Elevated responses to constant facial emotions in different faces in the human amygdala: an fMRI study of facial identity and expression
© Gläscher et al; licensee BioMed Central Ltd. 2004
Received: 12 July 2004
Accepted: 17 November 2004
Published: 17 November 2004
Human faces provide important signals in social interactions by inferring two main types of information, individual identity and emotional expression. The ability to readily assess both, the variability and consistency among emotional expressions in different individuals, is central to one's own interpretation of the imminent environment. A factorial design was used to systematically test the interaction of either constant or variable emotional expressions with constant or variable facial identities in areas involved in face processing using functional magnetic resonance imaging.
Previous studies suggest a predominant role of the amygdala in the assessment of emotional variability. Here we extend this view by showing that this structure activated to faces with changing identities that display constant emotional expressions. Within this condition, amygdala activation was dependent on the type and intensity of displayed emotion, with significant responses to fearful expressions and, to a lesser extent so to neutral and happy expressions. In contrast, the lateral fusiform gyrus showed a binary pattern of increased activation to changing stimulus features while it was also differentially responsive to the intensity of displayed emotion when processing different facial identities.
These results suggest that the amygdala might serve to detect constant facial emotions in different individuals, complementing its established role for detecting emotional variability.
Facial expressions and facial identities are important cues for the evaluation of social contexts [1, 2]. Two main types of information have to be processed while seeing other persons' faces: a face has to be identified as belonging to a unique individual, establishing facial identity, while facial expressions have to be interpreted for emotional context, which is crucial for the social interaction . Facial expression itself conveys two levels of information: first, facial emotions of others signal information about the emotional state and the benevolence or hostility of that person towards oneself; second, they convey information about the person's evaluation of the environment. The same emotional expression on several other persons' faces often signals a high degree of consistency in their evaluation of the current environment; they are therefore a particularly valid cue for one's own situational appraisal of this environment. In particular, the existence of specialized brain systems for the perception of fear expressions as a form of threat related to physical attack , points to the importance the brain attaches to social signals of potential environmental threats [5, 6].
Visual analysis of human faces has been suggested to be achieved by a core system comprising the fusiform gyrus together with the inferior occipital gyrus, the superior temporal sulcus, and the amygdala [7, 8]. The amygdala in particular has been shown to play a pivotal role in the processing and recognition of emotional facial expression, especially fear [9, 10]. Since early functional imaging studies [11, 12] showed amygdala activation to fearful face stimuli, a variety of imaging studies have led to a more detailed understanding of amygdala function in facial expression. Temporal properties of amygdala responses to stimulus repetition (i.e. habituation, e.g. [11, 13–15]), the influence of attentional state  and awareness (e.g. [17–19]), as well as the amygdala's response to different types of emotional expression (most recently e.g. ) have been explored. Despite many studies investigating facial affect, only few have addressed the interaction of facial affect and identity. Preceding experiments on face processing were not explicitly designed to investigate differences in processing of facial identity and facial expressions, rather they examined the effects of changes in appearance, viewing angle, selective features and physical stimulus properties on face processing (e.g. [14, 21, 22]).
Recent advances to assess the effects of facial identity and emotional expression include one report on increased left amygdala activation to blocks of multiple novel vs. single identical faces displaying neutral or emotionless expressions . Animated emotional expressions or identities, morphing either from neutral to emotional expression or from one neutral face to another, have been shown to elicit stronger activations in bilateral amygdala than static displays of these same stimuli . Notably, these studies were not designed to investigate the interaction of identity and emotional expression as identity was only varied within the neutral expression condition. Most recently, Winston and colleagues have employed a factorial design to explore fMRI-adaptation to repeated presentations of two facial expressions in the context of an event-related fMRI experiment . In contrast to earlier findings (e.g. [11, 12]), the authors did not observe any significant signal changes in the amygdala in the context of their factorial design, suggesting that sparse stimulus presentations interleaved with a checkerboard might not be potent enough to evoke a task-related fMRI activation in this region.
Given that numerous reports demonstrate amygdala activation in the processing of sequential facial expressions [11, 12, 27–29] and given the importance of different facial identities as a valid cue for making social inferences (see above), we hypothesized that constant emotional expressions displayed in different faces would elicit strong responses in the amygdala despite its well-described habituation to repeated stimuli [13, 15, 23]. Furthermore, we expected that constant fearful expressions (shown in different identities) would elicit the strongest amygdala activation given the evidence from lesion and imaging studies that this expression is a particularly potent activator of the amygdala [11, 12, 17, 30, 31]. Because we expected to find a stimulus repetition effect in the fusiform gyrus [14, 32] we systematically examined responses to changing versus constant stimulus features, i.e., facial expression and identity in this area.
Analysis of the behavioral data showed that subjects maintained a high degree of task attention (98.03 % hits to oddball stimuli (facial stimuli at reduced luminance)). There was a trend of a main effect of identity in the reaction times (RTs) to the oddball stimuli (F1,10 = 4.859, p < 0.06), with longer RTs in the conditions of changing identities, but no significant effects of emotion or an interaction (emotion: F1,10 = 0.755, p > 0.40; interaction: F1,10 = 1.656, p > 0.20).
For the analysis of the imaging data we chose a statistical threshold of p < 0.05, corrected for a reduced search volume of interest for regions with a priori hypotheses (amygdala, fusiform gyrus). For other brain areas the threshold was set to a threshold of p < 0.05, corrected for the entire brain. Post-hoc contrasts of specific experimental conditions for detailed characterization of the experimental effects were carried out at an uncorrected significance threshold.
Location of significant voxels in amygdala and lateral fusiform gyrus.
