Neuroticism, alexithymia and medically unexplained symptoms

Person. indiuid. DijJ Vol. 22, No. 4, pp. 551-564, 1997 0 1997 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0191-8869/97 $17.OO+O.C43 SO191-8869(96)-00229-2

Pergamon PII:

NEUROTICISM, ALEXITHYMIA AND MEDICALLY UNEXPLAINED SYMPTOMS Ian J. Deary, *’ Shonagh

Scott’ and Janet A. Wilson’

’ Department of Psychology, University of Edinburgh, Edinburgh EH8 9JZ, Scotland, and 2 Department of Otolaryngology

and Head and Neck Surgery,

University

of Newcastle,

Newcastle,

England

(Received 28 May 1996) Summary-The contribution that alexithymia can make to the understanding of medically unexplained physical symptoms (MUS) was studied in 244 subjects drawn from a range of medical and non-medical situations. People’s histories of MUS-also called somatisation-were assessed using physical symptom lists derived from the DSM-III-R somatisation criteria. Two subscales from the Toronto Alexithymia Scale-20 had significant correlations with reported MUS, but also with neuroticism, negative emotion health coping, anxiety, depression, general psychological distress and dysphoric mood. Despite there being a large general latent trait (negative affectivity) underlying most of the measured variables, the best model of the data in men and women was a two-factor model that emphasised that alexithymia could make a contribution to MUS variance beyond that made by negative affectivity. It is suggested that, for the purposes of studying MUS, alexithymia might be reconstructed as a single component construct, related to a confusion among feelings and between feelings and bodily symptoms, rather than its present threecomponent structure. A two parameter model for the occurrence of medically unexplained physical symptoms is proposed in which negative affectivity acts as a threshold factor (influencing symptom detection) and alexithymia acts as an interference factor (influencing symptom discrimination/recognition). 0 1997 Elsevier Science Ltd. All rights reserved.

INTRODUCTION The purpose of this study was to examine whether Toronto Alexithymia Scale (TAS and its newer

the concept

of alexithymia,

as measured

by the

form, the TAS-20; Bagby, Parker & Taylor, 1994a; Bagby, Taylor & Parker, 1994b; Taylor, Bagby, Ryan & Parker, 1990) could add to our understanding of the phenomenon of medically unexplained symptoms (MUS). We sought to discover whether TAS-20 scores accounted for variance in MUS that is independent of other psychological constructs, such as neuroticism and coping styles. The experience of physical symptoms that do not result from organic disease is very common and is time-consuming and costly to health services (Kroenke, 1992; Mayou, Bass & Sharpe, 1995). Up to 20% of general practice consultations in the U.K. relate to physical complaints for which no physical reason will be found (Bridges & Goldberg, 1985), and such problems are among the commonest source of referrals to hospital outpatient clinics in the U.K. (Bradlow, Coulter & Brookes, 1992). The medical phenomenon described above has attracted a number of names. These include hysteria, somatisation, somatoform disorders, psychosomatic symptoms (and/or syndromes), functional somatic symptoms and MUS. In preferring the term MUS, we are emphasising that there is no recognised theory of the generation of these symptom reports. Other terms are employed in this report where the investigators themselves have used the particular label, such as ‘psychosomatic symptoms’ and ‘functional somatic symptoms’. Therefore, the literature survey that we offer must be read with the forewarning that the symptoms discussed would not always meet today’s accepted categories of MUS (Mayou et al., 1995). The term alexithymia was first introduced by Sifneos (1973). The concept has since attracted many researchers and clinicians, resulting in a plethora of studies. These studies include: attempts to explain the aetiology of alexithymia, its relationships with other personality traits, the influence of alexithymia on therapeutic processes and the construction of a variety of measures to assess the construct (see Bagby et al., 1994a; Lesser, 1981; Norton, 1989; and Taylor, Bagby, Ryan & Parker,

*To whom all correspondence

should be addressed. 551

552

Ian J. Deary et al.

1990 for discussions of these topics). Despite over 20 years of research it appears that only the broad definition of the concept has been agreed upon, and alexithymia is generally accepted as consisting of the following dimensions: difficulty identifying and describing feelings, difficulty in distinguishing between feelings and the bodily sensations of emotional arousal, constricted imaginative fantasy life, and the tendency to focus on the concrete details of external events (‘pensee operatoire’) (Lesser, 1981; Nemiah, Freyberger & Sifneos, 1976; Taylor, 1984). The idea that patients with a ‘psychosomatic’ disorder may be psychologically different from those without one has long been discussed in the literature (Freedman & Sweet, 1954; Marty & de M’Uzan, 1963; Reusch, 1948). Indeed, the very concept of alexithymia rose from clinical observations that many patients with psychosomatic disorders had difficulty talking about feelings and fantasies when assessed in psychodynamically oriented interviews. However, investigation of the relationship between alexithymia and psychosomatic disorders has been hampered by the lack of a valid and reliable measure. Moreover, medical disorders that were originally termed psychosomatic, such as rheumatoid arthritis, would not be included today under the banner of MUS or functional somatic disorders. A number of measures has been devised to assess the concept of alexithymia, including: the Beth Israel Hospital Psychosomatic Questionnaire (BIQ; Sifneos, 1973) a semistandardised clinical interview; the Schalling-Sifneos Personality Scale (SSPS; Apfel & Sifneos, 1979) a self-report measure based on the BIQ; and an alexithymia score derived from the Minnesota Multiphasic Personality Inventory (MMPI-A; Kleiger & Kinsman, 1980) another self-report scale. Finally, there are two projective tests of alexithymia: the SAT9 test (Cohen, Auld, Demers & Catchlove, 1985) and the Rorschach Alexithymia Index (Acklin & Bernat, 1987). Several authors have identified significant flaws with all of the above measures including unacceptably low inter-rater reliability coefficients for the BIQ (Taylor, Doody & Newman, 1981). Also, initial findings of a high level of agreement between SSPS and BIQ scores have not been replicated in subsequent studies (Demers-Desrosiers, Cohen, Catchlove & Ramsay, 1983; Federman & Mohns, 1984). Results from a study examining and comparing the psychometric properties of the SSPS, the MMPI-A and the SAT9 found that the SSPS and MMPI-A scales has dissimilar factor structures and therefore assess different domains, and that the inter-rater reliability of the SAT9 was so low that it raised serious doubts as to its measurement accuracy (Norton, 1989). Because of the lack of an adequate alexithymia measure, Taylor and colleagues constructed the Toronto Alexithymia Scale (TAS), initially a 26-item self-report scale but more recently a 20-item scale (TAS-20), using a measurement-based approach to validate the construct (Taylor, Ryan & Bagby, 1985). Results from a series of studies showed that especially the TAS-20 demonstrated internal consistency, good testretest reliability, and a stable factor structure theoretically congruent with the alexithymia construct (Bagby et al., 1994a, 1994b; Taylor et al., 1990). A study of 30 patients with a DSM-III diagnosis of psychogenic pain disorder found that in comparison to 30 age- and sex-matched controls the psychogenic pain patients had significantly higher scores on both the Toronto Alexithymia Scale (26-item version) and the Beth Israel Hospital Questionnaire (Sriram, Chaturvedi, Gopinath & Shanmugam, 1987). Another study of 207 patients, referred to an out-patient behavioural medicine service with a diagnosis of pain found that 36% of the sample scored in the ‘alexithymic’ range of the MMPI-A scale (Papciak, Feuerstein, Belar & Pistone, 19861987). A heterogeneous group of 107 psychosomatic patients was significantly more alexithymic, as measured by the BIQ, than a control group of patients with somatic complaints (Keltikangas-Jarvinen, 1985). Scores on the TAS were found to correlate more highly with functional somatic symptoms than did scores on the Schalling-Sifneos Personality Scale or the MMPIAlexithymia Scale (Bagby, Taylor & Atkinson, 1988). Cohen, Auld and Brooker (1994) found an association between alexithymic features and a tendency to report physical signs and symptoms, although a small study by Bach, Bach, Bohmer and Nutzinger (1994a) found no more somatisation (based on responses to the Symptom Checklist 90) in a group of alexithymics. As recently as 1994, Wise and Mann stated that the relationship between alexithymia and somatosensory amplification remained unknown. While psychosomatic patients appear to demonstrate greater levels of alexithymia than controls or patients with somatic disease, patients with psychiatric disorder have levels just as high (Lesser, Ford & Friedman, 1979) or greater (Rubino, Grasso, Sonnino & Pezzarossa, 1991) than psychosomatic patients. Other studies have also reported alexithymic traits in individuals with substance

