Elsevier

Schizophrenia Research

Volume 40, Issue 3, 21 December 1999, Pages 189-199
Schizophrenia Research

Visual scanpaths in schizophrenia: is there a deficit in face recognition?

https://doi.org/10.1016/S0920-9964(99)00056-0Get rights and content

Abstract

There is substantial evidence that schizophrenics have deficits in face processing. We hypothesised that this difficulty is due to abnormalities in the visual scanning of faces. The specificity of these abnormalities to recognisable faces, the effect of task difficulty and their relationships to three primary symptom dimensions were examined. An infrared corneal reflection technique was used to record the visual scanpaths of 63 schizophrenics and 60 non-psychiatric controls while viewing non-degraded (‘recognisable’) and degraded (‘not recognisable’) faces. In the concurrent recognition task, subjects were asked to select the previously viewed face from among seven (exposure 1) or three (exposure 2) options. Both groups were unable to accurately recognise degraded faces, but schizophrenics were less accurate than controls for non-degraded faces in the more difficult task condition. Schizophrenics maintained a relatively ‘restricted’ scanpath style across both faces, but scanpath disturbances were most apparent for non-degraded faces. Analysis of fixation distribution to non-degraded faces showed that, unlike controls, schizophrenics did not concentrate their fixations on salient features. Scanpath aberrations showed only minimal associations with symptom dimensions. These results suggest that schizophrenic individuals have a specific deficit in the visual scanning of faces, that is a distinct case of a fundamental problem in complex object processing.

Introduction

Given the importance of face perception to social interaction, the study of disturbed face processing in schizophrenia may provide clues to understanding the schizophrenics' often reported difficulties with social communication (Cramer et al., 1992, Feinberg et al., 1986).

A number of studies have indicated that schizophrenics have a deficit in various aspects of face processing, including recognition, familiarity and judging attributes such as age (Archer et al., 1992, Ellis and Young, 1996, Feinberg et al., 1986, Gessler et al., 1989, Neufeld, 1976, Walker et al., 1984). Frith et al. (1983) suggested that schizophrenics may have a specific impairment in face perception (rather than a generalised impairment in complex stimulus processing), due to an inability to integrate features that involve affective (or socially meaningful) appraisal.

This study addresses the proposal that schizophrenics have a deficit in visuo-spatial processing strategies for effective face perception. The most direct, real-time method for assessing these processing strategies is to record visual scanpaths (Noton and Stark, 1971). A scanpath is the pattern of eye movements that occur when an individual processes a complex stimulus. Information about eye fixations can also be extracted to provide information about more detailed processing. In face recognition tasks, the visual scanpaths of normal subjects follow a regular sequence, with fixations focusing on the primary facial features (Walker-Smith et al., 1977). The extent of fixation to features will depend upon how much detail is required to perform the task. Walker-Smith et al.'s visual scanning model is compatible with cognitive models that emphasise an interaction between overall structure and local features in face perception (see Farah et al., 1998 for a review).

Our group has found that schizophrenics have a comparatively reduced fixation duration to features of the face, and fewer fixations overall (Gordon et al., 1992). Phillips and David, 1997, Phillips and David, 1998 also found a reduced focus on relevant features in deluded schizophrenics when viewing single faces and, in their later study, a trend towards fewer fixations.

This study extended previous schizophrenia research on visual scanpaths to faces in three ways. We investigated scanpath disturbances in relation to their specificity to recognisable faces, their relationship to three primary symptom dimensions, and the effect of difficulty on a concurrent face recognition task.

There is a confluence of perceptual, lesion and neuroimaging evidence that face processing is subserved, at least in part, by specific network activation (Ellis and Young, 1989, Ellis and Young, 1996, Geschwind, 1979, Puce et al., 1995). By including both non-degraded and degraded (control) face stimuli (see Methods for details) we investigated if schizophrenics would show distinct scanpath aberrations to faces.

Our method for degrading face images relied on spatial frequency analysis (Bruce, 1988, Harmon, 1973). Structural information is conveyed primarily by low spatial frequencies, whereas the finer-grained component information about features is conveyed by high spatial frequencies (Bruce, 1988). Faces were degraded by simultaneously filtering out high frequencies and masking lower frequencies (see Fig. 2b).