Left Fusiform Gyrus
Right Fusiform Gyrus
Post-hoc contrasts of the experimental conditions revealed that amygdala responses to VICE were only significantly more active than responses to CICE (left amygdala (-15 0 -15): T = 3.75, p < 10-4; right amygdala (18 0 -18): T = 4.76, p < 10-4), and also more active than the other two experimental conditions (CIVE and VIVE): left amygdala (-15 0 -15): VICE vs. CIVE: T = 1.77, p < 0.05; VICE vs. VIVE: T = 2.40, p < 0.01; right amygdala (18 0 -18): VICE vs. CIVE: T = 2.19, p < 0.01; VICE vs. VIVE: T = 3.30, p < 0.001.
In a further post-hoc analysis of the response to VICE we decomposed this condition with respect to different emotional expressions. This analysis revealed that the VICE activation is primarily caused by the response to maximally fearful expressions and (to a lesser degree) to other emotional expressions, but only marginal to neutral expressions (Figure 2, bottom graphs). Post-hoc pair-wise comparisons between maximally fearful and the other emotional expressions exceeded statistical thresholding (left amygdala (-15 0 -15): Fear Max vs. Fear Min: T = 1.95, p < 0.05, Fear Max vs. Neutral: T = 2.45, p < 0.01, Fear Max vs. Happy Min: T = 2.53, p < 0.01, Fear Max vs. Happy Max: T = 1.91, p < 0.05; right amygdala (18 0 -18): Fear Max vs. Neutral: T = 1.85, p < 0.05, Fear Max vs. Happy Max: T = 2.17, p < 0.05).
Pair-wise post-hoc contrasts of the four experimental conditions revealed that the lateral fusiform gyrus was significantly more activated by all experimental conditions with at least one changing stimulus feature (VICE, CIVE, VIVE) than by the condition with repeated presentation of the same picture (CICE): left lateral fusiform gyrus (-39 -54 -24): VICE vs. CICE: T = 2.75 p < 0.01, CIVE vs. CICE: T = 3.09, p < 0.01, VIVE vs. CICE: T = 3.77, p < 0.001; right lateral fusiform gyrus (36 -54 -24): VICE vs. CICE: T = 2.98, p < 0.01, CIVE vs. CICE: T = 2.79, p < 0.01, VIVE vs. CICE: T = 3.01, p < 0.01.
A further post-hoc analysis of the condition VICE revealed an effect of emotion intensity with the maximal intensity expressions eliciting stronger activations than those with reduced intensities and the neutral expression. Pair-wise contrasts of the emotion-specific VICE condition revealed that maximally fearful expressions elicited significantly larger activation than neutral and emotional expressions of minimal intensity (left lateral fusiform gyrus (-39 -54 -24): Fear Max vs. Fear Min: T = 2.6, p < 0.01; Fear Max vs. Neutral: T = 2.79, p < 0.001; Fear Max vs. Happy Min: T = 2.38, p < 0.01; right lateral fusiform gyrus (36 -54 -24): Fear Max vs. Fear Min: T = 1.72, p < 0.05; Fear Max vs. Neutral: T = 1.94, p < 0.05; Fear Max vs. Happy Min: T = 2.27, p < 0.01; indicated by asterisks (*) in the bottom panels of Figure 4). However, only in the left lateral fusiform gyrus the same post-hoc tests were significant when contrasting maximally happy expressions with all others (Happy Max vs. Happy Min: T = 1.84, p < 0.05; Happy Max vs. Neutral: T = 2.54, p < 0.01; Happy Max vs. Fear Min: T = 1.99, p < 0.05, indicated by pluses (+) in the bottom panels of Figure 4).
We investigated the interaction of facial identity and expression in 2 × 2 factorial blocked fMRI study which uniquely enabled us to compare all combinations of facial identity and expression within the same experiment, therefore the results of the present study complement and extend the results of previous studies on facial identity and expression processing in the human brain. In support for our hypotheses we found distinct response patterns in bilateral amygdala and bilateral lateral fusiform gyrus. While the lateral fusiform gyri can be characterized by a binary response pattern corresponding to an effect of changing stimulus feature, the amygdala responded maximally to constant emotional facial expressions in combination with changing facial identity. In addition, in the VICE experimental condition we found an effect of emotion intensity in the fusiform gyrus, with stronger activations to maximally intense expression irrespective of valence when compared with neutral or modestly intense expressions. While we found a (non-significant) trend for the same modulation of the VICE activation by emotion intensity in the left amygdala the general pattern when including the right amygdala rather conforms to a specific sensitivity for maximally fearful expression which elicited the strongest activation.
Fusiform gyrus activations
The binary response pattern conforming to sensitivity for changes in stimulus features in the lateral fusiform gyrus might be explained by reference to the proposed network of face processing in humans . Haxby and colleagues hypothesize that this part of the distributed network is especially sensitive to facial stimulus configurations (and the variability of these configurations). Evidence supporting this conclusion has been reported in a study by Vuilleumier and colleagues who demonstrated a significant repetition suppression effect in the fusiform gyrus in response to the second compared to the first presentation of a given facial identity . Similarly, Rotshtein and colleagues demonstrated stronger activation in face related voxels in the lateral occipital complex (LOC) for blocks with different identities (resembling the condition VICE of the present study) compared to blocks of the repetition of the same identity (our condition CICE) . Another study investigating the influence of facial expression and identity on fusiform activation found reduced activation to facial identity but not to expression and no interaction of these two factors . Our data support and extend these findings with respect to identity in the fusiform gyrus, as we found evidence for an effect of changing stimulus features in the lateral fusiform gyrus irrespective of the dimension of variation (facial identity, emotional expression collapsed over all emotions, or both).
