Assessment of Alexithymia: Self-Report and Observer-Rated Measures Graeme J. Taylor, R. Michael Bagby, and Olivier Luminet
Assessment of Alexithymia: Self-Report and Observer-Rated Measures Graeme J. Taylor, R. Michael Bagby, and Olivier Luminet
Assessment of Alexithymia: Self-Report and Observer-Rated Measures Graeme J. Taylor, R. Michael Bagby, and Olivier Luminet
301-
319). San Francisco, CA: Jossey Bass. 2000
Freyberger, & Sifneos, 1976), there has been controversy over its measurement. Several
projective techniques, and a Q-sort (Block, 1961/1978). Some of the self-report scales, such as
the Schalling-Sifneos Personality Scales (Apfel & Sifneos, 1979; Sifneos, 1986) and the
MMPI Alexithymia Scale (Kleiger & Kinsman, 1980), were constructed hastily and with little
concern for standard methods of test construction. Not surprisingly, subsequent investigations
found that these scales lack adequate reliability and validity (see Taylor, Bagby, & Parker,
1997). Although several investigators have used the Rorschach and/or the Thematic
Apperception Test (TAT) to assess various facets of the alexithymia construct, there is little
Since most of the measures of alexithymia have been reviewed previously (Linden et al.,
1995; Taylor & Bagby, 1988; Taylor et al., 1997), we will focus this chapter on the self-report
Twenty-Item Toronto Alexithymia Scale (Bagby, Parker, & Taylor, 1994a), which has become
the most widely used measure of the construct. Recognizing the need for a multi-method
approach for assessing a construct, we will also describe the observer-rated Beth Israel Hospital
will conclude the chapter with a description of two new measures -- the self-report Bermond-
Vorst Alexithymia Questionnaire (Bermond & Vorst, 1998) and the self- and observer-rated
California Q-set Alexithymia Prototype (Haviland & Reise, 1996a) -- which were not included
in previous review articles. Given the strong overlap of the alexithymia construct and the
emotional intelligence construct (which we discussed in Chapter 3), some of the measures of
alexithymia reviewed in this chapter can also be considered potential methods for identifying
The Twenty-Item Toronto Alexithymia Scale (TAS-20) was developed by Bagby et al.
(1994a) and is a revised version of the earlier 26-item Toronto Alexithymia Scale (TAS; Taylor,
Ryan, & Bagby, 1985). The TAS-20 has demonstrated good internal consistency and test-retest
reliability. In the initial validational study, exploratory factor analysis of the TAS-20 with a
student sample yielded a three factor structure congruent with the theoretical construct of
alexithymia: (F1) difficulty identifying feelings and distinguishing between feelings and the
bodily sensations of emotional arousal; (F2) difficulty describing feelings to others; (F3)
externally-oriented thinking. Despite the absence of items on the TAS-20 directly assessing
daydreaming and other imaginal activity, which were included on the TAS-26, the third factor,
together with factor 2, seem to reflect the pensée opératoire (operatory thinking) component of
the alexithymia construct, viz., a cognitive style that shows a preference for the external details of
everyday life rather than thought content related to feelings, fantasies, and other aspects of a
person's inner experience (Marty & de M'Uzan, 1963; Nemiah et al., 1976). Subsequent
research showed that the externally-oriented thinking factor correlates significantly and
negatively (r = -0.45, p < 0.01) with the fantasy subscale of the openness to experience
dimension in the NEO Personality Inventory (Bagby, Taylor, & Parker, 1994b). Given that high
scores on the fantasy subscale reflect a vivid imagination and capacity to create an interesting
inner world, this finding indicates that the externally-oriented thinking factor of the TAS-20
adequately assesses the constricted imaginal processes facet of the alexithymia construct.
The replicability of the three-factor structure of the TAS-20 has been demonstrated with
both clinical and nonclinical populations by the use of confirmatory factor analysis (Bagby et
al., 1994a; Parker et al., 1993). Although the first two factors correlate highly, a three-factor
model provided a better fit to the data obtained from several different samples than either a one-
or two-factor model. The TAS-20 has also been translated into many languages using the
method of back translation to establish cross-language equivalence. The validity of the three
factor structure has been demonstrated in these translated versions by confirmatory factor
analyses for the following languages: German (Bach et al., 1996; Parker et al., 1993), Hindi
3
(Pandey et al., 1996), Italian (Bressi et al., 1996), Korean (Lee, Rim, & Lee, 1996), Lithuanian
(Beresnevaite et al., 1998), Portuguese (Prazeres, Parker, & Taylor, in press), Spanish (Páez et
Notwithstanding the evidence for the replicability of the three-factor structure of the
TAS-20, and the considerable empirical and theoretical justification that have produced the scale,
some clinicians and researchers still question the suitability and usefulness of its three factor-
derived scales (e.g., Deary et al., 1997; Erni et al., 1997; Haviland & Reise, 1996a; Sifneos,
1996). Loas et al. (1996) for example, conducted principal components analysis on data
collected from students at a French university and obtained a two factor solution; the items
assessing difficulty identifying feelings and difficulty describing feelings constituted a single
factor, and the items assessing externally-oriented thinking comprised a second factor. It must
be remembered, however, that exploratory factor analyses are theory weak compared to
confirmatory factor analyses, as they generate different factor solutions from which the
researcher selects the most sensible, rather than evaluating a priori models (Kline, 1991). When
confirmatory factor analysis was applied to the same French data, the original three factor
structure of the scale was found to provide a better fit than a two factor solution (Loas et al.,
1997).
