OUP user menu

The human amygdala plays an important role in gaze monitoring
A PET study

Ryuta Kawashima, Motoaki Sugiura, Takashi Kato, Akinori Nakamura, Kentaro Hatano, Kengo Ito, Hiroshi Fukuda, Shozo Kojima, Katsuki Nakamura
DOI: http://dx.doi.org/10.1093/brain/122.4.779 779-783 First published online: 1 April 1999

Summary

Social contact often initially depends on ascertaining the direction of the other person's gaze. We determined the brain areas involved in gaze monitoring by a functional neuroimaging study. Discrimination between the direction of gaze significantly activated a region in the left amygdala during eye-contact and no eye-contact tasks to the same extent. However, a region in the right amygdala was specifically activated only during the eye-contact task. Results confirm that the left amygdala plays a general role in the interpretation of eye gaze direction, and that the activity of the right amygdala of the subject increases when another individual's gaze is directed towards him. This suggests that the human amygdala plays a role in reading social signals from the face.

  • PET
  • regional cerebral blood flow
  • amygdala
  • gaze direction discrimination
  • social contact
  • ANCOVA = analysis of covariance
  • rCBF = regional cerebral blood flow
  • SPM = statistical parametric map

Introduction

Humans, but not other primates, have the ability of making inferences about another individual's state of mind (Premack and Woodruff, 1978). Determination of the direction of the other person's gaze is essential for such inferences, as is the observation of another person's eyes for the judgment of his mental state (Baron-Cohen, 1995; Kobayashi and Kohshima, 1997). Neural mechanisms underlying these processes have been investigated by neurophysiological and lesion studies in monkeys which suggest that the superior temporal sulcus area (Rolls, 1984) and amygdala (Leonard et al., 1985) are involved in the determination of the direction of gaze. Autistic patients often exhibit impaired judgement of gaze direction and mental state (Courchesne, 1997) and neuropathological studies have indicated abnormalities of the amygdala, leading to the hypothesis that damage of the amygdala is related to the development of the autistic syndrome (Hoon and Reiss, 1992; Courchesne, 1997). However, data regarding the role of the amygdala in humans have largely been confined to emotion-processing (Gloor, 1990; Adolphs et al., 1994; Cahill et al., 1996; Zald and Pardo, 1997; LaBar et al., 1998; Morris et al., 1998a, b; Whalen et al., 1998), and the neural mechanisms underlying social judgment of other individuals, such as the determination of gaze direction and reading of gaze, are largely unknown.

Methods

Subjects

Eight right-handed male volunteers (aged 20–53 years) participated in the present study. Written informed consent was obtained from each subject on forms approved by the ethical committee of National Institute for Longevity Sciences and the Declaration of Helsinki (1975). All the subjects were healthy, with no past history of psychiatric or neurological illness, and none was on any medication. High resolution MRI of the brain was performed for each subject.

Task procedures

The experimental task in our study contained three different conditions (Fig. 1). Digitized cinematic images of a woman's face against a uniform background, tilted 30° rightwards from the front, were presented on the same location through a head-mounted display (Mediamask; Olympus, Tokyo, Japan) for all the conditions. In the eye-contact condition, the woman's gaze was directed towards the eyes of a subject who sat just behind the video camera, and then moved randomly either upwards or downwards for a random duration of time (mean 2000 ms; range 1500–2500 ms). After she had looked upwards or downwards, her gaze returned to the eye-contact position. Subjects were instructed as follows: `Imagine she is looking at you. If you feel her gaze is directed at your head or trunk, press a key with the index finger or middle finger of your right hand, respectively.' The subjects, at an interview after the PET scans, said that they felt that the woman had looked directly at them, and that her attention had been directed towards them during the eye-contact condition. In the no eye-contact condition, the woman's gaze was directed towards a subject who sat next to the video camera, and then she moved her eyes in the same manner as for the eye-contact condition. In this case, the subjects were instructed as follows: `Imagine she is looking at someone who is on your left. If you feel her gaze is directed at that person's head or trunk, press a key with the index finger or middle finger of your right hand, respectively.' The difference between the two eye-direction discrimination conditions was in the direction of gaze of the woman in the image, whether it was directed towards the subjects themselves (the eye-contact condition) or not (the no eye-contact condition), while all other details were exactly the same. In the control condition, her right or left eye winked for a random duration (mean 2000 ms; range 1500–2500 ms) in a random order. Subjects were instructed to press a key with either the index finger or middle finger of their right hand when her left or right eye winked, respectively (Fig. 1). Under all conditions, the subjects were instructed to attend to the eyes of the presented image, and perform discrimination tasks on the basis of the information obtained from the speed and direction of eye motion and the shape of the eyelids.

