شواهد الکتروفیزیولوژیک برای پردازش غیرآگاهانه از حالات ترس صورت

Abstract Non-conscious processing of emotionally expressive faces has been found in patients with damage to visual brain areas and has been demonstrated experimentally in healthy controls using visual masking procedures. The time at which this subliminal processing occurs is not known. To address this question, a group of healthy participants performed a fearful face detection task in which backward masked fearful and non-fearful faces were presented at durations ranging from 16 to 266 ms. On the basis of the group's behavioural results, high-density event-related potentials were analysed for subliminal, intermediate and supraliminal presentations. Subliminally presented fearful faces were found to produce a stronger posterior negativity at 170 ms (N170) than non-fearful faces. This increase was also observed for intermediate and supraliminal conditions. A later component, the N2 occurring between 260 and 300 ms, was the earliest component related to stimulus detectability, increasing with target duration and differentiating fearful from non-fearful faces at longer durations of presentation. Source localisation performed on the N170 component showed that fear produced a greater activation of extrastriate visual areas, particularly on the right. Whether they are presented subliminally or supraliminally, fearful faces are processed at an early stage in the stream of visual processing, giving rise to enhanced activation of right extrastriate temporal cortex as early as 170 ms post-stimulus onset.

. Introduction An ongoing debate in the field of emotional face processing and awareness concerns the extent to which facial expressions, in particular fear, may be processed automatically without necessarily reaching conscious awareness. In patients lacking visual cortex, de Gelder et al. (1999) first observed affective blindsight, in which a patient with right hemianopia was capable of guessing the emotional expression on a face at a level above chance, even though the stimuli did not reach consciousness due to damage to the left primary visual cortex. Later, similar results were found in a patient with complete cortical blindness (Pegna et al., 2005). When emotional and neutral expressions were contrasted in an fMRI paradigm in this patient, the former were found to activate the right amygdala despite the bilateral destruction of the primary visual cortex. Similar findings were observed in healthy controls, in whom awareness of emotional faces was eliminated by using backward masking paradigms. Indeed, emotional faces that were masked, and thus not accessible on a conscious level, nevertheless produced patterns of activation that differed from neutral faces, showing that the stimuli were processed despite the fact that masking had disrupted awareness (Morris et al., 1998b, Whalen et al., 1998, Morris et al., 1999 and Liddell et al., 2004). However, the view that emotional faces could be processed without awareness has been called into question. For example, Pessoa et al. (2006) presented masked fearful faces at 33 or 67 ms durations and assessed detection using an objective measure taken from signal detection theory, in addition to a confidence rating. The authors observed that, independently of presentation times, when fearful faces were not detected, no activation was found in the amygdala or in the fusiform gyrus. By contrast, when subjects reported seeing a fearful face (whether or not one had actually appeared), an increase in activation was observed in both the amygdala and fusiform gyrus. Thus, fearful faces may not actually be processed automatically and non-consciously. The authors argue that the divergence with previous results stems in part from the fact that the degree of awareness might not be sufficiently controlled. Indeed, in a separate behavioural study, the authors (Pessoa et al., 2005) found a variable threshold for conscious perception of a masked stimulus across subjects. These findings demonstrate the importance of using objective measures of detection when studying subliminal processing. The extent to which fearful faces might be processed non-consciously remains therefore to be established. Moreover, special care must be taken in ensuring that the stimuli are consciously undetectable. In addition, studies using event-related potential (ERP) measures, which possess a high temporal resolution, are necessary to investigate whether the processing of emotional expressions occurs early or late in time. Until now, most ERP studies of emotional face processing have used tasks in which the stimuli are largely visible on a conscious level. Faces are generally claimed to give rise to a specific response, the N170 (e.g., see Bentin et al., 1996 and Itier and Taylor, 2004) but it was previously thought that this response was insensitive to emotional expressions (Munte et al., 1998, Bobes et al., 2000, Krolak-Salmon et al., 2001, Eimer and Holmes, 2002 and Herrmann et al., 2002; see also Eimer and Holmes, 2007 and Vuilleumier and Pourtois, 2007 for recent reviews). Recently however, evidence has begun emerging that the N170 might nevertheless be modulated by emotional expression (Batty and Taylor, 2003, Miyoshi et al., 2004, Stekelenburg and de Gelder, 2004, Blau et al., 2007 and Hendriks et al., 2007), although it is not known whether this modulation can be extended to non-conscious processing. Indeed, little has been done in the way of ERPs and subliminal processing. Liddell et al. (2004) investigated the ERP responses to subliminal and supraliminal presentations of fearful and neutral faces and found that fear enhanced an N2 component on fronto-central sites, while supraliminal presentations affected later (N4 and late P3) components. This result therefore upheld the view that the earliest difference for non-conscious processing of fear occurred after 200 ms. However very recently, Kiss and Eimer (2008) and Eimer et al. (2008) reported earlier effects of subliminally-presented fearful faces, occurring between 140 and 180 ms over anterior sites. In these studies, no effect was seen over posterior electrodes that could suggest an N170 modulation. We therefore investigated the processing of masked fearful faces using an ERP paradigm, in which stimulus duration was varied parametrically and awareness established using the behavioural d′ measure from signal detection theory ( Green and Swets, 1966 and Macmillan and Creelman, 1991). The aim of the experiment was to determine whether electrical brain responses, in particular the N170, N4 and the late positive component or late P3, varied for fearful faces at stimulus durations that were below the threshold of conscious awareness.

