Facial affect decoding in schizophrenic disorders: A study using event-related potentials

Transcription

1 Psychiatry Research 141 (2006) Facial affect decoding in schizophrenic disorders: A study using event-related potentials Martin J. Herrmann *, Andreas Reif, Burkhard E. Jabs, Christian Jacob, Andreas J. Fallgatter Department of Psychiatry and Psychotherapy at the University of Würzburg Füchsleinstr. 15, Würzburg, Germany Received 13 July 2005; accepted 29 September 2005 Abstract Deficits in emotional processing are evident in schizophrenia, but the underlying processes are still under debate. In this study we tried to replicate findings of diminished prefrontal electroencephalographic response during facial affect recognition in healthy controls and subsequently in schizophrenic patients. As a first step, we analysed the event-related potentials (ERPs) of 36 healthy subjects during emotional expression decoding compared with neutral face viewing. Subsequently, the ERPs of 22 patients with schizophrenia were compared with the ERPs of 22 healthy subjects matched for age and sex. The hypothesised increase in the negative component at 200 ms over frontal brain regions during facial affect decoding was not found in this study. Instead we found increased positive amplitudes at 300 ms over parietal brain areas for the active affect-decoding task compared with the passive neutral face-viewing task. Interestingly, schizophrenic patients had higher amplitudes in the neutral condition than did healthy controls. This effect was more pronounced in the paranoid subgroup of patients. D 2005 Elsevier Ireland Ltd. All rights reserved. Keywords: Event-related potentials; Visual; Face; Facial emotion recognition; Paranoid; Psychosis 1. Introduction Schizophrenic patients exhibit well-known deficits in the decoding of facial expressions (Hellewell and Whittaker, 1998; Edwards et al., 2002) with consequences for social competence (Mueser et al., 1996) and ward behaviour (Ihnen et al., 1998). Based on these deficits, remediation strategies to address impairments in facial affect recognition have been proposed (Frommann et al., 2003), although the pathophysiological processes contributing to these deficits are largely * Corresponding author. Tel.: ; fax: address: Martin.Herrmann@mail.uni-wuerzburg.de (M.J. Herrmann). unknown. However, findings have been reported in recent years that shed light on the processes underlying the observed face-encoding deficits in schizophrenia. These studies found that the volume of the fusiform gyrus, a region that is involved in face processing (Kanwisher et al., 1997), is reduced in schizophrenic patients (Lee et al., 2002; Onitsuka et al., 2003). Further support of a deficit in face processing was provided by reports of reduced functional activation of the fusiform gyrus during face processing (Quintana et al., 2003) and reduced amplitudes of the face-specific event-related potential (ERP) N170 component (Herrmann et al., 2004). With regard to emotion-decoding processes, the picture is less clear. In a study using functional magnetic resonance imaging (fmri), Phillips et al. (1999) observed reduced brain activation in schizophrenic /$ - see front matter D 2005 Elsevier Ireland Ltd. All rights reserved. doi: /j.psychres

2 248 M.J. Herrmann et al. / Psychiatry Research 141 (2006) patients compared with healthy controls during the decoding of facial expressions of anger (inferior frontal gyrus, putamen, and cerebellum), fear (superior temporal gyrus, amygdala, and putamen), and disgust (globus pallidus and insula). Of the above-reviewed findings, only the reduction in activation of the amygdala during the decoding of facial expressions by schizophrenic patients was replicated in subsequent studies (Gur et al., 2002; Hempel et al., 2003). Otherwise the findings of neuroimaging in this area remain inconsistent