OUP user menu

Specific brain processing of facial expressions in people with alexithymia: an H215O‐PET study

Michiko Kano, Shin Fukudo, Jiro Gyoba, Miyuki Kamachi, Masaaki Tagawa, Hideki Mochizuki, Masatoshi Itoh, Michio Hongo, Kazuhiko Yanai
DOI: http://dx.doi.org/10.1093/brain/awg131 1474-1484 First published online: 8 April 2003

Summary

Alexithymia is a personal trait characterized by a reduced ability to identify and describe one’s own feelings and is known to contribute to a variety of physical and behavioural disorders. To elucidate the pathogenesis of stress‐related disorders and the normal functions of emotion, it is important to investigate the neurobiology of alexithymia. Although several neurological models of alexithymia have been proposed, there is very little direct evidence for the neural correlates of alexithymia. Using PET, we studied brain activity in subjects with alexithymia when viewing a range of emotional face expressions. Twelve alexithymic and 12 non‐alexithymic volunteers (all right‐handed males) were selected from 247 applicants on the basis of the 20‐item Toronto Alexithymia Scale (TAS‐20). Regional cerebral blood flow (rCBF) was measured with H215O‐PET while the subjects looked at angry, sad and happy faces with varying emotional intensity, as well as neutral faces. Brain response in the subjects with alexithymia significantly differed from that in the subjects without alexithymia. The alexithymics exhibited lower rCBF in the inferior and middle frontal cortex, orbitofrontal cortex, inferior parietal cortex and occipital cortex in the right hemisphere than the non‐alexithymics. Additionally, the alexithymics showed higher rCBF in the superior frontal cortex, inferior parietal cortex and cerebellum in the left hemisphere when compared with the non‐alexithymics. A covariance analysis revealed that rCBF in the inferior and superior frontal cortex, orbitofrontal cortex and parietal cortex in the right hemisphere correlated negatively with individual TAS‐20 scores when viewing angry and sad facial expressions, and that no rCBF correlated positively with TAS‐20 scores. Moreover, the anterior cingulate cortex and insula were less activated in the alexithymics’ response to angry faces than their response to neutral faces. These results suggest that people with alexithymia process facial expressions differently from people without alexithymia, and that this difference may account for the disorder of affect regulation and consequent peculiar behaviour in people with alexithymia.

  • Keywords: ACC, alexithymia, facial expression, lateralization, PET
  • Abbreviations: ACC = anterior cingulate cortex; fMRI = functional MRI; rCBF = regional cerebral blood flow; SPM = statistical parametric mapping; TAS‐20 = 20‐item Toronto Alexithymia Scale

Introduction

Alexithymia is a subclinical phenomenon characterized by the following: reduced ability to identify and describe one’s feelings, difficulty in distinguishing feelings from the bodily sensations of emotional arousal and impaired symbolization, along with a tendency to focus on external events rather than inner experiences (Nemiah et al., 1976; Taylor, 1994; Taylor et al., 1999). People with alexithymia tend to avoid internal emotional conflict by relying on action to express emotion. They are socially conforming but humourless (Haviland et al., 1996). This condition is considered a disorder of affect regulation (Taylor et al., 1999).

Alexithymia has been reported as a typical personality problem among patients with psychosomatic disorders (Sifneos, 1973). Previous studies have reported a high rate of alexithymia in patients with irritable bowel syndrome, somatization, panic disorder, eating disorders, post‐traumatic stress disorders and substance abuse problems, suggesting that alexithymia might have adverse effects on mental and physical health (Taylor et al., 1991; Parker et al., 1993; Taylor, 2000). Therefore, investigating the role of alexithymia in the development of these disorders is of great interest. It is especially important to study the brain activity of people with alexithymia in order to elucidate the pathogenesis of psychosomatic disorders and the normal functions of emotion.