Medically unexplained symptoms

553

abuse disorders (Krystal & Raskin, 1970) and antisocial personalities (Keltikangas-Jarvinen, 1982) suggesting that it might be a common concomitant of various types of psychological distress. Despite an apparently substantial literature on alexithymia and MUS, most previous studies: have used measures of alexithymia that are now known to have poor psychometric properties; have often used inadequate control groups; and have often failed to use formal diagnostic criteria, such as the Diagnostic and Statistical Manual for mental disorders published by the American Psychiatric Association, to classify the physical complaints of the target groups of patients with psychosomatic disorders. Furthermore, if people who are alexithymic have difficulties in understanding and describing their emotions, they may be unaware of such a deficit and so be unable to report it themselves (Kirmayer, Robbins & Paris, 1994). Self-report measures of alexithymia may therefore measure a construct different from that identified by clinicians, and it has been suggested that the development of a reliable and valid interview-rated instrument may be more productive (Codispoti, Codispoti & Battacchi, 1995). In addition to examining the relationship between alexithymia and functional somatic illness, attempts have been made to examine the relationship between alexithymia and other individual difference characteristics, and to discover whether alexithymia is a distinct construct or whether it is partly measured by existing constructs such as neuroticism or depression. Parker, Bagby and Taylor (1991) found a significant moderate correlation between the Toronto Alexithymia Scale and the Beck Depression Inventory (Beck, Rush, Shaw & Emery, 1979) in both a sample of undergraduates and a sample of psychiatric out-patients, but results from a factor analysis showed that alexithymia was a distinct construct from depression. A study investigating the relationship between alexithymia and DSM-III-R psychiatric syndromes and personality disorders (measured by the Personality Diagnostic Questionnaire-Revised; Hyler 8z Rieder, 1987) found no significant correlation between TAS scores and DSM-III-R diagnoses, but schizotypal, dependent and avoidant personality dimensions, as well as a lack of histrionic features, emerged as significant predictors of alexithymia (Bach, de Zwaan, Ackard, Nutzinger & Mitchell, 1994b). The authors concluded that the data supported the validity of alexithymia as a personality construct independent from other psychiatric disorders. However, schizotypal, dependent and avoidant personality disorders are all closely correlated with neuroticism, which could be the link between these concepts and alexithymia (Deary, Peter, Austin & Gibson, 1995). Indeed, others have found that, apart from having low correlations with somatic complaints, “alexithymia cannot be discriminated from neurotic introversion” (Groot & de Leeuw, 1995), in contradiction to earlier work by Parker, Bagby and Taylor (1989). Also, Myers (1995) found that people with high trait anxiety (Taylor Manifest Anxiety Scale) scored significantly higher on the TAS-20. The question as to whether alexithymia adds anything beyond that captured by the construct of neuroticism has been the subject of lively and unresolved debate (Rubino, 1993; Rubino et al., 1991; Taylor, Bagby & Parker, 1993). Alexithymia is, though, distinct from the construct of health locus of control (Wise & Mann, 1993). The present study Recently, Smith and Williams (1992) and Marshall, Wortman, Vickers, Kusulas and Hervig (1994) have suggested that there is a great deal of construct overlap in health psychology and that one of the first questions that should be asked of a new measure is whether it is redundant with respect to better-established constructs. In this regard, it is important to ask whether alexithymia is yet another surrogate for the very general tendency to experience negative emotions and bodily complaints that has been called negative affectivity or somatopsychic distress, terms that are closely related to neuroticism (Deary, Blenkin, Agius, Endler, Zealley & Wood, 1996; Watson & Pennebaker, 1989). With respect to our measurement of MUS, we have taken this to be the phenomenon variously called functional somatic symptoms/medically unexplained symptoms/ psychosomatic symptoms or syndromes (Kellner, 1985, 1991; Kirmayer, Robbins & Paris, 1994; Mayou et al., 1995). We have used the DSM-III-R list of symptoms for somatisation disorder as a relatively comprehensive and well-validated coverage of the functional somatic symptom domain. Therefore, we are interested in the degree to which alexithymia can account for variance in people’s past histories of bodily disturbances which had no discoverable physical origins. The purposes of the present study were, therefore, to examine the distinctiveness of alexithymia from other constructs related to psychological distress, and to determine the usefulness of alexi-

554

Ian J. Deary et al.

thymia in predicting and explaining the occurrence of medically unexplained (psychosomatic or functional) somatic symptoms. In particular, we sought to find out whether alexithymia, as assessed by the TAS-20, could add anything to well-validated constructs such as neuroticism, anxiety and depression.

METHODS

Subjects

Participants in the study were drawn from a range of medical and non-medical settings, chosen to ensure a spread of experience of MUS. Therefore, out-patient and general practice patients were recruited as well as Ss who were not attending medical clinics. A total of 244 Ss were tested. These comprised 20 patients with globus pharyngis (14 women, six men), 60 patients with dysphonia (38 women, 22 men), 68 patients attending an otolaryngology out-patient clinic (48 women, 20 men), 51 patients attending their general practitioner (29 women, 22 men) and 45 healthy controls (22 women, 23 men). Globus pharyngis is a condition in which patients have the feeling of something being stuck in their throats, but in which no cause can be found for the sensation (Deary, Wilson & Kelly, 1995). Therefore, globus is a pure functional symptom and a good example of MUS (Kellner, 1991). Dysphonia patients are people complaining of hoarseness (Wilson, Deary, Scott & MacKenzie, 1995). In the present sample some had physical causes for the symptom (such as chronic laryngitis, although major physical causes such as laryngeal cancer were not included) and some had ‘functional’ dysphonia in which no physical cause can be found for the complaint (Wilson et al., 1995). Therefore, this group comprised a mixture of physical and MUS problems. The patients attending the otolaryngology clinic had a mixture of conditions, some of which were purely physical (such as hearing conditions), and some of which were known to be partly determined by psychological factors (such as tinnitus and vasomotor rhinitis). Patients attending their general practitioner had a range of disorders. Healthy controls had not recently attended any medical practitioner. As a whole, the mean (SD) age of the women was 45.0 (13.9) years and of the men, 47.8 (14.8) years. Measures