Haviland and Reise (1996b) conducted confirmatory factor analyses on data sets from
medical students and psychoactive substance-dependent inpatients, and reported that the three
factor solution provided a poor fit to the data in both samples. In addition, in the substance-
dependent sample, the correlations between factors 1 and 3, and between factors 2 and 3, were
nonsignificant. Examination of the results for the medical student sample, however, reveals that
one of the goodness of fit indices met its criterion standard and two other indices were just
extremely unstable population for a factor analytic study, as they were recently abstinent from
alcohol or psychoactive drugs and completed the TAS-20 within their first week of
Given that the results of factor analysis of a scale can be influenced by the type of
subjects selected (Nunnally, 1978), and that the original derivation sample in the construction of
4
the TAS-20 was a homogeneous group of undergraduate university students, although cross-
validated with a clinical sample, it is possible that a different factor solution might emerge with a
more heterogeneous population. Moreover, some investigators have found TAS-20 scores to be
associated with male gender and age (Lane et al., 1998; Salminen et al., 1999). As Nunnally
(1978) points out, such variables also might influence the factor structure of a scale.
These issues were addressed in a recent study by Parker, Taylor, and Bagby (in
preparation), who administered the TAS-20 to a community sample of 1,933 adults (880 men
and 1053 women). The results of confirmatory factor analyses replicated the three-factor model
of the TAS-20, and the three factor model provided a better fit to the data than both a one-factor
model and a two-factor model. Multi-sample confirmatory factor analyses showed that the
three-factor model of the TAS-20 fit well for both men and women, and equivalent models were
obtained when the sample was divided at the median age of 32 years into younger and older
respondents.1
Some researchers have criticized the TAS-20 (or the earlier 26-item TAS) on the
grounds that the factor scales do not always show similar relationships with other constructs
(e.g., Deary et al., 1997; Haviland et al., 1994; Hendryx, et al.,1991; Kirmayer & Robbins, 1993;
Lane et al., 1998). When evaluating the convergent validity of a measure of a multifaceted
construct, however, one can expect the full scale and its factors to be related in similar ways only
to measures of closely related constructs (see Carver, 1989). Some facets are likely to relate to
other constructs better than the broad construct, and some may even be unrelated to other
constructs.
As we summarized in Chapter 3, strong support for the convergent validity of the TAS-
20 and for each of its three factors was provided by the findings of significant negative
correlations with the overlapping constructs of psychological mindedness, need for cognition,
and affective orientation, as well as with the openness to experience dimension in the five factor
model of personality (Bagby et al., 1994b; Taylor et al., 1997). In addition, the TAS-20 and its
1
Copyright on the TAS-20 is held by G.J. Taylor, R.M. Bagby, and J.D.A. Parker.
Information for ordering the scale may be found at www.gtaylorpsychiatry.org
5
three factors were shown to correlate significantly and negatively not only with the total score on
the BarOn Emotional Quotient Inventory (EQ-i; Bar-On, 1997), but also with the four second-
order factors assessing intrapersonal intelligence, interpersonal intelligence, adaptability skills, and
Other studies have examined relationships between the TAS-20 and measures of less
closely related constructs. Not surprisingly, the patterns of correlations that emerge for the three
factors depend on the nature of the other construct. For example, on theoretical grounds, one
would expect factors 1 and 2 of the TAS-20 to be associated with constructs related to emotional
experience or expression, and factor 3 to be associated with creative interests and imaginal and
analytical capacities. This discriminability among the factors has been demonstrated in studies
examining the relationship between alexithymia and the five factor model of personality. In a
university student sample, for example, only factors 1 and 2 correlated positively with the
neuroticism dimension of the NEO Personality Inventory (Bagby et al., 1994b). All three factors
correlated negatively with the positive emotions subscale of the extraversion dimension, but only
factors 1 and 2 correlated negatively with the full dimension of extraversion. While all three TAS-
20 factors correlated negatively with the receptivity to feelings subscale of the openness to
experience dimension, only factor 3 correlated negatively with the openness to aesthetics and
openness to ideas subscales. None of the TAS-20 factors were related significantly to the
Given their difficulty in identifying feelings and distinguishing between feelings and the
bodily sensations that accompany states of emotional arousal, alexithymic individuals are
considered prone to functional somatic symptoms (Taylor et al., 1997). This tendency, however, is
most likely assessed by factor 1 of the TAS-20, and to some extent by factor 2, with minimal
contribution from the externally-oriented thinking factor. This was demonstrated in a student
sample by Martínez-Sánchez (1996), who found that the full scale TAS-20 and factors 1 and 2
common physical symptoms and bodily sensations; Pennebaker & Skelton, 1978), whereas factor
3 did not. In a mixed clinical and nonclinical sample, Deary et al. (1997) also found that factors 1
and 2 of the TAS-20 correlated positively with a checklist of medically unexplained symptoms,
6
whereas factor 3 did not. However, such findings are not always consistent across studies. Bach,
Bach, & Zwaan (1996), for example, found that factor 2 of the TAS-20 does not always correlate
significantly with functional somatic symptoms, and factor 3 may sometimes show a significant
positive correlation.