There are several neurological models for alexithymia. One of them, proposed by McLean (1949), postulates a discommunication between the limbic and neocortical areas. In this model the limbic system provides the physiological sensation of emotions, while the neocortex offers the symbolic representation of emotions. The second model points to reduced interhemispheric communication evidenced by split‐brain patients being more alexithymic than the controls (Hoppe and Bogen, 1977; TenHouten et al., 1986). The third model involves dysfunction of the right hemisphere, because highly alexithymic individuals, as well as patients with right hemisphere lesion, show less accuracy in recognizing facial expressions of emotions (Parker et al., 1993; Mann et al., 1994; Lane et al., 1995), photographs of emotional scenes and the emotional nuance of sentences (Lane et al., 1995). The fourth model postulates a deficiency in the activation of the anterior cingulate cortex (ACC) during emotional processing as a cause of alexithymia, suggesting that people with alexithymia have a deficit in the conscious awareness of emotion (Lane et al., 1987). In addition, a PET study revealed positive correlation between high scores on the Level of Emotion Awareness Scale and increased activity in the right ACC during emotional processing when watching films or remembering personal experiences (Lane et al., 1997). Recently, a functional MRI (fMRI) study showed that alexithymia can be associated with deactivation and activation in ACC in response to highly negative emotional stimuli and highly positive emotional stimuli, respectively (Berthoz et al., 2002).

Many clinical reports and experimental studies have supported the hypothesis that alexithymia might be associated with variations in brain activity. However, most of these studies have relied mainly on indirect methods to characterize the brain functions of people with alexithymia. Even the proposed dysfunction in ACC demonstrated by the fMRI study might be insufficient to account for all of the features of alexithymia. Therefore, the underlying neural structure of people with alexithymia still remains to be elucidated.

Facial expressions are non‐verbal communicative displays that convey affective messages (King and Brothers, 1992). Because previous studies have shown a correlation between alexithymia and a lack of ability to recognize emotions in photographs of facial expressions (Parker et al., 1993; Mann et al., 1994; Lane et al., 1995, 2000), in the present study we examined how the brain activity of people with alexithymia is affected when they look at pictures of various facial expressions, and compared the results with those of people without alexithymia. Moreover, to clarify the effect of valence (positive or negative) and emotional intensity (high or low arousal), we adopted as stimuli four emotional categories (anger, sadness, happiness and neutral) and a range of three levels of emotional intensity (mild, moderate and intense) for each emotional category. Using PET, direct observation of the brain activity of subjects viewing facial expressions is expected to elucidate the specific neural structure in people with alexithymia. The question is whether the neural variations in people with alexithymia correspond with previous neurological models for alexithymia.

Material and methods

Subjects

Two hundred and forty‐seven male volunteers were screened for levels of alexithymia, using the 20‐item Toronto Alexithymia Scale (TAS‐20), which is the most psychometrically valid measurement of alexithymia (Bagby et al., 1994a, b). The TAS‐20 is a self‐report questionnaire with a maximum score of 100 that measures participants’ ability to describe and identify feelings, and their tendency to exhibit externally oriented thinking. Participants answer questions on a five‐point scale indicating ‘strongly disagree’ to ‘strongly agree.’ The Japanese version of the TAS‐20 has been found to be psychometrically valid (Fukunishi et al., 1997). The averaged TAS‐20 score of the 247 volunteers was 46.5 ± 8.5 (mean ± SD). Individuals with a TAS‐20 total score of >61 were considered alexithymic, and those with a score of <51 were considered non‐alexithymic (Taylor et al., 1999). In this study, the PET subjects consisted of 12 alexithymic and 12 non‐alexithymic subjects who were selected according to these cut‐off points (TAS‐20 total score). The alexithymics and non‐alexithymic subjects were aged 23.2 ± 2.4 and 22.8 ± 1.7 years, respectively (mean ± SD). All subjects were evaluated as right‐handed based on the Edinburgh inventory (Oldfield, 1971). None of the subjects had a history of psychiatric or neurological disorders, nor were any of them on medication. Informed consent was obtained from all subjects according to the Declaration of Helsinki, and all experiments were performed in compliance with relevant laws and institutional guidelines.