. Results 3.1. Behaviour For each of the 5 durations, the mean percentage of correct hits for fearful faces (± standard deviation – SD –) was, from shortest to longest, 52% (± 5), 63% (± 8), 74% (± 9), 86% (± 9) and 96% (± 3). In this paradigm, performance was established on the basis of the d′ detection measure from signal detection theory ( Green and Swets, 1966 and Macmillan and Creelman, 1991). This detection (or sensitivity) was computed for each subject using the hit and false alarm rates. The resulting d' values are shown in Fig. 2. Mean detection rate. The average d′ values (black squares) computed from each ... Fig. 2. Mean detection rate. The average d′ values (black squares) computed from each subjects' hit and false alarm rate is represented on the y axis as a function of stimulus duration represented on the x axis. The boxes around the means indicate the standard error and the whiskers represent the standard deviation. Figure options At 16 ms, subjects' d′ values were normally distributed with a mean of 0.08 and a SD of 0.52. This value was not statistically different from zero (expected in case of random performance; t(17) = .65; p > .05). It should be noted that, on an individual level, none of the participants obtained a hit rate that was different from chance (for all subjects' ps > .05, binomial two-tailed test). Thus, presentation could be considered subliminal at 16 ms for all participants with respect to the emotional faces (an observation of particular importance to our study as this was a prerequisite for further analysis). In the 33 ms condition, d′ rose slightly to 0.9, along with a small increase in SD to 0.69. This d′ value proved already to be significantly above zero (t(17) = 5.53; p < .01) suggesting a non-random response mode on average. However, a detailed look at the individual hit rates, revealed that 10/18 subjects performed above chance (p < .05, two-tailed binomial test), with the remaining situated below the .05 threshold. Thus, a certain amount of heterogeneity was observed among participants in the 33 ms condition. Intermediate levels of detection were obtained in the 66 ms condition, (mean d′ = 1.46, SD = 0.65), while at 133 ms a high d′ was observed with a value at 2.76 although a great amount of variability was observed across subjects resulting in a high SD (SD = 1.31) with 3 participants obtaining a value above 4 and 6 others scoring between 1 and 2. Finally, at 266 ms, mean d′ was at 3.87 (SD = 0.99), with all subjects above 2.5. A repeated-measures analysis of variance using duration as a within-subjects factor confirmed that on average, d′ increased significantly across durations (F(4,68) = 70.2; p < .0001). In order to ensure that the ERPs would correspond to conditions that were homogenous in perceptual terms, and in view of the fact that about half the subjects were above and the other half below chance in the 33 ms presentation condition, the 33 ms presentation times were removed. Along similar lines, large variations in performance were observed in the 133 ms condition showing up as a greater standard deviation. This condition was therefore also dismissed in the ERP analysis. The electrophysiological analysis therefore focused on durations of presentation of 16 (condition C1), 66 (condition C2) and 266 (condition C3) ms which reflected subliminal, intermediate and supraliminal levels of presentation at both the group and individual levels, while maintaining identical stimulus parameters within each condition. 3.2. Event-related potentials The aim of our study being to characterise the earliest ERP differences induced by facial expressions, a preliminary analysis consisted in a time-point by time-point t-test on each of the recording sites comparing fearful vs. non-fearful expressions in the 3 conditions (C1, C2 and C3). This analysis (not detailed) indicated that the first significant differences involving more than 3 neighbouring electrodes appeared at around 150 ms onwards and were mainly found on posterior sites. Nevertheless, we examined the effect of our experimental manipulations specifically on the P100 amplitude. The P100 mean amplitudes were computed over the 6 left and 6 right temporal and occipital leads. The mean values were entered into a 3 (durations) × 2 (expressions) × 2 (hemispheres) ANOVA. No effect of expression was observed and only a hemisphere effect was observed due to the fact that the P100 amplitude, as seen in Fig. 3, was greater on the