The finding of strong negative correlations between the TAS-20 and the EQ-i suggests
that the TAS-20 could be used in the assessment of emotional intelligence, at least as a brief
screening device for identifying individuals with low emotional intelligence. To further support
this recommendation, we summarize some additional findings from a study by Parker, Taylor,
and Bagby (submitted) in which they constructed latent models to examine the relationship
between the alexithymia construct (represented by TAS-20 scores) and Bar-On’s (1997) broad
concept of emotional intelligence (represented by EQ- i scores). Given that some researchers (e.g.,
Deary et al., 1997) have questioned whether the three factors of the TAS-20, as a whole, best
represent the alexithymia construct, the same model was tested separately for each of the three
factors from the TAS-20. It was predicted that the TAS-20 total score and the scores for each of
its three factors would be independent of, but strongly and inversely associated with, the total
For each of the models tested, all of the goodness-of-fit indices met pre-established
criteria standards. The parameter estimates between the EQ-i and the TAS-20 and its factors
were -0.94 (p < 0.001) for the total TAS-20, -0.78 (p < 0.001) for factor 1 (difficulty identifying
feelings), -0.70 (p < 0.001) for factor 2 (difficulty describing feelings), and –0.55 (p < 0.001)
for factor 3 (externally-oriented thinking). As with the Pearson correlations between the TAS-20
and the EQ- i (see Chapter 3), the strength of these parameter estimates indicates considerable
overlap of the two constructs, and suggests that the TAS-20 could be a useful measure in the
assessment of emotional intelligence. For each of the latent models tested, however, a two-factor
solution was superior to a one-factor solution, indicating that the construct measured by the
TAS-20 is also independent of the construct measured by the EQ-i. As Mayer, Salovey, and
Caruso (in press) point out, the EQ-i is based on a model of emotional intelligence which
combines mental abilities concerned with understanding emotions with a diverse set of
7
personality characteristics that Bar-On (1997) relates to the potential to succeed in life.
Although the TAS-20 may correlate with many of these personality characteristics, they are not
part of the definition of the alexithymia construct; nor are they included in the models of
emotional intelligence proposed by Salovey and Mayer (989/1990) and Mayer, Salovey, and
Caruso (in press). Future research might investigate the relationship between the TAS-20 and
measures of emotional intelligence that restrict themselves to mental abilities concerning the
One of the more challenging problems in the assessment of a personality trait that is
ensure that its measurement is not confounded by the state effects of the illness. Some
investigators have argued that the presence of alexithymia may merely reflect a concomitant state
reaction to an illness, which may be predicted by state anxiety, a depressed mood, or lowered
quality of life, and lessens over time as the illness improves (Haviland et al., 1994; Keltikangas-
Järvinen, 1987; Wise et al., 1990). However, several longitudinal studies have yielded strong
support for alexithymia being a stable trait that is independent of psychological distress or other
effects of a medical or psychiatric illness. Salminen et al. (1994), for example, followed a group
of patients with anxiety and depressive disorders and found that the mean TAS score was
unchanged after one year, even though there was a significant decrease in the mean score on a
al. (1988) found no significant change in the mean TAS score over a three-week treatment
period despite a significant drop in the mean score on the Beck Depression Inventory (BDI).