Facial expression images

Static images of Japanese models (expressers) in their 20s were selected from the ATR (ATR International) face database (Kamachi et al., 2001). This database is comprised of 60 males and 60 females showing basic facial expressions. These faces had previously been rated on all expression categories, making it possible to select models who were good exmples of each of the required expressions. Four facial expressions—anger, sadness, happiness and neutrality—each expressed by 20 models (10 females and 10 males), were selected and used as stimuli in the experiment. For each emotional category and individual face used in this experiment, a range of three levels of emotional intensity (33% mild, 67% moderate, 100% intense) was produced by computer graphic manipulation to enable the measurement of various levels of emotional intensity. The 33 and 67% emotional intensity faces were interpolations created using computer morphing procedures (Kamachi et al., 2001). These involved the shifting of the shape and the shading of the neutral face (0%) towards the angry, happy or sad prototype (100%).

Experimental task

In order to investigate implicit or automatic processing of facial expressions under conditions free of bias and the decision‐making process, the subjects viewed static images of emotional faces on a computer monitor screen and were simply required to determine the gender of each face (male or female) by pressing the left or right response button. Attention to each face stimulus was required, but recognition or categorization of the emotional expression was not required during this task.

Several previous studies adopted this gender discrimination task as the first step in examining implicit facial expression processing (Morris et al., 1996; Phillips et al., 1997; Sprengelmeyer et al., 1998; Blair et al., 1999; Critchley et al., 2000a, b).

Experimental design

For each subject, 10 separate PET scans were required. The scans consisted of one scan for the neutral facial expression, three scans for anger (mild, moderate and intense), three scans for sadness and three scans for happiness. During each of the 10 PET scans, 20 photographs (10 females and 10 males) with the same emotional category and intensity of emotional expression were presented, one at a time, on a computer monitor screen. Each photograph was displayed for 3 s, followed by a 2 s interval. The order of emotional faces presented was counterbalanced across the subjects.

Subject debriefing

Immediately after each scanning session, the subjects were informed that all 20 individuals in the photographs had displayed the same emotional category, and were asked to recall and identify the emotions being expressed by the 20 individuals. They evaluated the intensity of anger, sadness, happiness, disgust, fear and surprise on a 10‐point scale. A score of 10 corresponded to maximal intensity, while a score of 1 corresponded to minimal intensity. The rating scores were used to reveal differences between alexithymics and non‐alexithymics in their ability to accurately recognize the emotions shown in facial expressions.

PET scan acquisition

Scans of the distribution of H215O were obtained using an SET‐2400W PET scanner (Shimadzu, Kyoto, Japan) operated in high sensitivity 3D mode, with an average axial resolution of 4.5 mm at maximum strength and sensitivity for a 20 cm cylindrical phantom of 48.6 kcps/kBq/ml in the 3D mode (Fujiwara et al., 1997). The subjects received ∼5 mCi (185 MBq) of H215O intravenously through the antecubital vein for each scan and engaged in the task during measurements of regional cerebral blood flow (rCBF). The task started 30 s before the PET scan and finished 15 s after the end of the scan.

PET data analysis

Statistical parametric mapping (SPM) software (SPM99; Wellcome Department of Cognitive Neurology, London, UK) was used for image realignment, normalization and smoothing, and to create statistical maps of significant rCBF changes (Friston et al., 1995a, b). All rCBF images were stereotaxically normalized using nonlinear transformation into the standard space of Talairach and Tournoux (1988). The normalized images were smoothed using a 12 × 12 × 12 mm Gaussian filter, and the values of rCBF were expressed as ml/dl/min, adjusted using ANCOVA (analysis of covariance) and scaled to a mean of 50. Group and covariate effects were estimated according to the general linear model at each voxel.