Patients were tested on a range of self-report measures related to negative affectivity/somatopsychic distress. State as well as trait measures were included. The main outcome variable was patients’ reported history of MUS, and the principal independent variables were neuroticism and the alexithymia subscales as measured by the Toronto Alexithymia Scale-20 (TAS20). Toronto Alexithymia Scale-20. (TAS-20; Bagby et al., 1994a, 1994b). The TAS-20 comprises 20 statements designed to assess three aspects of the alexithymia construct: TAS-DIF, difficulty in identifying and describing feelings to others; TAS-DDF, the capacity to distinguish between feelings and bodily symptoms of emotion; and TAS-EOT, externally oriented thinking. Individual subscale scores were used in the analyses. The total TAS-20 score was not used in analyses because some of the subscales were entirely uncorrelated and because the TAS-DIF score always performed better than the TAS-20 total score. Coping with health, injuries and problems. (CHIP; Endler, Parker 8c Summerfeldt, 1992, 1993). This instrument has 32 statements, which cover four dimensions of coping with illnesses: distraction (CHIP D), palliative (CHIP P), instrumental (CHIP I) and negative emotion (CHIP NE). Negative emotion coping is of particular interest because it involves the expression of negative feelings. Eysenck Personality Questionnaire-Revised, short form. (EPQ-R; Eysenck, Eysenck & Barrett, 1985). The EPQ-R short form has 48 questions, 12 related to each of the three major personality dimensions of neuroticism, psychoticism and extraversion. There are also 12 questions in the Lie scale. Neuroticism was the variable of particular interest with respect to somatisation. DSM-III-R Somatisation scale. (DSM-Som; Dear-y, Clyde 8z Frier, in press). This scale has 31 questions in the male version, and 34 in the female version. The questions ask about Ss’ past experiences of certain physical symptoms. The symptom descriptions are taken verbatim from the somatisation disorder section of the Structural Clinical Interview for DSM-III-R (SCID; Spitzer, Williams, Gibbon & First, 1990). That is, all of the items are known MUS, and are indicative of a

Medically unexplained symptoms

555

Ss tendency toward somatisation (Kellner, 1991; Mayou ef al., 1995). Ss must answer yes to having had any of the symptoms only ‘if no physical cause for the symptom has ever been discovered by doctors’. It is further stressed that the symptom must not have been due to medication, drugs, alcohol or have occurred during a panic attack. Moreover, the symptom must have been severe enough to cause them to see a doctor, or have significantly altered their lifestyle in some way. Detailed instructions regarding these criteria are given on the test form, and reminders are given at the end of the scale. The total number of symptoms endorsed was taken as the S’s MUS score. General Health Questionnaire-28. (GHQ-28; Goldberg & Williams, 1988). GHQ-28 total score was used as a measure of recent general psychological distress. Because the scores were skewed (the majority of Ss had low scores), logarithmic transformation was applied. Hospital Anxiety Depression scale. (HAD; Zigmond & Snaith, 1983). The HAD was used to assess recent anxiety (HAD-A) and depression (HAD-D). None of the questions on the anxiety or depression scales refers to physical symptomatology, in contrast with the GHQ. UWIST Mood Adjective Checklist. (UMACL; Matthews, Jones & Chamberlain, 1990). This instrument measures current mood states on three scales: hedonic tone (UMACL-HT), tense arousal (UMACL-TA), and energetic arousal (UMACL-EA). Hedonic tone and tense arousal were of particular interest because they are negative emotions related to depression and anxiety, respectively. Statistical analyses

In order to answer the questions of(i) whether alexithymia was more than another expression of neuroticism/negative affectivity/somatopsychic distress and (ii) whether it could contribute anything novel to the prediction of MUS, various planned analyses were carried out. Pearson’s r correlations were computed between neuroticism and alexithymia subscales and MUS scores and between these and all the other variables. Second, stepwise multiple regression analyses were carried out to determine the principal personality- and personal distress-related predictors of MUS. Third, principal components analyses were carried out to assess the degree to which the measured variables (personality and various measures of personal distress) were reflecting a single underlying variable. Fourth, confirmatory factor analyses were constructed using structural equation modelling to discover the best latent trait model for the measured variables. Finally, the EPQ-R neuroticism and TAS-20 items were submitted to a joint factor analysis to discover whether there was an item content-based explanation for any neuroticism-alexithymia overlap.

RESULTS Although the associations among variables are the main results, the experience of the Ss with respect to MUS is of interest. Women reported a mean of 6.3 (SD = 5.3) previous medically unexplained symptoms, whereas men reported 4.5 (SD = 4.4). Because there were slightly different numbers of questions in the form of the questionnaire completed by each sex, a statistical comparison of the rates of these complaints across the sexes is not appropriate. However, the separate analyses conducted in men and women were used to test the replicability of the multivariate models. Correlations between neuroticism, alexithymia subscales and MUS and other variables are shown in Table 1. Correlations are very similar in women and men. TAS-DIF and TAS-DDF intercorrelated highly (>0.6, P < 0.001) in men and women, TAS-DDF and TAS-EOT have modest intercorrelations (>0.3, P c O.OOl),but those between TAS-DIF and TAS-EOT are low. TAS-DIF has strong correlations, well above 0.4 (P c 0.001) with MUS in both men and women. There are modest (above 0.2) correlations between TAS-DDF and MUS in both men and women, but the correlations between MUS and TAS-EOT are negligible. Therefore, the best correlations with MUS are with the DIF subscale, and the EOT subscale fails to correlate with these symptoms. There are strong correlations (above 0.4, all P < 0.001) in women and men between TAS-DIF and neuroticism, negative emotion coping, HAD-anxiety and -depression, and GHQ (Table 1). The correlations between TAS-DIF and all three mood scales from the UMACL are above 0.4 in women and close to 0.4 in men (all P < 0.001). The same variables correlated significantly with TAS-DDF, but mostly at slightly lower levels. TAS-EOT failed to correlate significantly across the sexes with any of these variables. MUS, apart from its significant correlations with TAS-DIF and -DDF in

Ian J. Deary et al.

556 Table 1. Correlations

between alexitbymia,

TAS-DIF Women Men

MUS and other variables

TAS-DDF Women Men

(N = 151 for women and 93 for men)

TAS-EOT Women Men

MUS” Women

Men

MUS (DSMSomatisation) TAS-DIF TAS-DDF TAS-EOT

o&*** 0.13

0.69*** 0.24’

o&l*** 0.27”’ -0.03

0.34;”

0.36***

6.3(5.3)

4.5 (4.4)

0.49*** 0.22’ 0.05

15.6(6.9) 12.3(4.9) 20.3(4.5)

16.0(6.6) 13.5(4.9) 21.4(4.5)

0.36*** -0.04 0.14

6.3(3.4) 8.2(3.2) 1.9(1.6)

5.1(3.5) 7.3 (3.7) 2.4(1.7)

EPQ-Neuroticism EPQ-Extraversion EPQ-Psychoticism

0.45*** -0.01 0.12

0.45*** -0.19 0.10

0.47*** 0.22** 0.07

0.33”’ -0.25. 0.18

0.13 -0.17 0.06

-0.09 -0.07 0.04

0.36*** -0.03 0.21”