Similar findings were reported by Pinard et al., (1996), who administered the TAS-20 and the
BDI to a group of substance dependent patients before and after a 4-6 week period of treatment,
and by Porcelli et al. (1996), who had a group of patients with inflammatory bowel disease
complete the TAS-20 and the Hospital Anxiety and Depression Scale before and after six
Notwithstanding the positive findings from these longitudinal studies, none have
examined the distinction between absolute and relative stability, an important concept in trait
psychology. Absolute stability refers to the extent to which personality scores change over time,
whereas relative stability indicates the extent to which the relative differences among individuals
remain the same over time (Santor et al., 1997). Evidence of relative stability in the context of
Treatment studies represent an ideal way to test for relative stability since profound
changes are expected. In a recent study, Luminet, Bagby, and Taylor (submitted) evaluated the
absolute and relative stability of alexithymia in a sample of patients who entered a treatment
program for major depression. Depression was used as the “disease” state on which to
examine the stability of alexithymia as several studies with clinical or nonclinical populations
have reported positive and significant relationships between the TAS-20 and measures of
depression. It was hypothesized that depression and alexithymia would be correlated at both
treatment initiation (baseline) and at follow-up (treatment completion), and that both constructs
would show significant reductions from baseline to treatment completion. However, it was
hypothesized also that alexithymia scores at baseline and treatment completion would be highly
correlated, and that alexithymia assessed at baseline would still predict alexithymia assessed at
treatment completion even after controlling for the effects of depression. Such a finding would
In a sample of 46 outpatients who met DSM-IV criteria for major depressive disorder
and were treated with antidepressant medication for 14 weeks, there was no significant
0.09); at treatment completion, however, the two measures did correlate significantly (r = 0.38, p
< 0.01). As expected, there was a significant and substantial reduction in depressive symptoms
over 14 weeks of treatment, mean change score = -11.31, t(45) = 10.54, p < 0.0001, as
measured by the Hamilton Rating Scale for Depression (HRSD). There was also a smaller but
still significant change in alexithymia as measured by the TAS-20, mean change score = -5.11,
9
t(45) = 2.61, p < 0.01. Despite this large reduction in depression severity, the relative stability of
alexithymia was demonstrated by a positive and significant correlation between TAS-20 scores
Luminet et al. (submitted) used a hierarchical regression analysis with TAS-20 scores at
treatment completion as the criterion variable. TAS-20 scores at baseline and HRSD scores at
baseline and treatment completion served as the predictor variables. HRSD scores were forced
into the model first and accounted for approximately 14% of the variance [F(1, 44) = 7.31, p <
0.01]. TAS-20 scores at baseline were then entered into the model and accounted for an
additional 34% of the variance [F(2,43) = 26.41, p < 0.001]. These results indicate that although
alexithymia scores may change in the context of a marked reduction in depression, there is
strong evidence for the relative stability of alexithymia despite such changes. Such findings,
together with results from the various longitudinal studies, support the view that the TAS-20 is
The Beth Israel Hospital Psychosomatic Questionnaire (BIQ) was developed by Sifneos
observer. Of these 17 questions, Sifneos (1973) selected eight "key" items that best assess
alexithymic characteristics; higher scores on these items indicate higher degrees of alexithymia.
The method of interviewing, as outlined by Nemiah et al. (1976), involves an initial period of
As studies evaluating the psychometric properties of the BIQ have been reviewed
elswhere (Linden et al., 1995; Taylor et al., 1997), we will summarize only the essential findings,
and then describe the development and preliminary testing of a modified version of the
questionnaire. Factor analysis of the BIQ, and correlations with total word count in response to
the TAT or Rotter Sentence Completion Test (as indices of the capacity to fantasize), have
10
provided partial support for the construct validity of the BIQ (Gardos et al., 1984; Lesser et al.,
1979; Sriram et al., 1987). In addition, the BIQ has demonstrated concurrent validity with the
26-item TAS in both clinical and nonclinical populations, with the magnitude of the correlations
ranging from 0.39 to 0.77 (p < 0.01) (Fukunishi, Saito, & Ozaki, 1992; Jimerson et al., 1994;
Kauhanen, Julkunen, & Salonen, 1992; Sriram Pratap, & Shanmugham, 1988). There is less
support, however, for the reliability of the scale. Although Sriram et al. (1988) demonstrated
adequate internal consistency and test-retest reliability, there is evidence that interrater reliability
is influenced by the experience, bias, and style of the interviewer (Taylor & Bagby, 1988).
Moreover, studies that used independently rated audio- or videotapes of a single interview have
reported high interrater reliability of the BIQ, whereas unacceptably low interrater reliabilities
have been reported in studies that used separate interviews to rate the same patients (see Taylor
et al., 1997).
Notwithstanding these limitations, Linden et al. (1995) concluded that the BIQ shows
attempt to improve the reliability and validity of the BIQ, some researchers have recommended
several modifications to the questionnaire. These include revising the item content, using
guidelines for rating responses. An important step in this direction was initiated by Sriram et al.
(1988), who proposed a set of guidelines and probes for rating the eight key items of the BIQ.
Following Sriram et al. (1988), Bagby et al. (1994b) recently developed a modified
version of the BIQ by adding four new items for rating alexithymia and eliminating nine of the
original 17 items that are less relevant to the construct. In addition, the rating scale was changed
from a dichotomous format to a 7-point Likert type format to enhance the reliability. The
resulting 12 item questionnaire comprises six items pertaining to the ability to identify and
verbally communicate feelings (i.e., affect awareness), and six items pertaining to imaginal
2
The modified BIQ is published as an Appendix in Taylor et al., (1997).
11
Even without probe questions and rating guidelines, Bagby et al. (1994b) obtained
concurrent validity of the modified BIQ was demonstrated by significant positive correlations
with the TAS-20 (r = 0.53, p < 0.01) and its three factor scales. The BIQ subscale assessing
affect awareness correlated with factor 1 (r = 0.43, p < 0.01) and factor 2 (r = 0.52, p < 0.01)
of the TAS-20; and the BIQ subscale assessing operatory thinking correlated with factor 2 (r =
0.58, p < 0.01) and factor 3 of the TAS-20 (r = 0.30, p < 0.05) (Bagby et al., 1994b).