and difficult to interpret. For example, Lane (2003) argued that the activity observed in the amygdala might reflect an emotional response in the viewer induced by the perception of facial expressions. Subsequently, the emotional activation might aid in the identification of facial expressions. Nevertheless, this interpretation is still speculative and hampered by the basic problem of unsatisfactory temporal resolution of neuroimaging techniques. Using ERPs, Streit et al. (2000) found that intact facial expression decoding, compared with blurred face recognition (low-pass-filtered), leads to increased negative amplitudes over frontal electrode positions beginning at 160 ms and peaking at around 240 ms. This negative deflection was less negative during facial affect decoding in schizophrenic patients than in controls (Streit et al., 2001). These reports are of interest, but understanding is hampered by the fact that experimental conditions differed not only with regard to task demands (facial expression decoding vs. face recognition) but also with respect to stimulus qualities (intact vs. low-pass-filtered). The aim of the present study is to replicate the increased negative amplitudes over Fz in the time window around 230 ms during facial affect decoding by presenting neutral instead of filtered faces. We hypothesised that those evoked during facial expression decoding around 230 ms over Fz would be more negative than the ERPs evoked during neutral face viewing. Furthermore, we expected that schizophrenic patients would not show increased negative amplitudes during expression decoding. 2. Methods 2.1. Participants In this study, 36 healthy subjects and 22 patients with schizophrenia diagnosed according to DSM-IV (American Psychiatric Association, 1994) were studied after giving written informed consent. All subjects were right-handed. After a first analysis of our hypothesis in the healthy control group (19 males, 17 females; mean age F S.D. = 31.4 F 11.6 years), we compared the 22 patients (17 males, 5 females; mean age = 31.7 F 8.4 years) with 22 healthy subjects (out of the 36 healthy controls) matched to the patients for age and sex (17 males, 5 females; mean age=31.9 F11.0 years). Subtype diagnoses in the schizophrenic group were as follows: paranoid, n =9, hebephrenic, n =9, and catatonic, n =3. Diagnoses were made by a senior psychiatrist in a comprehensive psychopathological assessment and confirmed by the first author. The mean level of education was 10.3 years for the patients and 9.9 years for the controls. Sixteen of the patients were being treated with neuroleptic medication (3 typical antipsychotics, 8 atypical antipsychotics, and 5 a combination of both types); the mean daily dosage was F mg in chlorpromazine equivalents. Scores on the Positive and Negative Syndrome Scale (PANSS) in the patients were as follows: negative symptoms, mean = 24.1, S.D. = 4.7; positive symptoms, mean = 10.7, S.D. = 3.2; and general psychopathology, mean =17.1, S.D.= Procedures Electroencaphalographic activity was acquired from 21 scalp electrodes positioned according to the international system against linked mastoids. Three additional electrodes were placed at the outer canthi of both eyes and below the right eye to monitor eye blinks and movements. The electroencephalogram was sampled continuously at a rate of 256 Hz with a bandpass from 0.1 to 70 Hz. Impedances were kept at 5 kv or below. During the recording period, the subjects had to look at faces of four different persons showing neutral and emotional (anger, disgust, and fear) expressions (Ekman and Friesen, 1976). The faces were shown repeatedly (eight times for each picture), and each was displayed for 500 ms with an interstimulus interval of 1500 ms (visual horizontal angle = 4.68 and vertical angle =6.38). The task consisted of two runs. In the first run, the subjects looked at the neutral faces; in the second run, they were asked to judge the three emotional expressions. After approximately eight presentations, the subjects had to verbalize the emotional expression of the last face (in the neutral condition, the subjects were instructed to say bneutralq) Data analysis Epochs ( 100 ms before stimulus onset to 650 ms) with amplitudes or a voltage step exceedingf50

