The present study revealed new insights into abnormalities in the perception of facial emotion at an automatic processing level in schizophrenia patients. None of the participants included in our analyses had subjective awareness of emotional primes. Therefore, the emotional expression in prime faces was most likely processed automatically, without time for overt attention. As hypothesized, in response to angry, happy, and neutral faces versus no facial expression primes, we observed increased right amygdala activation in schizophrenia patients compared to healthy controls during the initial phase. For the left amygdala, we found a heightened response in patients only to neutral faces compared to controls. These results are consistent with those of Rauch et al.  and indicate a stronger automatic amygdala response to facial expression in schizophrenia. Initial hyperactivation of the amygdala in schizophrenia may indicate greater automatic encoding and an increased allocation of resources for processing facial expressions. It is well known that the amygdala modulates vigilance to enhance subsequent information processing throughout the brain , such that it represents an integral part of a cerebral network for unconscious emotional vigilance .
As hypothesized, patients showed an initial bilateral amygdala hyperresponsivity to masked neutral faces compared to healthy controls. Several previous studies have demonstrated that schizophrenia patients exhibit amygdala hyperactivation to neutral faces in paradigms using unmasked face stimuli [15, 24]. The present findings suggest that this potential neural correlate of a generalized “fear of faces”  also emerges under automatic processing conditions. Therefore, previous results in studies using neutral faces as a baseline condition may be distorted [3, 15]. Against this background, it is not surprising that Das et al. , who used neutral faces as a baseline, observed amygdala hypoactivation to masked fearful faces in schizophrenia patients compared to healthy controls.
Our hypothesis that amygdala activation in response to all facial primes would be reduced in the second compared to the first phase of the experiment was confirmed by the observed main effect of phase. These effects of time could be due to habituation upon the second presentation of the stimuli. Repetition suppression has previously been observed in the amygdala for subliminally presented facial emotions , and repetition-related reductions in amygdala responses to emotional stimuli have been demonstrated [19, 20, 27]. According to our data, the amygdala response to facial expressions rapidly habituates, even when faces are presented outside of subjects’ conscious awareness.
Our fMRI results corroborate the hypothesis that schizophrenia patients manifest a stronger decrease in the (right) amygdala response to facial stimuli over time than healthy individuals. During the first phase, patients showed stronger activation of the amygdala in response to faces than did controls, whereas in the second phase of the experiment, healthy individuals manifested stronger amygdala activation than patients. The decreases in amygdala activation observed in patients can be viewed in the context of their high levels of initial activation. The mechanisms underlying these effects of repetition and time may reflect an initial hypersensitivity to facial emotions and a subsequent down-regulation of activity. It appears that the processing of emotional faces is automatically suppressed over time in patients. This suppression could represent a compensatory mechanism for initial limbic hypersensitivity. The switch of patients from amygdala hyperactivation to hypoactivation (compared to healthy subjects) might reflect inhibitory processes. Blocking of amygdala responsivity might protect schizophrenia patients from overstimulation during social interactions.
Our findings on the automatic processing of emotional faces are consistent with observations from research on controlled emotion processing in schizophrenia. According to Holt et al.  and Salgado-Pineda et al. , schizophrenia patients exhibit amygdala activation that is similar or higher than in healthy individuals during the first stage of processing emotional information, and this activation subsequently diminishes more strongly than in healthy controls. Our data are in agreement with the assumption of Holt and Phillips  that emotional processing in schizophrenia could be characterized by an initial hyperresponsivity followed by hyporesponsivity of medial temporal lobe structures such as the amygdala.
To draw stronger conclusions, our finding of a switch from amygdala hyperactivation to hypoactivation in schizophrenia should be replicated. Further studies should examine the time-course of limbic activation during automatic as well as controlled processing of emotionally relevant stimuli in more detail. In addition, re-analyses of prior studies with a focus on time-course could provide valuable information. In particular, studies reporting amygdala under-recruitment during emotional face processing should be reconsidered [5, 28], as their results might be explained at least in part by the occurrence of suppression processes over time.
According to our behavioral data, patients’ ratings were negatively biased by angry facial expressions. In the initial phase, patients’ evaluative ratings of neutral mask faces tended to be more negative when preceded by angry faces than when preceded by no facial expression primes, compared to healthy individuals. These findings corroborate previous results from affective priming studies, suggesting that schizophrenia patients exhibit a stronger negative evaluative bias in response to covert negative facial expressions than do healthy individuals [12, 13]. The fact that affective priming due to masked angry faces decreased from the first to the second phase, independently of group, substantiates previous findings on the time-course of affective priming, with strong effects of masked emotional faces in initial presentations followed by substantial decreases in priming with repetition .
Some limitations of our study need to be acknowledged. First, our operationalization of time-course by dividing the experiment into two equivalent halves, with prime stimulus repetition between halves, was rather coarse. We analyzed ten events per condition and phase, but amygdala blocking might occur even earlier, possibly after only two or three presentations. More fine-grained experimental designs and time-sensitive methods such as EEG-informed fMRI might help uncover precisely when blocking processes begin. All patients in our study received atypical antipsychotics. Future research must clarify the effects of antipsychotic medication on the time-course of activation of the amygdala during emotional face processing. There is evidence from longitudinal fMRI research that treatment with atypical antipsychotics is associated with a reduction of the amygdala response during emotional processing in schizophrenia . Against this background, antipsychotic medication in our sample could have counteracted the initial hyperresponsiveness effects observed in the patients in our study, which might therefore be even stronger in unmedicated patients. Furthermore, our sample size was too small to warrant subgroup analyses based on predominant negative or positive symptoms. A promising future direction may be to compare the neural underpinnings of automatic emotion processing in patient groups suffering primarily from anhedonia, affective flattening, or paranoid ideation. A plausible hypothesis is that the initial amygdala hyperactivation is most pronounced in patients suffering predominantly from positive symptoms such as persecutory delusions. Moreover, one could assume that the observed blocking of the amygdala response might be strongest in patients exhibiting severe affective negative symptoms.