CHIP-Distraction CHIP-Palliative CHIP-Instrumental CHIP-Negative emotion

0.23** 0.12 0.06 0.40***

0.29.’ 0.19 0.06 0.45***

0.04 0.09 0.01 0.27***

0.17 0.23; -0.06 0.38’:’

-0.04 0.24** -0.01 0.03

0.04 0.03 -0.11 -0.07

0.12 0.09 0.21’ 0.35***

0.20 0.16 0.16 0.26*

UMACL-Tense arousal UMACL-Hedonic tone UMACL-Energetic Arousal

0.44***

0.36***

0.36”’

0.24’

0.08

0.10

0.43***

0.36***

-0.43***

-0.38***

-0.36***

-0.33***

-0.03

-0.43***

-0.38***

-0.39***

-0.39***

-0.14

HAD-Anxiety HAD-Depression

0.50*** 0.52:;’

0.58”’ 0.47***

0.38”’ 0.39***

0.41*** 0.41”’

log (General health Questionnaire)

0.54”;

0.46***

0.38***

0.38”’

Now. “Medically

unexplained

symptoms

(DSM-Somatisation),

0.04

-0.06

Means (SD) Women Men

25.4(6.5) 24.4(4.8) 31.2(6.2) 23.1(7.1)

23.9(6.4) 24.4(5.2) 29.8(6.2) 22.6(7.2)

9.2(6.2)

7.5(5.3)

-0.29***

-0.31**

l&2(5.7)

18.0(5.8)

-0.29***

-0.30”

15.0(4.8)

14.8(4.8)

-0.00 0.08

0.06 0.10

0.51*** 0.42”’

0.47*** o&l***

7.4(4.9) 3.5(3.3)

5.9 (3.9) 3.5 (3.4)

0.13

0.08

0.46*‘*

0.47”’

1.3 (0.24)

1.3 (0.26)

*P < 0.05;

l

* P -c 0.01; ***P < 0.001.

both sexes, was also significantly correlated in both sexes with neuroticism, negative emotion coping, all three UMACL mood scales, HAD-anxiety and -depression, and GHQ. As expected, therefore, there were substantial correlations among several variables that are associated with negative affectivity/somatopsychic distress. To discover the principal independent predictors of MUS scores, stepwise multiple regression analyses were carried out. In both men and women nine predictor variables were entered, namely, TAS-DIF, TAS-DDF, neuroticism, negative emotion coping, UMACL-tense arousal and -hedonic tone, HAD-anxiety and -depression, and GHQ. In the female sample two variables accounted for 29.8% (adjusted R2) of MUS variance; HAD-anxiety was entered first, accounting for 25.5% of the variance, and TAS-DIF was entered second, adding a further 4.3% variance. In the male sample three variables accounted for 32.7% (adjusted R2) of MUS variance; TAS-DIF accounted for 22.7%, GHQ a further 7.2%, and TAS-DDF a further 2.8%. It is interesting to note that neuroticism and coping variables did not contribute variance independent of psychological distress, but that alexithymia variables did. In men, two alexithymia subscales were retained in the best fit multiple regression model. The multiple regression models may be seen as similar in men and women, because HAD-anxiety and GHQ correlated 0.72 in women and 0.63 in men. Therefore, in both sexes alexithymia subscales contributed to the prediction of MUS variance independent of psychological distress variance. The amount of common variance in the nine variables used to predict MUS was assessed by principal components analysis. In women principal components analysis of the correlations among TAS-DIF, TAS-DDF, neuroticism, negative emotion coping, UMACL-tense arousal and -hedonic tone, HAD-anxiety and -depression, and GHQ revealed only one component with an eigenvalue of greater than one, accounting for 58.0% of the total variance. In men, principal components analysis of the same nine variables revealed two components with eigenvalues greater than one. However, the first unrotated principal component accounted for 54.7% of the total variance. The correlations, multiple regression and PCA all point to a common conclusion: that there is much redundancy in variables intended to assess personality, coping, MUS’ anxiety, depression and general psychological distress.

Medically unexplained symptoms

557

Table 2. Models of association among 10 key variables in the study tested using the EQS structural equation modelling program One factor model Women Men Average off-diagonal standardised residuals Chi square (d.f.) P-value Bentler-Bonnet normed fit index Bentler-Bonnet nonnormed fit index Comparative fit index

Two factor model Women Men

0.046

0.047

0.038

0.041

86.6(34)
56.8 (34) 0.008 0.887

73.4 (3 I) to.001 0.918

42.3 (31) 0.084 0.916

0.919

0.934

0.928

0.964

0.950

0.950

0.975

0.939

Next, two structural equation models of the correlations among these variables were tested. First, a highly economical, single factor model assumed that only one latent variable could account for all the inter-test covariance. Models were tested as confirmatory factor analyses by the EQS structural equation modelling package (Bentler, 1995; Bentler & Wu, 1995). The single factor model had moderately good fit statistics, as shown in Table 2. The average of the off-diagonal standardised residuals was low in both men and women, indicating that a single latent factor could account for most of the inter-test covariance. The chi square of the models was significant, indicating that the covariance remaining differed significantly from zero. However, the P-value of the chi square test is not particularly important in assessing the fit of the model, and a rule of thumb is that a model might be considered acceptable if the chi square is not much larger than twice the number of degrees of freedom. The model meets this criterion in men, but less well in women. The fit indices for the single factor model are almost all greater than 0.9, indicating an acceptable model. Therefore, a single factor model, which states that a latent variable akin to negative affectivity/somatopsychic distress is all that is needed to account for the covariance among 10 different constructs, fits the data relatively well in both men and women. An additional analysis was run to discover how similar this single latent trait model was across the sexes. EQS was used to test a multi-sample confirmatory factor analysis in which the loadings of all 10 manifest variables on the latent variable was assumed to be equal in men and women. Additionally, the correlation of the error terms for the two TAS subscales was assumed to be equal. In testing the model none of these 11 constraints was rejected (P > 0.05 in all cases), indicating that a highly constrained model applied equally to men and women. Therefore, not only does a single factor model fit quite well in the two sexes, the same single factor model fits, with all the loadings of all pathways being very similar. The model is shown graphically in Fig. 1, where it may be seen that the factor loadings are very similar for the two sexes. Second, in competition with the single factor model, a two-factor model assumed that alexithymia made a contribution to the prediction of MUS variance that was relatively independent of negative affectivity/somatopsychic distress. In this model the two alexithymia subscales that have significant correlations with DSM-somatisation (MUS) were assumed to be reflecting a single latent variable that was also associated with MUS and negative emotion coping. The only association that was allowed with the large latent variable that captured the covariance among all of the other variables was a correlation between the two latent variables themselves. Figure 2 shows the model diagrammatically. Table 2 shows that this model has good fit statistics by most criteria; in particular, all fit indices are well above 0.9. In men the chi square of the model was not-significant, indicating a very well fitting model. In women the chi square was marginally in excess of twice the degrees of freedom. The only relaxations needed to this highly constrained two factor model to achieve the fits shown in Table 2 were a small association between HAD-anxiety and factor 2 in men (parameter weight = 0.16) and between UMACL-tense arousal and factor 2 in women (parameter weight = - 0.17). Note that these associations are very small and inconsistent across the sexes. The two factor model can be shown to make a significant improvement over the single factor model by comparing the chi square values of the models with respect to the number of degrees of freedom by which they differ. For males, the loss of three degrees of freedom improves the chi

558

Ian J.