3 M.J. Herrmann et al. / Psychiatry Research 141 (2006) AV were excluded from further analysis. After average reference calculation, the artifact-free trials (at least 20 epochs) were averaged separately for each subject and condition. The ERPs were determined according to the grand mean ERP with the segment borders for the positive peaks over Fz and Cz at 227, 301, 426, and 648 ms; over Pz at 234, 422, and 648 ms. Over Fz and Cz, we additionally analysed the negative peaks between 160 and 270 ms, between 270 and 371 ms, and between 371 and 583 ms. We labelled the positive components simply as P3, P4 and P5, and the negative components as N2, N3 and N4. These labels are not related to the classical components with the same names and should not be mistaken for them. To test the study hypotheses, we used alpha =0.05; for additional tests, we used an adjusted significance level (a adjusted =0.01/14=0.0007). Greenhouse Geisser correction was used for the degrees of freedom. Statistical analyses were calculated using SPSS 11. Values are presented as mean FS.D. 3. Results 3.1. Does emotion decoding lead to increased negative frontal components? For the 36 healthy subjects, we did not find any significant differences in the N2 (from 160 to 270 ms) amplitudes over Fz [( F(3, 105) = 1.12, P = 0.35], and therefore we had to reject the first hypothesis of this study (Table 1). Only the differences in the amplitude over Pz for the P3 [( F(2.6,89.3)=9.23, P b ] and for P4 [ F(3, 105) = 9.63, P b ] reached the adjusted significance level (a adjusted for 14 tests =0.01/ 14 = ). Post hoc t-tests revealed that these differences were caused by higher amplitudes elicited by the emotional facial expressions bfearq [P3: mean = 2.70F1.07; t(35) =3.71, P b0.0001; P4: mean =2.62 F1.24; t(95) =4.33, P b0.0001], bangerq [P3: mean = 2.83 F1.11; t(35) = 4.58, P b ; P4: mean = 2.45 F 1.06; t(35) = 3.85, P b ], and bdisgustq [P3: mean =2.63F1.01; t(35) =3.08, P b0.004; P4: mean = 2.50F1.02; t(35) =4.24, P b0.0001] compared with the neutral faces (P3: mean =2.09F1.22; P4: mean =1.89 F 0.75). No amplitude differences between the emotional faces were found for the amplitudes of the P3 [ F(2, 70) = 1.16, P = 0.32] or the P4 component [ F(2,70)=0.77, P =0.47] Differences between schizophrenic patients and healthy controls For the time window P3 as well as P4, we found significant interaction effects of condition diagnosis [see Fig. 1; P3: F(1, 42) = 5.0, P b 0.05; P4: F(1,42)=12.1, P b0.001]. Furthermore a main effect of the diagnosis bschizophreniaq on the P4 time segment was observable [ F(1,42]=5.1, P b0.05]. Post hoc tests revealed that for the healthy controls the amplitudes in the emotion-decoding condition were significantly higher than in the neutral face-viewing condition for both time segments [P3: t(21] =3.37, P b0.01; P4: t(21) = 3.78, P b 0.001]. In contrast, the amplitudes of the schizophrenic patients did not differ between Table 1 Amplitudes to the neutral and emotional faces for different time windows over the electrode positions Fz, Cz and Pz Time Neutral Disgust Anger Fear F(3, 105) P M FS.D. M FS.D. M FS.D. M FS.D. Fz N2 2.59F F F F N3 2.18F F F F N4 1.72F F F F a 0.18 P3 0.84F F F F P4 0.15F F F F P5 0.64F F F F b 0.37 Cz N2 1.68F F F F c 0.37 N3 1.46F F F F N4 1.19F F F F P3 0.14F F F F d 0.27 P4 0.73F F F F P5 1.46F F F F Pz P3 2.09F F F F e P4 1.89F F F F a df =2.5, 87.1; b df =2.5, 86.5; c df =1.8, 64.6; d df =2.4, 82.9; e df =2.6, 89.3.

