Activation of the amygdala and anterior cingulate during nonconscious processing of sad versus happy faces

Transcription

1 Activation of the amygdala and anterior during nonconscious processing of sad versus happy faces William D.S. Killgore a,b, * and Deborah A. Yurgelun-Todd a a Cognitive Neuroimaging Laboratory, McLean Hospital/Harvard Medical School, Belmont, MA 02478, USA b Division of Neuropsychiatry, Walter Reed Army Institute of Research (WRAIR), Silver Spring, MD 20910, USA Received 29 April 2003; revised 24 December 2003; accepted 30 December NeuroImage 21 (2004) Previous functional neuroimaging studies have demonstrated that the amygdala activates in response to fearful faces presented below the threshold of conscious visual perception. Using a backward masking procedure similar to that of previous studies, we used functional magnetic resonance imaging (fmri) to study the amygdala and anterior gyrus during preattentive presentations of sad and happy facial affect. Twelve healthy adult females underwent blood oxygen level dependent (BOLD) fmri while viewing sad and happy faces, each presented for 20 ms and masked immediately by a neutral face for 100 ms. Masked happy faces were associated with significant bilateral activation within the anterior gyrus and amygdala, whereas masked sadness yielded only limited activation within the left anterior gyrus. In a direct comparison, masked happy faces yielded significantly greater activation in the anterior and amygdala relative to identically masked sad faces. Conjunction analysis showed that masked affect perception, regardless of emotional valence, was associated with greater activation within the left amygdala and left anterior. Findings suggest that the amygdala and anterior are important components of a network involved in detecting and discriminating affective information presented below the normal threshold of conscious visual perception. D 2004 Elsevier Inc. All rights reserved. Keywords: fmri; Neuroimaging; Faces; Emotion; Affect; Amygdala; Sadness; Happiness; Limbic system; Backward masking Neuroimaging and brain lesion studies suggest the existence of two discrete neural systems for emotional perception, one operating at the explicit conscious level, and another operating implicitly, below the level of conscious awareness (de Gelder et al., 1999; Driver et al., 2001; LeDoux, 1996; Morris et al., 1998; Whalen et al., 1998). LeDoux (1996) proposed that emotional processing involves simultaneous activation of these separate neural pathways, each providing different levels of analysis to permit rapid * Corresponding author. Division of Neuropsychiatry, Department of Behavioral Biology, Walter Reed Army Institute of Research (WRAIR), 503 Robert Grant Avenue, Silver Spring, MD Fax: address: william.d.killgore@us.army.mil (W.D.S. Killgore). Available online on ScienceDirect ( responsiveness and maximal flexibility to changing environmental threats. Along one pathway, information about biologically relevant stimuli are first transmitted from primary sensory cortices via the thalamus directly to the amygdala to provide a rapid though unrefined assessment of the potential emotional and motivational significance of the stimulus. Simultaneously, a second relatively longer pathway sends information through a series of higher-order cortical processing areas to provide a more extensive analysis of the situation and generate potential response alternatives (LeDoux, 1996). It is hypothesized that the former pathway operates subcortically, below the threshold of conscious awareness, whereas the latter involves explicit cognitive operations within the cortex that are more directly accessible through conscious attention. The existence of two distinct neural pathways for processing emotional information is consistent with long-held notions that the brain may indeed process some aspects of experience at a nonconscious level (Freud, 1915). A particularly effective method for investigating nonconscious emotional processing involves a procedure known as backward stimulus masking. With this technique, a visual stimulus is presented very rapidly (e.g., <40 ms) and immediately replaced by an alternate visual stimulus for a slightly longer duration. Under such conditions, subjects are unaware of the initial target stimulus and only report perceiving the second masking stimulus (Esteves and Ohman, 1993; Esteves et al., 1994; Morris et al., 1999; Ohman and Soares, 1993; Soares and Ohman, 1993a,b). This technique has been used to demonstrate evoked physiological responses to presentations of emotional stimuli that are not consciously perceived. For instance, target facial expressions that have been previously paired with an aversive stimulus can elicit conditioned skin conductance responses even when presented below the threshold of conscious perception (Esteves et al., 1994). Similarly, distinct patterns of facial muscle activity that mimic the emotion depicted in a masked target photograph can be measured in subjects viewing such stimuli, despite an absence of conscious awareness of the emotional stimulus (Dimberg et al., 2000). Whalen et al. (1998) used backward masking of happy and fearful facial expressions during functional magnetic resonance imaging (fmri) and found increased activity within the amygdala in response to masked faces expressing fear and decreased activity to similarly masked faces showing happiness. Others (Morris et al., 1998) have found a dissociation between conscious and noncon /$ - see front matter D 2004 Elsevier Inc. All rights reserved. doi: /j.neuroimage

2 1216 W.D.S. Killgore, D.A. Yurgelun-Todd / NeuroImage 21 (2004) scious perception of emotional facial expressions. Conscious perception of negatively valenced face stimuli has been associated with greater left amygdala activation whereas nonconscious perception of the same stimuli appears to be associated with right amygdala activation, suggesting parallel visual pathways specialized for processing affective stimuli at different levels of conscious awareness (Morris et al., 1999). Because previous studies have only explored responses to a few categories of nonconsciously presented affect (i.e., fear, anger, happiness), it is not clear whether the right amygdala is responsive to negative affect in general, or whether the findings are specific only to aggressive and fear-related affects. To clarify the specificity of the amygdala to nonconscious processing, the present study used fmri to study amygdala activity during masked presentations of sad versus happy facial expressions. Given previous evidence of right amygdala activation during nonconscious processing of facial emotion (Morris et al., 1998, 1999), we