Factor analytic studies have shown that positive and negative symptoms of schizophrenia consistently form three independent dimensions. These dimensions can be differentiated by both neuropsychological and brain imaging indices, suggesting that they may be manifestations of distinct pathophysiological processes (Buchanan and Carpenter, 1997, Liddle, 1987a, Liddle, 1987b, Liddle, 1996). We therefore used a factor analytic approach to explore whether distinct symptom dimensions would be associated with different scanpath aberrations to neutral non-degraded and degraded face stimuli. Previous scanpath studies (using facial affect, simple geometric and complex picture stimuli) have taken individual positive and negative symptoms into account (Gaebel et al., 1987, Kojima et al., 1990, Streit et al., 1997). These studies suggest that positive symptoms are associated with an ‘extensive’ scanpath strategy, and negative symptoms with ‘restricted’ scanpaths and longer fixation duration. Both of these abnormal scanning strategies could result in the reduced attention to feature areas demonstrated for the specific positive and negative symptoms of delusions and affective flattening (Phillips and David, 1997, Phillips and David, 1998, Streit et al., 1997). There is generally a paucity of data concerning which symptom profiles are producing the different scanpath patterns.

In previous scanpath studies of neutral face stimuli, schizophrenics were no less accurate than controls on the accompanying face perception tasks (Gordon et al., 1992, Phillips and David, 1997). However, as noted by Phillips and David, the restricted-choice tasks may have been too simple to produce group differences. Even with aberrant scanpaths, schizophrenics may still take up enough information to make simple decisions about faces. In this study, we assessed face recognition under low (seven options) and high (three options) constraint conditions.

We drew on Frith et al.'s (1983) proposal to predict that, although both groups would show differential scanpath strategies for non-degraded compared with degraded faces, schizophrenics would be relatively more impaired for non-degraded faces. Schizophrenics should also be less accurate than controls in the low constraint condition. We further hypothesised that the use of a factor analytic approach would elucidate the associations between different complexes of positive and negative symptoms and aberrant scanpath strategies.

Section snippets

Subjects

Sixty-three subjects with schizophrenia were recruited from hospitals and community centres in Sydney, and 60 non-psychiatric control subjects were drawn from the general population. Inclusion criteria for both groups were age of 18–60 years and normal vision (assessed by Snellen chart), and exclusion criteria were recent history of substance abuse, epilepsy or other neurological disorders, and mental retardation or head injury [assessed using Section M from the Composite International

Accuracy data

Mean accuracy data for control and schizophrenia groups are presented in Fig. 1. The schizophrenia group was significantly less accurate than the control group [t(121)=4.4, p<0.0001] in the seven-option condition for non-degraded faces, but improved significantly for three options [t(62)=3.9, p<0.0001]. For degraded faces, the control group's increase in accuracy under the three-option condition was also significant [t(59)=2.4, p<0.05].

Scanpath data

Preliminary MANOVAs revealed no significant interactions

Discussion

In this study, both schizophrenia and control groups displayed different scanpath strategies to degraded and non-degraded faces. However, schizophrenic individuals were relatively impaired for both stimuli. Scanpath disturbances in schizophrenia were most apparent for the non-degraded face, suggesting that there is a distinct impairment in strategies used for face recognition. Neither age nor sex covaried significantly with group differences in scanpaths. The schizophrenia group was also less

Acknowledgements

The technical input of Chris Lisle in stimulus development is gratefully acknowledged.

The study was conducted with funding support from the Australian Research Council and the Schizophrenia Fellowship of NSW.

References (35)

  • P. Cramer et al.

    Schizophrenics and social judgement. Why do schizophrenics get it wrong?

    Br. J. Psychiatry

    (1992)
  • H.D. Ellis et al.

    Are faces special?

  • H.D. Ellis et al.

    Problems of person perception in schizophrenia

  • M.J. Farah et al.

    What is ‘special’ about face perception?

    Psychol. Rev.

    (1998)
  • T.E. Feinberg et al.

    Facial discrimination and emotional recognition in schizophrenia and affective disorders

    Arch. Gen. Psychiatry

    (1986)
  • C.D. Frith et al.

    Integration of schematic faces and other complex objects in schizophrenia

    J. Nerv. Ment. Dis.

    (1983)
  • N. Geschwind

    Specializations of the human brain

    Scient. Am.

    (1979)
  • Cited by (134)

    View all citing articles on Scopus
    View full text