حالت چهره عاطفی به عنوان پتانسیل های مغزی مرتبط با رویداد مشهود

Abstract As human faces are important social signals in everyday life, processing of facial affect has recently entered into the focus of neuroscientific research. In the present study, priming of faces showing the same emotional expression was measured with the help of event-related potentials (ERPs) in order to investigate the temporal characteristics of processing facial expressions. Participants classified portraits of unfamiliar persons according to their emotional expression (happy or angry). The portraits were either preceded by the face of a different person expressing the same affect (primed) or the opposite affect (unprimed). ERPs revealed both early and late priming effects, independent of stimulus valence. The early priming effect was characterized by attenuated frontal ERP amplitudes between 100 and 200 ms in response to primed targets. Its dipole sources were localised in the inferior occipitotemporal cortex, possibly related to the detection of expression-specific facial configurations, and in the insular cortex, considered to be involved in affective processes. The late priming effect, an enhancement of the late positive potential (LPP) following unprimed targets, may evidence greater relevance attributed to a change of emotional expressions. Our results (i) point to the view that a change of affect-related facial configuration can be detected very early during face perception and (ii) support previous findings on the amplitude of the late positive potential being rather related to arousal than to the specific valence of an emotional signal.

1. Introduction Facial expressions are among the most significant social signals in personal communication, as they convey information about ourselves and influence the affective states of others. The influential model of face processing by Bruce and Young (1986) postulates several subsequent stages of face recognition. According to this model, structural encoding of a face-specific configuration is followed by the recognition of a familiar face by activating stored face representations and by retrieving semantic knowledge about the person perceived. Evidence from imaging research suggests that processing of facial affect relies on the interplay of several distinct brain areas. The inferior occipitotemporal cortex, especially the fusiform gyrus, plays a key role for the detection of facial configurations (Kanwisher et al., 1997). Further analysis of facial affect has been shown to be related to activation of the superior temporal sulcus, the amygdala, the orbitofrontal cortex and the insular cortex (for a review, see Haxby et al., 2000). Research on event-related potentials (ERPs) has repeatedly demonstrated that the stage of structural encoding is related to a face-specific negativity (N170; Bentin et al., 1996) observed at occipitotemporal electrodes in response to faces, and to its positive frontocentral counterpart in the same latency range, the vertex positivity (Bötzel and Grüsser, 1989). The Bruce and Young (1986) model postulates that processing of facial expressions is independent of face recognition. This issue has been addressed by a series of studies measuring ERPs, a majority of which reported the N170 component to be unaffected by the presence or type of emotional facial expression (Carretié and Iglesias, 1995, Eimer and Holmes, 2002, Herrmann et al., 2002 and Streit et al., 1999; but see also Marinkovic and Halgren, 1998). However, recent ERP research indicates that the N170 may not be the earliest component which is sensitive to faces (Itier and Taylor, 2004), and several studies showed ERP correlates of processing emotional expressions preceding the N170. Eimer and Holmes (2002) reported a frontocentral positivity as early as 120 ms after the presentation of fearful faces as compared to neutral faces, and concluded that processing of facial affect starts before face identification. Using magnetoencephalographic recordings, Streit et al. (1999) observed an activation of the superior temporal cortex at 140–170 ms during affect recognition, but not during identity recognition. Eger et al. (2003) demonstrated that emotional expression even affected visually evoked potentials at 85 ms. Pizzagalli et al. (2002) detected differences between faces judged as ‘likable’ or ‘not likable’, starting 112 ms after stimulus presentation. In this study, stimuli were not standardized with respect to facial expression, and a number of faces in the ‘not likable’ condition were characterized by facial distortions. Consequently, this finding may provide further support for the view that a configuration of facial features which deviates from a