predicted significant right lateralized activation of the amygdala during presentations of the masked face stimuli. We were also interested in the role of the gyrus during nonconscious face perception, given prior evidence suggesting involvement of the during conscious perception of sad (Blair et al., 1999) and happy (Dolan et al., 1996; Phillips et al., 1998) facial expressions of emotion. Methods Subjects Twelve healthy right-handed female adults, ranging in age from 21 to 28 years (M = 23.7, SD = 2.1), underwent functional magnetic resonance imaging (fmri) using blood oxygenation level dependent (BOLD) contrast. Participants, recruited from the local community and the staff of McLean Hospital, were without history of psychiatric or neurological illness according to a structured clinical interview, and all had normal or corrected normal vision. Subjects were naïve to the face stimuli and hypotheses of the study. All subjects provided written informed consent before participation and were compensated for their time. Stimulation paradigms Each subject participated in two masked facial affect tasks conducted as separate scanning runs. The presentation order of masked sad and masked happy tasks was counterbalanced across subjects. Participants were unaware of the backward masking nature of the study, and were informed only that they would see a series of briefly presented photographs of faces and make a gender discrimination for each face by pressing a small hand-held key pad. Each facial affect task was identical in design except for the primary emotion displayed by the target faces (i.e., sadness versus happiness) and the order of presentation of the stimuli, which was reversed across the two tasks. Each run lasted for 5 min, and presented 25 alternating 12-s epochs of fully masked faces, unilaterally presented (left or right) partially masked faces, or a resting baseline condition consisting of a single fixation cross. The unilateral presentations are not the focus of the present study and are not included in the current data analysis, but will be presented in a subsequent manuscript. Face stimuli were obtained from the picture set developed by the Neuropsychiatry Section of the University of Pennsylvania Medical Center (Erwin et al., 1992) and consisted of black and white photographs of four males and four females posing each of three different emotional states (i.e., happy, sad, neutral), yielding a total of 24 expressions. The same eight posers were used for all stimuli. Emotional faces were used as target stimuli and matching neutral faces from each subject were used as masking stimuli. Each masked stimulus trial consisted of an emotional target face presented for 20 ms, followed immediately by a neutral masking face of the same poser displayed for 100 ms (see Fig. 1). Trials were separated by a 3-s interstimulus interval. To control for possible morphological differences in the asymmetry of the facial expressions (Borod et al., 1997), we presented each target face at two different times during the run, once in its normal orientation and again as a mirror-reversed image. Stimuli from the eight posers were pseudorandomly distributed throughout the entire run, with the stipulation that an equal number of male and female subjects appeared across every 12 presentations and that each half of the scanning run included an equal number of normal and mirror-reversed images. The stimuli were presented from a Power Macintosh G3 computer using Psyscope software (Macwhinney et al., 1997) and were backprojected onto a screen placed at the foot of the scanning bed. The stimuli were easily viewed via a mirror mounted on the head coil. Immediately upon completion of the functional scan, subjects were presented with a posttest that included all 24 facial expression stimuli and were asked to indicate for each expression whether it had been seen during the study. Consistent with previous studies using similar masked affect paradigms, subjects failed to recognize masked affective stimuli while easily identifying the neutral masks ( F[2,22] = 53.30, P < ). Specifically, the neutral masks were correctly identified with 82% accuracy, significantly above chance expectations (t[11] = 4.76, P = ). In contrast, masked happy faces were recognized only 9% of the time and masked sad faces were identified 16% of the time, findings which were significantly below chance (happy t[11] = 11.66, P = ; sad t[11] = 5.25, P = ), suggesting that subjects were generally unaware of having been exposed to these stimuli. Neuroimaging methods Data were acquired on a 1.5-T GE LX MRI scanner equipped with a quadrature RF head coil (TR = 3 s, TE = 40 ms, flip angle = 90j). Comfortable placement of foam padding and a tape strap across the forehead were used to minimize head motion. Echoplanar images were collected over 20 coronal slices (7 mm, 1-mm gap), with a 20-cm field of view and a acquisition matrix, providing an in-plane resolution of mm. Data were collected during two 100-scan (i.e., 300 s) runs. Three dummy images were taken at the outset of each functional scan to reduce nonsteady state effects. To aid in anatomical localization of regions of activation, matched T1-weighted high-resolution images were obtained for every subject at the outset of the scanning session. Image processing and analysis Images were realigned and corrected for motion using an intrarun realignment algorithm in SPM99. Functional MRI data were convolved into three-dimensional space and smoothed using an isotropic Gaussian kernel (full width half maximum [FWHM] = 10 mm), and resliced to mm within MNI space using

3 W.D.S. Killgore, D.A. Yurgelun-Todd / NeuroImage 21 (2004) Fig. 1. Each stimulus trial consisted of two rapidly presented stimuli: a target face depicting either a sad or a happy emotional expression from one of eight posers and a mask face consisting of a photograph of the same poser expressing a neutral emotion. During each trial, the target face was presented visually for 20 ms and was immediately replaced by the neutral mask photograph for 100 ms. Each trial was separated by a 3-s interstimulus interval. Due to the brief duration of the target and its temporal proximity to the lengthier mask presentation, subjects lacked explicit awareness of the target stimulus, while consistently reporting perceptual awareness of the mask. Fig. 2. The boundaries for the ROI analysis were based on the anatomical atlas of Tzourio-Mazoyer et al. (2002). (A) Coronal, axial, and sagittal views showing the ROI boundaries for the left (white) and right (green) anterior gyri. (B) Coronal, axial, and sagittal views showing the ROI boundaries for the left (white) and right (green) amygdala.