right than on left sites (F(1, 17) = 14.9, p < .005, mean amplitude = 5.19 vs. 3.66 respectively). N170 component. A–C: ERP response of electrodes T5 (left column) and T6 (right ... Fig. 3. N170 component. A–C: ERP response of electrodes T5 (left column) and T6 (right column) in response to fearful (black) and non-fearful faces (red). The illustrations depict the response for durations of presentation of 16 ms (A), 66 ms (B) and 266 ms (C). The centre column shows the mean topography of the activity between 160 and 180 ms at each of the durations in response to fearful and non-fearful expressions along with the difference map between the two. D: The mean response (in µV) between 160 and 180 ms over the 6 left and 6 right posterior electrodes (illustrated in the inset) are plotted respectively on the left and right graphs as a function of stimulus duration for fearful (black lines) and non-fearful faces (red lines). Central inset in D shows (above) the colour code for the scalp topographies in A–C, and (below) the location of electrodes T5 and T6 (yellow) that are represented in A–C, as well as that of the 12 leads (red and yellow) used to compute the statistics and whose means are displayed in the left and right graphs. Figure options 3.2.1. N170 component Visual observation of the N170 peak in Fig. 3A–C shows that its amplitude is more marked for fearful than non-fearful expressions for all 3 durations. Mean amplitude was computed between 160 ms and 180 ms and averaged for 6 left and 6 right temporo-occipital leads (including electrodes T5 and T6, see inset in D for electrode layout) where the response was maximal. As shown in Fig. 3D illustrating the mean amplitude on left and right sites for fear and non-fear in C1, C2 and C3, a greater negative amplitude is found for fearful expressions, particularly over the right electrodes. A 3 × 2 × 2 ANOVA using condition, expression and hemisphere as within-subjects factors showed a significant interaction between facial expression and hemisphere (F(1,17) = 6.78, p < .05) due to the fact that fearful expressions affected the right leads. In addition, a main effect of emotional expression was also found (F(1,17) = 11.3, p < .005) as the N170 for fearful faces was globally more negative than for non-fearful ones. In order to further reveal the effect of fearful expressions on the N170 at each presentation time, post-hoc tests were used to compare responses for fearful and non-fearful faces on each side and at each stimulus duration. LSD post-hoc tests showed no significant difference between fearful and non-fearful faces on the left for any of the durations. On the right, the difference between fear and non-fear was significant at 16 ms (p < .01, mean difference = − 0.51 μV), failed to reach significance at 66 ms (p > .05) and was highly significant at 266 ms (p < .0005, mean difference = − 0.70 μV). This finding therefore shows that fearful expressions produce a significant difference in N170 over right posterior electrodes, even when the subjects' behavioural performance is equal to 0. 3.2.2. N2 component At longer durations of presentation, a later difference appears in the ERP response between fear and non-fear, notably at the N2 component (see superimposition in Fig. 3 and Fig. 4). This response, found both on anterior and posterior sites, differentiates fearful and non-fearful faces, but appears to show an even greater modulation with increasing stimulus visibility. To illustrate this effect, the mean response of a group of 4 anterior and 4 posterior electrodes on left and right sites are presented in Fig. 4A for fearful (solid lines) and non-fearful faces dotted lines with the 3 durations (C1, C2 and C3) superimposed (in respectively blue, green and red). The peak of the N2 component is indicated by the arrow (Fig. 4A, lower right panel). N2 component. A. Average ERP response of 4 left anterior (top left panel) and ... Fig. 4. N2 component. A. Average ERP response of 4 left anterior (top left panel) and right anterior electrodes (top right panel), and 4 left posterior (bottom left panel) and right posterior electrodes (bottom right panel) for fearful (solid lines) and non-fearful faces (dotted lines). The 3 durations of presentation are superimposed and represented by blue (16 ms), green (66 ms) and red (266 ms) lines (see inset). The central inset shows the position of the electrodes selected for analysis in grey. B. The mean responses (in µV) between 260 and 300 