بازشناسی حالت عاطفی چهره دچار اختلال در افراد الکلی: آیا این نقایص مخصوص نشانه های عاطفی است؟

Abstract Previous studies have repeatedly linked alcoholism is to impairment in emotional facial expression decoding. The present study aimed at extending previous findings while controlling for exposure times of stimuli. Further, a control task was added on the decoding of non-emotional facial features. Twenty-five alcoholic participants were compared to 26 control participants matched for age, sex and educational level. Participants performed two computer tasks consisting of presentation of photographs of faces for either 250 or 1000 ms. The first task required “yes” or “no” responses as rapidly as possible to questions regarding non-emotional features of the face (gender, age range and cultural identity). The second task involved a different set of photographs implicating emotional facial expression decoding, with the same exposure times. Again, rapid “yes” or “no” responses to trials combining 32 emotional facial expressions by eight emotional labels (happiness, sadness, fear, anger, disgust, surprise, shame, and contempt) were required from participants. Reaction times were recorded for both tasks. Alcoholic and control participants showed similar results in both tasks in terms of response accuracy. Yet, in the emotional facial expression task, alcoholic participants' responses matched more negative emotional labels, especially sadness. Further, alcoholics were slower than control participants specifically to answer emotional questions on emotional facial expression. No differences appeared on reaction times in the control task. Contrary to expectations, no interaction of stimulus time exposure and group was observed. Overall, these findings replicate and extend previous results on emotional facial expression decoding ability: Alcoholics are specifically impaired on emotional non-verbal behavior information processing: They are slower to correctly identify an emotion.

Introduction Alcoholics have impairment in cognitive processing of emotional signals. Indeed, studies of recovering alcoholics found deficits in the recognition of emotional facial expressions (EFE) (Oscar-Berman et al., 1990, Philippot et al., 1999, Kornreich et al., 2001a, Kornreich et al., 2001b and Frigerio et al., 2002) as well as in the identification of affective prosody, a non-verbal aspect of speech (Monnot et al., 2001, Monnot et al., 2002 and Uekermann et al., 2005). More specifically, studies on the ability to decode emotional facial expressions have systematically revealed that alcoholics decode emotional facial expressions less accurately than normal controls (Oscar-Berman et al., 1990, Philippot et al., 1999, Kornreich et al., 2001a, Kornreich et al., 2001b and Frigerio et al., 2002) and, to a lesser degree, than opiate dependent patients (Kornreich et al., 2003). In addition, recovering alcoholics overestimate the intensity of emotional facial expression (Philippot et al., 1999, Kornreich et al., 2001a and Kornreich et al., 2001b). They also need a greater intensity of nonverbal signals to perceive an expression as being present (Frigerio et al., 2002), and they display different patterns of interpretation of emotion as compared to controls, with a specific bias towards perceiving expressions as hostile (Philippot et al., 1999, Frigerio et al., 2002 and Townshend and Duka, 2003). Such difficulties in the ability to recognize the emotions felt by others may have an important impact on sociability. Indeed, satisfying and successful interpersonal relationships are partly determined by the ability to accurately interpret non-verbal signals from interaction partners (Carton et al., 1999), among which emotion is a very important factor (Patterson, 1999). Concerning alcohol dependence specifically, it is well known that alcoholics are confronted with severe interpersonal problems in their daily functioning (Nixon et al., 1992 and Duberstein et al., 1993), which are partly mediated by emotional facial