4 1218 W.D.S. Killgore, D.A. Yurgelun-Todd / NeuroImage 21 (2004) Fig. 3. Region of interest SPMs showing clusters (k z 10) of significant activation for each of the statistical contrasts. The top row presents coronal views between y = 28 and y = 38 at the location of the anterior gyrus (BA 24/32), whereas the bottom row shows a coronal view of the amygdala at y = 2. Activation maps are masked to only display activation within the amygdala and anterior gyrus regions. Masked happy faces were associated with significant bilateral activation within the anterior gyrus and amygdala. Masked sad faces were associated with significant activation within the left anterior gyrus (BA 32), but no significant activation within either amygdala. The conjunction of the group activations for masked sadness and happiness yielded significant activation within the left anterior gyrus and left amygdala. A paired t test revealed significant activation within the bilateral anterior gyrus and right amygdala during happiness relative to sadness. In contrast, the sad happy comparison did not yield any suprathreshold activation in hypothesized regions of interest. sinc interpolation. Data were filtered for low-frequency confounds using a high pass filter set at the SPM99 default of twice the longest interval between two occurrences of the most frequently occurring stimulus event (i.e., 78 s). To account for temporal autocorrelations, data were also filtered using a low pass filter based on the hemodynamic response function. For each subject, a statistical parametric map was generated using the general linear model within SPM99 (Friston et al., 1995a,b), employing a design matrix that treated each masked stimulation period as a separate event. To determine the functional activation due to the nonconscious perception of specific affects, linear contrasts were constructed comparing masked happy and masked sad conditions to the resting fixation baseline condition. While data were analyzed in a whole-brain fashion using SPM99, we constrained our study to specific hypotheses concerning the amygdala and anterior gyrus during nonconscious affect perception. Specific hypotheses for activation within these regions were tested using a region of interest (ROI) approach on the obtained SPMs. These regions were specified using an established anatomical atlas (Tzourio-Mazoyer et al., 2002) to define and create anatomically based ROIs using the Masks for Regions of Interest Analysis (MARINA) software program (Bertram Walter Bender Institute of Neuroimaging, University of Giessen, Germany). According to the atlas, the ROI for the anterior extended from the rostral boundary of the para sulcus to terminate caudally at the white matter of the Fig. 4. Whole-brain corrected ( P < 0.05) group SPMs showing clusters (k z 10) of significant activation for happy, sad, and conjunction analyses. Coronal views are shown ranging from y = 60 to y = 70 at the location of the visual association cortex (BA 18/19) and cerebellum. Following whole-brain correction for multiple comparisons, significant activation was evident within the right inferior temporal gyrus and left inferior occipital gyrus during the masked happiness condition, and right lingual gyrus during the masked sadness condition. Conjunction analysis yielded significant activation within the left and right cerebellar crus, as well as the inferior temporal gyrus and superior occipital gyrus on the right.

