;Goerlich, 2018).It has been argued that a comprehensive assessment of alexithymia needs to capture alexithymia across both negative and positive emotions, as evidence from empirical work has shown that one's capacity to process negative emotions is not necessarily equivalent to one's capacity to process positive emotions (Barrett et al., 2001).This is in line with meta-analytic findings revealing that alexithymia for positive and negative emotions is characterised by different neural circuits (van der Velde et al., 2013).The TAS-20 was not designed to provide valence-specific scores, or scores for each facet of alexithymia (whilst subscale scores are often derived, the TAS-20 authors recommend against the use of subscale scores; Bagby et al., 2007).Indeed, recent work suggests that the TAS-20 only assesses alexithymia with respect to negative emotions (see Chan et al., 2022;Preece et al., 2020), and when subscales are extracted from the TAS-20, its EOT items have unacceptably low reliability (alpha <.70; e.g., Kooiman et al., 2002).There are also some concerns about the discriminant validity of the TAS-20 against measures of distress, as some of its items refer to somatic symptoms (e.g., "I have physical sensations that even doctors don't understand"), and these items have been found to overlap with measures of depression and anxiety in factor analysis (e.g., Leising et al., 2009;Marchesi et al., 2014;Preece et al., 2020).
In an attempt to enhance the comprehensiveness of alexithymia assessments, Preece et al. (2018a) recently developed the Perth Alexithymia Questionnaire (PAQ).The PAQ is a 24-item self-report measure designed to assess the DIF, DDF and EOT facets of alexithymia. 2One novelty of the PAQ relative to older alexithymia tools is that it was designed to enable facet-level assessments (i.e., analysis across DIF, DDF and EOT) and valence-specific assessments (i.e., analysis across both negative and positive emotions) of alexithymia.As such, separate subscale scores for positive and negative emotions can be derived for the DIF and DDF facets.This results in the PAQ having five subscales: Negative-Difficulty identifying feelings (N-DIF, four items; e.g., "When I'm feeling bad, I can't tell whether I'm sad, angry, or scared"), Positive-Difficulty identifying feelings (P-DIF, four items; e.g., "When I'm feeling good, I can't make sense of those feelings"), Negative-Difficulty describing feelings (N-DDF, four items; e.g., "When I'm feeling bad, I can't talk about those feelings in much depth or detail"), Positive-Difficulty describing feelings (P-DDF, four items; e.g., "When something good happens, it's hard for me to put into words how I'm feeling") and General-Externally orientated thinking (G-EOT, eight items; e.g., "I tend to ignore how I feel").These subscales can also be combined into a total scale score, as an overall marker of alexithymia.
The psychometric properties of the PAQ have so far been examined in several studies, mainly consisting of Australian and American adults (e.g., Fynn et al., 2022;Greene et al., 2020;Preece et al., 2018aPreece et al., , 2020;;Preece et al., 2020).In all these studies, the analyses supported the theoretically congruent factor structure of the PAQ, consisting of five factors corresponding to the five intended subscales.Moreover, all subscales and composite scores of the PAQ in these studies have displayed good reliability and concurrent validity.For example, as would be expected given the clinical relevance of alexithymia, significant associations have been found between elevated alexithymia (as operationalised by the PAQ) and greater depression, anxiety and stress symptoms, as well as greater emotion regulation difficulties, and other measures of overall alexithymia, like the TAS-20 (e.g., Chan et al., 2022).
In addition to the psychometric studies with English-speaking samples, researchers have started to examine the PAQ's utility among other cultures too (Bilge & Bilge, 2020;Chan et al., 2022;Larionow et al., 2022;Lashkari et al., 2021;Mousavi Asl, Mahaki, et al., 2020), including Persian-speaking samples.To date, two studies have examined the psychometric properties of the PAQ among Iranian adults (Lashkari et al., 2021;Mousavi Asl, Mahaki, et al., 2020).Both of these studies (using either an army sample or a university student sample) found support for the intended factor structure, and reported good reliability and validity for the PAQ; however, the generalisability of their findings is constrained by the samples they recruited.
These results are promising, but further psychometric studies of the PAQ are needed with more diverse participant groups.For example, to date, the PAQ has been examined only among adults, and no published studies have examined its utility for adolescents, despite the pivotal conceptual role of alexithymia in the development of different forms of internalising and externalising symptoms during adolescence (Bordalo & Carvalho, 2022;Prino et al., 2019;Sfeir et al., 2020).This is an important endeavour, as the developers of another popular alexithymia measure, the TAS-20, recommend against the use of the TAS-20 in adolescent samples based on psychometric problems highlighted in this population type (Parker et al., 2010).Moreover, no studies have directly examined the measurement invariance of the PAQ between samples from Middle Eastern and Western samples or different demographic categories (e.g., gender, age).Measurement invariance testing determines if a measure assesses the same construct in the same way across different groups and is ideally needed to facilitate confident comparisons of measure scores across different groups (Putnick & Bornstein, 2016).
To address these gaps in the PAQ literature, the current study was designed to examine the psychometric properties of the Persian version of the PAQ (Lashkari et al., 2021) among Iranian adults and adolescents, as well as an American adult sample.We tested its factor structure, measurement invariance (across different culture, gender and age groups), internal consistency, test-rest reliability and concurrent validity.For concurrent validity, relationships were examined between alexithymia and emotion regulation ability; emotion regulation strategy use; and depression, anxiety and stress symptoms.In addition, the relationship between the PAQ and another alexithymia measure, the TAS-20, was computed as part of concurrent validity analysis.