Additionally, we found an effect of emotion intensity in the fusiform gyrus activity in condition VICE with higher activation to maximally fearful and happy expressions relative to with neutral expressions, paralleling earlier findings which report stronger activation to fearful than to neutral expressions [22, 24, 27, 33, 35]. Rotshtein and colleagues show significantly larger activation for aversive versus happy expressions in repeated presentations (CICE), suggesting an identity repetition × emotion interaction , while others fail to find an effect of negative (or positive) expression in the fusiform gyrus . This apparent negativity bias might be a confound of stimulus selection as researchers use negative material more frequently than positive stimuli. In fact, the underlying mechanism might be an enhanced attentional processing of arousing facial expressions , which are usually also the most negative faces. Our finding of an effect of emotion intensity supports this notion more convincingly as we also show increased activation in the fusiform gyrus to happy expressions. An underlying arousal dimension that exerts a modulatory effect on the activation in this region might also explain the lack of an effect of expression in the findings of Winston and colleagues , as they did not include a neutral (low arousal) expression in their stimulus set.
Two recent studies suggest that the low frequency information of facial stimuli might drive the modulatory effect of arousal, which is thought to be a feedback influence of the amygdala, which also displays a preference for low frequency information [22, 35].
Our 2 × 2 factorial blocked fMRI design allowed us to compare between the possible conditions of constant and variable facial identity and expression. We show for the first time a maximum of activation of the amygdala to variable facial identities displaying the same emotion (VICE) compared to all three other main conditions (CICE, CIVE and VIVE). Furthermore, within this condition (VICE) maximally fearful expressions elicited the strongest amygdala activity. The result of the present study replicates and extends, in part, the earlier findings that also utilized blocked presentations of facial expression with stimulus configurations resembling one or more of our experimental conditions (VICE blocks: [11, 12, 28, 37, 38]; CICE and VICE blocks: ; VICE and VIVE blocks: ). In light of our findings, the significant amygdala activation to fearful expression in the aforementioned studies might have been obtained because they employed a stimulus presentation conforming to our condition VICE. Interestingly, in a study comparing dynamically morphed and static presentations of facial stimuli, LaBar and colleagues  also report amygdala activation when neutral faces are morphed into each other (a dynamic version of our condition VICE) as compared to static and repeated presentation of the same facial stimuli (the condition CICE in the present study). A closer inspection of our findings in Figure 2 reveals a similar trend, as the neutral expression in the VICE condition still elicited a larger signal change than the condition CICE. Thus, our comprehensive factorial design can relate these earlier findings in the broader context of constancy and variability of identity and expression.
Many imaging and lesion studies have documented activations or behavioral impairments following the presentation of fear-related stimuli (for review see ). Our decomposition of the amygdala activation in condition VICE supports this claim. However, recent studies, which used different sensory modalities and carefully controlled for the often confounded dimensions of valence and intensity, argued for an effect of emotion intensity (arousal) in the amygdala [40–42]. Further support for this interpretation comes from a comprehensive study investigating the effects of different expressions of basic emotions during direct and incidental stimulus processing . The authors found no specific effect for a particular emotional expression; they rather report amygdala activations when comparing high vs. low intensity exemplars of the facial stimuli. Although we also find trends for an effect of emotion intensity (Figure 2, bottom left panel), the question of specific amygdalar fear sensitivity or a more general arousal sensitivity remains equivocal based on our findings.
Interestingly, Winston and colleagues , employing a similar design as in the present study, failed to observe significant signal changes in the amygdala. The divergent findings might be explained by the difference between our blocked and their event-related design, in which they sequentially presented pairs of facial stimuli. Sparse stimulus presentations in that study might not be sufficient to trigger activation in the amygdala, as this structure might decode a more sustained stimulus train for assessing the biological relevance of the current situation (see below). Furthermore, in their study, the pairs of faces presented were separated by an face-outlined checkerboard  which might have prevented the induction of an emotional state that would be encoded by the amygdala.
The temporal properties of amygdala responses to facial expressions are crucial. Many studies show habituation of the amygdala response to blocks of directly adjacent fearful vs. neutral or happy expressions with constant identities over the course of 30 to 80 sec [13, 15, 38]. This is comparable to the CICE condition in the present study in terms of presented stimuli but not with respect to block length and block order. Interestingly, when the CICE condition in the present study is analyzed for the comparison of maximum fear vs. neutral blocks a non-significant trend for left ventro-lateral amygdala activity (data not shown) is seen, indirectly supporting those findings.
Other studies used fearful and neutral VICE conditions showing within-block and across-block habituation with fixed alternating block order  or in the comparison of VICE vs. CICE blocks [14, 23] in designs with pseudo-randomized block order. Likewise, we also show a highly significant difference between the VICE and CICE conditions in the amygdala. The response profile of the amygdala for maximal fearful expressions in the VICE condition trends toward a within-block habituation effect (Figure 3). In contrast, the response profile for neutral expressions exhibits a sustained response during the entire block albeit at a much lower overall level. This aspect parallels findings by Breiter and colleagues , although direct comparison is limited because that study showed habituation on a larger timescale (between blocks). The within-block increase of amygdala activation to happy expressions, however, suggests that the temporal nature of signal changes in the amygdala might be more complex. This delayed onset of activation to happy faces is in accordance with an evolutionary interpretation of our findings. The amygdala is located in a critical position on the efferent pathway that is involved in the preparation of autonomic responses to threatening situations . Because happy expressions are usually a valid signal for non-threatening situations that do not require fight or flight responses, amygdala activation is not needed at an early stage. The explanation also holds for the early but overall attenuated response to neutral expressions that are inherently ambiguous, which thus prompt for sustained perceptual processing and, potentially, for preparation of defensive behavior . Additionally, Wright and colleagues  showed greater amygdala activation to neutral VICE blocks compared with neutral CICE blocks. Clearly, further research is needed to characterize the temporal evolution of amygdala activation during this type of blocked stimulus presentation in more detail.