5 W.D.S. Killgore, D.A. Yurgelun-Todd / NeuroImage 21 (2004) corpus collosum (see Fig. 2). The amygdala included the subcortical gray nuclei at the rostral boundary of the hippocampus and caudal boundary of the uncus (see Fig. 2). Given that we were interested in four small a priori defined regions, activation within these ROIs was tested at P < 0.05, uncorrected. To facilitate the generation of future hypotheses, we also report data for activation clusters that fell outside of the hypothesized ROIs. Clusters of activation falling outside of the a priori hypothesized ROIs were statistically corrected for multiple comparisons across the entire brain, P < Statistical analyses followed three stages: (1) single condition group SPMs contrasting the affect condition with its resting fixation baseline were created for the masked happy and sad face conditions separately; (2) a conjunction analysis was performed to determine the regions that were significantly activated during both the masked happy and masked sad conditions; (3) contrasts were then conducted between the masked happy and masked sad conditions to isolate the cerebral regions that were uniquely involved in processing nonconsciously perceived stimuli of each affective valence. For visualization, SPM {t} maps were displayed on an average template brain in the standardized coordinate space of the Montreal Neurological Institute (MNI). Results Masked happiness Presentation of happy faces masked by neutral faces yielded significant activation within the left ( P < 0.005) and right ( P < 0.05) amygdala (see Fig. 3 and Table 1). Several clusters of suprathreshold activation were also observed within the left ( P < 0.01) and right ( P < 0.01) anterior gyrus. Masked sadness Presentations of masked sad faces were not associated with any suprathreshold clusters of activation within either the left or right amygdala relative to the fixation baseline (see Fig. 3 and Table 1). There was, however, a cluster of significant ( P < 0.05) activation within the left anterior gyrus. No clusters of suprathreshold activation were evident in the right anterior ROI. Masked happy and masked sad conjunction analysis To test for brain activation common to both masked happy and masked sad faces, the two experimental conditions were entered into a conjunction analysis using SPM99. As evident in Fig. 3, masked presentations of facial affect, regardless of valence condition, resulted in significant suprathreshold activation within the left amygdala ( P < 0.01) and the left anterior gyrus ( P < 0.005, BA 32). No significant regions of activation were detected in the right amygdala or right anterior gyrus ROIs. Table 1 presents standard coordinates and z scores for significant local maxima within the hypothesized ROIs. Affect condition comparisons To test whether masked happy and sad face presentations were associated with distinct patterns of activation, the two conditions Table 1 Activation within hypothesized regions of interest for each affect condition Regions of activation Brodmann s area Voxels x y z z score Masked happiness L. amygdala *** R. amygdala * L. anterior ** * * R. anterior ** * Masked sadness L. amygdala R. amygdala L. anterior * R. anterior Conjunction of masked sadness and masked happiness L. amygdala ** R. amygdala L. anterior *** ** R. anterior Masked happiness masked sadness L. amygdala R. amygdala * L. anterior 23/ ** * * * R. anterior **** ** Masked sadness-masked happiness L. amygdala R. amygdala L. anterior R. anterior Note. L = left hemisphere, R = right hemisphere. Atlas coordinates are from the MNI standard atlas, such that x reflects the distance (mm) to the right or left of midline, y reflects the distance anterior or posterior to the anterior commissure, and z reflects the distance superior or inferior to the horizontal plane through the AC PC line. * P values reflect small volume correction for multiple comparisons: P < ** P values reflect small volume correction for multiple comparisons: P < *** P values reflect small volume correction for multiple comparisons: P < **** P values reflect small volume correction for multiple comparisons: P < were subjected to a second-level random effects paired t test analysis in SPM99. During the masked happiness vs. masked sadness comparison (i.e., masked happiness masked sadness), there was a small cluster of significantly greater activation within

6 1220 W.D.S. Killgore, D.A. Yurgelun-Todd / NeuroImage 21 (2004) the right amygdala ( P < 0.05), whereas the left amygdala showed no clusters of significant activation. Furthermore, during the masked happiness condition, there were several clusters of significantly greater activation within both the left ( P < 0.01; BA 10/23/ 24) and right ( P < 0.001; BA 24) anterior gyri relative to the masked sadness condition. In contrast, when compared to the masked happiness condition (i.e., masked sadness masked happiness), there were no clusters of suprathreshold voxels within any of the ROIs (see Fig. 3 and Table 1). Nonhypothesized regions To report unbiased results for nonhypothesized regions and to provide data that might generate further hypotheses regarding regions important in nonconscious affect processing, we also conducted a whole-brain SPM analysis with the voxelwise height threshold set for P < and whole-brain correction for multiple comparisons at P < Table 2 summarizes the regions of activation surviving this correction. For individual masked affect conditions, masked happy faces were associated with a significant cluster of activation within the right inferior temporal gyrus and left inferior occipital gyrus, whereas masked sadness was associated with a significant cluster of activation within the right lingual gyrus (see Fig. 4). Conjunction analysis to determine regions of activation common to both masked affect Table 2 Nonhypothesized activation clusters for each affect