To determine whether there are specific brain regions correlated with alexithymia, we performed three types of analysis. In all analyses, the estimates were compared using linear compounds or contrasts, and the resulting set of voxel values for each contrast constituted a parametric map of the t‐statistic. In the first analysis, we compared rCBF changes between alexithymics and non‐alexithymics when they looked at the same category of emotional faces (anger, sadness or happiness). In order to maximize the sensitivity of the analysis, mild, moderate and intense levels of emotional intensity were combined in each emotional category. In the second analysis, we performed a correlation analysis between the rCBF changes combined in the three intensities of each emotional category and the individual TAS‐20 scores for all 24 subjects. The relationship between the relative rCBF values at peak in each brain area and the TAS‐20 scores was evaluated by the Pearson’s correlation method. Finally, the third analysis was a conjunction analysis performed to assess differences between alexithymic and non‐alexithymic subjects in cerebral regional activation in response to emotional faces compared with neutral faces. Brain responses in the three intensities of emotional expression were combined in each emotional category. The conjunction analysis reveals areas in the brain where there is a significant main effect of two contrasts (Price and Friston, 1997).

Results

TAS‐20 scores

The TAS‐20 scores were normally distributed with an average of 46.5 ± 8.5 (mean ± SD). The means ± SD for TAS‐20 scores between the alexithymic and the non‐alexithymic group were 64.2 ± 3.6 (61–70) and 40.5 ± 5.7 (31–49), respectively.

Behavioural data and subject debriefing

In the gender discrimination task, the alexithymics and non‐alexithymics correctly identified 96% and 98% at the levels of the images, respectively, and the task performance was not significantly different between the two groups.

The subjects, rating scores for stimuli indicated the degree to which the subjects recognized each emotional category and its intensity levels. For example, the rating score after the subjects viewed 0 (neutral), 33, 67 and 100% of angry expressions correlated highly with the original degree of anger prototype in the morphed face (r = 0.59, P < 0.0001 in non‐alexithymics, r = 0.68, P < 0.0001 in alexithymics). Significant correlations were similarly observed for sad (r = 0.45, P < 0.002 in non‐alexithymics, r = 0.41, P < 0.004 in alexithymics) and happy (r = 0.67, P < 0.0001 in non‐alexithymics, r = 0.77, P < 0.0001 in alexithymics) expressions.

Unpaired t‐tests showed that alexithymics significantly rated intense sad faces as disgust (t = 2.36, P < 0.03). However, there were no significant differences between the alexithymics and the non‐alexithymics in any other ratings.

Differences between alexithymics and non‐alexithymics

Areas of significantly lower rCBF among the alexithymics when compared with the non‐alexithymics were mostly localized in the neocortex of the right hemisphere (P < 0.0001, uncorrected; Fig. 1). The coordinates and Z‐scores for these areas are given in Table 1. Lateralization was commonly observed when viewing all three continua of an emotional face. Areas where rCBF was lower in the alexithymics than in the non‐alexithymics were: the orbitofrontal cortex, middle frontal gyrus, inferior parietal gyrus, cuneus and cerebellum culmen of the right hemisphere. In contrast, areas of higher rCBF in the alexithymics when compared with the non‐alexithymics were almost localized in the neocortex of the left hemisphere (P < 0.0001, uncorrected; Fig. 2). The coordinates and Z‐scores for these areas are given in Table 2. Lateralization was commonly observed in all types of emotional faces. rCBF in the superior frontal cortex, inferior parietal cortex and cerebellar lingual in the left hemisphere was higher in the alexithymics regardless of the emotional face used.

Fig. 1 SPMs of a comparison of 12 alexithymic versus 12 non‐alexithymic subjects while viewing facial expressions. Each figure depicts the brain areas where the rCBF was significantly (P < 0.0001, uncorrected) lower in the alexithymics than in the non‐alexithymics while the subjects looked at (A) angry, (B) sad, (C) happy and (D) neutral faces. The lower rCBF responses in the alexithymic group were mainly lateralized in the neocortex of the right hemisphere.