One might argue that the responses in the conditions with variable emotional expression (CIVE, VIVE) were reduced because within those blocks variable emotional and neutral expressions were intermixed, possibly leading to smaller signal changes within those blocks. However, subjects were presented with the same number of stimuli of each emotional expression in every experimental condition. Therefore, differences cannot be attributed to a varying number of emotional expressions seen in different conditions. Thus, if the conditions with variable emotions elicit a smaller signal change because of the neutral faces within them, this effect should also apply to the blocks of constant neutral expressions in different faces (VICE condition). This effect is in fact shown in the lower bar graphs in Figure 2, especially in comparison to the other emotional VICE conditions. Because we found an overall elevated response in the amygdala to the condition with constant emotions in changing identities (VICE) compared to conditions with variable emotions (CIVE, VIVE; see Figure 2 upper bar graphs), we argue that the observed differences between our experimental conditions represent a true effect of the sequential stimulus configuration within this experimental design. Although we can not ultimately exclude the possibility of confounding order effects within the categories of variable emotion conditions (CIVE, VIVE), this does not diminish the main point of this study, namely that the human amygdala is most responsive to the sequential presentation of faces with constant emotion (fearful) and varying identity. To further confirm and generalize this finding future studies are needed using additional emotions (e.g. anger, sadness, etc.).
Novelty effects of changing facial identity and stimulus order of the different conditions might also be claimed as an explanation for the strong responses to the condition VICE . But in contrast to the latter study, subjects in our study were familiarized with the stimuli before the experiment and the same facial identities were presented repeatedly throughout the experiment, thus minimizing stimulus novelty. Furthermore, if novelty detection were the main factor driving the amygdala response, the condition VIVE should have also elicited a strong amygdala response which it did not (see Figure 2). There was also no systematic sequence of blocks of constant facial identities followed by variable facial identities in the present study which has been shown to relatively increase the activation to the multiple identity condition presented secondly .
It is interesting to note that the peak activation within the amygdala is located in the medio-dorsal part of this structure. Previous imaging studies of the processing of facial affect have predominantly reported their activations in the dorsal amygdala (for review see , Figure 3). The dorsal amygdala has also been associated with the representation of ambiguous stimuli, such as fearful expressions that do not signal a potential threat directly [10, 29, 44]. However, given the resolution and post scan smoothing of the functional images in this and other functional imaging studies, the localization of amygdala activity should be discussed with caution.
The potential biological relevance of constant facial emotion
Sensitivity for constant facial emotions in different identities can be seen as a process that integrates stimuli with respect to the conveyed emotion over a certain amount of time. Thus, environmental stimuli (such as facial expressions) are compared to each other to detect changes to stability in the emotion, a concept that can be described as emotion constancy. In this context, constant facial emotions in different individuals signal a high degree of consistency in others' appraisals of the environment, and constitute a more valid cue for one's own appraisal.
Furthermore, facial expressions differ in their degree of saliency which often reflects the biological relevance of the stimulus that evoked the expression [18, 44]. For example, a fearful face as a reaction to a threatening stimulus is more salient and calls more immediately for appropriate action than a neutral facial expression. We found elevated response in the amygdala especially to these salient (fearful) expressions because these constant salient facial emotions call for an immediate situational appraisal and subsequent action.
Our findings gain support and extend the findings of earlier imaging studies on the processing of human facial emotions which employed an experimental design similar to our condition VICE and found significant activation in the left dorsal amygdala with constant fearful expressions [11, 12]. Taken together, these results, as well as our own findings in this experiment, suggest that the same emotional expressions displayed in many different faces are potent stimuli that activate the amygdala and might serve the detection of emotion constancy. Conceptually, conditions with variable emotionality (CIVE, VIVE) can be seen as noise in the context of the detection of constant emotions among others and thus, it is not surprising that the amygdala shows less signal change in these conditions.
The perception of emotional facial expressions in others yields insights into their evaluations of the environment and guides one's own emotional and behavioral reactions. Encountering the same emotional expression in many different people (emotion constancy) is an especially valid and readily available cue for making subsequent inferences about the potential harmfulness of an environmental situation. These inferences would have direct implications for evolutionary survival and must have been a central feature of human ancestral cognitive abilities . They also remain essential for safe locomotion in our current complex environment as they reduce the time spent in a potentially threatening situation, because the perceiving subject can avoid energy-intensive and time-consuming search for potential threats in the environment . For example, encountering the same fearful expression in several different people (emotion constancy) strongly implies an activation of the fear system [45, 47] and that precaution and avoidance behavior are adequate reactions in this situation. Our data, in particular the large response to constant maximally fearful expressions in different individuals, suggest that the amygdala plays a central role in the neurobiological realization of this environmental evaluation.
Emotional facial expressions are an important cue for the appraisal of an environmental situation. Our study has demonstrated a new perspective on the functional characterization of the amygdala involved in the perceptual processing of human faces by incorporating the dimensions of constancy and variability detection in these stimuli. These are essential for the assessment of the temporal dynamics of social situations.
A 2 × 2 factorial design (identity × emotional expression) across 4 different block types was used: (a) constant identity, constant expression (CICE), (b) variable identity, constant expression (VICE), (c) constant identity, variable expression (CIVE), (d) variable identity, variable expression (VIVE). Figure 1 schematically displays our experimental design showing three consecutive stimuli from one block within each cell of the factor table.
