Second-order facial information processing in schizophrenia PDF

Preview

Summary

Neuropsychology Copyright 2008 by the American Psychological Association 2008, Vol. 22, No. 3, 313–320 0894-4105/08/$12.00 DOI: 10.1037/0894-4105.22.3.313 Second-Order Facial Information Processing in Schizophrenia Jean-Yves Baudouin Mathilde Vernet Université de Bourgogne Centre de Neurosciences C...

Description

Neuropsychology 2008, Vol. 22, No. 3, 313–320

Copyright 2008 by the American Psychological Association 0894-4105/08/$12.00 DOI: 10.1037/0894-4105.22.3.313

Second-Order Facial Information Processing in Schizophrenia Jean-Yves Baudouin

Mathilde Vernet

Universite´ de Bourgogne

Centre de Neurosciences Cognitives

Nicolas Franck Centre de Neurosciences Cognitives and Centre Hospitalier le Vinatier and Universite´ Claude Bernard This study investigated the processing of second-order relational face information in schizophrenia. Twenty-eight schizophrenic patients and 28 controls were asked to say whether the space between the eyes was the same in 2 side-by-side faces. The 2 faces were derived from the same original face, but the spacing between the eyes was either the same or differed by various distances. The results showed that schizophrenic patients needed a space that was twice as great as controls to see a difference. The authors conclude that schizophrenic patients have a deficit in processing second-order relational face information. Keywords: schizophrenia, configural processing, second-order information

of face stimuli. It might reflect a dysfunction in the networks that synchronize the local and global processing of face stimuli. In this case, the abnormal visual scanpath would reflect overreliance on sequential visual search strategies, perhaps to compensate for an earlier problem in configural processing. Configural processing is hypothesized to be the source of face-processing expertise in healthy individuals (for a review, see Maurer, Le Grand, & Mondloch, 2002). Consistent with a configural explanation for the abnormal visual scanning observed in schizophrenia, schizophrenic patients seem to use the same types of eye movements for upright and upside-down faces, whereas controls do not (Schwartz, Rosse, Johri, & Deutsch, 1999; for different exploration patterns for upright and upside-down faces in infants, see also Gallay, Baudouin, Durand, Lemoine, & Le´cuyer, 2006). Because inverting a face is assumed to disrupt its configural processing (Diamond & Carey, 1996; Yin, 1969), similar exploration of a face in the two orientations suggests that schizophrenic patients explore faces in a local way, even when the face is upright. Direct testing of configural processing in schizophrenia has not given rise to congruent observations, however. Using inversion and composite paradigms, Schwartz, Marvel, Drapalski, Rosse, and Deutsch (2002) reported that schizophrenic patients were impaired on inverted faces to the same extent as controls. Schizophrenic patients were also deficient when they had to process the identity of a half face if a counterpart half from another person was aligned. Again, this composite effect (Young, Hellawell, & Hay, 1987) was similar to that observed for controls. Consequently, Schwartz et al. (2002) concluded that schizophrenic patients rely on configural information to recognize a face. By studying the inversion effect in a facial emotion recognition task, Chambon, Baudouin, and Franck (2006) also found an inversion effect similar to that reported for controls. However, in that study, controls adapted their decision criterion to face orientation: They adopted a more conservative criterion for negative emotions and a more liberal criterion for neutrality on upside-down faces (i.e., they tended to not recognize negative emotions and to respond “neutral”). Schizophrenic patients did not modulate their criteria and, for both orientations, adopted the same pattern of decision criteria as controls for upside-down emotions. Chambon et al. suggested

One aspect of schizophrenia is a strong deficit in face processing. More specifically, many researchers have reported that schizophrenic patients exhibit a deficit in facial– emotion recognition (for reviews, see Edwards, Jackson, & Pattison, 2002; Mandal, Pankey, & Prasad, 1998; Morrison, Bellack, & Mueser, 1988). This deficit also extends to other aspects of face processing, including face recognition (e.g., Archer, Hay, & Young, 1992; Feinberg, Rifkin, Schaffer, & Walker, 1986; Martin, Baudouin, Tiberghien, & Franck, 2005; Salem, Kring, & Kerr, 1996). Although some authors have suggested that the deficit is more pronounced for facial emotion than for other kinds of facial information, particularly during certain phases of the disease (e.g., Gaebel & Wo¨lwer, 1992; Penn et al., 2000), schizophrenic patients appear to be impaired in the perceptual processing of faces; they do not process faces in a way that allows them to extract facial information in an efficient manner. In line with this idea of a deficit in the perceptual processing of faces, many researchers have reported abnormal visual scanning of faces in schizophrenics (e.g., Loughland, Williams, & Gordon, 2002a, 2002b; Manor et al., 1999; Streit, Wolwer, & Gaebel, 1997; Williams, Loughland, Gordon, & Davidson, 1999). In particular, schizophrenic patients tend to make shorter saccades as well as fewer and longer fixations, looking less frequently at features in comparison to healthy controls. Lougland et al. (2002b) suggested that this particular kind of visual scanning results from a breakdown in the neurocognitive strategies that underlie the processing