4 250 M.J. Herrmann et al. / Psychiatry Research 141 (2006) µv 3 emotional decoding neutral viewing µv 3 2 P4 P5 2 P4 P ms ms -1 Healthy subjects -1 Schizophrenic patients Fig. 1. Grand mean event-related potentials for the healthy controls and the schizophrenic patients over Pz. Vertical lines indicate segment borders for the time windows. the two conditions [P3: t(21) = 0.59, P = 0.56; P4: t(21) =1.38, P =0.18]. Comparing each condition between groups revealed, that schizophrenic patients had higher amplitudes in the neutral condition than did the healthy controls [ P3: t(34) = 1.96, P = 0.06; P4: t(31) =3.38, P b0.01] but did not differ in the emotion-decoding condition [P3: t(42) = 0.33, P = 0.75; P4: t(42) =0.77, P =0.45]. Comparisons of the amplitude differences between the emotion-decoding condition and the neutral viewing condition showed higher differences for the controls than for the patients [P3: controls: mean =0.46F0.64; patients: mean = 0.14F1.08; t(42) =2.23, P b0.05; P4: controls: mean =0.61F0.75; patients: mean = 0.27F0.92; t(42) =3.47, P b0.001] (Table 2). The behavioural data revealed that controls significantly more often (mean = 3.86 F 0.47) reported the correct answer bneutralq in the neutral viewing condition than schizophrenic patients (mean = 2.50 F 1.66; Z = 3.51, P b0.001). Schizophrenic patients showed a trend to misidentify (mean=0.68f1.29) the neutral face as an emotional face compared with the healthy controls (mean = 0.14 F 0.47; Z = 1.87, P b 0.1). The emotion-decoding condition revealed no differences in the number of errors for facial affect recognition between the controls (mean=3.45f1.99) and the patients (mean =4.32F2.39; Z = 1.15, P =0.25) Comparisons of paranoid and non-paranoid patients In the neutral viewing condition, paranoid patients had significantly higher amplitudes [P3: t(19) = 2.15, P b0.05; P4: t(19) =2.21, P b0.05] than non-paranoid patients, but the two subgroups did not differ in the emotion-decoding condition [P3: t(19) = 0.28, P = 0.78; P4: t(19) = 1.67, P = 0.11]. Neither paranoid [P3: t(8)=1.82, P =0.11; P4: t(8)=1.64, P =0.14] nor nonparanoid patients [P3: t(11)= 1.78, P =0.10; P4: t(11) = 0.15, P = 0.89] showed significant betweencondition performance differences. Amplitude differences between the emotional condition and the neutral condition were higher for non-paranoid (mean = 0.36 F 0.70) than for paranoid patients for P3 [mean = 0.76F1.26; t(19) =2.61, P b0.05], but not for P4 [non-paranoid: mean= 0.03F0.79; paranoid: mean = 0.59F1.07; t(19) =1.36, P =0.19] (Table 3). The behavioural data did not reveal differences in the neutral condition between the subgroups. Neither the number of correctly labelled neutral faces (nonparanoid: mean =2.58F1.68; paranoid: mean =2.22F 1.72; Z = 0.41, P =0.70) nor the number of erroneously labelled neutral faces (non-paranoid: mean = 0.50F1.17; paranoid: mean =1.00F1.50; Z = 0.94, P = 0.46) differed significantly between the subgroups. Table 2 Amplitudes (in AV) over Pz and statistical analyses Healthy controls Schizophrenic patients ANOVA Neutral Emotional Neutral Emotional C D BD P3 2.14F F F F * P4 1.72F F F F * 12.1*** df =1,42; C = condition; D = diagnosis and interaction CD = conditiondiagnosis; *P b0.05; ***P b0.001.

5 M.J. Herrmann et al. / Psychiatry Research 141 (2006) Table 3 Amplitudes (in AV) over Pz for paranoid and non-paranoid schizophrenic patients Non-paranoid (n =12) Paranoid (n =9) Neutral Emotional Neutral Emotional P3 2.35F F F F1.50 P4 2.38F F F F0.99 In contrast to the previous studies, paranoid patients (mean=5.67f2.50) made more errors in the emotiondecoding conditions than non-paranoid patients (mean = 3.42F1.93; Z =2.23, P b0.05). No significant correlations between PANSS scores and ERP parameters were found. 4. Discussion We were unable to replicate a previous finding of an increased N2 component during facial affect decoding in healthy subjects (Streit et al., 2000). In both the earlier study and our study, ERPs derived from emotional faces which had to be categorised were compared with a condition in which no emotion decoding was necessary. In the study of Streit et al. (2000), blurred faces had to be processed, while in our study intact neutral faces had to be viewed. Under the assumption that the facial decoding condition in both studies leads to similar ERPs, the different results most likely are caused by different control conditions. Therefore, the reduced N2 found by Streit et al. (2000) might reflect the task demands of face recognition, which were absent in our study, or reflect the fact that the faces were blurred. However, these interpretations do not support the conclusion that the N2 component signifies a neural correlate of facial affect decoding (Streit et al., 2000). With a strictly adjusted alpha level, the differences of the N3 component over Fz and the N3 and N4 components over Cz did not reach significance. One may argue that the earlier described effect (Streit et al., 2000) is slightly later or more centrally localized in our study and therefore was not identified as significant. However, when the data were analysed in a descriptive manner, we found the most negative values for the neutral condition for all components, contradicting the hypothesis of increased negative amplitude being indicative of facial affect decoding. Not surprisingly, we found higher amplitudes beginning at 234 ms over the parietal brain region for the active condition (facial affect decoding) than for the less neutral face-viewing condition. These increased amplitudes cannot be explained as a neural correlate of a specific process involved in facial expression analysis but rather as a sign of nonspecific neural activity related to task demands. Interestingly, we found a significant interaction effect of condition diagnosis, which was explained by reduced differences between conditions in the schizophrenic patients. Even more interestingly, this was attributable to the increased amplitudes in the schizophrenic patients during the neutral condition and not by diminished amplitudes in the facial expression analysis. The increased amplitudes of the schizophrenic patients in the neutral condition were accompanied by significantly higher error rates in the neutral condition and by a tendency for more errors in identifying neutral faces as expressing an emotion. These behavioural alterations have been described in other studies (Heimberg et al., 1992; Schneider et al., 1995). Our results indicate that schizophrenic patients are not characterized by a deficit in activation during facial affect decoding as reported in previous studies but also show increased activation to neutral faces. The increased brain activation in response to neutral faces might be interpreted as a tendency of schizophrenic patients to look actively for information in a neutral situation. This interpretation was supported by the additional findings of higher amplitudes in the neutral condition of the paranoid subgroup compared with the non-paranoid subgroup, without any differences in the active facial affect decoding condition. On a behavioural level, this effect was not found. Paranoid patients performed the same as non-paranoid patients in the neutral condition, yet they made more errors in the emotiondecoding condition. This is in contrast to previous results showing better emotion recognition in paranoid patients (Kline et al., 1992; Lewis and Garver, 1995). Similarly, the failure to find behavioural deficits in the emotion-decoding condition for schizophrenic patients compared with the controls is in contrast to findings reported in the literature (Hellewell and Whittaker, 1998; Edwards et al., 2002). It might be argued that our task, with its repeated presentation of face stimuli, leads to a training effect. For the paranoid group, this repeated presentation may led to an aggravation of facial affect decoding caused by the tendency to look for further information. In summary, our study provides evidence against a frontal N2 effect for facial affect decoding. For the schizophrenic patients, we found neither significant differences in behavioural data nor in brain activation during facial affect decoding with this task. The failure