expression decoding deficits (Kornreich et al., 2002). In everyday life, facial expressions of emotion rarely last more than 1 s (Ekman, 1984). Further, even with 30 to 50 ms exposure time, emotional facial expression decoding accuracy rates in a normal population remains above chance level (Kirouac and Doré, 1984). In sum, in real life, facial expressions of emotion are displayed for a very short time and are recognized very rapidly. While results from earlier studies systematically evidenced impairments in the decoding of emotional information conveyed by the face, the true depth and nature of this deficit in alcoholics may have been hidden by the design used in these studies: Indeed, participants had the possibility to pace themselves and look at stimuli as long as they pleased, which does not correspond to real life conditions. Further, although the emotional facial expression decoding deficit observed in alcoholism has been well documented, its scope remains nevertheless unclear. It may be the result of a more general impairment in facial or visual perception. Few of the past studies used a control recognition task. Townshend and Duka (2003) included cognitive control tasks on pattern and spatial recognition in their methodology. Alcoholic patients made more errors than control participants in the pattern recognition task, but not in the spatial recognition task. Frigerio et al. (2002) used a control task with animated facial stimuli changing from masculine to feminine: Participants had to indicate when the gender changed. No difference emerged between control and alcoholic participants on this ‘perception of face gender’ task. Apart from this specific study, none of the previous studies included a control-decoding task on facial perception. In order to extend and replicate previous data, the present study used a different approach to control these methodological difficulties and to approach real life conditions: It investigates emotional facial expression decoding in alcoholics while controlling for the exposure time of the stimuli (1000 or 250 ms). Reaction times of the participants were recorded. Further, a new control task on the perception of non-emotional features of the face was designed: Each face stimulus was associated with three questions on gender, ethnicity and age. Again exposure times of stimuli were controlled for (1000 or 250 ms) and reaction times responses of the participants were recorded. We hypothesized that compared to normal controls, recovering alcoholics would demonstrate difficulties in the processing of emotional features of the face, but would not display problems in the processing of non-emotional features.

. Results 3.1. Data analysis All statistical analyses were conducted using the Statistical Package for the Social Sciences (SPSS), version 12.00 for PC. All statistical tests were two-tailed. Unless otherwise specified, P < 0.05 was assumed to define statistical significance. Results of ordinal variables are expressed as means and standard deviations. Correlations for preliminary analyses used the Pearson product–moment correlation coefficient (r). Moreover, between-group comparisons and within-group comparisons were conducted by using either one-way analysis of variance (ANOVA) for ordinal variables, or chi-square statistical test for categorical variables. Statistical differences between the two groups (alcoholic and control participants) on (1) decoding accuracy and (2) reaction times for accurate answers in both tasks were evaluated by repeated measure analyses of variance using a multivariate approach (MANOVA). In the context of the present article, only main effects or interactions involving the factor ‘Group’ are of interest. 