Deary et al.

Fig. 1. Single latent trait model of the associations among 10 key variables in the study. Loadings for the male sample (IV = 93) are shown outside parentheses and female (N = 151) loadings inside. Not shown is a correlation of 0.58 between the error terms of TAS-DIF and TAS-DDF. There were no other significantly correlated error terms. For adequacy of fit of this model see Table 2. HAD = Hospital Anxiety Depression Scale; GHQ = General Health Questionnaire; UMACL = UWIST Mood Adjective Checklist; CHIP= Coping Inventory for Health, Illnesses and Problems; EPQ-R = Eysenck Personality QuestionnaireRevised; TAS = Toronto Alexithymia Scale; DIF = Difficulty Identifying Feelings; DDF = Difficulty Describing Feelings.

/

/1(74

( 6X\

+ ‘-- ‘-

ITAS_DIF)

-q&q /

(98)

emotion

piq

Fig. 2. Two-factor model of the associations among 10 key variables in the study. Loadings for the male sample (N = 93) are shown outside parentheses and female (N = 151) loadings inside. Not shown are two small but significant loadings: a loading of - 0.17 between UMACL-tense arousal and F2 in women, and a loading of 0.16 between HAD-anxiety and F2 in men. For adequacy of fit of this model see Table 2. For key to test acronyms see Fig. 1.

square by 14.5 (P < O.Ol), and a similar change in the female model improves the chi square by 13.2 (P < 0.01). Therefore, in both men and women the two factor model is significantly preferable to the single factor model. As was done with the single factor model, the two factor model was subjected to multisample confirmatory factor analyses using the two sexes as cross-validation samples. Again, this exercise tests not just that the same structural model will fit in the two sexes, but the additional constraint that the loadings will have the same values. Therefore, this is a much

Medically unexplained symptoms

o!

0

I

4

8

Component Fig. 3. Eigenvalues Form) neuroticism

559

12

16

number

of the first 16 components of the 12 Eysenck Personality Questionnaire-Revised (Short questions and the 20 Toronto Alexithymia Scale questions after principal components analysis (N = 228).

more powerful test of model stability in the two sexes. All factor loadings seen in Fig. 2 were constrained to be equal. Additionally, the correlation between the two latent factors was constrained to be equal. EQS was used to test these constraints, and all but two of them were supported. The loadings of HAD-anxiety and UMACL-tense arousal on latent variable Fl were found to be significantly different between the sexes. Therefore, except in these two instances, the very same two factor model fits well in both sexes, the parameter weights may be assumed to be equal across the sexes, and the two factor model fits significantly better than the single factor model. The last analysis involved only neuroticism and alexithymia and was conducted at the item level. The 12 neuroticism questions from the EPQ-R short form and the 20 questions from the TAS-20 were subjected to a joint principal components analysis. There were 228 complete records for this analysis. Sixteen Ss had missing data for one item across all the scales and had had scores prorated for earlier analyses. In the present analysis only Ss with complete item data were used. The scree slope of the first 16 components’ eigenvalues for the 32 neuroticism/alexithymia questions is shown in Fig. 3. Four components emerged above the scree and were extracted and obliquely rotated in a subsequent analysis. Table 3 shows the pattern matrix from this analysis. Factor 2 has substantial loadings for all neuroticism items, and none of the TAS-20 items has a loading of more than 0.2 on this factor. Therefore, neuroticism items and TAS-20 items are well separated on this analysis. The correlation between factor scores from factor 2 and MUS scores was -0.27 (N = 228, P < 0.001). Factor 1 has very high loadings ( > 0.6) for five DIF items and one DDF item. There are substantial loadings (0.4 to 0.6) for two DIF items and three DDF items. It is not surprising that DIF and DDF items co-segregate, given the large intercorrelation found between these subscales of the TAS in men and women in this study (Table 1). However, it should be stated that Table 3 does not represent a test of the factor structure of the TAS-20, because the presence of EPQ-R neuroticism items could distort factor structures in the TAS. Nevertheless, it is clear that factor 1 holds the items that contribute to the associations between alexithymia and MUS in this study. The correlation between factor scores for factor 1 and MUS was 0.30 (N = 228, P < 0.001). The correlation between factor 1 and factor 2 was 0.27 (N = 228, P < 0.01). The three EPQ-R neuroticism items that have modest loadings (0.34 to 0.40) on this factor are all related to unstable, unexplained or low moods: EPQ-R item 1, “Does your mood often go up and down?“; EPQ-R item 5, “Do you ever feel ‘just miserable’ for no reason?“; and EPQ-R item 17, “Do you often feel ‘fed up’?” When factor scores from factor 1 and 2 were used to predict MUS scores the multiple R was 0.47 (adjusted RZ = 0.22). Factor 1 contributed 8.5% of the variance and factor 2 13.2%. It is instructive to inquire about the nature of these two factors, because they are relatively distinct and contribute independent and substantial amounts of the variance accounted for in MUS. The texts of the three highest loading items on factor 1, all of which have low loadings on factor 2 are: TAS-20 item 13, “I don’t know what’s going on inside me”; TAS-20 item 9, “I have feelings I can’t quite identify”;

560

Ian J. Deary et al. Table 3. Principal components analysis of EPQ-R neuroticism items and TAS-20 items (N = 228): Factors are the obliquely rotated components (pattern matrix) Subscale

Factor

EPQ-R item number I 5 9 13 17 21 25 30 34 38 42 46

Neuroticism Neuroticism Neuroticism Neuroticism Neuroticism Neuroticism Neuroticism Neuroticism Neuroticism Neuroticism Neuroticism Neuroticism

0.40 0.38 0.15 -0.09 0.34 -0.09 -0.01 0.10 -0.00 -0.03 0.25 0.22

TAS-20 item number 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 I7 18 19 20

DIF DDF DIF DDF EOT DIF DIF EOT DIF EOT DDF DDF DIF DIF EOT EOT DDF EOT EOT EOT

0.69 0.68 0.57 -0.32 0.07 0.59 0.67 0.35 0.75 0.10 0.56 0.45 0.76 0.72 0.27 -0.01 0.46 -0.02 0.13 0.11

I

Factor 2

Factor 3

Factor 4

0.44 0.41 0.39 0.51 0.45 0.78 0.70 0.69 0.54 0.78 0.38 0.56

0.03 -0.08 -0.09 0.02 -0.14 0.08 0.05 0.07 -0.05 0.00 -0.08 0.02

-0.22 -0.27 -0.24 0.06 -0.25 0.16 -0.00 0.07 -0.07 0.07 -0.13 -0.19

0.20 0.16 -0.00 -0.07 0.05 0.18 -0.06 -0.08 -0.09 0.09 0.11 0.06 -0.05 0.09 -0.05 -0.15 -0.00 -0.02 -0.11 0.03

-0.07 -0.17 0.28 0.50 0.50 0.18 0.37 -0.13 0.18 0.65 -0.17 -0.13 -0.01 0.02 0.01 - 0.02 -0.17 0.55 0.62 -0.00