13 Subjects (8 females, 10 right-handed) participated in this fMRI study. The mean age was 25.6 (SD 7.8). The data sets of two subjects were excluded from further image analysis because of radio frequency artifacts caused by the scanner leaving a total of 11. All subjects were fully informed about the experimental procedure and signed a consent statement which was approved by the local ethics committee.
The procedure was completed in one imaging session with 4 runs each containing 20 blocks of stimulus presentation. Within one run the blocks were interleaved with 15 s of central fixation (rest period). Each run started with a rest period of 20 s. 20 stimuli were presented per block at a rate of 1 Hz. Stimuli were shown for 900 ms interposed with a 100 ms gray blank of mean luminance to make transitions between stimuli less abrupt. Numbers and types of both facial expressions and facial identities were counterbalanced across the entire experiment. Conditions varied only in terms of the sequential configuration of the stimuli within the blocks. Stimulus order within each block was pseudo-randomized and fixed. The sequence of blocks within each run was counterbalanced, pseudo-randomized and fixed to minimize stimulus order effects. Additionally, the sequence of the four runs was pseudo-randomized across subjects assuring different orders of runs in each subject.
Stimuli consisted of 4 different faces (facial identities) drawn from the Ekman series of facial affect  with 5 different emotional expressions ranging from fearful to happy expressions, including neutral. Two intermediate expressions displayed fearful and happy emotions at reduced (50 %) intensities. Those face pictures were interpolations using computer morphing procedures  similar to those in other studies. Subjects were instructed to fixate on a small red fixation dot presented in the middle of the viewing monitor while simultaneously attending to the entire stimulus presentation.
In order to maintain and control for attentional effects during the procedure we included an oddball task as we were seeking to avoid confounds by an emotional judgment or a gender differentiation task (latter being a property of facial identity, ). As oddball stimulus we occasionally presented the facial stimuli with reduced luminance of the entire face while leaving the stimulus visible which was not expected to affect activation in high-order visual areas . We chose to manipulate the entire facial stimulus for the oddball task in order to keep the subjects attention on the entire stimulus rather than on some small feature. Subjects were instructed to respond with a button press whenever an oddball target appeared. The number of oddball targets per block ranged from 2 to 4 to avoid subjects' expectancy effects. In order to prevent a systematic effect of the number of oddballs in the data analysis, the number of oddball stimuli was counterbalanced across blocks and conditions. Subjects were familiarized with the oddball task in a practice session prior to the first scanning run.
Imaging was performed on a 1.5 T Magnetom Vision (Siemens, Erlangen, Germany) scanner. 43 transversal slices of echo-planar (EPI) T2* weighted images in each volume with a slice thickness of 2 mm and 1 mm gap (TR = 3.5 s, TE = 40 ms, flip angle 90°, FoV 192 × 192 mm2, matrix 64 × 64) were acquired. A total of 204 volumes were collected per run.
Image processing and statistical analysis were carried out using SPM99 for the single subject analysis and SPM2 for the group analysis [http://www.fil.ion.ucl.ac.uk/spm/] All volumes were realigned to the first volume, spatially normalized to a standard EPI template  using sinc interpolation and finally smoothed with a 11 mm isotropic full width at half maximum (FWHM) Gaussian filter to account for anatomical differences between subjects and to allow statistical inference using Gaussian Random Field theory.
The data of 11 subjects were included in the statistical analysis. Data analysis was performed using the mass univariate general linear model as implemented in SPM99 and commenced by specifying the design matrix for each subject using a boxcar and an exponential decay regressor for modeling the hemodynamic response to each experimental condition. The boxcar regressor models the mean activation within the block while the exponential decay regressor (time constant 4 s) models decreases and increases (through negative contrast weights) within the block. The conditions with constant emotional expressions (CICE, VICE) allowed a decomposition into emotion-specific components. Thus each of these two conditions involved 10 regressors (boxcar and exponential decay for 5 different expression), while each of the other conditions (CIVE, VIVE) were specified with two regressors. Data were high-pass filtered at 1/120 Hz. Serial autocorrelation was controlled by superimposing a known autocorrelation in form of temporal smoothing using a low-pass filter at 4 sec filter width. Successively, contrasts for each experimental condition were computed by averaging the same block type across runs and multiplying the design matrix with the contrast vectors.
These single-subject contrast images were then taken to the second level oneway ANOVA [52, 53] in SPM2 allowing for an appropriate non-sphericity correction . This correction is equivalent to the Greenhouse-Geisser procedure in multivariate ANOVA analyses and allows for correct assessment of the error covariance matrix, hence securing valid inference in the group comparisons.
In order to detect voxels that show elevated responses to the same expression displayed in different individuals we constructed the interaction contrast of our 2 × 2 design. Hence, we created the contrast [(VICE > CICE) > (VIVE > CIVE), p < .01 for the purpose of visualization] for the amygdala and masked it with the contrast [(VICE > CIVE), p < .05] to exclude regions showing higher activations to CIVE than to VICE. For our hypothesis in the fusiform gyrus we used a contrast that compared conditions with at least one changing stimulus feature (facial identity, emotional expression, or both) with the condition in which the same stimulus was shown for the entire block [(VICE + CIVE+ VIVE) > 3 × CICE].