condition after correction for whole-brain volume, P < 0.05 Regions of activation Masked happiness R. infer. temporal gyrus L. infer. occipital gyrus Brodmann s area Voxels x y z z score **** *** Masked sadness R. lingual gyrus * Conjunction of masked happiness and masked sadness L. cerebellum crus *** R. cerebellum crus **** R. inferior temp **** gyrus ** R. superior occip. gyrus * Masked happiness masked sadness Masked sadness masked happiness Note. L = left hemisphere, R = right hemisphere. Atlas coordinates are from the MNI standard atlas, such that x reflects the distance (mm) to the right or left of midline, y reflects the distance anterior or posterior to the anterior commissure, and z reflects the distance superior or inferior to the horizontal plane through the AC PC line. * P values corrected for total brain volume: P < **P values corrected for total brain volume: P < *** P values corrected for total brain volume: P < **** P values corrected for total brain volume: P < conditions yielded significant activation within the left and right cerebellar crus, right inferior temporal gyrus, and right occipital gyrus. At this level of correction, no regions survived direct paired t test comparisons between the masked happy and sad conditions. Discussion Subjects were presented with photographs of happy and sad faces in a manner that minimized or eliminated conscious visual perception of the presented affect. Despite being presented below the threshold of conscious awareness, these unseen emotional stimuli yielded specific patterns of activation within the amygdala and anterior gyrus that differed as a function of the emotional valence of the masked facial expression. Masked happiness was associated with significant bilateral activation of the amygdala and the anterior gyrus (BA 24/32), whereas masked sadness yielded significant activation only within the left and no significant activation within either amygdala. When masked affect was analyzed irrespective of valence, common regions of activation included the left anterior gyrus and the left amygdala. To our knowledge, these are the first published data showing regional brain responses to nonconsciously perceived sadness in facial expressions. Relative to other emotional phenomena, sadness has been understudied in the neurosciences, and the neural pathways leading to its experience, expression, and perception are poorly understood (Barr-Zisowitz, 2000). While the ability to rapidly perceive some facial expressions of emotion, such as those communicating fear or safety, have clear survival value, the role of perceptual mechanisms for detecting the expression of sadness in the face of another individual is less obvious. Sadness is generally accepted as an emotion that occurs in response to an aversive experience, usually an actual or perceived loss of something of value to the individual (Ellsworth and Smith, 1988). At a primitive survival level, sadness is unique in its tendency to slow down behavior and thought processes (Potts et al., 1989), which may serve an adaptive function by limiting continuation of the initial behavior that potentially caused the loss, thereby preventing further loss and allowing time to revise plans for future courses of action (Izard and Ackerman, 2000). At a more complex social level, the display of sadness can have a strong regulatory effect over social interactions by leading others to inhibit aggression and exhibit prosocial behavior (Eisenberg et al., 1989a,b; Miller and Eisenberg, 1988). Facial displays of sadness may be adaptive because they engender the support and resources of others in proximity to the individual. The ability to perceive sadness in others is likely to be adaptive to the individual because it increases awareness of potential losses to the self, and is adaptive to the species through elicitation of altruistic or prosocial behavior toward conspecifics who appear to be suffering. The limbic system in general, and the amygdala in particular, have been suggested to be key brain regions involved in social behaviors such as affiliation, attachment, and emotional intelligence (Joseph, 1996). Our findings are consistent with the hypothesis that the detection of sad affect in conspecifics may serve an adaptive survival function by slowing down the perceiver s cognition and behavior to prevent similar losses from befalling the perceiver, perhaps paralleling the inhibition of thought and action that has been

7 W.D.S. Killgore, D.A. Yurgelun-Todd / NeuroImage 21 (2004) observed during the actual experience of sadness (Potts et al., 1989). Recent fmri studies of consciously perceived facial emotion provide support for this hypothesis, as sad expressions have yielded limited cerebral activation relative to other emotional expressions (Kesler-West et al., 2001; Phillips et al., 1998). Our findings suggest that this suppression of cognitive activity during the perception of sad faces may also occur during nonconscious perception of sadness, as we found only minimal activation within the left anterior gyrus and no activation within any of our other hypothesized regions during the masked sad face condition. Direct comparisons between the masked happy and masked sad faces indicated that none of the regions we examined were significantly more active during the presentation of sadness relative to identical masked presentations of happiness. During both masked affect conditions, the left anterior gyrus was activated regardless of the valence of the masked expressions, consistent with recent suggestions that the rostral anterior serves an indirect affective function by integrating emotional and attentional processing (Yamasaki et al., 2002), and thus may activate generally without regard to emotional valence. The anterior gyrus appears to be involved in a wide range of cognitive processes, including directed attention (Faw, 2003), conflict