6 252 M.J. Herrmann et al. / Psychiatry Research 141 (2006) to find deficits in this paradigm supports the idea that training might be able to overcome problems in facial affect decoding (Frommann et al., 2003). The increased activation in the neutral condition points to the fact that disadvantageous information-processing strategies, more than brain deficits, might cause social disturbances in schizophrenic patients. References American Pychiatric Association, DSM-IV: Diagnostic and Statistical Manual of Mental Disorders. (4th ed.). APA, Washington, DC. Edwards, J., Jackson, H.J., Pattison, P.E., Emotion recognition via facial expression and affective prosody in schizophrenia: a methodological review. Clinical Psychological Review 22, Ekman, P., Friesen, W.V., Pictures of Facial Affect. Consulting Psychologists Press, Palo Alto. Frommann, N., Streit, M., Woelwer, W., Remediation of facial affect recognition impairments in patients with schizophrenia: a new training program. Psychiatry Research 117, Gur, R.E., McGrath, C., Chan, R.M., Schroeder, L., Turner, T., Turetsky, B.I., Kohler, C., Alsop, D., Maldjian, J., Ragland, J.D., Gur, R.C., An fmri study of facial emotion processing in patients with schizophrenia. American Journal of Psychiatry 159, Heimberg, C., Gur, R.E., Erwin, R.J., Shtasel, D.L., Gur, R.C., Facial emotion discrimination: III. Behavioral findings in schizophrenia. Psychiatry Research 42, Hellewell, J.S.E., Whittaker, J.F., Affect perception and social knowledge in schizophrenia. In: Mueser, K.T., Tarrier, N. (Eds.), Handbook of Social Functioning in Schizophrenia. Allyn and Bacon, Needham Hengths. Hempel, A., Hempel, E., Schonknecht, P., Stippich, C., Schröder, J., Impairment in basal limbic function in schizophrenia during affect recognition. Psychiatry Research: Neuroimaging 122, Herrmann, M.J., Ellgring, H., Fallgatter, A.J., Dysfunction of early stage face processing in schizophrenias. American Journal of Psychiatry 161, Ihnen, G.H., Penn, D.L., Corrigan, P.W., Martin, J., Social perception and social skill in schizophrenia. Psychiatry Research 80, Kanwisher, N., McDermott, J., Chun, M.M., The fusiform face area: a module in human extrastriate cortex specialized for face perception. Journal of Neuroscience 17, Kline, J.S., Smith, J.E., Ellis, H.C., Paranoid and nonparanoid schizophrenic processing of facially displayed affect. Journal of Psychiatric Research 26, Lane, R.D., The neural substrates of affect impairment in schizophrenia. American Journal of Psychiatry 160, Lee, C.U., Shenton, M.E., Salisbury, D.F., Kasai, K., Onitsuka, T., Dickey, C.C., Yurgelun-Todd, D., Kikinis, R., Jolesz, F.A., McCarley, R.W., Fusiform gyrus volume reduction in first-episode schizophrenia: a magnetic resonance imaging study. Archives of General Psychiatry 59, Lewis, S.F., Garver, D.L., Treatment and diagnostic subtype in facial affect recognition in schizophrenia. Journal of Psychiatric Research 29, Mueser, K.T., Doonan, R., Penn, D.L., Blanchard, J.J., Bellack, A.S., Nishith, P., DeLeon, J., Emotion recognition and social competence in chronic schizophrenia. Journal of Abnormal Psychology 105, Onitsuka, T., Shenton, M.E., Kasai, K., Nestor, P.G., Toner, S.K., Kikinis, R., Jolesz, F.A., McCarley, R.W., Fusiform gyrus volume reduction and facial recognition in chronic schizophrenia. Archives of General Psychiatry 60, Phillips, M.L., Williams, L., Senior, C., Bullmore, E.T., Brammer, M.J., Andrew, C., Williams, S.C., David, A.S., A differential neural response to threatening and non-threatening negative facial expressions in paranoid and non-paranoid schizophrenics. Psychiatry Research: Neuroimaging 92, Quintana, J., Wong, T., Ortiz-Portillo, E., Marder, S.R., Mazziotta, J.C., Right lateral fusiform gyrus dysfunction during facial information processing in schizophrenia. Biological Psychiatry 53, Schneider, F., Gur, R.C., Gur, R.E., Shtasel, D.L., Emotional processing in schizophrenia: neurobehavioral probes in relation to psychopathology. Schizophrenia Research 17, Streit, M., Wölwer, W., Brinkmeyer, J., Ihl, R., Gaebel, W., Electrophysiological correlates of emotional and structural face processing in humans. Neuroscience Letters 278, Streit, M., Wölwer, W., Brinkmeyer, J., Ihl, R., Gaebel, W., EEG correlates of facial affect recognition and categorisation of blurred faces in schizophrenic patients and healthy volunteers. Schizophrenia Research 49,