3.2. Preliminary analyses Whether in all participants or in the alcohol or control group only, no correlation reached statistical significance between age, and BDI, STAI-B, Sad-Q scores on the one hand, and any computed scores (including accuracy scores and reaction times for accurate responses) in both tasks — control and emotional facial expression task, on the other hand. Similarly, concerning alcoholic participants: length of consumption, length of abstinence, daily alcohol consumption, and number of previous inpatient detoxification stays did not have any impact on the dependent variables. Finally, no gender, educational level, history of coma and familial alcoholism antecedent significant main effect or interaction was observed on the computed dependent variables in the whole group of participants or in each group independently. Therefore, all subsequent analyses were collapsed across these factors. 3.3. Decoding accuracy 3.3.1. Control task In order to compare alcoholic and control participants' performances on the control task, a repeated MANOVA was conducted on the answers of the participants with Exposure Time of the stimulus (250 versus 1000 ms) as the within-subject factors, and Group (alcoholic and control participants) as the between-subject factor on accuracy scores. This analysis did not reveal any main effect or interaction involving the factor Group or the factor Exposure Time of the stimulus. Alcoholic and control participants showed similar performances in their decoding abilities of non-emotional cues in faces without modulation by exposure time of stimulus (see Table 2). Table 2. Decoding accuracy scores in the control task as a function of Group: F-values for MANOVA Sources df F-values Power η2 Group 1, 49 2.845 0.380 0.055 Exposure Time 1, 49 2.117 0.297 0.041 Exposure Time × Group 1, 49 3.335 0.433 0.064 Note.Exposure time of the stimuli (250 and 1000 ms) as the within-subject factor and Group (Alcoholic and Control) as the between-subjects factor. Table options 3.3.2. Emotional facial expression task In order to assess whether there was a difference between alcoholic and control participants' performances in their ability to decode emotional facial expression, a repeated MANOVA was conducted with Exposure Time of the stimulus (250 and 1000 ms), emotional facial expression (happiness, anger, disgust and sadness) and Intensity (30% and 70%) as the within-subject factors, and Group (alcoholic and control participants) as the between-subjects factor on emotional facial expression accuracy scores. No main effect or interaction involving the factor Group was observed (see Table 3). A main effect of “Exposure Time of stimulus” revealed that overall participants demonstrated better accuracy scores with the 1000 ms than with the 250 ms exposure time of the stimulus (respectively, m1000 = 0.25; S.D.1000 = 0.15; m250 = 0.17; S.D.250 = 0.13). Table 3. Decoding accuracy scores in the emotional facial expression decoding task as a function of Group: F-values for MANOVA Sources df F-values Power η2 Group 1, 49 1.939 0.276 0.038 Exposure time 1, 49 30.301⁎⁎⁎ 1.000 0.382 Emotion 3, 47 50.920⁎⁎⁎ 1.000 0.765 Intensity 1, 49 19.631 0.991 0.286 Exposure Time × Group 1, 49 0.120 0.063 0.002 Emotion × Group 3, 47 1.660 0.407 0.096 Intensity × Group 1, 49 1.083 0.175 0.022 Exposure Time × Emotion 3, 47 3.469⁎ 0.741 0.181 Exposure Time × Emotion × Group 3, 47 0.408 0.125 0.025 Exposure Time × Intensity 1, 49 0.004 0.050 0.000 Exposure Time × Intensity × Group 1, 49 0.432 0.099 0.009 Emotion × Intensity 3, 47 0.220 0.088 0.014 Emotion × Intensity × Group 3, 47 0.595 0.164 0.037 Exposure Time × Emotion × Intensity 3, 47 1.229 0.308 0.073 Exposure Time × Emotion × Intensity × Group 3, 47 0.563 0.157 0.035 Note.