configuration perceived as ‘normal’ or ‘neutral’ may be already detected during the early stage of visual processing (cf. Halgren et al., 2000). In recent ERP research, the priming technique has proven to be a valuable method for the temporal localisation of processing stages. Repetition priming has been employed by several studies investigating the recognition of familiar faces. Most notably, two repetition-sensitive ERP components have been shown, that is, alterations of the ERP waveform in response to primed as compared to unprimed faces. The ‘early repetition effect’ or N250r is characterized by attenuated potentials at frontal and temporal sites about 200–350 ms after the presentation of primed as compared to unprimed target faces (Schweinberger et al., 1995). Recent findings indicate that this priming effect reflects the activation of face recognition units (Schweinberger and Burton, 2003, for review). The second repetition-sensitive component, the so-called ‘late repetition effect’, consists in an enhanced late positive potential (LPP) following the presentation of repeated (primed) as compared to unprimed faces. This effect also appears when semantically related faces are presented consecutively and is considered to result from a diminished superimposed N400 component when semantically related or identical faces follow each other. Similar results were reported by several studies employing matching tasks, in which participants compared two consecutive portraits with respect to congruence of facial expressions or identities (Bobes et al., 2000 and Münte et al., 1998). Priming of stimuli which evoke similar affective attitudes has up to now only been demonstrated by behavioral studies showing that reaction times to words and pictures are shortened if their affective valence is primed by a preceding stimulus of similar valence (e.g. Fazio et al., 1986). As these affective priming effects decay rapidly, they can be observed best for stimulus onset asynchronies (SOAs) shorter than 300 ms (Hermans et al., 2001). In another line of research, ERP studies on processing of emotional words and pictures reported augmented LPP amplitudes in response to affectively relevant stimuli as compared to neutral ones. Pleasant and unpleasant stimuli elicited larger LPPs than neutral ones (e.g., Schupp et al., 2003), independent of the explicit task performed by the participants (Crites et al., 1995). This finding was interpreted as a P300 effect, reflecting the increased attentional engagement evoked by affective stimuli. The fact that pleasant and unpleasant pictures elicited comparable LPPs (Schupp et al., 2003 and Vanderploeg et al., 1987) indicates that increased LPP amplitudes might not be related to a specific stimulus valence, but rather to the arousal level evoked by the stimulus, which may alter its perceived subjective relevance. To our knowledge, there are no ERP studies directly aimed at investigating priming effects in response to faces showing the same or a different facial expression so far. However, the results of a recent experiment by Campanella et al. (2002) on the categorical perception of different emotional expressions can be taken as a first, tentative evidence for priming of facial affect. They employed a delayed same–different matching task for facial expressions (happy or fearful). At 150 ms after presentation of the second face, ERP amplitudes were reduced in response to identical pairs of faces and to faces with different intensities of the same expression, as compared to faces with different expressions. This finding can be interpreted as an early priming effect because identical pairs and within-category pairs were perceived by the participants as the same expression. Additionally, LPP amplitudes were smaller for identical and within-category pairs than for between-category pairs. Unfortunately, as participants perceived between-category pairs as rarer events, it was not possible do decide whether the late effect reflected priming in response to identical and within-category pairs or whether it reflected an augmented P300 in response to subjectively infrequent between-category pairs. Taken together, previous research suggests that processing of emotional expressions (i) precedes the initial stages of face identification as indexed by the N170 and (ii), affect the amplitude of the LPP, which may be related to the emotional significance of the information perceived. By applying the priming method, the present study investigated at which point in time event-related potentials are modulated by the repetition of the same expression, and attempted to localize the anatomical sources of these priming effects with the help of a dipole source analysis.