Participants and procedure
Ethics approval for this project was granted by the University of Western Australia Human Research Ethics Committee and Babol University of Medical Sciences in Iran.All participants provided informed consent for their data to be used.For the Iran adolescent sample, parents completed the consent form for their adolescents to participate in the study.Three samples of participants were recruited.

Iran adolescent sample
The adolescent sample included 672 participants who were recruited using a convenience sampling method from three elementary schools in two cities (Gilan and Tehran) in Iran.They completed the Persian version of the questionnaire via Porsline (https://survey.porsline.ir/), an online survey platform.The same data cleaning and participant exclusion criteria were used as with the PAQ original psychometric study (Preece et al., 2018a).Specifically, participants with incorrect responses to an attention check question (which asked them to select a specific scale response) and those who completed the questionnaires too quickly for attentive responding (i.e., less than 2 s for each item; see Becerra et al., 2020;Preece et al., 2018a) were excluded.The final sample consisted of 557 adolescents (52.96% female) that were between 12 and 17 years old (mean age = 14.94,SD = 1.29).Adolescent participants were given bonus course credits for completing the survey.

Iran adult sample
The Iranian adult general population sample consisted of 1040 participants that were selected using a convenience sampling approach.They completed the Persian version of the online survey that was posted on different social media platforms.One hundred and forty participants were excluded from quality screening because they failed an attention check question, or completed questionnaires impossibly quickly, indicative of inattentive responding (i.e., same criteria as indicated above for adolescent sample).The final sample consisted of 926 participants (61.98% female; mean age = 32.52,SD = 9.65, range = 18-65).Iranian adult participants were given access to five monetary prize draws for their participation.

United States adult sample
The American sample was recruited using Amazon Mechanical Turk (MTurk; Litman et al., 2017).They completed the English version of the questionnaires as part of an online survey.Two hundred and sixtyeight participants completed the survey.After excluding participants with inattentive responses (i.e., failed attention check or completed too quickly) using the same exclusion criteria with Iranian samples, the final sample consisted of 242 participants (40.08% female; mean age = 40.69,SD = 11.91,.American participants' marital status was not recorded.They were compensated US$3 for participation.

Perth Alexithymia Questionnaire (PAQ)
The PAQ is a 24-item self-report tool that measures alexithymia across both positive and negative emotions (Preece et al., 2020).Participants rate each item on a 7-point Likert scale (1 = strongly disagree; 7 = strongly agree), with higher scores indicating greater alexithymia.In addition to the total scale score, the PAQ provides scores for five dimensions of alexithymia: N-DIF, P-DIF, N-DDF, P-DDF and G-EOT.Good convergent and discriminant validity, and good internal consistency have previously been reported for PAQ (Preece et al., 2018a(Preece et al., , 2020)).A standard translation procedure was followed for the Persian version of the PAQ (Wild et al., 2005).First, a native Persian speaker psychologist (the first author) translated the English version of the PAQ, which then was back-translated into English by an independent translator and checked by the developers of the original PAQ.Copies of both Persian and English versions of the PAQ with scoring instructions are provided in the supplementary materials.