T-statistics for the assessment of significant regional activation were assembled into Statistical Parametric Maps (SPMs) which refer to the probabilistic behavior of Gaussian random fields . Our threshold was set at p < .05 (corrected). Because we had region-specific hypotheses for the amygdala and the lateral fusiform gyrus, we applied a reduced search volume to our amygdala activation which was derived by an anatomical mask created with MRIcro  on the template brain of the Montreal Neurological Institute (MNI, ). With additional visual reference to a high-resolution anatomical atlas  we outlined the amygdala on each slice of the MNI template brain. Thus, the amygdala search volume comprised 77 voxels, or 2071 mm3 on the right side and 74 voxels, or 2005 mm3 on the left side. Similarly, we applied a 10 mm radius sphere to our activation peaks in the lateral fusiform gyrus centered on coordinates reported by Vuilleumier and colleagues when contrasting faces vs. houses in a functional localizer task ; see Table 1). For additional brain areas not included in our volumes of interest we corrected for the entire brain volume.
For the emotion-specific analyses of condition VICE (Figure 2 and 4, bottom panels) we referred to specific contrasts for each emotional expression created at the single-subject level. These contrast images were raised to another second level one-way ANOVA to test, for example, for significant activations to maximally fearful faces within this condition. These statistical comparisons were carried out at an uncorrected significance threshold.
List of abbreviations
functional magnetic resonance imaging
lateral occipital complex
statistical parametric map
variable identity, variable emotion
variable identity, constant emotion
constant identity, constant emotion
constant identity. variable emotion
- Fear/Happy Max:
maximally fearful/happy expression
- Fear/Happy Min:
minimal fearful/happy expression.
This research was supported by grants from the BMBF and the Volkswagenstiftung to C. B., and a doctoral fellowship by the Studienstiftung des Deutschen Volkes to J. G. We thank Dr. James Root for helpful comments on the manuscript.
- Adolphs R: Social cognition and the human brain. Trends Cogn Sci. 1999, 3: 469-479. 10.1016/S1364-6613(99)01399-6.View ArticlePubMedGoogle Scholar
- Adolphs R: The neurobiology of social cognition. Curr Opin Neurobiol. 2001, 11: 231-239. 10.1016/S0959-4388(00)00202-6.View ArticlePubMedGoogle Scholar
- Posamentier MT, Abdi H: Processing faces and facial expressions. Neuropsychol Rev. 2003, 13: 113-143. 10.1023/A:1025519712569.View ArticlePubMedGoogle Scholar
- Gray JA: The psychology of fear and stress. Problems in the Behavioral Sciences Series. 1987, Cambridge, Cambridge University Press, 5: 432.Google Scholar
- Anderson AK, Christoff K, Panitz D, De Rosa E, Gabrieli JD: Neural correlates of the automatic processing of threat facial signals. J Neurosci. 2003, 23: 5627-5633.PubMedGoogle Scholar
- Hadjikhani N, de Gelder B: Seeing fearful body expressions activates the fusiform cortex and amygdala. Curr Biol. 2003, 13: 2201-2205. 10.1016/j.cub.2003.11.049.View ArticlePubMedGoogle Scholar
- Haxby JV, Hoffman EA, Gobbini MI: The distributed human neural system for face perception. Trends Cogn Sci. 2000, 4: 223-233. 10.1016/S1364-6613(00)01482-0.View ArticlePubMedGoogle Scholar
- Perrett DI, Oram MW, Ashbridge E: Evidence accumulation in cell populations responsive to faces: an account of generalisation of recognition without mental transformations. Cognition. 1998, 67: 111-145. 10.1016/S0010-0277(98)00015-8.View ArticlePubMedGoogle Scholar
- Adolphs R: Recognizing Emotion From Facial Expressions: Psychological and Neurological Mechanisms. Behavioral and Cognitive Neuroscience Reviews. 2002, 1: 21-61. 10.1177/1534582302001001003.View ArticlePubMedGoogle Scholar
- Calder AJ, Lawrence AD, Young AW: Neuropsychology of fear and loathing. Nat Rev Neurosci. 2001, 2: 352-363. 10.1038/35072584.View ArticlePubMedGoogle Scholar
- Breiter HC, Etcoff NL, Whalen PJ, Kennedy WA, Rauch SL, Buckner RL, Strauss MM, Hyman SE, Rosen BR: Response and habituation of the human amygdala during visual processing of facial expression. Neuron. 1996, 17: 875-887. 10.1016/S0896-6273(00)80219-6.View ArticlePubMedGoogle Scholar
- Morris JS, Frith CD, Perrett DI, Rowland D, Young AW, Calder AJ, Dolan RJ: A differential neural response in the human amygdala to fearful and happy facial expressions. Nature. 1996, 383: 812-815. 10.1038/383812a0.View ArticlePubMedGoogle Scholar
- Fischer H, Wright CI, Whalen PJ, McInerney SC, Shin LM, Rauch SL: Brain habituation during repeated exposure to fearful and neutral faces: a functional MRI study. Brain Res Bull. 2003, 59: 387-392. 10.1016/S0361-9230(02)00940-1.View ArticlePubMedGoogle Scholar
