تشخیص حالت چهره احساسات در اسکیزوفرنی

Abstract The aim of the present study was to examine spatial processing of facial emotion in schizophrenic patients suffering from affective symptoms. A face-in-the-crowd task using schematic stimuli was administered to schizophrenic patients with flat affect (n=30), schizophrenic patients suffering from anhedonia (n=30), schizophrenic patients not suffering from anhedonia or flat affect (n=28), and a group of healthy controls (n=30). Participants searched displays of neutral schematic faces for a face with a positive or negative mouth expression. Schizophrenic patients manifested a general slowing of response speed compared to normal subjects. All patient groups as well as normal subjects found negative faces more quickly than positive faces amongst neutral faces. Unexpectedly, with increasing anhedonia as assessed by psychiatric rating, a more efficient spatial detection of positive facial expression was observed. For flat-affect patients only, efficiency of search for negative facial expression did not differ from that in the neutral face control condition. This response pattern indicates that, in flat-affect schizophrenic patients, spatial search of negative facial expression might be slowed after the initial engagement of search processes. Potential explanations of the face processing effects found in anhedonia are discussed.

1. Introduction A number of studies have examined the perception of facial emotions in schizophrenia, and most have yielded evidence that schizophrenic patients are less accurate than normal subjects in their ability to identify and discriminate facial emotion Feinberg et al., 1986 and Walker et al., 1980. Controversy exists as to whether these impairments represent specific deficits or are part of a global cognitive impairment Penn et al., 2000 and Salem et al., 1996. Research on emotion decoding impairments of schizophrenic patients has frequently been based on static photographs of facial expressions and has assessed primarily controlled or attentive processes. Spatial processing, as well as automatic processing, of facial emotion have been neglected until now, although automatic processes appear to be central to the elicitation of spontaneous affective reactions in everyday life (Scherer, 2001). According to a two-stage model of perception, object recognition starts with the visual field scanned in parallel for biologically relevant features (preattentive vision). The subsequent integration of features is thought to be a resource-limited serial process, requiring focused attention (attentive vision) (Treisman, 1986). From adaptive and evolutionary aspects, threat-indicating facial expressions (i.e., expressions of fear or anger), in particular, should be instantly recognizable without processing resources having to be drawn upon Krebs et al., 1993 and Ohman, 1996. Using schematic stimuli, it has been shown that negative faces are detected faster than positive or happy faces in a crowd of neutral faces Fox et al., 2000, Ohman et al., 2001 and White, 1995. Search functions for detection of a negative face are flatter than for detection of a positive face. Detection latencies for negative faces, but not for positive, faces were found to be largely independent of display size (number of faces); that is, negative faces appeared to “pop out” from a crowd, suggesting a nonserial search for negative faces by normal subjects (White, 1995). This is in line with reports that automatic processing (“parallel search”) occurs only for negatively valenced stimuli Hansen and Hansen, 1988 and Niedenthal, 1990. Search slopes of less than about 10 ms per item are generally considered to reflect automatic or preattentive visual searching (Nothdurft, 1993). In contrast, happy faces appear to be found during a serial or controlled search. Serial searching or scanning processes imply sequential shifts of the attentional “spotlight” (engage, move, and disengage functions; see Posner and Peterson, 1990). Emotion processing deficits in schizophrenia have rarely been examined as a function of affective symptoms. Among the most prominent affective symptoms in schizophrenia are flat affect (a diminished expression of emotion) and anhedonia (a lowered ability to experience pleasure) (Andreasen, 1987), both of which, according to factor-analytic studies, have to be seen as constituents of relatively independent symptom dimensions Mueser et al., 1994 and Sayers et al., 1996. The affectively flat exterior of schizophrenic patients may mask an emotionally volatile interior: Schizophrenic patients with diminished facial expressiveness were repeatedly found not to differ from normal controls with respect to subjective emotional experience elicited by affect-evoking stimuli (Kring et al., 1993). However, a lack of emotional expressivity might also reflect a slowed or impaired processing of emotion information (Neale et al., 1998). According to Meehl (1962), anhedonia is the expression of a genetic defect in the limbic brain system involved with reward. Social isolation is thought to be an observable consequence of this pleasure deficit. In Meehl, 1962 and Meehl, 1990 theory, anhedonia represents a contributor to—or, in some cases, the result of—an “aversive drift” in schizophrenia (i.e., the tendency for activities and people to take on a threatening, negative affective meaning). The aversive drift was interpreted as the consequence of an enduring imbalance between appetitive and aversive brain centers. The purpose of the present study was to examine for the first time preattentive and attentive spatial processing of facial emotion in schizophrenic patients suffering from affective negative symptoms. To this end, a face-in-the-crowd task using schematic stimuli was administered to three groups of schizophrenic patients (i.e., patients with a flattened affect expression, patients suffering from anhedonia, and patients not suffering from anhedonia or flat affect), as well as to a group of healthy controls. Since reaction time differences between (chronic) schizophrenic patients and normal subjects were expected a priori, our analysis focused on the qualitative pattern of detection latencies and especially on search slopes and the difference between detection latencies for positive faces and those for negative faces. It was hypothesized that anhedonic patients should exhibit an efficient detection of negative faces but should be impaired in the detection of positive faces. Ratings of anhedonia were therefore expected to be positively correlated with the latency difference score “positive face detection−negative face detection

. Results The error rate in the face-in-the-crowd task was 8.6%. Only 0.2% of latencies were extreme outliers and thus excluded from further analysis. Since latency variances differed significantly between groups for most of the display conditions (Levene tests; p≤0.05), a 3×3 ANOVA was carried out on response latencies with two intrasubject variables [display size (number of faces): 2, 4, or 6; valence condition: negative among neutral vs. positive among neutral vs. all neutral] for each group separately. For anhedonic patients, patients without affective negative symptoms, and normal subjects, significant main effects of display size and valence condition and significant interactions between display size and valence condition were obtained (ps≤0.01). For flat-affect patients, significant main effects of display size [F=7.7; df=2,25; p≤0.01] and valence condition [F=15.7; df=2,25; p≤0.001] were found but the interaction display size×valence condition was not significant. According to Tukey post hoc tests, response latencies in all study groups were significantly faster in the negative face condition than in the valence conditions “positive among neutral” and “all neutral” (ps≤0.05). Response latencies in the positive face condition were significantly faster than in the “all neutral” condition for anhedonic patients and normal subjects, but not for flat-affect patients and patients without affective negative symptoms (Table 2). Table 2. Mean response times (in ms) for correct answers in the face-in-the-crowd task as a function of valence condition and display size Flat-affect schizophrenic patients Anhedonic schizophrenic patients Schizophrenic control patients Healthy controls patients Mean S.D. Mean S.D. Mean S.D. Mean S.D. Valence condition All neutral Two faces 1141.6 (385.7) 1106.5 (296.8) 1047.8 (247.0) 873.0 (195.5) Four faces 1228.4 (370.6) 1276.2 (374.4) 1174.7 (310.7) 986.6 (200.6) Six faces 1304.6 (396.4) 1367.0 (376.4) 1326.0 (353.3) 1100.0 (246.8) Positive among neutral Two faces 1095.7 (302.3) 1070.3 (337.8) 1103.2 (288.9) 838.9 (134.8) Four faces 1174.3 (328.9) 1139.9 (301.4) 1147.5 (264.5) 916.6 (198.6) Six faces 1267.9 (411.7) 1194.6 (281.1) 1203.6 (319.8) 948.2 (167.7) Negative among neutral Two faces 1005.9 (269.5) 1035.0 (285.7) 935.9 (259.2) 772.8 (127.7) Four faces 1043.6 (366.0) 1028.5 (295.4) 1005.5 (355.7) 802.5 (155.5) Six faces 1144.9 (437.2) 1118.2 (344.2) 1079.4 (381.3) 823.7 (137.7) Table options For purposes of intergroup response speed comparison, Kruskal–Wallis tests were performed on sum scores of latencies for valence conditions. For all valence conditions (“all neutral,” ”positive among neutral,” and ”negative among