monitoring (Fan et al., 2003), working memory (Osaka et al., 2003), self-monitoring (Carter et al., 2001), as well as emotional modulation and control (Hariri et al., 2003). Anterior activation has been demonstrated in functional imaging studies of consciously experienced sadness (George et al., 1995; Levesque et al., 2003) and during the perception of negatively valenced facial expressions such as sadness and anger (Blair et al., 1999). The activation we observed within the left anterior gyrus is similar to findings from a recent fmri study that revealed that activation within this region (BA 24/32) was correlated with self-ratings of sadness when subjects viewed sad films (Levesque et al., 2003). Similarly, in a recent PET study, conscious perception of sadness in facial expressions was associated with increased activation of the anterior gyrus and left amygdala, and activation within the latter was correlated with the intensity of the sadness expressed in the target faces (Blair et al., 1999). The anterior is believed to play an important role in the evaluation of the motivational significance of emotional stimuli, regardless of whether they are generated internally or externally (Devinsky et al., 1995). We found that the anterior was comparably activated, particularly on the left, for both the masked happy and sad conditions, suggesting that even nonconscious presentations of facial affect may serve to increase activity within brain regions important for directed attention. The amygdala, in contrast, was activated only in response to masked faces expressing happiness, whereas nonconsciously perceived sadness was not associated with activation within either amygdala. Previous research has suggested that the right amygdala is specifically involved in the detection and early processing of stimuli that are not explicitly available to conscious awareness (Morris et al., 1998, 1999; Whalen et al., 1998). Morris et al. (1998, 1999) found that fear-conditioned angry faces presented below conscious awareness expressly activated a network of right hemisphere subcortical structures including the amygdala, pulvinar, and superior colliculus, suggesting that this network is involved specifically in processing nonconsciously perceived stimuli. Our data were not consistent with this interpretation, as we only found right amygdala activation in response to masked presentations of happy rather than sad faces. The lack of amygdala activation in our masked sad condition argues against the hypothesis that the right amygdala is activated specifically by preattentive or nonconsciously perceived stimuli, and suggests instead that the amygdala may be responsive to some other attribute of affective faces, such as prior conditioning, degree of arousal displayed in the expressions, or some category-specific feature that was not as salient in the sad faces. Given that sadness tends to be lower in the degree of expressed arousal than fear, anger, or happiness (Watson and Tellegen, 1985), the differences in activation may reflect a responsiveness by the amygdala to the arousal dimension of affect (Yang et al., 2002). The increased activation of the amygdala in response to masked happiness found in our study contrasts with the results of Whalen et al. (1998), who found reduced activation of the amygdala during presentations of masked happy faces. However, the two studies are not directly comparable, as Whalen et al. included alternating blocks of both happy and fearful masked faces within the same experimental run, whereas we included only a single target affect condition, either happy or sad, within any single run. Given that happy faces may serve as signals of safety from threat, the copresentation of happy and fearful faces in the Whalen et al. study may have lead to decreased activation of the amygdala when the happy face condition was perceptually contrasted with recently appearing fearful faces. In the present study, happy faces were presented in the absence of contemporaneous masked negative affect, and it is therefore possible that the activation of the amygdala we observed was less influenced by the immediate contrast effects between these two affective valences. Several additional methodological issues should be kept in mind when interpreting these results. First, we limited our primary analyses to only four relatively small regions of interest to address specific hypotheses and to reduce the need for highly stringent corrections for multiple comparisons. Clearly, other regions of the brain are also important in the nonconscious processing of facial affect, and we therefore included a whole-brain analysis, corrected for multiple comparisons, to permit presentation of such data. As evident in Fig. 4 and Table 2, several nonhypothesized regions of the visual system, including the occipital lobes and inferior temporal lobes, as well as bilateral cerebellar regions were significantly activated even after statistical correction for multiple comparisons. These regions should be more closely examined in future studies of nonconscious emotional processing. Secondly, without correction for multiple comparisons, some of the data observed within the hypothesized ROIs could reflect spurious activation of some voxels due to Type I error. However, given the large body of evidence supporting the role of the amygdala in the processing of both consciously and nonconsciously perceived emotional facial stimuli, we believe that a noncorrected threshold is justified for these small structures. Further, to reduce the likelihood of chance detection of spike activations, we designated an extent threshold of at least 10 contiguous voxels for each cluster. To maintain comparability across regions, we also analyzed the anterior gyrus at the same uncorrected height threshold. It should be acknowledged, however, that the anterior was selected as an ROI based on evidence from substantially fewer studies than was the case for the amygdala, and these studies also used less stringent statistical thresholds (Blair et al., 1999). Thus, greater caution should be maintained when interpreting the findings for the anterior gyrus until these results have been replicated.