As several other measures were completed by the same group of behavioural medicine
outpatients, it was possible to examine the relationships of the modified BIQ and its subscales
with general intelligence and with neurotic psychopathology. General intellectual functioning
was assessed with the Shipley Institute of Living Scale (SILS; Zachary, 1986), which provides
subscores for vocabulary and abstract concept formation. Neurotic psychopathology was
measured with the Crown-Crisp Experiential Index (CCEI; Crown & Crisp, 1979). The
modified BIQ and its two subscales were unrelated to the total score and vocabulary score on
the SILS; however, the subscale assessing affect awareness was significantly and negatively
related to the abstract thinking score (r = -0.32, p < 0.05). The modified BIQ and its two
subscales were unrelated to the total score on the CCEI; this result is consistent with findings
from a study with anorexia nervosa patients in which the 26-item TAS also was unrelated to the
The psychometric properties of the modified BIQ were subsequently evaluated in Japan
by Fukunishi et al. (1997) in samples of college students and psychiatric outpatients. Principal
components factor analysis yielded a two-factor solution, which accounted for approximately
41% of the total variance in both samples, and corresponded to the two subscales identified by
Bagby et al. (1994b). In addition, and in both samples, the modified BIQ and its two subscales
demonstrated adequate internal consistency (Cronbach alpha coefficients ranged from 0.72 to
0.85) and test-retest reliability over a three month interval (r = 0.71 for college students and
The level of agreement between observer ratings on the modified BIQ and self-report
Spain. In this study, the modified BIQ and its two subscales correlated positively and
significantly with the TAS-20 total score; the magnitude of the correlations ranged from 0.47 to
0.51, p < 0.001). Significant positive correlations were obtained also with the three factors of
the TAS-20, except for the correlation between the affect awareness subscale of the BIQ and the
Although further studies are needed to evaluate the reliability and construct validity of
the modifed BIQ, the preliminary findings are promising. Investigators who use the modified
version are advised to employ a semistructured form of interview, as outlined by Nemiah et al.
(1976). Interrater reliability should be established by using different raters who interview
interview and rate different subjects if they are randomly assigned to conduct the actual
interview while the other clinician/researcher(s) observe and rate the live interview.
1999; Vingerhoets et al., 1995) initially developed a 20-item scale -- the Amsterdam
Alexithymia Scale -- which was subsequently extended into the 40-item BVAQ with the
purpose of having two parallel versions. Items were written to assess five elements of
alexithymia, which were confirmed by exploratory factor analysis of data collected from
samples of university students (Bermond & Vorst, 1998). The factors were defined as:
identifying, verbalizing, analyzing, fantasizing, and emotionalizing. Whereas the first three
factors assess facets of the alexithymia construct that correspond to those assessed by the three
TAS-20 factors, the fantasizing factor attempts to assess directly the facet involving constricted
imaginal processes, and the emotionalizing factor assesses the degree to which a person can be
emotionally aroused by emotion-inducing events (Bermond & Vorst, 1998). Each factor of the
BVAQ is comprised of 8 items which are rated on five-point likert scales; half of the items are
negatively keyed. High scores are indicative of higher degrees of alexithymia. Originally
13
developed in Dutch, the BVAQ has been translated into several languages including English
and French. Acceptable levels of internal consistency have been demonstrated for the Dutch,
English, and French translations (Vorst & Bermond, 1999; Zech et al., in press), but estimates
The replicability of the factor structure of the BVAQ was evaluated recently by Zech et
al. (in press) for both English and French language versions by means of confirmatory factor
analysis. These investigators found that although the parameter estimates of all items of the 40-
item BVAQ in both language versions loaded significantly on each of the a priori factors (p's <
0.001), most of the indices of goodness of fit were just below the standard criteria. Examining
the two parallel versions of the questionnaire separately, however, the second 20-item version
(BVAQ-20B) yielded much better indices of fit for both English and French speaking samples.
Although the first four factors of the BVAQ correspond to the four salient features in Nemiah et
al’s (1976) definition of the alexithymia construct, and also to the four factors of the original
TAS, emotionalizing is not part of the original definition of the construct and should be
with the TAS-20 total score for both the English version (r = 0.62; p < 0.0001) and the French
version (r = 0.61, p < 0.0001). Bermond and Vorst (1998) found a similar correlation between
the TAS-20 and the Dutch version of the 40-item BVAQ (r = 0.64). The moderate magnitude of
these correlations might be due to the presence in the BVAQ of two factors that are not shared
with the TAS-20, viz., fantasizing and emotionalizing. This was confirmed when Zech et al. (in
press) computed correlations between total TAS-20 scores and BVAQ-20B scores excluding
these two factors. The correlation coefficients were 0.82 for the English translations of both
There is evidence also that the three conceptually similar factors in the TAS-20 and the
two 20-item versions of the BVAQ correlate positively and strongly (Zech et al., in press).