Fig. 2 SPMs of a comparison of 12 alexithymic versus 12 non‐alexithymic subjects while viewing facial expressions. Each figure depicts a significantly (P < 0.0001, uncorrected) higher rCBF in the alexithymics when compared with the non‐alexithymics while the subjects looked at (A) angry, (B) sad, (C) happy and (D) neutral faces. The higher rCBF in the alexithymics was mainly lateralized in the neocortex of the left hemisphere.

View this table:
Table 1

The coordinates and Z‐scores for the brain areas where the rCBF responses were lower in the alexithymics than in the non‐alexithymics while the subjects viewed facial expressions

Area (BA)AngrySadHappy
SideZ‐scoreTalairach coordinatesZ‐scoreTalairach coordinatesZ‐scoreTalairach coordinates
x y z x y z x y z
1 Orbitofrontal cortex (11)R4.433250–204.93652–184.13048–22
2 Middle frontal gyrus (9)R4.78442424.53440425.544044
3 Inferior parietal lobe (40)R5.0954–48324.2850–48344.4854–4834
4 Cuneus (19)R5.1920–78345.2520–78324.5120–7630
5 Cerebellum culmenR4.4420–52–224.2316–52–265.218–52–24
6 Inferior frontal gyrus (44/45)R5.476216125.75621810
7 Superior occipital gyrus (19)R4.1236–74384.2236–7236
8 Lingual gyrus (19)L4.2–20–5804.28–18–540
9 Cerebellum tuber R5.0644–82–284.3138–72–26
10 Superior temporal gyrus (40/42)R5.5270–1824
11 Middle frontal gyrus (46)R4.26503026
12 Postcentral gyrus (2)L4.36–50–2640
13 Fusiform gyrus (37)R4.7946–54–4
14 Middle occipital gyrus (19)R4.6540–9410
15 CerebellumL4.12–8–80–38
16 Anterior cingulate gyrus (24)R4.210438
17 Inferior parietal cortex (40)R4.4172–22–26
18 CuneusR4.488–6610

BA = Brodmann area; L = left; R = right.

View this table:
Table 2

The coordinates and Z‐scores for higher rCBF areas in the alexithymics when compared with the non‐alexithymics while the subjects looked at facial expressions

Area (BA)AngrySadHappy
SideZ‐scoreTalairach coordinatesZ‐scoreTalairach coordinatesZ‐scoreTalairach coordinates
x y z x y z x y z
1 Superior frontal gyrus (10)L5.6–1868164.34–1668145.16–186816
2 Inferior parietal lobe (40)L4.92–62–44485.3–62–40506.15–62–4248
3 Cerebellar lingualL5.48–4–46–125.18–4–46–124.84–4–44–10
4 Corpus callosumL/R4.68–2–38164.162–3618
5 Middle occipital gyrus (19)L5.35–48–8644.49–50–820
6 CerebellumL4.37–60–50–264.26–58–50–26
7 Middle temporal gyrus (21)L4.51–56–8–105.14–52–10–10
8 Corpus callosumR4.2916–1430
9 Superior temporal gyrus (38)R4.882812–36
10 ThalamusL4.15–18–328
11 Cerebellum tonsilL4.77–28–36–32
12 Superior frontal gyrus (8)R4.34422850
13 InsulaL4.29–421010
14 Superior temporal gyrus (21)L4.38–60–64
15 Superior parietal gyrus (7)L4.21–46–6652
16 Occipital gyrusL4.27–22–9230

BA = Brodmann area; L = left; R = right.

Differences between alexithymics and non‐alexithymics in response to each stimulus are listed in Tables 3 and 4 (P < 0.001, uncorrected). The less responsive areas in the alexithymics were localized in the right hemisphere, while the more responsive areas were mostly localized in the left hemisphere. Differences between alexithymics and non‐alexithymics could frequently be identified even at mild levels of emotional intensity.