- Rotshtein P, Malach R, Hadar U, Graif M, Hendler T: Feeling or features: different sensitivity to emotion in high-order visual cortex and amygdala. Neuron. 2001, 32: 747-757. 10.1016/S0896-6273(01)00513-X.View ArticlePubMedGoogle Scholar
- Wright CI, Fischer H, Whalen PJ, McInerney SC, Shin LM, Rauch SL: Differential prefrontal cortex and amygdala habituation to repeatedly presented emotional stimuli. Neuroreport. 2001, 12: 379-383. 10.1097/00001756-200102120-00039.View ArticlePubMedGoogle Scholar
- Pessoa L, Ungerleider LG: Neuroimaging studies of attention and the processing of emotion-laden stimuli. Prog Brain Res. 2004, 144: 171-182.View ArticlePubMedGoogle Scholar
- Morris JS, Ohman A, Dolan RJ: Conscious and unconscious emotional learning in the human amygdala. Nature. 1998, 393: 467-470. 10.1038/30976.View ArticlePubMedGoogle Scholar
- Whalen PJ: Fear, vigilance, and ambiguity: Initial neuroimaging studies of the human amygdala. Curr Dir Psychol Sci. 1998, 7: 177-188. 10.1111/1467-8721.ep10836912.View ArticleGoogle Scholar
- Phillips ML, Williams LM, Heining M, Herba CM, Russell T, Andrew C, Bullmore ET, Brammer MJ, Williams SC, Morgan M, Young AW, Gray JA: Differential neural responses to overt and covert presentations of facial expressions of fear and disgust. Neuroimage. 2004, 21: 1484-1496. 10.1016/j.neuroimage.2003.12.013.View ArticlePubMedGoogle Scholar
- Winston JS, O'Doherty J, Dolan RJ: Common and distinct neural responses during direct and incidental processing of multiple facial emotions. Neuroimage. 2003, 20: 84-97. 10.1016/S1053-8119(03)00303-3.View ArticlePubMedGoogle Scholar
- Morris JS, deBonis M, Dolan RJ: Human amygdala responses to fearful eyes. Neuroimage. 2002, 17: 214-222. 10.1006/nimg.2002.1220.View ArticlePubMedGoogle Scholar
- Vuilleumier P, Armony JL, Driver J, Dolan RJ: Distinct spatial frequency sensitivities for processing faces and emotional expressions. Nat Neurosci. 2003, 6: 624-631. 10.1038/nn1057.View ArticlePubMedGoogle Scholar
- Wright CI, Martis B, Schwartz CE, Shin LM, Fischer HH, McMullin K, Rauch SL: Novelty responses and differential effects of order in the amygdala, substantia innominata, and inferior temporal cortex. Neuroimage. 2003, 18: 660-669. 10.1016/S1053-8119(02)00037-X.View ArticlePubMedGoogle Scholar
- LaBar KS, Crupain MJ, Voyvodic JT, McCarthy G: Dynamic perception of facial affect and identity in the human brain. Cereb Cortex. 2003, 13: 1023-1033. 10.1093/cercor/13.10.1023.View ArticlePubMedGoogle Scholar
- Winston JS, Henson RN, Fine-Goulden MR, Dolan RJ: fMRI-Adaptation Reveals Dissociable Neural Representations of Identity and Expression in Face Perception. J Neurophysiol. 2004, 92: 1830-1839. 10.1152/jn.00155.2004.View ArticlePubMedGoogle Scholar
- Ganel T, Goshen-Gottstein Y: Perceptual integrality of sex and identity of faces: further evidence for the single-route hypothesis. J Exp Psychol Hum Percept Perform. 2002, 28: 854-867. 10.1037//0096-15188.8.131.524.View ArticlePubMedGoogle Scholar
- Morris JS, Friston KJ, Buchel C, Frith CD, Young AW, Calder AJ, Dolan RJ: A neuromodulatory role for the human amygdala in processing emotional facial expressions. Brain. 1998, 121: 47-57. 10.1093/brain/121.1.47.View ArticlePubMedGoogle Scholar
- Phillips ML, Young AW, Scott SK, Calder AJ, Andrew C, Giampietro V, Williams SC, Bullmore ET, Brammer M, Gray JA: Neural responses to facial and vocal expressions of fear and disgust. Proc R Soc Lond B Biol Sci. 1998, 265: 1809-1817. 10.1098/rspb.1998.0506.View ArticleGoogle Scholar
- Whalen PJ, Shin LM, McInerney SC, Fischer H, Wright CI, Rauch SL: A functional MRI study of human amygdala responses to facial expressions of fear versus anger. Emotion. 2001, 1: 70-83. 10.1037//1528-35184.108.40.206.View ArticlePubMedGoogle Scholar
- Adolphs R, Tranel D, Damasio H, Damasio A: Impaired recognition of emotion in facial expressions following bilateral damage to the human amygdala. Nature. 1994, 372: 669-672. 10.1038/372669a0.View ArticlePubMedGoogle Scholar
- Calder AJ, Yound AW, Rowland D, Perrett DI, Hodges JR, Etcoff NL: Facial Emotion Recognition after Bilateral Amygdala Damage: Differentially Severe Impairment of Fear. Cognitive Neuropsychology. 1996, 13: 699-745. 10.1080/026432996381890.View ArticleGoogle Scholar
- Vuilleumier P, Henson RN, Driver J, Dolan RJ: Multiple levels of visual object constancy revealed by event-related fMRI of repetition priming. Nat Neurosci. 2002, 5: 491-499. 10.1038/nn839.View ArticlePubMedGoogle Scholar
- Vuilleumier P, Armony JL, Driver J, Dolan RJ: Effects of attention and emotion on face processing in the human brain: an event-related fMRI study. Neuron. 2001, 30: 829-841. 10.1016/S0896-6273(01)00328-2.View ArticlePubMedGoogle Scholar