While all three factors of the TAS-20 correlate positively with the total score on the BVAQ-20B
(magnitude of the correlations range from 0.42 to 0.65, p < 0.01), in agreement with their
conceptual specificity, the factors assessing fantasizing and emotionalizing within the BVAQ-
14
20B remained statistically uncorrelated with the total TAS-20 and showed low or nonsignificant
correlations with the three TAS-20 factor scales. Moreover, the emotionalizing and fantasizing
factors showed nonsignificant or low magnitude correlations with the other factors of the
BVAQ-20B. Although the results of preliminary tests of reliability and validity of the BVAQ
are encouraging, further research is needed to determine whether responses to items on the
fantasizing factor of the BVAQ are influenced by social desirability, as was found during the
development of the TAS-20. Additional tests of convergent validity are needed also, and will
help determine whether the emotionalizing factor should be retained within the BVAQ.
Haviland and Reise (1996b) recently developed a new self- and observer–rated measure
of alexithymia using the Q-sort method (Block, 1961/1978). In contrast to the development of
most self-report scales in which items are written to compare one person to another person on a
variety of attributes, the Q-sort method is a person-centered approach in which attributes are
compared to other attributes within the same person. The goal in developing such an instrument
cognitions and behaviors highly representative of the construct. Haviland and Reise (1996b)
asked 17 experts in the field of alexithymia to sort 100 personality statements from the
California Q-set (CAQ) into a forced, nine-category, quasinormal distribution ranging from
returned by 13 judges, who gave each CAQ item a score ranging from 1 to 9. Individual items
were then ranked and converted to a normal nine-points score distribution. The average
interjudge correlation was satisfactory (r = 0.58), and the judge-prototype correlation was high
(r = 0.77). The normal distribution of the 100 items led to 13 "most characteristic" and 13
difficulties experiencing and expressing emotion, lack of imagination, lack of insight, being
literal and utilitarian, being humorless, and experiencing meaninglessness. This description is
reasonably consistent with Nemiah et al’s (1976) definition of the alexithymia construct. The
personal fantasy and daydreams, having insight into one’s own motives and behaviors, being
warm and compassionate, having the capacity for close relationships, being introspective and
To date, there are only sparse data on the validity of the CAQ-AP. In the development of
the measure, Haviland and Reise (1996b) found that high scores correlated moderately and
positively with a Q-sort prototype for ego control, and strongly and negatively with a Q-sort
prototype for ego resiliency. In a subsequent study with a college student sample, Haviland
(1998) correlated Q-sorts similarity scores given by observers with self-ratings on measures of
depression (r = 0.38), and anxiety (r = 0.17), and significant negative correlations with
to studies using the TAS-20, however, a significant negative correlation was found with
agreeableness (r = -0.24).
Although the preliminary validity data for the CAQ-AP are consistent with theoretical
assumptions about the alexithymia construct, there are several uncertainties about the measure.
First, there are no data available on concurrent validity; as the CAQ-AP is claimed to be both a
self- and observer-rated measure of the alexithymia construct, future studies should investigate
its relationship with the TAS-20 and the modified version of the BIQ. Second, completing the
100 items of the CAQ-AP can be a tedious task, especially if judges have to rate several people.
Finally, one can question whether the CAQ-AP can represent equally the most characteristic and
contrast to the broad focus of the CAQ, which was designed to cover a large domain of
personality. Consequently, while the most characteristic attributes of alexithymia selected by the
judges are likey to be very characteristic of the construct, it may be argued that the most
uncharacteristic attributes of alexithymia in the CAQ can be more heterogeneous with respect to
CONCLUSIONS
The TAS-20 is currently the best validated measure of the alexithymia construct and can
be recommended for both clinical and research purposes. Results from various studies that use
the TAS-20 as a dependent variable are directly comparable to one another; and investigators
using the scale to create "alexithymic" and "nonalexithymic" subjects in between group
experimental designs can be assured that their designation is similar to that used in other
studies. Although alexithymia is considered a dimensional construct, upper and lower cutoff
scores have been established empirically for the TAS-20, which allows for comparisons of rates
Although some researchers have criticised the use of self-report measures to assess
alexithymia and/or emotional intelligence (Lane et al., 1997, 1998; Mayer, Salovey, & Caruso, in
press), the findings of consensual agreement between TAS-20 scores and observer ratings on
the modified BIQ in two different cultures suggest that even when used as a sole measure, the
TAS-20 can assess the construct adequately. Nonetheless, if researchers are able to include the
modified BIQ in their studies, empirical research is always enhanced by the use of a multi-
method assessment approach. In clinical situations, when in-depth evaluations of patients may
be required, we have found it useful to administer not only the TAS-20 and the modified BIQ,
but also the EQ-i (Bar-On, 1997) and the Revised NEO Personality Inventory (Costa &
McCrae, 1992). The EQ-i provides a more comprehensive assessment of intrapersonal and
interpersonal intelligence, and alexithymic individuals generally score in the low range on the
openness to fantasy and the receptivity to feelings facets of the openness dimension of
personality.