A digital video recording was made for subsequent analysis of each subject's performance, and the reaction time was determined by frame-by-frame analysis. An electro-oculogram was also recorded for each subject during the PET measurements.

PET

The regional cerebral blood flow (rCBF) was measured using a Siemens ECAT EXACT HR PET scanner in 3D mode, after a bolus injection of H215O (15 mCi per scan). Attenuation-corrected data were reconstructed into 47 image planes with a resulting resolution of 6 mm at full-width half-maximum. All tasks were started immediately after the bolus injection. Each PET measurement was commenced immediately after radioactive counts were identified on the PET camera, and was continued for a period of 90 s. In the present study, one PET scan was obtained for each task. The order of the three tasks was counterbalanced across the subjects.

Image data analysis

A human brain atlas system (Roland et al., 1994) was used for image realignment and transformation into standard stereotaxic space. Then statistical parametric mapping (SPM96, Wellcome Department of Cognitive Neurology, London) software was used for smoothing and statistical analysis (Friston et al., 1995). In this study, an isotropic Gaussian filter with a full-width half-maximum of 12 mm was used to increase the signal-to-noise ratio and to compensate for individual differences in gyral anatomy. Differences in global flow were analysed using ANCOVA (analysis of covariance). Comparisons across conditions were made by means of t-statistics, and thereafter transformed into normally distributed Z-statistics. In the case of the amygdala, for which we had a region-specific hypothesis, correction for multiple comparisons was based on the size of the amygdala as done by others (Breiter et al., 1996; Whalen et al., 1998) and the smoothness of the underlying SPM (Worsley et al., 1996). For the rest of the brain, where we had no a priori regional hypothesis, correction was made for the entire volume analysed. Thus, in all cases the threshold for significant activation was set at P < 0.05 (corrected). Finally, each area of activation was superimposed on to the average transformed MRI of the same eight subjects involved in this study. Anatomical localization of areas of activation in each comparison was made in relation to the mean anatomically standardized MRI, since the standard brain of the human brain atlas has a size and shape different from that of SPM96 and is also situated in a slightly different stereotaxic coordinate system from that of Talairach and Tournoux (1988).

Results

Behaviour

Electro-oculogram recordings showed no evidence of saccadic eye movements during any task. The mean (standard deviation) response times were 438 (31), 408 (31) and 398 (31) ms during the eye-contact, no eye-contact and control conditions, respectively.

Brain activation

Eye-direction discrimination tasks during the eye-contact condition activated the amygdala bilaterally (P < 0.05, corrected for multiple comparisons) when compared with the control condition (Fig. 2). The corresponding area in the left amygdala was activated to the same extent (P < 0.05, corrected for multiple comparisons) during the no eye-contact condition (Fig. 2) as during the eye-contact condition (Fig. 3) when compared with the control condition. The right amygdala, however, did not show any increase in neural activity during the no eye-contact condition (Fig. 3). This field shows increases in rCBF during the eye-contact condition when compared with the no eye-contact condition (P < 0.003, uncorrected). In addition, areas in the left insula (Fig. 4A) and posterior cingulate cortex (Fig. 4B) showed increases in rCBF during the eye-contact condition when compared with the control condition. Activation in these fields did not reach statistical significance (P < 0.05, corrected) although it did achieve an uncorrected significance level of P < 0.001. The stereotactic coordinates of rCBF changes according to the atlas of Talairach and Tournoux (1988) are given in Table 1. These two fields also showed increases in the rCBF when compared with the no eye-contact condition (P < 0.001, uncorrected). Deactivation was not observed for either the eye-contact or the no eye-contact condition when compared with the control condition.

Discussion

The experimental protocol was designed such that significant activation as evaluated by both eye-contact condition versus control and no eye-contact condition versus control comparisons reflected the neural responses involved in determining the eye gaze direction, and that significant increases in rCBF during the eye-contact condition compared with either the no eye-contact or control conditions identified areas activated specifically when another individual's gaze is directed towards the subject (Keating and Keating, 1982; Baron-Cohen, 1995). Our results indicate that the human amygdala plays a role in reading social signals from the face (Leonard et al., 1985; Heywood and Cowey, 1992; Nakamura et al., 1992; Seeck et al., 1993; Young et al., 1995).