Toronto Alexithymia Scale (TAS-20)
The TAS-20 is a self-report measure of alexithymia (Bagby et al., 1994).Participants respond using a fivepoint Likert scale (from 1 = strongly disagree to 5 = strongly agree) with higher scores indicating greater alexithymia.It consists of 20 items and was originally developed to only measure overall alexithymia levels via a total scale score; it is now standard practice for separate subscale scores to also be derived (DIF, DDF, EOT), though in response to criticism of low reliability among the EOT items (e.g., Kooiman et al., 2002) the developers of the TAS-20 caution that it is best used only as a total scale score (Bagby et al., 2007).Various studies have shown adequate psychometric properties of the TAS-20 (Taylor et al., 2003).The Persian version of TAS-20 demonstrated good validity and reliability in the Persian-speaking population, although the EOT subscale has shown low internal consistency (Besharat, 2008;Khosravani et al., 2019).

Perth Emotion Regulation Competency Inventory (PERCI)
The PERCI is a 32-item measure of emotion regulation ability across both positive and negative emotions (Preece et al., 2018a).Participants rate items on a 7-point Likert scale (1 = strongly disagree to 7 = strongly agree), with higher scores indicating more emotion regulation difficulties or poorer emotion regulation ability.In addition to a total score, PERCI provides separate emotion regulation ability scores for positive (16 items, e.g., "When I'm feeling bad, I'm powerless to change how I'm feeling") and negative emotions (16 items, e.g., "When I'm feeling good, I have no control over whether that feeling stays or goes").The PERCI has shown good validity and internal consistency in previous studies (Preece et al., 2018b(Preece et al., , 2021)), and the Persian version of the PERCI has indicated the same factor structure and good psychometric properties (Mazidi et al., 2023).

Emotion Regulation Questionnaire (ERQ)
The ERQ is a 10-item questionnaire that measures habitual use of two emotion regulation strategies, reappraisal (six items, e.g., "I control my emotions by changing the way I think about the situation I'm in") and suppression (four items, e.g., "I keep my emotions to myself") (Gross & John, 2003).Items are rated on a 7-point Likert scale, and higher scores on each subscale indicate higher usage of the corresponding strategy.The ERQ has shown good validity and reliability in different studies (Melka et al., 2011;Preece et al., 2020).The Persian version of the ERQ has also shown good psychometric properties (Hasani, 2016).The adolescent sample completed the modified version of the ERQ for children and adolescents (ERQ-CA; Gullone & Taffe, 2012).The modifications for the ERQ-CA include simplification of the item wording and reduction of the response scale from seven to five points (Gullone & Taffe, 2012).The Persian version of the ERQ-CA has shown good validity and internal consistency among Persian adolescents (Khatibi et al., 2021;Lotfi et al., 2019).

Depression Anxiety and Stress Scale-21 (DASS-21)
The DASS-21 is a 21-item self-report measure of depression, anxiety and stress symptoms over the past week (Lovibond & Lovibond, 1995).Participants were asked to answer the items using a 0 (did not apply to me at all) to 3 (apply to me very much) scale.Three separate subscale scores can be derived for each symptom category, and all items can also be summed into a total scale score as an overall marker of psychological distress.Both the English and Persian versions of the DASS-21 showed acceptable construct and convergent validity as well as internal consistency (Antony et al., 1998;Habibi et al., 2017;Kami et al., 2019).