17
REFERENCES
Apfel, R.J., & Sifneos, P.E. (1979). Alexithymia: Concept and measurement.
Bach, M., Bach, D., & de Zwaan, M. (1996). Independency of alexithymia and som
Bach, M., Bach, D., de Zwaan, M., Serim, M., & Böhmer, F. (1996). Validierung der
Bagby, R.M., Parker, J.D.A., & Taylor, G.J. (1994a). The Twenty-Item Toronto
Alexithymia Scale -- I. Item selection and cross-validation of the factor structure. Journal of
Bagby, R.M., Taylor, G.J., & Parker, J.D.A. (1994b). The Twenty-Item Toronto
Systems.
Beresnevaite, M., Taylor, G.J., Parker, J.D.A., & Andziulis, A. (1998). Cross validation of
the factor structure of the 20-item Toronto Alexithymia Scale. Acta medica Lituanica, 5, 146-149.
Bermond, B., & Vorst, H.C. (1998). Validity and reliability of the Bermond-Vorst
Netherlands.
Bermond, B., Vorst, H. C., Gerritsen, W., & Vingerhoets, Ad. J. (1994). Psychometric
Bermond, B., Vorst, H.C.M., Vingerhoets, Ad. J.J.M., & Gerritsen, W.(1999).The
Amsterdam Alexithymia scale: Its psychometric values and correlations with other personality
Bourke, M.P., Taylor, G.J., Parker, J.D.A., & Bagby, R.M. (1992). Alexithymia in
Bressi, C., Taylor, G., Parker, J., Bressi, S., Brambilla, V., Aguglia, E., Allegranti, I.,
Bongiorno, A., Giberti, F., Bucca, M., Todarello, O., Callegari, C., Vender, S., Gala, C.,
Invernizzi, G. (1996). Cross validation of the factor structure of the 20-item Toronto
Alexithymia Scale: An Italian multicenter study. Journal of Psychosomatic Research, 41, 551-
559.
Carver, C.S. (1989). How should multifaceted personality constructs be tested? Issues
Costa, P. T. Jr., & McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO-
PI-R) and NEO Five-Factor Inventory (NEO-FFI) professional manual. Odessa, FL:
Crown, S., & Crisp, A.H. (1979). Manual of the Crown-Crisp Experiential Index.
Deary, I.J., Scott, S., & Wilson, J.A. (1997). Neuroticism, alexithymia and medically
Erni, T., Lötscher, & Modestin, J. (1997). Two-factor solution of the 20-Item Toronto
Fukunishi. I, Nakagawa, T., Nakamura, H., Kikuchi, M., & Takubo, M. (1997). Is
alexithymia a culture-bound construct? Validity and reliability of the Japaneses versions of the
20-item Toronto Alexithymia Scale and modified Beth Israel Hospital Psychosomatic
Fukunishi,I., Saito, S., & Ozaki, S. (1992). The influence of defense mechanisms on
Gardos, G., Schniebolk, S., Mirin, S.M., Wolk, P.C., & Rosenthal, K. (1984).
Haviland, M.G., Hendryx, M.S., Shaw, D.G., & Henry, J.P. (1994). Alexithymia in women
and men hospitalized for psychoactive substance dependence. Comprehensive Psychiatry, 35, 124-
128.
Haviland, M. G., & Reise, S. P. (1996a). A California Q-set alexithymia prototype and
its relationship to ego-control and ego-resiliency. Journal of Psychosomatic Research, 41, 597-
608.
Haviland, M.G., & Reise, S.P. (1996b). Structure of the Twenty-Item Toronto Alexithymia
Haviland, M.G., Shaw, D.G., Cummings, M.A., & MacMurray, J.P. (1988).
164-170.
Hendryx, M.S., Haviland, M.G., & Shaw, D.G. (1991). Dimensions of alexithymia and
their relationships to anxiety and depression. Journal of Personality Assessment, 56, 227-237.
Jimerson,D.C., Wolfe, B.E., Franko, D.L., Covino, N.A., & Sifneos, P.E.(1994).
Alexithymia ratings in bulimia nervosa: Clinical correlates. Psychosomatic Medicine, 56, 90-93.
Kauhanen, J., Julkunen, J., & Salonen, J.T. (1992). Validity and reliability of the Toronto
Alexithymia Scale (TAS) in a population study. Journal of Psychosomatic Research, 36, 687-694.
Kirmayer,L.J., & Robbins, J.M. (1993). Cognitive and social correlates of the Toronto
Kleiger, J.H., & Kinsman, R.A. (1980). The development of an MMPI alexithymia scale.
Kline, R.B. (1991). Latent variable path analysis in clinical research: A beginner's tour
Lane, R. D., Ahern, G. L., Schwartz, G. E., & Kaszniak, A. W. (1997). Is alexithymia
Lane, R.D., Sechrest, L., & Riedel, R. (1998). Sociodemographic correlates of alexithymia.