8 1222 W.D.S. Killgore, D.A. Yurgelun-Todd / NeuroImage 21 (2004) Another potential weakness of our experimental design is that it included several unilateral hemifield presentations of masked faces as part of the larger experimental design. It is possible that the inclusion of these additional stimuli could have influenced the BOLD response in some unforeseen way. However, given that these unilateral presentations were identically positioned in all affect conditions, it is unlikely that they systematically affected the responses to one affect more than another. Furthermore, because the overall rates of recognition of the target sad and happy faces still fell significantly below chance expectations, clearly the impact of including the unilateral presentations was negligible in affecting conscious recognition. The data for the unilateral masked faces are planned for presentation in a forthcoming paper. Contrary to our hypotheses, we did not find greater right amygdala activation during nonconscious presentations of facial affect, which is inconsistent with findings by Morris et al. (1998, 1999), who reported right amygdala activation to nonconsciously perceived emotional stimuli and left amygdala activation to consciously perceived stimuli. The differences in the findings, however, may be attributable to several methodological factors, such as the type of scanning procedure used (PET versus fmri). Recent data also suggest that the requirement to engage in a gender discrimination task, as in the present study, may attenuate the activation of the amygdala relative to paradigms requiring only passive viewing of faces (Lange et al., 2003). Outcome differences between studies may also be attributable to sex-related processing of emotional stimuli. We chose to include only female participants to reduce possible sex-related variability. In light of recent findings suggesting that the responses of the amygdala to affective faces may differ as a function of both the sex of the subject and the valence of the emotion (Killgore and Yurgelun- Todd, 2001), future studies will need to explore whether men show similar activation in response to masked sad and happy faces. It will also be important to include a wider range of masked affects in future studies to more clearly delineate the neural pathways involved in nonconscious processing of various negative affects, such as fear, anger, sadness, and disgust within the same sample. Conclusion This is the first fmri study to use backward masking of facial stimuli to demonstrate differential regional activation associated with nonconscious processing of happy versus sad facial affect. While the gyrus demonstrated relative consistency across affect conditions, the amygdala showed clearly distinct patterns of activation depending on whether the nonconscious stimuli depicted happy or sad expressions. These findings suggest a role for the amygdala in processing nonconscious affective stimuli that may be influenced by valence or arousal factors, whereas the left anterior appears to remain activated regardless of valence condition, perhaps subserving an attention orienting function. The right anterior, on the other hand, responded differentially to the affective valence of the stimulus. These data suggest that the amygdala and anterior gyrus are important components of a network of cerebral structures involved in detecting motivationally relevant visual stimuli encountered below the threshold of conscious perception, perhaps functioning to reallocate attention and bring potentially important stimuli into the forefront of conscious awareness. Acknowledgments This material has been reviewed by the Walter Reed Army Institute of Research. There is no objection to its presentation and/ or publication. The opinions or assertions contained herein are the private views of the authors, and are not to be construed as official, or as reflecting true views of the Department of the Army or the Department of Defense. This work was supported by a grant from the National Institute of Child Health and Human Development (NICHD) to William D. Killgore, PhD, NIH grant number 1 R03 HD References Barr-Zisowitz, C., Sadness is there such a thing? In: Lewis, M., Haviland-Jones, J.M. (Eds.), Handbook of Emotions, Second ed. Guilford Press, New York. Blair, R.J., Morris, J.S., Frith, C.D., Perrett, D.I., Dolan, R.J., Dissociable neural responses to facial expressions of sadness and anger. Brain 122 (Pt 5), Borod, J.C., Haywood, C.S., Koff, E., Neuropsychological aspects of facial asymmetry during emotional expression: a review of the normal adult literature. Neuropsychol. Rev. 7 (1), Carter, C.S., MacDonald III, A.W., Ross, L.L., Stenger, V.A., Anterior cortex activity and impaired self-monitoring of performance in patients with schizophrenia: an event-related fmri study. Am. J. Psychiatry 158 (9), de Gelder, B., Vroomen, J., Pourtois, G., Weiskrantz, L., Non-conscious recognition of affect in the absence of striate cortex. NeuroReport 10 (18), Devinsky, O., Morrell, M.J., Vogt, B.A., Contributions of anterior cortex to behaviour. Brain 118 (Pt 1), Dimberg, U., Thunberg, M., Elmehed, K., Unconscious facial reactions to emotional facial expressions. Psychol. Sci. 11, Dolan, R.J., Fletcher, P., Morris, J., Kapur, N., Deakin, J.F., Frith, C.D., Neural activation during covert processing of positive emotional facial expressions. NeuroImage 4 (3 Pt 1), Driver, J., Vuilleumier, P., Eimer, M., Rees, G., Functional magnetic resonance imaging and evoked potential correlates of conscious and unconscious vision in parietal extinction patients. NeuroImage 14 (1 Pt 2), S68 S75. Eisenberg, N., Fabes, R.A., Miller, P.A., Fultz, J., Shell, R., Mathy, R.M., et al., 1989a. Relation of sympathy and personal distress to prosocial behavior: a multimethod study. J. Pers. Soc. Psychol. 