Jean-Yves Baudouin, SPMS, Universite´ de Bourgogne, Dijon, France; Mathilde Vernet, Centre de Neurosciences Cognitives, Lyon, France; Nicolas Franck, Centre de Neurosciences Cognitives, Lyon, France, and Centre Hospitalier le Vinatier and Universite´ Claude Bernard, Lyon, France. This work was supported by ACI “Jeunes Chercheuses et Jeunes Chercheurs, 2003” Grant 6056 awarded to Jean-Yves Baudouin by the French Ministry of Research. This research program was approved by local ethics committee (Comite´ de Protection des Personnes-CPP Sud-Est IV). Correspondence concerning this article should be addressed to JeanYves Baudouin, Universite´ de Bourgogne, Poˆle AAFE—Esplanade Erasme, BP 26 513, 21 065 Dijon Cedex, France. E-mail: [email protected] 313

314

BAUDOUIN, VERNET, AND FRANCK

that schizophrenic patients do extract configural information, but that it is of low quality because of inadequate face exploration. Thus, it is not clear whether or not schizophrenic patients rely on configural information when performing a face-processing task, or whether the configural information extracted, if any, is of good quality. One way to reconcile the diverging observations would be to consider different types of configural information. Maurer et al. (2002) defined three types. First-order relations refer to the relative positions of the features (e.g., the eyes above the nose). Secondorder relations refer to spacing between features. Finally, holistic information results from “gluing” the different aspects of the face together. These types of configural information are distinct from local (componential or featural) information, which includes the visual characteristics (shape, color, etc.) of individual features. Maurer et al. (2002) underlined that when a face is presented upright, participants tend to process it as a gestalt (i.e., holistically), making it harder to process individual features. Two effects illustrate this tendency: the composite effect (Young et al., 1987; see also Baudouin & Humphreys, 2006b; Calder, Young, Keane, & Dean, 2000), where the alignment of an incongruent counterpart makes it harder to process facial information contained in a half face, and the whole–part superiority effect, which involves better recognition of featural information when the features are displayed in the context of a whole face (Tanaka & Farah, 1993; Tanaka & Sengco, 1997). The observation of the usual composite effect in schizophrenia (Schwartz et al., 2002) suggests that the processing of this kind of configural information is preserved. Moreover, the automaticity and rapidity of its extraction further suggest that it does not involve or need active exploration of the face to be extracted. But what about second-order relations? It is actually quite difficult to know whether schizophrenic patients process this kind of configural information in the proper way. The composite paradigm does not allow one to study this. Finding an inversion effect in schizophrenia is not more informative: Inversion disrupts the processing of second-order relational information (Leder & Bruce, 1998, 2000) but also the processing of holistic information (see Maurer et al., 2002). Inversion can even disrupt the processing of local, featural information (see Baudouin & Humphreys, 2006a; Mondloch, Le Grand, & Maurer, 2002), which makes it difficult to draw any definitive conclusions. In the present study, we further tested the ability of schizophrenic patients to process faces in a configural way. More specifically, our aim was to see whether they are impaired in processing second-order relations in faces. Schizophrenic patients and

controls were asked to say whether the spacing between the eyes was the same in two side-by-side faces. The size of the space between the eyes was varied across trials to determine the minimal space participants were able to detect. Both faces were either upright or upside-down. If schizophrenic patients have a deficit in second-order processing, they should need a larger space than controls to detect a difference.