Factorial validity
Confirmatory factor analyses (CFA; robust maximum likelihood estimation with the Satorra-Bentler [SB] scaled χ2 statistic) were conducted using the lavaan package (Rosseel, 2012) for R version 4.0.2.We tested the five-factor model endorsed in previous studies (e.g., Lashkari et al., 2021;Preece et al., 2020;Preece et al., 2020), corresponding to the five intended subscale scores (see Figure 1).We also tested some simpler models as comparative baselines, to examine whether separating different valence domains and subscale categories adds statistical value within the latent structure of alexithymia (Preece et al., 2020(Preece et al., , 2021)).These simpler models consisted of a onefactor model, where all 24 items loaded on a single general alexithymia factor; a two-factor model that only differentiated between the attention (EOT) and appraisal (G-DIF/DDF stages of emotion processing); a three-factor model that distinguished between the DIF, DDF and EOT facets of alexithymia but did not distinguish between negative and positive valence; and an alternative three-factor model that distinguished positive and negative valences but combined the DIF and DDF items together.Model goodness-of-fit was judged based on three fit indices: the comparative fit index (CFI), the root mean square error of approximation (RMSEA) and the standardised root mean residual (SRMR).CFI values ≥0.90 were judged to indicate an acceptable fit (and ≥.95 excellent), as were RMSEA and SRMR values ≤0.08 (and ≤ .06excellent) (Bentler & Bonett, 1980;Browne & Cudeck, 1992;Marsh et al., 2004).The models were also directly compared using the Akaike Information Criterion (AIC), which penalises for model complexity, and lower values indicate a better fitting model (Byrne, 2016).Factor loadings ≥0.40 were considered meaningful loadings (Stevens, 1992).

Measurement invariance
To examine the measurement invariance of the PAQ across gender, age and culture, the best fitting factor model was tested separately for each group in the three samples (Joshanloo & Bakhshi, 2016).Then, the basic configural invariance model (equal form) was tested followed by progressively more restrictive measurement invariance tests: a metric invariance test (equal factor loadings) and a scalar invariance test (equal intercepts).Models were compared in terms of change in CFI.Invariance was indicated when an absolute difference in CFI (ΔCFI) was less than 0.01 between the configural, metric and scalar models (Cheung & Rensvold, 2002).

Internal consistency and temporal stability
Cronbach's alpha (α) and McDonald's Omega (ω) reliability coefficients were calculated for all PAQ subscales and the total score.Values ≥.70 were judged acceptable, ≥.80 good and ≥.90 excellent (Groth-Marnat, 2009).For temporal stability, 79 of adolescents completed PAQ again with a 2.5-week interval.The Intra-Class Correlation (ICC) was selected to measure the test-retest reliability as it is a more precise method compared to Pearson correlation (Portney & Watkins, 2009;Shrout & Fleiss, 1979).Model (3, 1), which is based on a two-way mixed effect ANOVA, is appropriate when the same subjects score themselves on two or more occasions.For ICC, values between 0.50 and 0.75 indicate moderate reliability, values between 0.75 and 0.90 indicate good reliability, and values greater than 0.90 indicate excellent reliability (Koo & Li, 2016;Portney & Watkins, 2009).

Relationships with other constructs/measures
Pearson correlations were calculated between PAQ scores and DASS-21, PERCI, ERQ and TAS-20 scores.Because models of alexithymia implicate it as a critical factor in the development and maintenance of affective disorders (Hendryx et al., 1991;Honkalampi et al., 2001;Scimeca et al., 2014), we expected a positive association between PAQ scores and greater depression, anxiety and stress symptoms.Similarly, we expected that greater PAQ scores would be associated with higher PERCI scores, which indicates poorer emotion regulation ability, as it has been shown that difficulties in identifying and describing emotions can compromise emotion regulation ability (Edwards & Wupperman, 2017;Preece et al., 2022;Swart et al., 2009).Regarding emotion regulation strategy use, we expected high PAQ scores would be associated with more use of suppression (i.e., an avoidant-type strategy that is usually linked to poorer mental health) and less use of reappraisal (i.e., an adaptive strategy that is usually linked to good mental health; Aldao et al., 2010;Cutuli, 2014;Gross & John, 2003).A positive association between PAQ and TAS-20 scores was also expected as both are designed as measures of alexithymia (Greene et al., 2020).