⁎⁎⁎P < 0.001; ⁎P < 0.05. Emotion (happiness, anger, disgust and sadness), Intensity (30% and 70%), and Exposure Time of the stimuli (250 and 1000 ms) as the within-subject factors, and Group (Alcoholic and Control) as the between-subjects factor. Table options A second repeated MANOVA was performed on the number of times each emotion was recognized as being present (correct identifications and false positives) with Exposure Time of the stimulus (250 and 1000 ms), emotional facial expression (happiness, anger, disgust and sadness), Intensity (30% and 70%) and Emotional Scale (i.e., happiness, anger, disgust, sadness, fear, shame, contempt and surprised) as the within-subject factors, and Group (alcoholic and control participants) as the between-subjects factor. This analysis revealed three significant interactions: (1) “Group × Emotional Scale”, (2) “Group × Emotion” and (3) “Group × Emotion × Emotional Scale” (see Table 4). Of the 32 (4 Emotions × 8 Emotional Scales) Bonferonni post hoc tests computed to specify this third level interaction, six revealed an effect of Group. More specifically, in alcoholics, emotional facial expressions of happiness were more often judged as reflecting surprise than in control participants (P ≤ 0.03, respectively, m = 0.37, S.D. = 0.31; m = 0.20, S.D. = 0.20). Emotional facial expressions of disgust were more frequently labeled as reflecting sadness by alcoholic participants than by control participants (P ≤ 0.02, respectively, m = 0.38, S.D. = 0.20; m = 0.24, S.D. = 0.21). Concerning emotional facial expressions of anger, alcoholic individuals were more prone than the control volunteers to identify them as disgust (P ≤ 0.003, respectively, m = 0.51, S.D. = 0.28; m = 0.30, S.D. = 0.20) or sadness (P ≤ 0.004, respectively, m = 0.38, S.D. = 0.25; m = 0.20, S.D. = 0.18). Finally, the alcohol group rated emotional facial expressions of sadness as expressions of disgust more frequently than control participants (P ≤ 0.05, respectively, m = 0.33, S.D. = 0.28; m = 0.19, S.D. = 0.21), and they more frequently inferred sadness correctly than individuals with no history of alcohol dependence (P ≤ 0.02, respectively, m = 0.65, S.D. = 0.20; m = 0.50, S.D. = 0.24). Table 4. Emotional facial expression decoding task : F-values for MANOVA on the number of times each emotion was recognized as being present (on correct identifications and false positives) Sources df F-values Power η2 Group 1, 49 3.989 0.330 0.047 Exposure Time 1, 49 1.147 0.183 0.023 Emotion 3, 47 42.585⁎⁎⁎ 1.000 0.731 Intensity 1, 49 15.081⁎⁎⁎ 0.968 0.235 Scale 7, 43 12.664⁎⁎⁎ 1.000 0.673 Exposure Time × Group 1, 49 0.166 0.068 0.003 Emotion × Group 3, 47 3.116⁎ 0.689 0.166 Intensity × Group 1, 49 0.108 0.062 0.002 Scale × Group 7, 43 2.437⁎ 0.807 0.284 Exposure Time × Emotion 3, 47 1.925 0.466 0.109 Exposure Time × Emotion × Group 3, 47 0.896 0.231 0.054 Exposure Time × Intensity 1, 49 2.785 0.373 0.054 Exposure Time × Intensity × Group 1, 49 0.404 0.096 0.008 Emotion × Intensity 3, 47 14.155⁎⁎⁎ 1.000 0.475 Emotion × Intensity × Group 3, 47 1.145 0.288 0.068 Exposure Time × Emotion × Intensity 3, 47 4.694⁎⁎ 0.869 0.231 Exposure Time × Emotion × Intensity × Group 3, 47 0.512 0.146 0.032 Emotion × Scale 21, 29 38.041⁎⁎⁎ 1.000 0.965 Emotion × Scale × Group 21, 29 2.109⁎ 0.916 0.604 Exposure Time × Scale 7, 43 2.589⁎ 0.834 0.296 Exposure Time × Scale × Group 7, 43 0.201 0.099 0.032 Emotion × Exposure Time × Scale 21, 29 1.678 0.821 0.549 Emotion × Exposure Time × Scale × Group 21, 29 1.444 0.743 0.511 Intensity × Scale 7, 43 10.244⁎⁎⁎ 1.000 0.625 Intensity × Scale × Group 7, 43 2.000 0.708 0.246 Emotion × Intensity × Scale 21, 29 13.181⁎⁎⁎ 1.000 0.905 Emotion × Intensity × Scale × Group 21, 29 0.997 0.535 0.419 Exposure Time × Intensity × Scale 7, 43 1.057 0.399 0.147 Exposure Time × Intensity × Scale × Group 7, 43 0.256 0.115 0.040 Emotion × Exposure Time × Intensity × Scale 21, 29 2.216⁎ 0.931 0.616 Emotion × Exposure Time × Intensity × Scale × Group 21, 29 1.499 0.763 0.520 Note.