57 (1), Eisenberg, N., Fabes, R.A., Schaller, M., Miller, P.A., 1989b. Sympathy and personal distress: development, gender differences, and interrelations of indexes. New Dir. Child Dev. (44), Ellsworth, P.C., Smith, C.A., From appraisal to emotion: differences among unpleasant feelings. Motiv. Emot. 12 (3), Erwin, R.J., Gur, R.C., Gur, R.E., Skolnick, B., Mawhinney-Hee, M., Smailis, J., Facial emotion discrimination: I. Task construction and behavioral findings in normal subjects. Psychiatry Res. 42 (3), Esteves, F., Ohman, A., Masking the face: recognition of emotional facial expressions as a function of the parameters of backward masking. Scand. J. Psychol. 34 (1), Esteves, F., Parra, C., Dimberg, U., Ohman, A., Nonconscious associative learning: Pavlovian conditioning of skin conductance responses to masked fear-relevant facial stimuli. Psychophysiology 31 (4), Fan, J., Flombaum, J.I., McCandliss, B.D., Thomas, K.M., Posner, M.I., Cognitive and brain consequences of conflict. NeuroImage 18 (1),

9 W.D.S. Killgore, D.A. Yurgelun-Todd / NeuroImage 21 (2004) Faw, B., Pre-frontal executive committee for perception, working memory, attention, long-term memory, motor control, and thinking: a tutorial review. Conscious Cogn. 12 (1), Freud, S., The unconscious. In: Strachey, J. (Ed.), The Standard Edition, vol. 14. Hogarth Press, London. Friston, K., Ashburner, J., Poline, J., Frith, C., Heather, J., Frackowiak, R., 1995a. Spatial registration and normalization of images. Hum. Brain Mapp. 2, Friston, K., Holmes, A., Worsley, K., Poline, J., Frith, C., Frackowiak, R., 1995b. Statistical parametric maps in functional imaging: a general approach. Hum. Brain Mapp. 5, George, M.S., Ketter, T.A., Parekh, P.I., Horwitz, B., Herscovitch, P., Post, R.M., Brain activity during transient sadness and happiness in healthy women. Am. J. Psychiatry 152 (3), Hariri, A.R., Mattay, V.S., Tessitore, A., Fera, F., Weinberger, D.R., Neocortical modulation of the amygdala response to fearful stimuli. Biol. Psychiatry 53 (6), Izard, C.E., Ackerman, B.P., Motivational, organizational, and regulatory functions of discrete emotions. In: Lewis, M., Haviland-Jones, J.M. (Eds.), Handbook of Emotions, Second ed. Guilford Press, New York. Joseph, R., Neuropsychiatry, Neuropsychology, and Clinical Neuroscience: Emotion, Evolution, Cognition, Language, Memory, Brain Damage, and Abnormal Behavior, Second ed. Williams & Wilkins, Baltimore, MD. Kesler-West, M.L., Andersen, A.H., Smith, C.D., Avison, M.J., Davis, C.E., Kryscio, R.J., et al., Neural substrates of facial emotion processing using fmri. Brain Res. Cogn. Brain. Res. 11 (2), Killgore, W.D.S., Yurgelun-Todd, D.A., Sex differences in amygdala activation during the perception of facial affect. NeuroReport 12 (11), Lange, K., Williams, L.M., Young, A.W., Bullmore, E.T., Brammer, M.J., Williams, S.C., et al., Task instructions modulate neural responses to fearful facial expressions. Biol. Psychiatry 53 (3), LeDoux, J., The Emotional Brain: The Mysterious Underpinnings of Emotional Life. Simon and Schuster, New York. Levesque, J., Eugene, F., Joanette, Y., Paquette, V., Mensour, B., Beaudoin, G., et al., Neural circuitry underlying voluntary suppression of sadness. Biol. Psychiatry 53 (6), Macwhinney, B., Cohen, J., Provost, J., The PsyScope experimentbuilding system. Spat. Vis. 11 (1), Miller, P.A., Eisenberg, N., The relation of empathy to aggressive and externalizing/antisocial behavior. Psychol. Bull. 103 (3), Morris, J.S., Ohman, A., Dolan, R.J., Conscious and unconscious emotional learning in the human amygdala. Nature 393 (6684), Morris, J.S., Ohman, A., Dolan, R.J., A subcortical pathway to the right amygdala mediating unseen fear. Proc. Natl. Acad. Sci. U. S. A. 96 (4), Ohman, A., Soares, J.J., On the automatic nature of phobic fear: conditioned electrodermal responses to masked fear-relevant stimuli. J. Abnorm. Psychol. 102 (1), Osaka, M., Osaka, N., Kondo, H., Morishita, M., Fukuyama, H., Aso, T., et al., The neural basis of individual differences in working memory capacity: an fmri study. NeuroImage 18 (3), Phillips, M.L., Bullmore, E.T., Howard, R., Woodruff, P.W., Wright, I.C., Williams, S.C., et al., Investigation of facial recognition memory and happy and sad facial expression perception: an fmri study. Psychiatry Res. 83 (3), Potts, R., Camp, C., Coyne, C., The relationship between naturally occurring dysphoric moods, elaborative encoding, and recall performance. Cogn. Emot. 3 (3), Soares, J.J., Ohman, A., 1993a. Backward masking and skin conductance responses after conditioning to nonfeared but fear-relevant stimuli in fearful subjects. Psychophysiology 30 (5), Soares, J.J., Ohman, A., 1993b. Preattentive processing, preparedness and phobias: effects of instruction on conditioned electrodermal responses to masked and non-masked fear-relevant stimuli. Behav. Res. Ther. 31 (1), Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard, O., Delcroix, N., et al., Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. NeuroImage 15 (1), Watson, D., Tellegen, A., Toward a consensual structure of mood. Psychol. Bull. 98 (2), Whalen, P.J., Rauch, S.L., Etcoff, N.L., McInerney, S.C., Lee, M.B., Jenike, M.A., Masked presentations of emotional facial expressions modulate amygdala activity without explicit knowledge. J. Neurosci. 18 (1), Yamasaki, H., LaBar, K.S., McCarthy, G., Dissociable prefrontal brain systems for attention and emotion. Proc. Natl. Acad. Sci. U. S. A. 99 (17), Yang, T.T., Menon, V., Eliez, S., Blasey, C., White, C.D., Reid, A.J., et al., Amygdalar activation associated with positive and negative facial expressions. NeuroReport 13 (14),