The number of teenagers with symptoms of psychological stress and associated behaviors is increasing in Japan. For example, teenagers may suddenly become violent, or complain of physical symptoms (e.g., headaches and stomachaches) that do not have a clear physical cause. It has been hypothesized that such phenomena may be related to a limited ability to recognize and verbalize feelings or emotions. Indeed, aggressive children report more difficulties with understanding their emotions 1, and such difficulties are associated with somatic symptoms 234.

A limited ability to recognize and verbalize emotions is generally referred to as alexithymia 5, literally "no words for feelings". This concept was derived from observations of adult, psychosomatic patients, and was first introduced by Sifneos 6. It is defined by four facets: Difficulties in identifying feelings and discriminating between emotions and physical sensations, difficulties in describing feelings, constricted imaginal capacity, and an externally-oriented cognitive style 67. A widely used and extensively validated self-report questionnaire for evaluating alexithymia in adults is the twenty-item Toronto Alexithymia Scale (TAS-20). The TAS-20 has a three-factor structure 89: Difficulty identifying feelings (DIF), difficulty describing feelings to others (DDF), and externally-oriented thinking (EOT).

It has been reported that alexithymia has a developmental aspect: Teenagers appear to have more difficulty identifying and describing their feelings than do adults 10 and their ability to recognize and understand emotion generally develops with age 11. However, only a few studies to date have assessed alexithymia in children and adolescents 1213. These studies found that difficulties with identifying and describing feelings are associated with the somatic complaints of children 12, and that adolescents showed alexithymic characteristics to the same degree as adults (i.e., the prevalence was similar) 13. However, these two studies have several limitations: First, neither factor loading nor internal consistency were reliable for EOT, similar to findings for adult subjects 7. Second, the TAS-20 does not assess a limited capacity for imagination. Third, an examination of the criterion-related validity of the TAS-20 was not done. Finally, no clinical samples were examined and compared to normal subjects.

Thus, the purpose of this study was to assess alexithymia in adolescents and to compare the results to the conventional findings for adults as assessed by TAS-20. Items were added to the TAS-20 that are intended to measure limited capacity for imagination. The criterion-related validity of this new questionnaire was examined by comparing scores with those on a measure of empathic ability. Because empathy has been generally defined as "the ability to understand and share in another's emotional state or context" 14, self-awareness is necessary for its development 15. Indeed, alexithymic individuals are relatively deficient in empathy 161718. As opposed to the deficit seen in people with alexithymia, there may be a strong correlation between the empathy and emotional awareness of people who are not. We thus hypothesized that there would be negative correlations between scores on our TAS-20-based alexithymia questionnaire and empathic ability. Third, we also examined the contribution of linguistic ability to the alexithymic trait, given that children's linguistic abilities are quite possibly related to their emotional understanding 11920. Finally, we examined patients with psychosomatic complaints who were of the same age as the healthy subjects to examine the discriminative validity of our questionnaire.

The purpose of this study was to assess alexithymia in adolescents and to compare the findings to previous findings for adults, taking into account the contribution of linguistic ability to alexithymic traits. We also sought to examine the criterion-related and discriminative validity of the self-reported TAS-20-based questionnaire, QAAA, for adolescents.

An exploratory factor analysis found the QAAA to be comprised of four factors: DIF, EOT, CIC, and DDF. These four factors correspond to the original four facets of the alexithymia construct 6: (a) difficulties in identifying feelings and in discriminating between emotions and physical sensations, (b) an externally oriented cognitive style, (c) constricted imaginal capacity, and (d) difficulties in describing feelings. In contrast to previous findings regarding EOT 12, we found that the factor loadings onto EOT were satisfactorily robust, thus confirming a four factor structure. This structure also has satisfactory reliability, although the MIC of DDF was slightly high, possibly because there were fewer DDF items in comparison with the other factors.

We found that DIF/DDF and EOT/CIC were negatively correlated: The higher scores should convey a deficit in cognition concerning emotions for all four factors related to alexithymia; DIF DDF, EOT and CIC. However, participants who had difficulties with identifying or describing their feelings had (paradoxically) a greater imaginal capacity and reported being more in tune with their inner experiences. These results were apparently different from the adult findings. Because positive correlations were reported between DIF/DDF and EOT among adults 828, this striking contrast in the current study may reflect differences in the developmental process between DIF/DDF and EOT, as we recently reported in a community sample 10. It is possible that alexithymic adolescents, lacking skill at introspection, may overestimate their imaginal capabilities and ability to get in touch with their internal experiences. These findings, however, do not suggest that DIF and DDF scores are of no meaning; as would be the case if the items of the QAAA were not understood by the subjects, who would then answer in a more or less random way, but in fact were answered in an unexpected reverse way.

Higher Japanese linguistic ability was found to be associated with lower EOT and CIC scores but with higher DDF scores, suggesting that subjects with lower linguistic ability do not look much into their inner experiences or cultivate much imagination. In addition, positive correlations between linguistic scores and DDF scores indicated that those with higher linguistic ability attempt to describe their feelings with appropriate words. A positive correlation between DIF scores and emotional understanding 119 was reported among primary school-age children; however, no such association was found in the present study, suggesting that this ability has no effect on the emotional awareness of adolescents. However, it is possible that this disparity may reflect the different educational system in Japan, or perhaps difficulties with the linguistic ability questionnaire itself.