Lee, Y-H., Rim, H-D. & Lee, J-Y. (1996). Development and validation of a Korean
version of the 20-item Toronto Alexithymia Scale (TAS-20K). Journal of the Korean
Lesser, I.M., Ford, C.V., & Friedmann, C.T.H. (1979). Alexithymia in somatizing
Linden, W., Wen, F., & Paulhus, D.L. (1995). Measuring alexithymia: Reliability,
Loas, G., Otmani, O., Verrier, A., Fremaux, D., & Marchand, M. P. (1996). Factor analysis
of the French version of the 20-Item Alexithymia scale (TAS-20). Psychopathology, 29, 139-144.
Loas, G., Parker, J.D.A., Otmani, O., Verrier, A., & Fremaux, D. (1997). Confirmatory
factor analysis of the French translation of the 20-item Toronto Alexithymia Scale. Perceptual and
Luminet, O., Bagby, R.M., & Taylor, G.J. An evaluation of the absolute and relative
submitted.
Martínez-Sánchez, F., Ato-García, M., Córcoles Adam. E., Huedo Medina, T.B., Selva
España. J.J. (1998). Stability in alexithymia levels: A longitudinal analysis on various emotional
Mayer, J.D., Salovey, P., & Caruso, D. (in press). Competing models of emotional
intelligence. In R.J. Sternberg (Ed.), Handbook of Human Intelligence (2nd Ed.) New York:
Nemiah, J.C., Freyberger, H., & Sifneos, P.E. (1976). Alexithymia: A view of the
Páez, D., Martínez-Sánchez, F., Velasco, C., Mayordomo, S., Fernández, I., & Blanco, A.
Pandey, R., Mandal, M.K., Taylor, G.J., & Parker, J.D.A. (1996). Cross-cultural
Parker, J.D.A., Bagby, R.M., Taylor, G.J., Endler, N.S., & Schmitz, P. (1993). Factorial
validity of the 20-item Toronto Alexithymia Scale. European Journal of Personality, 7: 221-232.
Parker, J.D.A., Taylor, G.J., & Bagby, R.M . The factorial validity and convergent validity
Parker, J.D.A., Taylor, G.J., & Bagby, R.M . The relationship between alexithymia and
Pinard, L., Negrete, J.C., Annable, L., & Audet, N. (1996). Alexithymia in substance
Porcelli, P., Leoci, C., Guerra, V., taylor, G.J., & Bagby, R.M. (1996). A longitudinal study
Prazeres, N., Parker, J.D.A., & Taylor, G.J. (in press). Portuguese adaptation of the 20-
Salminen, J.K., Saarijärvi, S., Äärelä, E., & Tamminen. (1994). Alexithymia -- state or trait?
Salminen, J.K., Saarijärvi, S., Äärelä, E., Toikka, T., & Kauhanen, J. (1999). Prevalence of
alexithymia and its association with sociodemographic variables in the general population of
Santor, D.A., Bagby, R.M., & Joffe, R.T. (1997). Evaluating stability and change in
personality and depression. Journal of Personality and Social Psychology, 73, 1354-1362.
22
Sifneos, P.E. (1996). Alexithymia: Past and present. American Journal of Psychiatry, 153,
137-142.
Simonsson-Sarnecki, M., Lundh, L.G., Torestad, B., Bagby, R.M., Taylor, G.J., & Parker,
J.D.A. (in press). A Swedish translation of the 20-item Toronto Alexithymia Scale: Cross-
Sriram, T.G., Chaturvedi, S.K., Gopinath, P.S., & Shanmugam, V. (1987). Controlled
Sriram, T.G., Pratap. L., Shanmugham, V. (1988). Towards enhancing the utility of the
Beth Israel Hospital Psychosomatic Questionnaire. Psychotherapy and Psychosomatics, 49, 205-
211.
Taylor, G.J., & Bagby, R.M. (1988). Measurement of alexithymia: Recommendations for
clinical practice and future research. Psychiatric Clinics of North America, 11, 351-366.
Taylor, G.J., Bagby, R.M., & Parker, J.D.A. (1997). Disorders of Affect Regulation:
Taylor, G.J., Bagby, R.M., Ryan, D.P., & Parker, J.D.A. (1990). Validation of the
Taylor, G.J., Ryan, D.P., & Bagby, R.M. (1985). Toward the development of a new
Vingerhoets, Ad. J.J.M., Van Heck, G.L., Grim, R., & Bermond, B. (1995). Alexithymia:
A further exploration of its nomological network. Psychotherapy and Psychosomatics, 64, 32-42.
Vorst, H.C.M., & Bermond, B. (1999). Validity and reliability of the Bermond-Vorst
Wise, T.N., Mann, L.S., Mitchell, J.D., Hryvniak, M., & Hill, B. (1990). Secondary
Zachary, R.A. (1986). Shipley Institute of Living Scale. Revised Manual. Los Angeles:
Zech, E., Luminet, O., Rimé, B., & Wagner, H. (in press). Alexithymia and its
measurement: Confirmatory factor analyses of the twenty-item Toronto Alexithymia Scale and