Method Participants Twenty-eight patients with schizophrenia (7 women, 21 men; mean age: 36.7 years, range 20 – 60) and 28 healthy participants (7 women, 21 men; mean age: 34.8 years, range 18 –57) volunteered to participate in the study. All patients were hospitalized at the Vinatier Psychiatric Hospital in Lyon, France. The patient participants were recruited if their current diagnosis according to Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM– IV; American Psychiatric Association, 1994) criteria was schizophrenia, with no other psychiatric comorbidity on DSM–IV Axis I. For both groups, exclusion criteria included visual difficulty, history of neurological illness or trauma, alcohol or drug dependence according to DSM–IV criteria, and age older than 65 years. All patients were receiving antipsychotic medication (principally levomepromazine, olanzapine, and risperidone) and were clinically stable at testing time. Scale for the Assessment of Positive Symptoms (SAPS; Andreasen, 1984) and Scale for the Assessment of Negative Symptoms (SANS; Andreasen, 1983) were used to obtain ratings for positive and negative symptoms in the schizophrenia sample (mean scores are presented in Table 1). None of the controls reported neurological diseases or psychiatric problems. They were matched with the schizophrenic patients on sex, age, and level of education. All participants reported normal or corrected-to-normal visual acuity. No one was paid for taking part in the study. Written informed consents were obtained from the patients.

Material We used neutral color photographs of two females. The size of the pictures was 500 ⫻ 500 pixels. Using a morphing software (Morpheus v1.85), two versions were created for each female: one where the eyes were close together, and the other where the eyes were far apart. Nineteen intermediate versions were generated for each female by morphing together the first two versions, resulting

Table 1 Demographic Characteristics of Patients and Control Participants Schizophrenic patients

Controls

Characterstic

M (SD)

Range

M (SD)

Range

t(54)

p

Age (years) Education (years) Duration of illness (years) SANS score SAPS score

36.7 (11.3) 11.3 (2.3) 10.6 (10.2) 40.7 (17.0) 37.6 (27.0)

20–59 2–17 1–49 13–72 0–82

34.7 (10.9) 12.5 (2.4) — — —

18–57 8–17 — — —

.69 1.92 — — —

⬎ .49 ⬎ .06

Note. SANS ⫽ Scale for the Assessment of Negative Symptoms; SAPS ⫽ Scale for the Assessment of Positive Symptoms.

315

FACIAL INFORMATION PROCESSING IN SCHIZOPHRENIA

in 21 pictures with different spacings between the eyes. These pictures ranged from 1 (close eyes) to 21 (far-apart eyes) according to the size of the space between them, with a regular step size between neighboring pictures. To make the material such that the spacing between the eyes was equally discriminable for the two female faces, 16 control group participants were shown pairs of faces presented side-byside. The two pictures were of the same female, and the distance between their eyes was either the same or different. The task was to state whether or not the eyes were spaced equally on the two faces. The pair of faces remained on the monitor until the participant had responded. Participants performed two sessions, one for each of the two faces. A session consisted of 184 pairs. Four pairs corresponded to two side-by-side presentations of the intermediate picture in which the space between the eyes was the mean of the space in the close and far versions (Picture 11). In four other trials (“Difference ⫹1” trials), two pictures with the smallest difference in spacing (i.e., two neighboring pictures, Pictures 11 and 12 in two trials, and Pictures 10 and 11 in the other two trials) were presented side-by-side. The picture where the eyes were closer together was presented on the left and the one where the eyes were farther apart was presented on the right. Four other trials consisted of these same pairs, but the picture with the eyes farther apart was presented on the left and the one with the closer eyes was on the right (“Difference ⫺1” trials). In four other trials, participants saw Picture 9 on the left and Picture 11 on the right (“Difference ⫹2”); in four other trials, Picture 11 was on the left and Picture 9 was on the right (“Difference ⫺2”). In a similar way, Pictures 8 and 11 and 9 and 12 were used for trials “Difference ⫺3” and “Difference ⫹3,” with four trials for each, and so on until the differences were ⫺20 and ⫹20. Finally, to increase the number of trials where the difference was 0, each picture other than Picture 11 (already used 4 times) was presented once in a pair with itself. From the pretest results, the percentage of “same” responses for each difference was used to compute the standard deviations for each female face (see Data Analysis for more details). For each female, we selected 2 pictures that where at the same distance in number of times the standard deviation. This was done to have 2 pictures whose discriminability (in terms of eye spacing) was similar for the two female faces. These 2 pictures were used to build the material for the experiment. Using the morphing software, we made 15 intermediate pictures between the 2 pictures selected for each female. For each female, this gave us 17 pictures with increasingly spaced eyes. The pictures are illustrated in Figure 1.