- Kanwisher N, McDermott J, Chun MM: The fusiform face area: a module in human extrastriate cortex specialized for face perception. J Neurosci. 1997, 17: 4302-4311.PubMedGoogle Scholar
- Winston JS, Vuilleumier P, Dolan RJ: Effects of low-spatial frequency components of fearful faces on fusiform cortex activity. Curr Biol. 2003, 13: 1824-1829. 10.1016/j.cub.2003.09.038.View ArticlePubMedGoogle Scholar
- Dolan RJ: Emotion, cognition, and behavior. Science. 2002, 298: 1191-1194. 10.1126/science.1076358.View ArticlePubMedGoogle Scholar
- Phillips ML, Young AW, Senior C, Brammer M, Andrew C, Calder AJ, Bullmore ET, Perrett DI, Rowland D, Williams SC, Gray JA, David AS: A specific neural substrate for perceiving facial expressions of disgust. Nature. 1997, 389: 495-498. 10.1038/39051.View ArticlePubMedGoogle Scholar
- Phillips ML, Medford N, Young AW, Williams L, Williams SC, Bullmore ET, Gray JA, Brammer MJ: Time courses of left and right amygdalar responses to fearful facial expressions. Hum Brain Mapp. 2001, 12: 193-202. 10.1002/1097-0193(200104)12:4<193::AID-HBM1015>3.0.CO;2-A.View ArticlePubMedGoogle Scholar
- Critchley H, Daly E, Phillips M, Brammer M, Bullmore E, Williams S, Van Amelsvoort T, Robertson D, David A, Murphy D: Explicit and implicit neural mechanisms for processing of social information from facial expressions: a functional magnetic resonance imaging study. Hum Brain Mapp. 2000, 9: 93-105.View ArticlePubMedGoogle Scholar
- Adolphs R, Russel JA, Tranel D: A role for the human amygdala in recognizing emotional arousal from unpleasant stimuli. Psychol Sci. 1999, 10: 157-161. 10.1111/1467-9280.00126.View ArticleGoogle Scholar
- Anderson AK, Christoff K, Stappen I, Panitz D, Ghahremani DG, Glover G, Gabrieli JD, Sobel N: Dissociated neural representations of intensity and valence in human olfaction. Nat Neurosci. 2003, 6: 196-202. 10.1038/nn1001.View ArticlePubMedGoogle Scholar
- Small DM, Gregory MD, Mak YE, Gitelman D, Mesulam MM, Parrish T: Dissociation of neural representation of intensity and affective valuation in human gustation. Neuron. 2003, 39: 701-711. 10.1016/S0896-6273(03)00467-7.View ArticlePubMedGoogle Scholar
- Lang PJ, Davis M, Ohman A: Fear and anxiety: animal models and human cognitive psychophysiology. J Affect Disord. 2000, 61: 137-159. 10.1016/S0165-0327(00)00343-8.View ArticlePubMedGoogle Scholar
- Davis M, Whalen PJ: The amygdala: vigilance and emotion. Mol Psychiatry. 2001, 6: 13-34. 10.1038/sj.mp.4000812.View ArticlePubMedGoogle Scholar
- Cosmides L, Tooby J: Evolutionary Psychology and the Emotions. Handbook of Emotions. Edited by: Lewis M and Haviland-Jones JM. 2000, New York, Guilford Publications, 91-115.Google Scholar
- Kameda T, Nakanishi D: Cost–benefit analysis of social/cultural learning in a nonstationary uncertain environment: An evolutionary simulation and an experiment with human subjects. Evol Hum Behav. 2002, 23: 373-393. 10.1016/S1090-5138(02)00101-0.View ArticleGoogle Scholar
- Ohman A, Mineka S: Fears, phobias, and preparedness: toward an evolved module of fear and fear learning. Psychol Rev. 2001, 108: 483-522. 10.1037//0033-295X.108.3.483.View ArticlePubMedGoogle Scholar
- Ekman P, Frieser W: Pictures of facial affects. 1976, Palo Alto, CA, Consulting Psychologists PressGoogle Scholar
- WinMorph: [http://www.debugmode.com/winmorph].
- Grill-Spector K, Kushnir T, Edelman S, Itzchak Y, Malach R: Cue-invariant activation in object-related areas of the human occipital lobe. Neuron. 1998, 21: 191-202. 10.1016/S0896-6273(00)80526-7.View ArticlePubMedGoogle Scholar
- Friston KJ, Ashburner J, Worsley KJ, Poline JP, Frith CD: Spatial registration and normalization of images. Hum Brain Mapp. 1995, 2: 165-189.View ArticleGoogle Scholar
- Friston KJ, Holmes AP, Price CJ, Buchel C, Worsley KJ: Multisubject fMRI studies and conjunction analyses. Neuroimage. 1999, 10: 385-396. 10.1006/nimg.1999.0484.View ArticlePubMedGoogle Scholar
- Friston KJ, Holmes AP, Worsley KJ: How many subjects constitute a study?. Neuroimage. 1999, 10: 1-5. 10.1006/nimg.1999.0439.View ArticlePubMedGoogle Scholar
- Glaser D, Penny W, Henson R, Friston KJ: Coping with the variance structure of neuroimaging data [abstract]. J Cogn Neurosci. 2002, Suppl 2002:E92.Google Scholar
- Worsley KJ: Local maxima and the expected euler characteristic of excursion set of chi², F, and t fields. Adv Appl Probab. 1994, 26: 13-42.View ArticleGoogle Scholar
- Rorden C, Brett M: Stereotaxic display of brain lesions. Behav Neurol. 2000, 12: 191-200.View ArticlePubMedGoogle Scholar
- Evans AC, Collins DL, Mills DR, Brown ED, Kelly RL, Peters TM: 3D statistical neuroanatomical model from 205 MRI volumes. Proc IEEE Nucl Sci Symp Med Imaging. 1993, 1: 1813-1817.Google Scholar
- Mai JK, Assheuer JK, Paxinos G: Atlas of the Human Brain. 1998, San Diego, CA, Academic Press, 1Google Scholar