View this table:
Table 3

The coordinates and Z‐scores for the brain areas where the rCBF were lower in the alexithymics than in the non‐alexithymics while they viewed each facial stimulus

Area (BA)SideNo. voxelsZ‐scoreTalairach coordinates
x y z
Mild anger
Inferior parietal gyrus (40)R753.7770–2222
Middle frontal gyrus (9)R1003.76483030
Temporal sub‐gyrusR563.548–56–8
Moderate anger
Inferior frontal cortex (44)R1114.0864812
Mild sadness
Inferior frontal gyrus (40)R203.6641010
Orbitofrontal cortex (11)R203.413660–16
Mild happiness
Cerebellum tuberR293.3140–76–26

BA = Brodmann area; R = right.

View this table:
Table 4

The coordinates and Z‐scores for the brain areas where the rCBF were higher in the alexithymics than in the non‐alexithymics while they viewed each facial stimulus

Area (BA)SideNo. voxelsZ‐scoreTalairach coordinates
x y z
Mild anger
Middle occipital gyrus (19)L753.7770–2222
Superior frontal gyrus (10)L203.44–226814
Intense anger
Inferior parietal lobe (40)L213.33–60–4650
Mild sadness
Inferior parietal lobe (40)L333.62–62–4250
Moderate sadness
Cerebellum culmenL213.35–30–32–28
Mild happiness
Precentral gyrus (6)R933.720–2478
Moderate happiness
Inferior parietal lobe (40)L1074.47–60–4452
Superior frontal gyrus (10)L353.61–26686
Intense happiness
Parietal sub‐gyrusL333.38–24–5624
Neutral R203.3144210

BA = Brodmann area; L = left; R = right.

Correlation with individual TAS‐20 scores

Most brain areas that negatively correlated with TAS‐20 scores were localized in the right hemisphere (P < 0.05, corrected; Fig. 3). The coordinates and Z‐scores for these areas are given in Table 5. Each brain area was observed while the subject viewed angry and sad faces. On the other hand, no significant correlation was observed between rCBF and TAS‐20 scores when the subjects viewed happy and neutral faces. Brain areas where the rCBF was lower in the alexithymics than in the non‐alexithymics also negatively correlated with TAS‐20 scores. No region in the brain positively correlated with TAS‐20 score.

Fig. 3 Brain activity was correlated negatively with TAS‐20 scores while the subjects looked at (A) angry and (B) sad faces. Areas of significant correlation (P < 0.05, corrected) are colour‐scaled according to the Z‐scores (scale given in figure).

View this table:
Table 5

The coordinates and Z‐scores for areas of rCBF were negatively correlated with TAS‐20 scores while the subjects looked at facial expressions

Area (BA)SideAngrySad
Z‐scoreTalairach coordinatesZ‐scoreTalairach coordinates
x y z x y z
1 Orbitofrontal cortex (11)R5.572854–225.213454–22
2 Inferior frontal gyrus (44/45)R4.646418105.71642210
3 Middle frontal gyrus (9)R4.74440464.5946–242
4 Middle temporal gyrus (21)L5.62–44–5485.17–38–5210
5 Cuneus (19)R4.8520–76324.6320–7630
6 Cerebellum tuberR5.5444–84–28
7 Superior temporal gyrus (40/42)R4.8168–2424
8 Inferior parietal lobe (40)R5.8650–4832
9 Fusiform gyrus (37)R5.1948–56–4
10 Precuneus (7)L4.69–16–8050
11 Middle frontal gyrus (6)R5.1648262

BA = Brodmann area; L = left; R = right.

Conjunction analysis

The coordinates and Z‐scores of the conjunction analysis are given in Table 6 (P < 0.001, uncorrected). Anger stimuli (angry faces compared with neutral faces) induced less activation in the alexithymics than in the non‐alexithymics in the bilateral insula, left precentral gyrus, right cingulate cortex (Fig. 4), right superior temporal gyrus, left temporal sub‐gyrus, middle occipital gyrus and left cerebellum. Sadness stimuli (sad faces compared with neutral faces) were associated with reduced activation in the right insula, left precentral cortex and left temporal sub‐gyrus in the alexithymics compared with the non‐alexithymics. Happiness stimuli (happy faces compared with neutral faces) were related to less activation in the alexithymics than in the non‐alexithymics in the right medial frontal gyrus, right insula and left precentral gyrus, left temporal sub‐gyrus, corpus callosum and left cerebellum culmen.