Descriptive statistics and reliability coefficients
Descriptive statistics and reliability coefficients for all PAQ subscales and the total scale score (for each sample) are displayed in Table 1.All PAQ subscales and the total score showed acceptable to excellent alpha and omega internal consistency reliabilities for the three samples (α range from 0.74 to 0.97, and ω range from 0.75 to 0.97).The ICC values for the test-retest reliability of the PAQ among the adolescents sample were also high for the N-DIF, P-DIF, N-DDF, P-DDF, G-EOT and total scale (0.80, 0.78, 0.85, 0.76, 0.73 and 0.88, respectively), indicating good test-retest reliability. 3

Factor structure
Fit indices for all CFA models for the three samples are displayed in Table 2.The intended five-factor model was the best fitting model, and a good fit to the data according to all fit indices.All items loaded highly on their intended subscale factor (i.e., >.40; see Table 3), except items 6, 8 and 15 for the adolescents sample that were just below the threshold, between 0.3 and 0.4.This should be noted that some researchers consider factor loadings above 0.3 also acceptable (e.g., Floyd & Widaman, 1995;Hair et al., 2010).All factors were significantly positively correlated for all the three samples (see supplementary material).The five-factor model was better fitting than the other tested models in all three samples, confirming the statistical value of distinguishing between the different valence categories and subscale components within the latent structure of alexithymia.

Measurement invariance
Next, the measurement invariance of the five-factor model was tested across gender, age and culture.To examine the measurement invariance across genders, the five-factor model was tested separately for males  , respectively, the values that have been considered problematic for CFA (Harrington, 2009).A table containing kurtosis and skewness for each variable is reported in the supplementary material.(n = 614) and females (n = 869) in a data-set that included Iranian adult and adolescent samples.The two samples were combined to increase power for the analysis. 4As the model showed acceptable fit indices in both groups, we proceeded with the configural invariance test.Equality constraints were then imposed on all factor loadings.As shown in Table 4, the ΔCFI (=0.001) indicated full metric invariance.Next, equality constraints were imposed on all item intercepts to test scalar invariance, indicating full scalar invariance.The same procedure was followed for the measurement invariance of age, which was carried out using Iranian adolescents (n = 557) and Iranian adults (n = 926) to control for potential effects of culture.The five-factor structure was also invariant across age categories, as the CFI values did not differ substantially (i.e., <.01) across the configural, metric and scalar models.Next, the measurement invariance for culture was tested across Iranian adults (n = 926) and American participants (n = 242).There was configural and full metric invariance for culture.However, at the scalar level, the ΔCFI exceeded the 0.01 criterion, indicating some level of non-invariance.Inspection of the modification indices suggested that freeing the equality constraints on the intercept for EOT item 11 would improve the fit of the model.As can be seen in Table 4, after doing so, the ΔCFI (=0.009) indicated that partial scalar invariance for culture was achieved.The intercept for item 11 was higher for the American sample (b = 2.76) than Iranian participants (b = 2.15).

Latent mean differences
Given that the five-factor model showed scalar measurement invariance, we proceeded with comparing latent means across participant groups, using the parameters of this model.In this model, the latent factor means in Iranian adult and adolescent samples were constrained to zero, whereas the latent means in the American sample were freely estimated.The results indicated that the American sample scored significantly lower than both Iranian adults and adolescent groups for the N-DIF, P-DIF, N-DDF and P-DDF factors (all unstandardised fitted means (American sample) ≥ −1.26, SEs ≤ 0.11, ps < 0.001).There were no significant group differences for the G-EOT factor.To compare the latent means between Iranian adult and adolescent samples, the latent factor means in the Iranian adult sample were constrained to zero, whereas the latent means in the adolescent sample were freely estimated.The results showed significant differences between the two groups only on the N-DIF factor, with adolescents reporting significantly lower scores than Iranian adults (unstandardised fitted mean (Iranian Adolescent Sample) = −0.30,SE = 0.08, p < 0.001).To compare the scores between male and female participants, the latent factor means of male participants were constrained, while the latent means of females were freely estimated for each sample separately.No significant difference was found for the American participants (all unstandardised fitted means (American females) ≥0.16, SEs ≤ 0.19, ps ≥0.34).For the Iranian participants, males reported significantly higher levels of alexithymia than females on the G-EOT, P-DIF, N-DDF and P-DDF factors (all unstandardised fitted means (Iranian females) ≥ −0.36, SEs ≤ 0.07, ps ≤0.01).
There were no significant group differences for the N-DIF factor.