⁎⁎⁎P < 0.001; ⁎⁎P < 0.01; ⁎P < 0.05. Emotion (happiness, anger, disgust and sadness), Intensity (30% and 70%), emotional Scale (happiness, sadness, anger, disgust, contempt, fear, shame, and surprise) and Exposure Time of the stimuli (250 and 1000 ms) as the within-subject factors, and Group (Alcoholic and Control) as the between-subjects factor. Table options Thus, alcoholic and control participants demonstrated similar accuracy scores. Yet, both groups slightly differed in their pattern of answers, alcoholics generally matching more negative emotional items, particularly sadness, with negative emotional facial expressions, be they actual sadness expressions or not. 3.4. Reaction time for accurate responses 3.4.1. Control task A repeated measures analysis of variance using a multivariate approach with 2 Exposure Times of the stimulus (250 versus 1000 ms) and 2 Answer Type (correct acceptance versus correct rejection) as the within-subject factors, and Group (alcoholic and control participants) as the between-subjects factor was computed on the reaction times for accurate answers on the control task (89, 64% of the total number of responses in the whole sample: 87, 75% and 89, 50% in alcoholic and control participants, respectively). There was no effect of ‘Group’ (see Table 5): Alcoholic (m = 2009.08; S.D. = 773.80) did not differ statistically from control participants (m = 1709.60; S.D. = 610.17) in their reaction time to questions relating to non-emotional cues of the face, regardless of the exposure time of the stimulus. A main effect of the factor “Accurate Answer” was observed: Overall, participants were faster for correct acceptances (m = 1772.70; S.D. = 633.14), compared to correct rejections (m = 1940.10, S.D. = 825.76). Table 5. Reaction times in the control task as a function of Group: F-values for MANOVA Sources df F-values Power η2 Group 1, 49 2.365 0.326 0.046 Exposure Time 1, 49 0.000 0.050 0.000 Answer Type 1, 49 7.772⁎⁎ 0.780 0.137 Exposure Time × Group 1, 49 1.573 0.233 0.031 Answer Type × Group 1, 49 0.032 0.053 0.001 Exposure Time × Answer Type 1, 49 0.061 0.057 0.001 Exposure Time × Answer Type × Group 1, 49 2.529 0.344 0.049 Note.⁎⁎P < 0.01. Answer Type (correct acceptance versus correct rejection) and Exposure Time of the stimuli (250 and 1000 ms) as the within-subject factors, and Group (Alcoholic and Control) as the between-subjects factor. Table options 3.4.2. Emotional facial expression task A repeated measures analysis of variance using a multivariate approach with Exposure Time of the stimulus (250 versus 1000 ms) and Answer Type (correct acceptance versus correct rejection) as the within-subject factors, and Group (alcoholic and control participants) as the between-subjects factor was computed on the reaction times for accurate answers on the emotional facial expression task (76.07% of the total number of responses in the whole sample: 74.50% and 77.58% in alcoholic and control participants, respectively). It revealed a main effect of ‘Group’ and a main effect of ‘Accurate Answer’ (see Table 6). Fig. 1 shows that, overall, correct acceptances (m = 1534.79; S.D. = 576.47) were made faster than correct rejections (m = 1671.85; S.D. = 808.38) and that alcoholics needed significantly more time (m = 1818.12; S.D. = 684.92) than control participants (m = 1375.56; S.D. = 585.18) to answer accurately to questions pertaining to the emotional decoding of the emotional facial expression task, regardless of the exposure time of the stimulus, and the type of answer expected (correct acceptance versus correct rejection). Table 6. Reaction times in the emotional facial expression decoding task as a function of Group: F-values for MANOVA Sources df F-values Power η2 Group 1, 49 6.171⁎ 0.683 0.112 Exposure Time 1, 49 1.769 0.256 0.112 Answer Type 1, 49 6.178⁎ 0.683 0.112 Exposure Time × Group 1, 49 0.606 0.119 0.012 Answer Type × Group 1, 49 0.823 0.144 0.017 Exposure Time × Answer Type 1, 49 1.465 0.221 0.029 Exposure Time × Answer Type × Group 1, 49 1.692 0.247 0.033 Note.⁎P < 0.05. Answer Type (correct acceptance versus correct rejection) and Exposure Time of the stimuli (250 and 1000 ms) as the within-subject factors, and Group (Alcoholic and Control) as the between-subjects factor. Table options Reaction times for accurate responses in the emotional facial expression ... Fig. 1. Reaction times for accurate responses in the emotional facial expression decoding task as functions of Groups (Control and Alcohol), Exposure Times of stimuli (250 and 1000 ms), and the types of answers expected (correct acceptance versus correct rejection).