Fig. 4 An SPM showing the right insula and ACC where rCBF was lower in alexithymic subjects than in non‐alexithymics in their responses to angry faces compared with their responses to neutral faces. Views of the brain are shown through orthogonal slices, sliced at the pixel of maximal activation within the right insula (x = 42, y = 2, z = 10).

View this table:
Table 6

The coordinates and Z‐scores for areas where rCBF were lower in alexithymics than in non‐alexithymics in response to emotional faces compared with neutral faces

Area (BA)AngrySadHappy
SideZ‐scoreTalairach coordinatesZ‐scoreTalairach coordinatesZ‐scoreTalairach coordinates
x y z x y z x y z
1 InsulaR3.8422103.7440203.6642210
2 Precentral gyrus (4)L3.53–46–14503.67–44–10523.71–48–458
3 Temporal sub‐lobarL3.5–46–22–123.65–46–22–123.57–46–24–12
4 Cerebellum culmenL4.13–30–38–203.96–26–38–24
5 Cingulate gyrus (32)R3.7414648
6 InsulaL3.46–46–212
7 Superior temporal gyrus (42)R4.7170–2614
8 Middle occipital gyrus (19)L3.88–34–8618
10 CerebellumL3.34–48–50–46
11 Corpus callosumL3.62–10–1624
12 Medial frontal gyrus (10)R3.4110508

BA = Brodmann area; L = left; R = right.

In contrast, anger stimuli induced a higher activation in the alexithymics than in the non‐alexithymics in the left temporal sub‐lobar (x = –28, y = –24, z = 22). Sadness and happiness stimuli did not lead to high activation in any area of the brain (data not shown).

Discussion

The present study clearly reveals differences in brain response to facial expressions between people with and without alexithymia. The salient results were as follows. (i) Differences in brain response were mainly observed in the cortex areas, but not in the subcortical regions. (ii) Brain regions where rCBF was lower in the alexithymics than in the non‐alexithymics were localized in the right hemisphere, while those where rCBF was higher in the alexithymics than in the non‐alexithymics were localized in the left hemisphere. (iii) The ACC and insula were less activated in the alexithymics in response to angry faces than neutral faces. (iv) There were differences in response to stimuli depending on the emotion shown, i.e. a negative correlation between rCBF and TAS‐20 scores was observed in the right hemisphere when subjects were viewing the angry and sad faces, but not the happy and neutral faces. (v) There was no clear difference in the effect of emotional intensity (high or low arousal) between people with and without alexithymia. These results are consistent with the suggestion that alexithymia is associated with a deficit in the cognitive comprehension of emotion (Reiman et al., 1997; Taylor, 2000). In addition, no difference between the alexithymics and the non‐alexithymics was observed in the limbic structure (i.e. the amygdala, the hippocampal formation and hypothalamus), which plays a central role in emotional responses to simple perceptual aspects of stimuli. This finding is supported by a recent fMRI study indicating that the limbic area is not associated with alexithymia (Berthoz et al., 2002).