neutral”), significant intergroup differences were found [χ2>13.0; df=3; ps≤0.005]. Results of Mann–Whitney U tests indicated that normal subjects responded faster than patients in all of the valence conditions. However, there were no latency differences between patient groups. The educational level was not correlated with sum scores of latencies for any of the valence conditions (p≤0.05). Search slopes were calculated by dividing the mean increase in overall response time by the number of additional items. A 4×3 ANOVA was carried out on slope scores with “group” as intersubject variable and “valence condition” as intrasubject variable. There was a highly significant main effect of valence condition [F=12.9; df=2,102; p≤0.001] but no group effect. The interaction of groups and valence condition was near-significant [F=1.8; df=6,206; p≤0.10]. To explore the latter interaction, Tukey HSD post hoc tests were applied. For all study groups, there were no slope differences between the positive and negative face condition. For flat-affect schizophrenic patients, slopes in the neutral condition were no different from those obtained in the positive or negative face condition (ps>0.90). In contrast, all other study groups manifested significant slope differences between these conditions (ps≤0.05) (Table 3). Table 3. Slope (ms/item) of study participants in the face-in-the-crowd task (baseline: two-face condition) Flat-affect schizophrenic patients Anhedonic schizophrenic patients Schizophrenic control patients Healthy controls Mean S.D. Mean S.D. Mean S.D. Mean S.D. Valence condition All neutral 40.7 (77.5) 65.1 (46.1) 69.5 (53.6) 56.8 (59.0) Positive among neutral 43.0 (61.1) 31.1 (45.0) 25.1 (42.1) 27.3 (39.7) Negative among neutral 34.7 (70.6) 20.8 (59.1) 35.9 (54.7) 12.7 (23.8) Table options A 4 (group)×3 (display size)×3 (valence condition) ANOVA based on error rates yielded main effects for display size [F=89.0; df=2,102; p≤0.001] and valence condition [F=105.3; df=2,102; p≤0.001] and an interaction for display size×valence condition [F=21.7; df=4,100; p≤0.001]. No other significant effects were found. Since atypical neuroleptics have been reported to have a special impact on affective symptoms (e.g., Falkai and Vogeley, 2000), we examined the effect of type of neuroleptic medication (typical vs. atypical) on face detection. 2×3×3 ANOVAs on response latencies (and on error rates) were conducted with the intersubject variable “typical vs. atypical neuroleptic treatment” and two intrasubject variables (display size: 2, 4, or 6; valence condition: negative among neutral vs. positive among neutral vs. all neutral). Both ANOVAs revealed no significant main or interaction effects for the intersubject variable “type of neuroleptic medication” (p≤0.05). In addition, a 2×3 ANOVA was carried out on slope scores with “type of neuroleptic medication” as intersubject variable and “valence condition” as intrasubject variable. No significant main or interaction effect for type of medication was obtained. The anhedonia subscale score of the SANS correlated negatively with the latency difference score “positive−negative” [rs=−0.28; p≤0.05] and positively with the latency difference score “neutral−positive” [rs=0.30; p≤0.01], but was not related to the latency difference score “neutral−negative” [rs=−0.01]. In other words, the stronger the anhedonic symptoms, the smaller was the difference between detection speed of positive faces compared with that of negative faces, and the greater was the difference between detection speed of positive faces compared with reaction latencies to neutral faces. The affective flattening subscale score of the SANS was not correlated with any of the latency difference scores. On the item level of the SANS anhedonia scale, the highest correlations with latency difference scores were: sexual interest and activity–latency difference score “positive−negative” [rs=−0.28; p≤0.05], and ability to feel intimacy and closeness–latency difference score “neutral−positive” [rs=0.31; p≤0.01]. The latency difference score “positive−negative” was also correlated with anticholinergic medication (BZT dosage) [rs=−0.24; p≤0.05] and gender [rs=−0.20; p≤0.05], but not with dosage of neuroleptic medication, the SAPS sum score, extrapyramidal symptoms, duration of illness or lifetime duration of psychiatric hospitalization, age, or education (p≤0.05). Results of a partial correlation analysis revealed that after removing the linear effects of anticholinergic medication and gender, the correlation between the anhedonia score of the SANS and the latency difference score “positive−negative” was still significant [r=−0.24; p≤0.05].