Female students scored higher for difficulties in identifying and describing feelings than did males, but showed more interest in their internal experiences and greater self-reported imaginative ability. No linguistic abilities affected these gender differences. It presents, however, doubt about whether females were not as good at identifying inner emotions, as we recently found higher DIF scores for female adult subjects 10. These lines of evidence clearly support the validity issue of DIF/DDF factors in the self-reported questionnaire. Furthermore, regarding EOT, male adolescents were more externally-oriented, as previously reported in studies of adult males in Asian countries 1029. Additional cross-cultural and developmental studies on gender differences will be needed to clarify possible cultural influences on alexithymic tendencies among adolescents.

As for criterion-related validity, the various factors of the multi-dimensional empathy scale for adolescents showed negative correlations with EOT and CIC scores, consistent with previous findings with adults 18. Students with lower levels of internal focus or with poor imaginations may have lower empathic abilities, and their attention toward psychological processes may be key in terms of the development of various aspects of emotional ability (e.g., emotional intelligence or emotional awareness) 20. This line of evidence suggests that the cultivation of interest in inner experiences or imagination during adolescence should not be neglected.

DIF and DDF scores were positively associated with scores on the multi-dimensional empathy scale for adolescents. Those who reported having difficulty identifying and describing their feelings were more empathic, inconsistent with earlier studies of adult subjects: While high 'personal distress' scores predicted high DIF and DDF scores, high 'empathic concern' predicted low DDF scores. High 'perspective taking' (corresponding to 'cognitive empathy' in the current study) predicted low DIF scores 18. Thus, the criterion-related validity of the DIF/DDF factors was not confirmed for early adolescent subjects using the present questionnaire, indicating that the present DIF/DDF items did not adequately assess difficulty in identifying/describing their feelings.

The items of the DIF and DDF subscales ask about one's meta-cognition: In this case, cognition about one's feelings 30. However, meta-cognitive skill is generally still developing during adolescence 31. Thus, as suggested by Müller et al. 32, it may be difficult to examine whether or not early adolescent subjects have identified and correctly described their feelings, especially by self-report. A high score on DIF may reflect a) actual difficulty identifying feelings that can be verified using objective measures, or b) subjective difficulty in identifying feelings that might be reported among those who have more objective talent. This view is supported by a previous report that alexithymia rates in general decreased from early to middle adolescence 13, consistent with the view that metacognitive ability develops during this time. The current unexpected results, such as the negative correlations between DIF/DDF and EOT/CIC, as well as the positive correlations between linguistic ability and DDF, may reflect inadequate metacognitive abilities among this age group. These results suggest that it is difficult to assess such characteristics in early adolescent subjects by self-reported questionnaire.

Test-retest reliability coefficients showed significant and high correlations, except for DDF. This is likely because this scale consisted of only two items. It will be necessary to develop further items for this scale for use in future studies.

The patients with psychosomatic and/or behavioral problems were significantly more likely to report difficulty in identifying feelings than the normative subjects, irrespective of linguistic ability. Consistent with previous studies, difficulty identifying feelings contributed in large part to the prediction of somatic complaints among children 12 and adults 2. These lines of evidence do appear to support the discriminative validity of the questionnaire. However, as noted earlier, the higher DIF scores for patients do not necessarily indicate that they are alexithymic. The number of somatic complaints reported was not associated with difficulty identifying emotions, but was instead associated with the content of reported emotion (e.g., fear) 33. Moreover, previous studies of adults and adolescents have suggested a relationship between DIF scores and depression and/or anxiety 2834, suggesting that somatic symptoms are affected by negative mood even when alexithymia scores are high 353637. The TAS questionnaire appears to reflect specific aspects of depression or general distress, and a tendency to access and express negative emotions 383940. The evidence suggests that our depressed patients were rather self-critical, and their responses on the questionnaires may have reflected this 41.

In summary, the present findings suggest that alexithymia in early adolescents is not adequately assessed by current self-report questionnaires based on the TAS-20. This is particularly the case for the DIF and DDF subscales. In the current study more than half of the items were eliminated after exploratory factor analysis due to low factor loadings, suggesting the necessity of modifying the content of these subscales or perhaps creating a new item pool.

On the basis of our observations, we conclude that self-report questionnaires for the measurement of alexithymia in early adolescents require further careful investigation; performance measures will be needed. Moreover, the present findings also suggest that the assessment of metacognitive ability and other emotional abilities (even among adult populations) merit further research attention. Finally, although alexithymia has been considered to be a universal characteristic, the present results suggest the possibility of different developmental trajectories of alexithymia as a function of culture.

The authors declare that they have no competing interests.

HN collected and analyzed the data, performed the statistical analysis and drafted the manuscript. GK conceptualized, designed and conceived the study. TI and YM participated in the design of the study and collected the normative sample data. SK and TA collected the clinical data. All authors have read and approved the final manuscript.