0.04 0.22 0.02 -0.16 -0.05 0.23 0.09 0.45 0.07 -0.07 0.37 0.46 0.26 0.05 0.45 0.73 0.37 0.06 -0.10 0.61

and TAS-20 item 14, “I often don’t know why I am angry”. All three items belong to the DIF subscale. The texts of the three highest loading items on factor 2, all of which have low loadings on factor 1 are: EPQ-R item 21, “Would you call yourself a nervous person?‘; EPQ-R item 38, “Do you suffer from ‘nerves’?“; and EPQ-R item 25, “Are you a worrier?” Factor 1 is clearly about the ability to understand and identify internal feelings and perhaps sensations, and factor 2 is related to trait anxiety. Factor 3 in Table 3 has five high loadings (> 0.5); four are EOT items and one is a DDF item. The content of these items is oriented toward a preference for examining, taking cognisance of, and recognising the importance of feelings. The two high loadings on factor 4 (both EOT items) relate to a liking for meaning/feelings in entertainments. Scores on neither of these factors related to MUS, and the latter may be a bloated specific, i.e. a spurious factor emerging from similar, narrowlyfocussed questions. DISCUSSION Aspects of the Toronto Alexithymia Scale-20 (TAS-20) proved successful in predicting Ss’ experiences of somatic symptoms that lacked a physical basis. Therefore, the medical problem that spawned the construct of alexithymia-medically unexplained symptoms, also called functional somatic symptoms or syndromes-bears at least a statistical association with a psychometric instrument designed to assess alexithymic traits. Not all subscales of the TAS-20 were equally effective in this regard. The TAS-DIF was the most effective correlate of MUS, with TAS-DDF a more modest correlate. Not only were these correlations highly stable across the sexes in this study, they were similar in magnitude to correlations we obtained in a study of insulin-dependent diabetics (Deary

Medically unexplained symptoms

561

et al., in press). In multiple regression models of MUS, in competition with a number of well validated relevant variables, TAS subscales, especially the TAS-DIF, played a prominent part. Therefore, among extant variables thought to be related to the experience of medically unexplained symptoms, alexithymia is equally effective when compared with neuroticism, negative emotion coping, anxiety, depression, general psychological distress and current mood state. However, does the alexithymia measure add anything new to the prediction of MUS? The fact that so many variables were unnecessary in the multiple regression equations once the TAS-20 subscales had been added raises the possibility that the variance in alexithymia measures might be overlapping with that of other constructs. Indeed, TAS-DIF and -DDF subscales have highly significant associations with neuroticism, negative emotion coping, tense arousal, hedonic tone, energetic arousal, anxiety, depression and general psychological distress. Moreover, the large percentage of the variance accounted for by the first unrotated principal component, when many of the variables used here were subjected to principal components analysis, confirms a large degree of redundancy in the measures. By acknowledging this result we are in a similar position to theorists in psychometric intelligence. It is possible to recognise this large source of common variance and make it the principal object of study. This is very similar to the approach taken by those who study Spearman’s g (general intelligence; see Carroll, 1993) as a determinant of cognitive ability. In the study of human distress a similar general factor has been suggested by Watson and Pennebaker (1989), who called it negative affectivity or somatopsychic distress, and its relationship to neuroticism is very strong. Such a powerful source of variance must clearly be of interest and must be acknowledged as a potential threat to the profusion of overlapping health related psychological constructs that threatens to choke progress in the field (Deary et al., in press; Kirmayer et al., 1994; Marshall et al., 1994; Smith & Williams, 1992). Nevertheless, recognising the primacy of negative affectivity/somatopsychic distress does not detract from the potential importance of narrower, individual constructs. Again, a comparison with intelligence research is useful. There are those in intelligence research who, rather than focusing on g, study specific abilities because, whereas it is recognised that all mental abilities are positively correlated, there are separable group factors and specific factors also (Carroll, 1993). In the field of human distress a similar approach has been taken. For example, Goldberg (Goldberg & Williams, 1988) in his work on the General Health Questionnaire, has acknowledged that, although a single general factor exists among its questions, anxiety and depression may still be usefully distinguished. Therefore, while acknowledging that most of the variables used in the present study were substantially intercorrelated, one may also inquire whether there are relatively separable group or specific factors among the variables. In the present study this was done by constructing structural equation models of the covariance among the variables. A single latent factor model, very similar across the sexes, accounted for much of the inter-test covariance among 10 constructs, giving strong support to a negative affectivity model of psychological distress, and suggesting that MUS is yet another reflection of a very general tendency toward negative reporting. However, a two-factor model fitted the data significantly better; again, a very similar model-in terms of model structure and parameter weights-could be constructed in the two sexes. The twofactor model gave support to the hypothesis that alexithymia might have a special place in the explanation of MUS. Factor 1 in the two factor model had high loadings from anxiety, depression, dysphoric mood, general psychological distress and neuroticism, and modest loadings from MUS (the DSM-somatisation scores) and negative emotion coping. Factor 2 had high loadings from TASDIF and -DDF, and modest loadings from MUS and negative emotion coping. An interpretation of this model, then, is that MUS are partly an expression of negative affectivity/somatopsychic distress and that alexithymia adds some independent variance to the explanation of MUS differences. It should be noted that factors 1 and 2 are correlated, reflecting alexithymia’s sharing in the very general factor, but it appears that there might be a special place for alexithymia in the study of unexplained medical symptoms. What is it about the alexithymia construct that makes this special contribution to accounting for functional somatic symptoms? The results from the joint factor analysis of neuroticism and alexithymia items might be of some help in answering this question. From the factor scores of this analysis it was possible to identify two separable and substantial contributions to the prediction of somatisation. One was clearly related to neuroticism items that focussed on anxiety proneness

562

Ian J. Deary et al.