ms are plotted for the anterior sites (two graphs on the left) and the posterior sites (two graphs on the left) as a function of duration of presentation. In each group, the graph on the left represents the group of electrodes over the left hemisphere. Fearful faces are represented by solid and non-fearful faces by dotted lines. Figure options N2 modulation was measured by computing the mean amplitude at 4 electrodes situated in the anterior left, anterior right, posterior left and posterior right sites in the 260–300 ms time window for fearful and non-fearful faces in C1, C2 and C3. These values are represented graphically in Fig. 4B. At the anterior sites, an increase in positivity can be seen over the left and right leads with increasing durations of presentation. In parallel, the posterior sites show an increase in negativity. To verify the statistical validity of these effects and their modulation by facial expression, a 3 × 2 × 2 ANOVA was performed using condition, expression and hemisphere as repeated measures, first on posterior, then on anterior sites. On posterior sites, a highly significant main effect was found for the duration condition (F(2, 34) = 24.9, p < .0001) due to an increase in negativity with longer presentations. Post-hoc LSD tests revealed that amplitudes differed across all 3 conditions (C1 vs. C2: p < .05; C1 vs. C3: p < .0001; C2 vs. C3: p < .001). Emotional expression also gave rise to a significant main effect (F(1, 17) = 5.6, p < .05) due to the fact that on average, responses to fear were more negative than responses to non-fear. However, two second-order interactions also proved significant: duration × hemisphere (F(2, 34) = 8.0, p < .005) and expression × hemisphere (F(1, 17) = 6.5, p < .05). Post-hoc LSD tests revealed that the interaction was in fact due to the absence of any significant effect of expression over left posterior sites at any of the durations, while changes occurred over the right leads. Indeed, on the right sites, mean amplitudes for fearful and non-fearful faces were not different in C1 (p = .12), appearing in C2 with a mean difference of − 0.52 μV (p < .003,) and increasing to − 0.64 μV at C3 (p < .0004; see Fig. 4B, right graph). At the anterior sites, the 3 × 2 × 2 ANOVA also revealed a highly significant main effect of duration of presentation (F(2, 34) = 15.2, p < .0001) due to the fact that the response became more positive with greater duration ( Fig. 4B, left graphs). Again, post-hoc LSD tests showed that amplitudes differed across all 3 conditions (C1 vs. C2: p = .04; C1 vs. C3: p < .0001; C2 vs. C3: p = .002). A hemisphere effect (F(1, 17) = 4.6, p < .05) was also observed due to globally greater responses on left than on right sites for all durations. A small interaction was also found between expression and hemisphere (F(1, 17) = 4.8, p < .05) due to the fact that the response to fear and non-fear did not differ at right sites (p = .49) but differed globally on the left (p < .005). Post-hoc LSD tests showed that the amplitude differences between fear and non-fear at these left sites were not significant at 16 ms (p > .05, mean difference = 0.16 μV), but reached the statistical level of significance with differences of 0.32 μV (p < .05) and 0.78 μV (p < .00001) at C2 and C3 respectively (see Fig. 4B, extreme left graph). Thus, a greater negativity was observed over the posterior leads, as well as a greater positivity over anterior sites at longer durations of stimulus presentation. The posterior negativity was significantly more marked on the right for fearful faces compared to non-fearful expressions at intermediate and long durations while the converse effect of an increase in positivity was found for fearful faces on the left anterior sites. 3.2.3. N4-LPC components Subsequent to the N2 component, a vertex negative deflection was seen at around 400 ms (N4 component), followed by a late positive component (LPC) which was particularly obvious for the stimuli presented for 266 ms. Fig. 5 shows these components for fear (Fig. 5A1) and non-fear (Fig. 5A2) in the 3 conditions, averaged across the 6 central electrodes illustrated in the middle inset of the figure. Fig. 5B illustrates the mean topographic maps computed during the N4 and LPC periods at each duration of presentation for fear (Fig. 5B1) and non-fear (Fig. 5B2). N4 and LPC. A: Average response for fearful faces (A1, solid lines) and ... Fig. 5. N4 and LPC. A: Average response for fearful faces (A1, solid lines) and non-fearful faces (A2, dotted lines) of the 6 central electrodes that are highlighted in grey in the central inset. The 3 durations of presentation are superimposed and shown respectively in blue (16 ms), green (66 ms) and red (266 ms). B: Mean scalp topographies between 350 ms and 400 ms for the N4 and between 450 ms and 650 ms for the LPC at the 3 durations of presentation for fearful (B1) and non-fearful faces (B2). C: Mean amplitude of the N4 computed between 350 and 400 ms over the 6 central electrodes (shown in inset in A) for the 3 conditions of presentation. Fearful faces are displayed using solid lines and full circles and non-fearful faces with dotted lines and open circles. D: Mean amplitude of the LPC between 450 and 650 ms over the 6 central electrodes (inset A) for the 3 conditions of presentation. Again, fearful faces are represented by black solid lines and full circles, and non-fearful faces by red lines and open circles. Figure options For statistical testing, the mean potential over the 350–400 ms time period of these 6 central electrodes were computed for the N4 component. The resulting values are shown in Fig. 5C. The individual values were subjected to a 3 (condition) × 2 (expression) ANOVA. Neither the condition × expression interaction (F(2, 34) = 2.17, p > .05), nor emotional expression alone (F(1, 17) = .03, p > .05) were found to be significant. Only stimulus duration affected the N4 component to a significant degree (F(2, 34) = 23.702, p < .00001). A post-hoc Fisher LSD test showed that all conditions differed between each other (all ps < .01). In order to assess the differences for the LPC, the mean potential of the same 6 electrodes were averaged over the 450–650 ms time window and the resulting values (represented in Fig. 5D) were compared statistically using a 3 (condition) × 2 (expression) ANOVA. For this later component, condition was again highly significant (F(2, 34) = 9.6, p < .0005) while expression alone was not (F(1, 17) = .14, p > .05). However, condition × expression interacted significantly (F(2, 34) = 4.0, p < .05). Post-hoc comparisons using the Fisher LSD test showed that in C3, fear and non-fear differed significantly (p < .01), while the two expressions yielded no difference in LPC amplitude in C1 and C2 (both ps > .05). 3.3. Source localisation Since only the effect of expression was significant in the N170 time window, source localisation was performed on the N170 response to fear and non-fear. The generators were therefore estimated on the basis of the N170 topographies obtained separately for fear and non-fear in the 160–180 ms time period of the 3 conditions together. Fig. 6A shows the distribution of active areas at this period. Both expressions showed similar patterns, although fear appeared to produce an overall greater level of activation than non-fear. Superior, middle and inferior temporal areas were activated bilaterally but more markedly on the right. In addition, slightly weaker activity was found in medial temporal areas, again more marked on the right, compatible with amygdala regions. Primary visual cortex was also found to be bilaterally active. Source localisation. A. LORETA solutions for the N170 topographies (average ... Fig. 6. Source localisation. A. LORETA solutions for the N170 topographies (average between 160 ms and 180 ms) of fearful (top) and non-fearful faces (bottom) faces (16, 66 and 266 ms conditions together). Inset below shows the level of the horizontal sections in the average MNI brain (Montreal, Canada) given in A and B. The strength of the source is represented by the colour code given on the right. B. Results of a paired t-test comparing the solutions for fear and non-fear over subjects. Significant pixels at p < .01 are shown in red. Figure options In order to establish which regions differed significantly when comparing the 2 expressions, the LORETA algorithm was computed on the mean 160–180 ms topographies of each participant. The N170-induced activation for fear was then compared to the N170 for non-fear using a paired t-test on the 3005 pixels included in the head model. Fig. 6B shows the pixels that are significant at p < .01. Areas that were more significantly activated by fearful expressions involved the right inferior, middle and superior temporal gyri, extending posteriorly into the extrastriate occipito-temporal visual areas both medially and laterally, as well as into the ventral temporal areas.