Recent neuroimaging studies in humans have demonstrated activation within the amygdala in response to overt (Adolphs et al., 1994; Morris et al., 1996, 1998a) or masked (Morris et al., 1998b; Whalen et al., 1998) emotionally expressive faces, or to arousal, threatening or fear-provoking stimuli (Gloor, 1990; Cahill et al., 1996; Zald and Pardo, 1997; LaBar et al., 1998). These data regarding the role of the amygdala in humans have been confined to emotion-processing. In these studies, the gazes of both emotionally expressive and control neutral faces used as stimuli were directed towards the subjects. Although subjects probably responded to the emotional expression in the eyes of the objectives, the amygdala activation in these studies could not be related to the direction of gaze. To our knowledge, this is the first brain imaging study investigating the involvement of the human amygdala in social judgement of other individuals on the basis of their eye gaze direction. Our findings are consistent with the results of neurophysiological studies in monkeys which demonstrated that the medial and lateral nuclei of the amygdala have cells sensitive to eye direction (Brothers et al., 1990), as well as human lesion studies investigating impairment of gaze direction interpretation after amygdalotomy (Young et al., 1995; Adolphs et al., 1998). They support the hypothesis proposing a role for the amygdala in social behavior (Dicks et al., 1968; Brothers et al., 1990; Adolphs et al., 1998).

Although, the area most closely related to determination of the gaze direction was found in the superior temporal sulcus in neurophysiological (Perrett et al., 1985) and lesion (Campbell et al., 1990) studies in monkeys, we did not find any significant activation in this region. This could well be due to activation of the superior temporal sulcus region during the control condition, in which subjects were required to attend to the presented images of the face, since the neurons in this region have been shown to respond primarily to faces (Rolls et al., 1992).

In the present study, eye-contact condition-specific activation was observed not only in the right amygdala but also in other areas in the limbic system, namely, the left insula and cingulate cortex. Human studies have suggested that limbic structures may respond more to stimuli that directly induce strong emotional responses (Adolphs et al., 1994; Morris et al., 1996; Phillips et al., 1997). Since we used cinematic images of faces as visual stimuli, three objectives in all the three conditions had some emotional expression. However, in human social life, it is reasonable to assume that most people would be hypersensitive to another person watching them (Baron-Cohen, 1995). We therefore suggest that the activation of these structures in the limbic system is related to the strong emotional responses evoked during the eye-contact task.

View this table:
Table 1

Locations of increased blood flow during eye direction discrimination tasks

AreaTalairach coordinatesZ-scoreVolume (mm3)
xyz
Stereotaxic coordinates (mm) identify the location of the maxima of rCBF change corresponding to the atlas of Talairach and Tournoux (1988). Activation in the amygdala is significant at P < 0.05 (correction for multiple comparisons based on the size of the amygdala and the smoothness of the SPM). The activations in the remaining brain are significant at P < 0.001 (uncorrected).
Eye-contact
Left amygdala–2014–223.82504
Right amygdala26 2–263.48256
Left insula–3610–123.77240
Posterior cingulate 0–24503.80 80
No eye-contact
Left amygdala–2014–223.64 80
Fig. 1

Visual stimuli used for the eye-contact (upper column), no eye-contact (middle column) and control (lower column) conditions. Subjects were asked to judge whether the eyes were directed upwards or downwards during the eye-contact and no eye-contact conditions, and whether the left or right eye blinked during the control condition.

Fig. 2

Horizontal (left) and coronal (right) sections of the mean spatially-normalized MRI (grey scale) through peak activation of the amgydala as evaluated in the eye-contact versus control (upper) and the no eye-contact versus control (lower) comparisons. Red areas on the MRI show the field of activation. The left sides of the figures correspond to the left hemisphere.

Fig. 3

Mean normalized rCBF values in ml/100 g/min during the eye-direction discrimination and control conditions are shown from voxels showing peak activation in the left and right amygdala. Error bars indicate standard deviation.

Fig. 4

Horizontal (left) and coronal (right) sections of the mean spatially-normalized MRI (grey scale) through peak activation of the left insula (A) and posterior cingulate cortex (B) as evaluated in the eye-contact versus control comparison. Red areas on the MRI show the field of activation.

Acknowledgments

We wish to thank Professor P. E. Roland for his advice on the manuscript and Ms A. Watanabe for technical assistance. This research was supported by Grants-in-Aid for Scientific Research on Priority Research from the Japanese Ministry of Education, Science, Sports and Culture (10164206, 09207102) and Research for the Future from Japan Society for the Promotion of Science (JSPS-RFTF 97L00202).

References

View Abstract