(factor 2 in Table 3). The other (factor 1 in Table 3) emphasised confusion about the origins and characteristics of feelings; here, in the contents of the high loading questions on factor 1, the key problem did not appear to be merely the giving of words to feelings so much as a difficulty in discriminating internal sensations. That is, the sensations created by emotional turmoil and bodily disturbances were unable to be distinguished and, thereafter, labelled. Speculatively one might suggest that two separate mechanisms contribute to the experiencing of MUS. Negative affectivity/somatopsychic distress/neuroticism level might act as a threshold for the liability to such symptoms by acting as a general threshold for most types of human distress, including anxiety, depression and MUS. One possible physical basis for individual differences in this construct, suggested by Eysenck (1967) is reactivity of the autonomic nervous system. Hepburn, Deary, MacLeod and Frier (1994) have provided some evidence to support this idea, by demonstrating that neuroticism is related to worry and preferentially to autonomic symptoms (rather than other types of symptoms) in diabetic patients. Therefore, higher levels of negative affectivity, in signal detection terminology, might cause a greater rate of detection of functional (medically unexplained) somatic symptoms. Whether this is caused by greater sensory sensitivity or altered response bias is a moot point; either way, their threshold is lower for symptom detection. Alexithymia, or the part of it that we have found to be most useful in this study, might be a parameter that affects signal recognition rather than signal detection. That is, its principal effect in signal detection terms affects not the decision about whether or not a signal (in this case reporting a significant physical symptom) has occurred, but in discriminating which signal has occurred. In understanding alexithymia further it might be useful to attempt to characterise the processes whereby emotions and physical sensations achieve distinctiveness and the sources of noise in discriminating such feelings. Therefore, a two-parameter model for the reporting of medically unexplained symptoms may be offered. Individual differences in negative affectivity might set a threshold, whereby higher levels will predispose a person to report more negative emotions and physical symptoms alike. Individual differences in alexithymia might act as an interference factor that tends to influence the confusion matrix among emotions and between physical symptoms and emotions. Somehow, the person with high alexithymia has a blurring across emotional and symptomatic representation, like cross-talk on a telephone line or a horizontal connection in a neural network. An interesting prediction from this model is that, given a real rise in emotional distress, a person with high alexithymia will report more physical symptoms and, vice versa, given a real rise in physical symptoms (especially autonomic symptoms as, say, in a gastrointestinal upset) will be likely to report more psychological distress. In the present study the approach to the measurement of alexithymia and MUS was to optimise their accuracy and to ensure that the sample studied was not restricted in range for MUS experience. Therefore, TAS-20 subscales were employed, rather than the total TAS-20 score, because better correlations were obtained; adding in the EOT subscale score appeared merely to add noise to the measurement of relevant variance contributed by the DIF and DDF subscales. For the assessment of medically unexplained symptoms, or somatisation, we used a self-report scale adapted from the SCID interview, which is based on the symptom list from the DSM-III-R somatisation disorder. Of the symptom checklists available, this would appear to be the most comprehensive and have the best clinical validation, criteria emphasised by Bach et al. (1994). With regard to the S sample tested in the present study, it was considered important to include a wide range of MUS experience, from none to individuals with proven, current functional disorders. In this field of research (e.g. Bach et af., 1994) and in other areas of health-personality research (e.g. in the study of Type A behaviour and heart disease; see Miller, Turner, Tindale, Posavac & Dugoni, 1991) it is common for null studies to result from attenuation of the health-related variable. Therefore, studies that include only those people with MUS, or only students, will be unlikely to find strong associations with a predictor variable because the range of experience of the symptoms is too small. TAS-EOT failed to correlate significantly with any of the variables in the study. Add to this the fact that it fails to correlate above a negligible degree with the other two subscales with which it is supposed to form a unitary construct (alexithymia). In the light of the results of this study, therefore, it is not possible to identify an empirical reason to retain this TAS-20 subscale in the study of alexithymia’s association with MUS. On the other hand, the correlations between TAS-DIF and -DDF are so large, and the questions so mingled in the analysis presented in Table 3, that it should

Medically unexplained

symptoms

563

be asked whether the alexithymia construct might not be captured by a single scale developed from the questions therein. This interpretation is supported by the F2 latent trait in Fig. 2, in which TASDIF and -DDF reflect a single underlying variable. In conclusion, it seems that the alexithymia construct can contribute to the study of MUS in a way that is incremental to many other validated constructs. Therefore, Taylor, Bagby and Parker (1991) were correct in suggesting that alexithymia might be a “potential paradigm for psychosomatic medicine”. However, the construct may require some pruning when applied to the study of medically unexplained symptoms. The aspects captured by the Externally Oriented Thinking subscale are unimportant, and the problems with the confusion among different feelings and between feelings and bodily sensations might better be captured by a single scale. Acknowledgements-We thank providing data for this study.

Drs Kenneth

MacKenzie,

Aileen White and lsobel Wilson

for their valuable

assistance

in

REFERENCES Acklin, M. W. & Bemat, E. (1987). Depression, alexithymia and pain prone disorder: a Rorschach study. Journal of Personality Assessmeni, 51,462479. Apfel, R. J. & Sifneos, P. E. (1979). Alexithymia: concept and measurement. Psychotherapy and Psychosomarics, 32. 18& 190. Bach, M., Bach, D., Bohmer, F. & Nutzinger, D. 0. (1994a). Alexithymia and somatization: relationship to DSM-III-R diagnoses. Journal of Psychosomatic Research, 38, 23-32. Bach, M., de Zwaan, M., Ackard, D. O., Nutzinger, D. 0. &Mitchell, J. E. (1994b). Alexithymia: relationship to personality disorders. Comprehensive Psychiafry, 35,239-243. 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 Psychosomatic Research, 38, 23-32. Bagby, R. M., Taylor, G. J. & Atkinson, L. (1988). Alexithymia: a comparative study of three self-report measures. Journal of Psychosomatic Research, 32, 107-l 16. Bagby, R. M., Taylor, G. J. & Parker, D. A. (1994b). The twenty-item Toronto Alexithymia Scale-II. Convergent, discriminant and concurrent validity. Journal of Psvchosomatic Research, 38, 3340. Beck, A. T., Rush, A. J., Shaw, B. F. & Emery, G. (1979). Cognitive therapy of depression. New York, NY: Guilford. Bender, P. M. (1995). EQS structural equations program manual. Encino, CA: Multivariate Software Inc. Bentler, P. M. & Wu, E. J. C. (1995). EQSfor Mucinrosh user’sguide. Encino, CA: Multivariate Software Inc. Bradlow, J., Coulter, A. & Brookes, P. (1992). Patterns of referral. A study of referrals to out-patient clinics from general practices in the Oxford region. Oxford: Health Services Research Unit. Bridges, K. W. & Goldberg, D. P. (1985). Somatic presentation of DSM-III psychiatric disorders in primary care. Journal of Psychosomatic Research, 29, 563-569. Carroll, J. B. (1993). Human cognitive abilities: A survey offactor-analytic studies. Cambridge: Cambridge University Press. Codispoti, M., Codispoti, 0. & Battacchi, M. W. (1995). What’s wrong with the operational definition of the ‘alexithymia construct’? European Journal of Psychological Assessment, II(Suppl. 1), 64. Cohen, K., Auld. F. & Brooker, H. (1994). Is alexithymia related to psychosomatic disorder and somatizing? Journal of Psychosomatic Research, 38, 119-127. Cohen, K., Auld, F., Demers, L. & Catchlove, R. (1985). Alexithymia: the development of a valid and reliable projective test (the objectively scored archetypal9 test). Journal of Nervous and Mental Disease, 173, 621426. Deary, I. J., Blenkin, H., Agius, R. M., Endler, N. S., Zealley, H. & Wood, R. (1996). Models of job-related stress and personal achievement among consultant doctors. British Journal of Psychology, 87, 3-29. Deary, I. J., Clyde, Z. & Frier, B. M. (in press). Constructs and models in health psychology: the case of personality and illness reporting in diabetes mellitus. British Journal of Health Psychology. Deary, I. J., Peter, A., Austin, E. J. &Gibson, G. J. (1995). Normal and abnormal personality traits: problems of assessment. European Journal of Psychological Assessment, II(Supp1. l), 25-26. Deary, I. J., Wilson, J. A. & Kelly, S. W. (1995). Globus pharyngis, personality and psychological distress in the general population. Psychosomatics, 36, 570-577. Demers-Desrosiers, L. A., Cohen, K. R., Catchlove, R. F. H. & Ramsay, R. A. (1983). The measure of symbolic function in alexithymia pain patients. Psychotherapy and Psychosomatics, 39, 65-76. Endler, N. S., Parker, J. D. A. & Summerfeldt, L. J. (1992). Coping with health problems: developing a reliable and valid multidimensional measure. York University (Canada) Department of Psychology Research Report No. 204. Endler, N. S., Parker, J. D. A. & Summerfeldt, L. J. (1993). Coping with health problems: conceptual and methodological issues. Canadian Journal of Behavioural Science, 25,38+399. Eysenck, H. J. (1967). The biological basis ofpersonality. Springfield, IL: Charles C. Thomas. Eysenck, S. B. G., Eysenck, H. J. & Barrett, P. (1985). A revised version of the psychoticism scale. Personality and Individual Differences, 6, 2 l-29. Federman, R. & Mohns, E. (1984). A valid study of the MMPI alexithymia scale conducted on migraine headache outpatients. Psychotherapy and Psychosomatics, 41, 29-32. Freedman, M. B. & Sweet, B. S. (1954). Some specific features of group psychotherapy and their implication for selection of patients. Inrernational Journal of Group Psychotherapy, 4, 355-368. Goldberg, D. & Williams, P. (1988). A user’s guide to the General Health Questionnaire. Windsor, UK: NFER-Nelson. Groat, C. & de Leeuw, R. (1995). Convergent and discriminant validity of alexithymia scales. European Journal of Psychological Assessment, IIfSuppl. I), 57.