Results 3.1. Performance Only 1.4% of the targets were classified incorrectly. Mean error percentages did not differ between the four conditions. Responses were faster for primed (601 ms) than for unprimed (617 ms) targets, F(1,23)=6.36, p<0.05. Moreover, reaction times to positive targets (601 ms) were shorter than to negative targets (618 ms), F(1,23)=9.56, p<0.01. Additionally, there was a priming by valence interaction, F(1,23)=5.45, p<0.05, which indicated that the behavioral priming effect was only present for positive targets (primed positive: 585 ms, unprimed positive: 617 ms), whereas reaction times to negative targets did not differ (primed negative and unprimed negative: 618 ms). 3.2. Event-related potentials Visual inspection of the ERPs shows that ERPs following primed targets as compared to unprimed targets (Fig. 1) revealed an attenuation of the ERP amplitudes at frontal and posterior sites between 100 and 200 ms, and a reduced amplitude of the parietal LPP. The N170 component following the prime was clearly obtained for all conditions (see Fig. 1), most marked at parieto-occipital electrodes (PO9/PO10). In accordance with earlier research, the N170 peak amplitude at these electrodes did not differ between positive and negative primes, F(1,23)=0.037, p=0.85. The early potentials evoked by the target started at a more positive level than those evoked by the prime and did not show the typical N170 pattern. This was presumably due to the short SOA of 250 ms, which led to an overlap of the prime-related late potentials and the early potentials related to the target. ERPs at selected electrodes in response to targets preceded by faces with the ... Fig. 1. ERPs at selected electrodes in response to targets preceded by faces with the same expression (primed: solid lines) or a different expression (unprimed: dashed lines). EPE: early priming effect, LPE: late priming effect. Figure options The ERP amplitudes in the selected time intervals were statistically analysed as described above. The ANOVA confirmed the visual analysis and revealed a significant early priming effect for the time periods of 100–150 ms, F(30,690)=4.71, p<0.001 and 150–200 ms, F(30,690)=6.48, p<0.001, respectively, and a late priming effect between 500 and 600 ms, F(30,690)=4.28, p<0.01. As illustrated in Fig. 2, there was no valence effect, neither within the early time period (100–150 and 150–200 ms: F's<1) nor within the late period (500–600 ms: F<1). Further, the priming by valence interaction was not significant for any of these time periods (F's<1). ERPs at selected electrodes in response to targets preceded by faces with a ... Fig. 2. ERPs at selected electrodes in response to targets preceded by faces with a happy expression (solid lines) or an angry expression (dotted lines). No significant valence effects were found. Figure options Correspondingly, the analysis for the five selected topographical regions (frontal, parietal, occipital, left temporal, right temporal) revealed a significant region by priming interaction for these three time periods [100–150 ms: F(4,92)=5.82, p<0.01; 150–200 ms: F(4,92)=6.55, p<0.001; 500–600 ms: F(4,92)=5.66, p<0.000]. No other interactions with region were significant. Subsidiary analyses of mean amplitudes for each region revealed for the two early time windows a main effect for priming in the frontal region [100–150 ms: F(1,23)=10.22, p<0.01; 150–200 ms: F(1,23)=22.34, p<0.001] and occipital region [100–150 ms: F(1,23)=21.23, p<0.001; 150–200 ms: F(1,23)=23.00, p<0.01]. For the 500–600-ms period, a priming effect was found at parietal electrode sites [F(1,23)=10.54, p<0.01]. To visualize these priming effects, difference curves were computed for primed minus unprimed conditions, and the amplitude differences were displayed topographically. As shown in Fig. 3, the early priming effect is characterized by a frontocentral positivity and an occipitotemporal negativity for primed faces, which starts between 100 and 150 ms, and increases in the following interval between 150 and 200 ms. The late effect, in turn, consists in a reduced LPP for primed targets, involving mainly parietal electrodes. Topographical maps of ERP differences between primed and unprimed conditions for ... Fig. 3. Topographical maps of ERP differences between primed and unprimed conditions for the two time segments of the early priming effect and for the late priming effect. Maps were obtained by using spherical spline interpolations. They are providing a two-dimensional projection, showing all electrode sites. Figure options 3.3. Dipole analysis To derive the source model for the ERP priming effects, we used difference waveforms (primed minus unprimed), averaged across subjects. For the analysis of the early ERP priming effect, the dipole source analysis was performed on the difference waves for the time interval from 150 to 200 ms relative to target onset, where the difference was maximal. In a first step, a PCA was applied to estimate the minimal number of dipoles active in this interval. Two principal components were needed to explain more than 95% of the variance. Therefore, two dipole pairs were fitted in this interval, accounting for 98.4% of the variance (Fig. 4). The first dipole-pair was localised in the inferior occipitotemporal cortex (x=±43 mm, y=−72 mm, z=−11 mm). The second dipole-pair was localised in a more anterior lateral region, which is likely to correspond to the insular cortex (x=±31 mm, y=8 mm, z=−5 mm). Source analysis of the late priming effect did not provide a sufficiently reliable dipole model and is therefore not reported. Dipole model for the early priming effect with the first dipole pair (1 and 2) ... Fig. 4. Dipole model for the early priming effect with the first dipole pair (1 and 2) being located in the inferior occipital cortex and the second dipole pair (3 and 4) in the insular cortex. Left: source waveforms, with the grey bar indicating the fitting interval (150–200 ms); right: schematic view of dipole source locations.