Concurrent validity
Pearson correlations between the PAQ and other measures are displayed in Table 5. Greater levels of alexithymia (PAQ scores) were associated with greater depression, anxiety and stress in all three samples.Higher Alexithymia was also associated with more difficulty in emotion regulation for both positive and negative emotions, and overall emotion regulation.All PAQ subscales and total scores showed significant positive correlations with TAS-20 subscales and total scores, except the PAQ N-DDF and P-DDF subscales that did not show a significant correlation with the EOT subscale of TAS-20.In terms of the emotion regulation strategies, greater alexithymia was associated with more suppression use among all three samples.There were also significant negative correlations between alexithymia and the use of reappraisal in the American sample, but no significant correlation was found between alexithymia and reappraisal among the Iranian samples.The complete Pearson correlation matrix is provided in Supplementary Materials.

Discussion
The current study examined the psychometric properties of the PAQ among Iranian and American adults, as well as a sample of Iranian adolescents.Overall, the PAQ performed well in all these sample types, indicating that it appears to function well as a marker of alexithymia across a diverse range of groups.
Regarding the factorial structure of the PAQ, the intended five-factor model (reflecting the intended five-subscale structure) was found to be the best fitting model in all three samples, thus replicating past work mainly in English-speaking adult samples (e.g., Preece et al., 2020).The superiority of the five-factor model over other models supports the significance of distinguishing between the processing of negative and positive emotions, as well as the importance of considering alexithymia as a multidimensional construct that includes separable DIF, DDF and EOT components.Importantly, this was also found to be the case among the adolescent sample, as well as both adult samples, suggesting that PAQ can be used to study alexithymia among adolescents too.The PAQ performed similarly across the Western and Middle Eastern cultures we examined here, underscoring its potential utility in cross-cultural work.It bears noting  that a few (three) of the PAQ items had lower loadings (below 0.4) in the adolescent sample, though this did not seem to impact the overall scale performance meaningfully: these items remained close to the 0.4 threshold, all subscales still had good reliability, and overall measurement invariance was supported across age groups.Currently, it is difficult to speculate about the reasons for this finding due to the scarcity of studies on PAQ with adolescents (i.e., currently no other published studies).Future replication of these patterns in other samples will be required before firm conclusions can be drawn.
Importantly, the current study also addressed the measurement invariance of the confirmed five-factor model of the PAQ across gender, age groups and the two studied cultures.Full metric and scalar invariance was obtained for gender and age, and full metric and partial scalar invariance was present across cultures.As such, this supports that the latent structure of the alexithymia construct, as assessed by the PAQ, manifests similarly across these demographic categories (Putnick & Bornstein, 2016).This is important because it shows that this instrument can be confidently employed to measure and compare alexithymia levels between individuals who differ in these demographic backgrounds, and paves the way for more robust cultural studies in this field using the PAQ to obtain valence-specific alexithymia information (Cheung & Rensvold, 2002).
In terms of reliability, acceptable to excellent internal consistency was found for all subscales and total scores of the PAQ amongst the three studied samples, similar to the reliability figures reported for the original English version (Preece et al., 2018a).Importantly, the PAQ did reliably assess the EOT component of alexithymia in all three samples, a component of alexithymia that has often had lower reliability when assessed with other common alexithymia tools (e.g., TAS-20; Bagby et al., 1994).Another novel contribution of our study was the examination of the PAQ's test-retest reliability among adolescents, as previous work has tended to focus just on internal consistency reliability.Thus, our finding that the PAQ demonstrated acceptable to good test-retest reliability scores for all subscales and the total scale score is consistent with the status of alexithymia as a relatively stable trait (Taylor et al., 1997).
In terms of concurrent validity, the PAQ showed expected correlations with other measures.Specifically, higher levels of alexithymia were associated with greater depression, anxiety and stress, more overall difficulty in emotion regulation, and more use of suppression strategies in emotion regulation.This is in line with previous theorising that alexithymia is a risk factor for psychopathology and that a primary pathway behind this link is the impairing impact of alexithymia on emotion regulation (Preece et al., 2022).Interestingly, difficulty in identifying and describing negative emotions showed stronger associations with difficulty in regulating negative emotions than positive emotions, while more difficulty in identifying and describing positive emotions was more strongly associated with difficulty in regulating positive emotions, thus showing some level of valence specificity across constructs.In terms of emotion regulation strategy use, it bears noting that in terms of reappraisal use, only the American participants showed a significant negative association between alexithymia (for both the PAQ and TAS-20) and use of the reappraisal strategy, as no significant association was found between alexithymia and reappraisal among Iranian participants.This might reflect cultural differences in the utility or appropriateness of certain emotion regulation strategies (e.g., Ford & Mauss, 2015;Matsumoto et al., 2008;Ramzan & Amjad, 2017) and could be an important focus of future work.Finally, as predicted, PAQ scores were highly positively associated with the scores from the TAS-20, reinforcing that these measures both assess a similar alexithymia construct.
Taken together, our results therefore highlight that the PAQ has the potential to serve as a valuable tool for researchers and clinicians in obtaining a more comprehensive facet-level and valence-specific profile for alexithymia.In recent years, clinical researchers have become increasingly interested in examining alexithymia at a component level to identify distinct psychopathology or brain injury categories that may have specific DIF, DDF and EOT profiles (e.g., Leweke et al., 2012;Williams & Wood, 2010).Such profiles can help to inform the targeting of interventions to improve specific aspects of emotion processing.For example, patients presenting with high alexithymia scores on the PAQ might benefit from interventions designed to enhance the developmental level of their emotion schema systems (i.e., those cognitive structures used in processing emotions) and reduce their use of experiential avoidance of emotions as an emotion regulation strategy (Preece et al., 2017).These are core mechanisms thought to underly alexithymia, and mechanisms that can be targeted in therapy (e.g., via psychoeducation about the different components of emotions, experiential exercises like mindfulness of emotions, and therapeutic discussions where the therapist guides the patient in labelling their emotional states) (see Preece et al., 2017;Samur et al., 2013).Previous studies assessing alexithymia in research or clinical settings have primarily relied on the use of the TAS-20, thus limiting the ability to draw confident conclusions about different alexithymia facets and emotional valences (as the TAS-20 was not designed to provide assessment at the facet level; Bagby et al., 2007).