Previous studies have reported inhibition in the right hemisphere and higher activation in the left hemisphere in people with alexithymia. Accordingly, conjugate lateral eye movements used as an index of hemisphere activation (Parker et al., 1992) and right visual field search of chimeric faces (Berenbaum and Prince, 1994; Jessimer and Markham, 1997) revealed left hemispheric dominance or right hemispheric weakness among alexithymic subjects. The right hemisphere contains essential emotional processing systems (Ross, 1981, 1984; Ross et al., 1994; Adolphs et al., 1996), and the right inferior frontal cortex has been associated with comprehension and production of emotion in facial and vocal expressions (Hornak et al., 1996). It has also been reported that lesions in the right inferior parietal cortex correlate with impaired recognition of emotion (Adolphs et al., 1996). Moreover, lesions in the orbitofrontal cortex in the right hemisphere are known to impair identification of facial and vocal expressions (Ross and Mesulam, 1979; Hornak et al., 1996). The orbitofrontal cortex represents an important site for capturing the emotional significance of events, and is used to guide behaviour through its rich interconnections with the limbic and other cortices (Damasio, 1994; Barbas, 2000). Dysfunction in these areas, as shown in the present study, might account for the reduced ability to identify emotion and consequent behavioural changes in people with alexithymia. In contrast to the right hemisphere, the left hemisphere in split‐brain patients has been reported to show sharp improvement in facial emotion discrimination when instructions are changed slightly to emphasize verbal labels for the facial expression (Stone et al., 1996). As the left hemisphere may be superior in processing information analytically through the language system, alexithymic subjects viewing facial expressions might possibly activate functions in their left hemisphere in compensation for the defect in the right hemisphere. As suggested in previous reports, people with alexithymia might rely on the cognitive processing style of the left hemisphere over the right hemisphere (Cole and Bakan, 1985; Hoppe, 1988). This hemisphericity might reflect a tendency to focus on external events and poor awareness of emotion in people with alexithymia.

In the subjects’ debriefing, the rating scores for the emotional faces were similar between the subjects with and without alexithymia. It has been shown that when rough rating methods are used, alexithymics often accurately judge affect‐laden external stimuli (Roedema and Simons, 1999). However, when using more targeted self‐report measures, such as having the subjects describe how the stimuli made them feel, affective self‐description is impoverished in subjects with alexithymia (Wehmer et al., 1995; Roedema and Simons, 1999). The peculiar brain activities in alexithymics might produce almost the same output as those in non‐alexithymics. Consequently, people with alexithymia do not deviate severely from the norm in their social interactions.

As anger was the most unpleasant and intense stimulus in the emotional faces used in this study, the alexithymic subjects showed less activation in the ACC and bilateral insula in response to the anger component of the facial expressions. The ACC has been associated with conscious awareness of emotion (Lane et al., 1997), and increasing intensity of the angry facial expressions has been associated with enhanced activity in the ACC (Blair et al., 1999). It has also been suggested that the ACC plays a crucial role in assessing emotional arousal or the attentive components of emotion, and that the insula, as well as the ACC, are activated by emotional recall/imagery and by emotional tasks with cognitive demand (Phan et al., 2002). This led to the assumption that emotional arousal might be insufficient in people with alexithymia because of the deficit in ACC activity. Our results are partially in line with the ACC deficit model of alexithymia (Lane et al. 1997; Berthoz et al., 2002). However, the ACC deficit may not be the only structure to account for alexithymia.

The difference in response depending on the emotion shown, which was demonstrated in our correlation analysis, might be associated with hemispheric specialization for emotional valence. Indeed, the left hemisphere is considered to be dominant for positive emotion and the right hemisphere for negative emotion (Davidson and Tomarken, 1991; Adolphs et al., 2001). Therefore, in alexithymic subjects, negative emotion processing might be more impaired in the right hemisphere.

In this study, the rating scores after each scan indicated that subjects recognized the emotional significance in facial expressions even though they were not asked to categorize the emotional expression in the task. However, the details of how the subjects processed the stimuli in the implicit facial expression processing task are still unclear. Further studies more targeted to specific emotional processing might be needed to reveal the details of brain activity in people with alexithymia.

In summary, people with alexithymia showed lower response in the right hemisphere and higher response in the left hemisphere when viewing a range of facial expressions, and less activation in the ACC, which is associated with emotional arousal. These findings may partly explain the disorders that affect regulation and the consequent peculiar behaviour associated with alexithymia.

Acknowledgements

We appreciate the technical assistance of M. Miyake, Y. Ishikawa and S. Watanuki in the PET studies. This work was supported by grants‐in‐aid from the Ministry of Education, Science, Sports and Culture, and the Ministry of Health and Welfare, Japan. This research was also supported in part by the Telecommunications Advancement Organization of Japan.

References

View Abstract