Procedure Participants were presented with pairs of side-by-side pictures of the same female. Their task was to say whether the eyes were at the same distance in the two pictures by pressing one of two keys on the keyboard. The pairs remained on the monitor until the participant responded. Each participant performed two sessions, one session with upright faces and the other with upside-down faces. The female face used was different in the two sessions. The session order and the assignment of a session to a given female were alternated across participants. Each session comprised 40 trials. In 2 trials, participants were presented twice with the picture

whose eyes were at the mean distance (Picture 9, Difference 0). In 4 other trials, they were presented with Pictures 8 and 10, with the closer eyes on the left (Difference ⫹2) in 2 trials and on the right in the other 2 trials (Difference ⫺2). Four other trials consisted of Pictures 7 and 11 with the same manipulations (Differences ⫺4 and ⫹4), and so forth until Pictures 1 and 17 (Differences ⫺16 and ⫹16). In 6 other trials (Difference 0), participants saw pairs of the same picture, respectively, Pictures 3, 5, 7, 11, 13, and 15. These trials were aimed at increasing the number of trials where the two pictures had the same eye spacing. Thus, the percentage of trials with the same eye spacing was 20% (8/40). The percentage of “same” response varied according to the participant’s threshold. Prior to each session, participants performed a practice session of 11 trials, 1 for each possible difference and 3 with the same spacing but different pictures.

Data Analysis Because our main purpose was to quantify the ability to process second-order facial information, we used a psychophysical procedure to compute the discriminability threshold for each participant. From the percentage of “same” responses for each distance, we drew a discriminability curve ranging from Difference ⫺16 to Difference ⫹16, centered on Difference 0. This curve should be normally distributed, with the highest percentage at 0 and a symmetrical decrease for values between 0 and ⫺16 or ⫹16. For each participant, we used these properties to compute the mean and standard deviation of the Gaussian curve for “same” responses. The formulas are as follows: M⫽ SD ⫽ 兵

冘

冘

共Diff k ⫻ SDiff.k 兲/

冘

关S Diff.k ⫻ 共Diffk ⫺ M兲2 兴/关共

SDiff.k and

冘

SDiff.k 兲 ⫺ 1兴其1/2 ,

where M is the mean, SD the standard deviation, Diffk the difference values (from ⫺16 to ⫹16), and SDiff.k the number of “same” responses for Difference k. Note that for Difference 0, only the two responses for the pair containing the central picture (Picture 9) were considered so as to equalize the possible number of “same” responses in the no-difference and difference trials. When the participants did not respond “same” for more than half of the eight trials with a difference of 0, and if the percentage of “same” responses for Difference 0 was not significantly different from the mean of the other differences (according to the t test), the participant was considered to be unable to see variations in spacing. All such participants were assigned the standard deviation value that would have been obtained by a participant who had responded “same” all the time (i.e., 9.95) and a mean of 0. This was the case for four schizophrenic patients in the upright condition and six schizophrenic patients in upside-down conditions (two of whom were attributed the maximal threshold for the upright condition also). When a participant responded “different” for all trials of two adjacent distances (e.g., ⫺4, ⫺2, ⫹2, and ⫹4), that person’s “same” responses for higher distances were considered keying errors (i.e., they did not reflect her or his discriminability threshold), providing there were no more than two such responses, in which case they were considered as “different” responses. The

316

BAUDOUIN, VERNET, AND FRANCK

Difference = 12 pictures (left: picture 3, right: picture 15)

Difference = 8 pictures (left: picture 5, right: picture 13)

Difference = 4 pictures (left: picture 7, right: picture 11)

Difference = 0 picture (left and right: picture 9)

Figure 1. Illustration of the continua and pairs used in the experiment: Spacing between eyes increases regularly from Picture 1 to Picture 17.

percentage of trials for which the response was changed was 0.2 for controls with upright faces, 0.0 for controls with upside-down faces, 0.3 for schizophrenic patients with upright faces, and 0.1 for schizophrenic patients with upside-down faces. The shape of the curves depended on two factors. First, the dispersion was a function of the discriminability threshold of the participant: The shorter the distance the participant was able to detect, the narrower the space under the curve. Second, the height of the curve depended on the participant’s bias toward “same” or “different” responses. A participant who had a tendency to respond “different,” even when the distance was the same, should have a percentage below 10...