Limitations and future directions
We think our study makes a useful contribution to the alexithymia field; however, the findings of the current study must be considered in the light of some limitations.First, the samples recruited for the current study were not from clinical populations, and the present findings may not be generalisable to clinical samples; this will be an important direction for future research.Second, we did not have access to an American adolescent sample, which would be ideal for future comparisons.Third, the PAQ's testretest reliability was not assessed for the Iranian and American adult samples, so it is unclear how it operates in adults in that context.Fourth, all the measures we employed were self-report questionnaires (Dang et al., 2020;Jeong et al., 2018); future studies could use behavioural and laboratory-based methods to further examine the validity of the PAQ and related constructs (e.g., Farhoumandi et al., 2021).Fifth, as is the case for many clinical measures, the model fit indices for the proposed five-factor model of PAQ revealed acceptable fit (e.g., CFI >.90) for both Iranian adolescents and adults, but they did not meet the stricter criteria sometimes used for excellent fit levels (e.g., CFI >.95).Therefore, future studies could explore ways to further enhance model fit for data from Iranian populations.It should also be noted that a limitation of the present study was the use of convenience sampling to recruit participants, which may limit the generalisability of the findings.Future studies could benefit from using more diverse or representative sampling methods to increase the generalisability of their results.

Conclusions
Overall, our results indicate that the PAQ has good psychometric properties across Middle Eastern and Western populations, as well as across adults and adolescent age groups.These findings reinforce the validity of the alexithymia construct across diverse groups, and indicate that the PAQ may be a useful option for comprehensive assessments of this clinically relevant construct.

Table 1 .
Descriptive statistics and Cronbach's alpha and McDonald's omega reliability coefficients for the administered measures.

Table 2 .
Goodness-of-fit index values from confirmatory factor analyses of the PAQ.

Table 3 .
Standardised item factor loadings for all PAQ items and subscales (Five-factor model).

Table 4 .
Measurement invariance for the five-factor model across gender, age and culture.