564

Ian J. Deary

et al.

Hepburn, D. A., Deary, I. J., MacLeod, K. M. & Frier, B. M. (1994). Structural equation modelling of symptoms, awareness and fear of hypoglycemia, and personality in patients with insulin-treated diabetes, Diabetes Care, 17, 1273-1280. Hyler, S. E. & Rieder. R. 0. (1987). PDQ-R: Personality Diagnostic Questionnaire-revised. New York, NY: New York State Psychiatric Institute. Kellner. R. (1985). Functional somatic symptoms and hypochondriasis. Archives of General Psychiatry, 42, 821-833. Kellner. R. (1991). Psychosomatic syndromes and somatic symptoms. Washington, D.C.: American Psychiatric Press. Keltikangas-Jarvinen. L. (1982). Alexithymia in violent offenders. Journal of Personality Assessment, 46.462467. Keltikangas-Jarvinen, L. (1985). Concept of alexithymia. Psychotherapy and Psychosomatics, 44, 132-138. Kirmayer, L. J., Robbins, J. M. & Paris, J. (1994). Somatoform disorders: personality and the social matrix of somatic distress. Journal of Abnormal Psychology, 103, 125-136. Kleiger. J. H. &Kinsman, R. A. (1980). The development of an MMPI alexithymia scale. Psychotherapy andPsychosomatics, 34, 17-24. Kroenke, K. (1992). Symptoms in medical patients: an untended field. American Journal of Medicine. 92, 3-6. Krystal, H. & Raskin, H. A. (1970). Drug dependence. Detroit, MI: Wayne State University Press. Lesser, I. M. (1981). A review of the alexithymia concept. Psychosomatic Medicine, 4.3. 531-543. Lesser, I. M.. Ford. C. V. & Friedman, C. T. H. (1979). Alexithymia in somatising patients. General Hospital Psychiatry, I, 256261. Marshall, G. N.. Wortman, C. B.. Vickers, R. R., Kusulas, J. W. & Hervig. L. K. (1994). The five factor model of personality as a framework for personality-health research. Journal of Personality and Social Psychology, 67. 278-286. Marty. P. & de M’Uzan, M. (1963). La ‘pensee operatoire’. Revue Francaise de Psychoanalyse, 27, 134551356. Matthews, G.. Jones, D. M. &Chamberlain. A. G. (1990). Refining the measurement of mood: the UWIST Mood Adjective Checklist. British Journal of Psychology, 81. 1742. Mayou, R.. Bass. C. & Sharpe, M. (Eds), (1995). Treatmenf offunctional somatic s_ymptoms. Oxford. UK: Oxford University Press. Miller, T. Q,, Turner, C. W., Tindale, R. S.. Posavac, E. J. & Dugoni, B. L. (1991). Reasons for the trend toward null findings in research on Type A behavior. Psychological Bulletin, 110,469485. Myers. L. (1995). Alexithymia and repression: the role of defensiveness and trait anxiety. Personality and Individual Differences, 19,489-492. Nemiah. J. C.. Frevberaer. H. & Sianeos, P. E. (1976). Alexithymia: a view of the psychosomatic process. In 0. W. Hill (Ed.), Modern tiendrin psychosomatic medicine (Vol. 3, pp. 430-439). London: Butterworths. Norton. N. C. (1989). Three scales of alexithymia: do they measure the same thing? Journal of Personalify Assessment, 53, 621637. . Paociak. A. S.. Feuerstein. M.. Belar. C. D. & Pistone, L. (19861987). Alexithymia and pain in an outpatient behavioural ‘medicine chnic. International Journal of Psychiatry in Medicine, 16, 347-357: Parker, J. D. A., Bagby. R. M. &Taylor, G. J. (1989). Toronto Alexithymia Scale. EPQ and self-report measures of somatic complaints. Personality and Individual Differences, IO, 599-604. Parker, J. D. A., Bagby, R. M. & Taylor, G. J. (1991). Alexithymia and depression: distinct or overlapping constructs? Comprehensive Psychiatry, 32. 3877394. Reusch, J. (1948). The infantile personality: the core problem of psychosomatic medicine. Psychosomatic Medicine, IO, 134 144. Rubino, 1. A. (1993). Is alexithymia a non-neurotic personality dimension? A counter-response to Taylor, Bagby and Parker. British Journal of Medical Psychology, 66, 2899294. Rubino, I. A., Grasso, S., Sonnino, A. & Pezzarossa, B. (1991). Is alexithymia a non-neurotic personality dimension? British Journal of Medical Psychology, 64, 385-391. Sifneos, P. E. (1973). The prevalence of ‘alexithymic’ characteristics in psychosomatic patients. Psychotherapy and Psychosomatics, 22, 255-262. Smith, T. W. & Williams, P. G. (1992). Personality and health: advantages and limitations of the five factor model. Journal of Personality, 60, 395423. Spitzer. R. L., Williams, J. B. W., Gibbon, M. & First, M. B. (1990). Structured clinical interview for DSM-III-R-nonpatient edition. Washington, D.C.: American Psychiatric Press. Sriram, T. G., Chaturvedi. S. K., Gopinath, P. S. & Shanmugam, V. (1987). Controlled study of alexithymia characteristics in patients with psychogenic pain-disorder. Psychotherapy and Psychosomatics, 47, 1 t-17. Tavlor. G. J. (1984). Alexithvmia: concept, measurement and implications for treatment. American Journal of Psychiatry, .141,725-732. . Taylor, G. J., Bagby. R. M. & Parker, J. D. A. (1991). The alexithymia construct: a potential paradigm for psychosomatic medicine. Psychosomatics, 32, 153-162. Tavlor. G. J.. Baabv. R. M. & Parker, J. D. A. (1993). Is alexithymia a non-neurotic personality dimension? A response to ‘R&no, Grasio: Sonnino and Pezzarossa. British Journal of medical Psycholog.v, 66, 281-287. Taylor, G. J.. Bagby, R. M., Ryan, D. P. & Parker, J. D. A. (1990). Validation of the alexithymia construct: a measurement based approach. Canadian Journal of Psychiatry, 35, 29&297. Taylor, G. J., Doody, K. & Newman, A. (1981). Alexithymic characteristics in patients with inflammatory bowel disease. Canadian Journal of Psychiatry, 26, 47W74. Tavlor. G. J.. Rvan. D. & Baabv, R. M. (1985). Toward the development of a new self-report alexithymia scale. Psychotherapy