DISCUSSION
This study investigated whether neuroticism affects ocular exploration of emotional faces and whether this can be modulated by antidepressant drug treatment. Using a gender-discrimination task of fearful and happy faces with different emotional intensity, we found that: (a) subjects with high neuroticism present a different pattern of ocular exploration of faces compared with subjects with low neuroticism, characterized by reduced gaze maintenance over the eyes region of the face, irrespective of emotional expression; (b) short-term repeated SSRI administration increased eye-gaze maintenance and scanning time over all facial expressions, irrespective of emotion type and intensity, compared with placebo in a high-risk sample with high neuroticism. The same treatment also reduced the spatial extension explored by eye movements over faces of medium emotional intensity, and this was significant for fearful faces. Moreover, (c) eye-movement pattern predicted performance on facial expression recognition on a separate task. These results occurred in the absence of any subjective modifications in mood or anxiety (Supplementary Material), consistent with the model by which antidepressants' action on the processing of emotional information precedes changes in affective state (Pringle et al, 2011).

Effects of High Neuroticism on Ocular Exploration of Face Stimuli
Our study indicates that high neuroticism is characterized by an attentional bias away from salient face regions, which is not affected by emotional valence of the facial expression and is indexed by an abnormal pattern of ocular face exploration. High Ns avoided holding their gaze over the eye region for the whole trial duration, and general face avoidance was significantly correlated with self-rated hostility. This suggests that High Ns may avoid engaging with socially salient stimuli such as faces, reflecting a self-perceived negative disposition towards interpersonal contact. Previous studies have reported a similar pattern in socially phobic patients (Horley et al, 2003). Although we did not collect measures of social anxiety, stress during social interactions is a well-recognized feature of the high neurotic phenotype, contributing to the higher risk for social anxiety (Clark et al, 1994; Watson et al, 2008). Interestingly, this avoidant ocular exploration pattern is consistent with reduced neural response to faces (irrespective of emotional expression) in the amygdala in the same population (Di Simplicio et al, 2013).
Differences in the task demands could explain the discrepancy between our findings and that by Perlman et al (2009), who reported that increased scanning over the eye region, in particular of fearful faces, is associated with high neuroticism during an overt emotion discrimination task. It is possible that when explicitly cued to respond to the emotional expression of the face, highly neurotic subjects can orient their attention to the eyes, which they otherwise avoid when less relevant to their contingent objective. If this were true, the increased eye region exploration over fearful eyes found by Perlman et al (2009) could represent a task-driven strategy, consistent with a recent meta-analysis showing that heterogeneity of results from attention orienting and maintenance data is often context dependent (Armstrong and Olatunji, 2012). Another explanation could lie in our sample focused on High N individuals, whereas Perlman et al (2009) using a continuous measure: ‘extreme' neuroticism, is more likely to present prominent socially anxious features, which could explain an avoidant pattern of eye region exploration (Gamble and Rapee, 2010).
Together, these data suggest that high neuroticism is associated with inadequate capacity to maintain attention over key areas of face stimuli, which could in turn underpin interpersonal difficulties often presented by this group. This may contribute to increased risk of developing anxiety consistent with clinical models (Clark and Wells, 1995) and to lower quality of life, influenced by hypersensitivity to interpersonal stressors and lack of social support (Kendler and Gardner, 2014; Lahey, 2009). Of note, neuroticism is a heterogeneous and multifaceted construct. Our task of covert attention towards emotional facial expression with a short stimulus presentation allows a global assessment of emotion-processing biases, more in particular to those related to general anxiety (hypervigilance—avoidance of negative stimuli) and social anxiety (avoidance of the eye region of faces). Future studies could employ larger samples and identify population subgroups with high neuroticism characterized by more prominent dysphoric, generally anxious, and socially anxious traits. This could be combined with tasks showing concurrent negative and positive facial expressions and longer stimulus presentation in order to investigate more extensively the presence of attentional biases associated with anxiety and depression, respectively.

Effect of SSRIs on Ocular Exploration of Face Stimuli
We also found that SSRIs can modulate attentional biases towards faces in a high-risk sample early in treatment, by directly increasing gaze maintenance over the internal features of faces. This suggests that an initial SSRIs' effect in anxious individuals could be to partly correct their dysfunctional pattern of ocular exploration of faces.
Initial attention allocation to salient stimuli is associated with amygdala activation (Vuilleumier, 2005). The present finding is therefore consistent with neuroimaging data showing that short-term repeated SSRI administration to High Ns increased the amygdala signal to facial expression irrespective of emotion type (Di Simplicio et al, 2013). The causal mechanism operating between amygdala activation and eye gaze movements remains unclear. The first neural response could precede attention deployment, guiding the eyes to explore the detected stimuli (Kennedy and Adolphs, 2011), but time spent gazing over the stimuli could equally increase the limbic neural signal (Pessoa and Adolphs, 2010). SSRI-driven modifications could occur at one or both stages of this process.
Notably, in the citalopram-treated group prolonged attention to faces (as measured by the ability to hold the gaze on the face for the entire trial and by the time spent scanning faces) was predictive of better performance in the gender discrimination task and to a more accurate discrimination of positive expressions in a separate facial expressions recognition task. High Ns find happiness recognition particularly difficult compared with non-vulnerable samples (Chan et al, 2007), and the current data suggest that recognition of positive emotional facial expressions improves early with SSRI treatment in this group. Although our data cannot make any causal inference, they suggest that modulation of basic eye-movement and face-exploration parameters by SSRIs are associated with correction of this negative bias in emotion recognition. Further investigations should confirm whether the capacity of different antidepressant agents to shift biases in facial expression recognition in unselected populations (Pringle et al, 2011) is similarly associated with effects on ocular exploration.
By contrast, short-term repeated SSRI administration did not modify the preference for exploring the mouth rather than the eye region seen in High Ns in Study 1. This suggests that antidepressants may produce a non-specific effect on gaze maintenance over faces. In a previous research, 2 weeks of citalopram administration in healthy volunteers increased eye contact compared with placebo in an interaction task with a stranger (Tse and Bond, 2002). Future studies with samples also characterized on social anxiety dimensions, more naturalistic experimental settings, and longer treatment duration could clarify whether the SSRI-mediated increase in face exploration after 1 week would also facilitate with time the reversal of the eye region avoidance when this is present.
Interestingly, citalopram-treated individuals also tended to spacially explore the face area less compared with placebo-treated ones, but this effect appeared prominent for fearful faces of medium intensity. Anxious individuals have been shown to spatially explore larger areas of faces while holding the eye gaze for shorter time duration over the face (Horley et al, 2003). This could reflect a ‘scanning style' of faces that avoids to properly stop over those cues holding potential danger and can impair accurate processing (Armstrong and Olatunji, 2012). This result should be treated with caution given the relatively small sample, with a medium effect size of the comparison. Accordingly, in our sample the spatial extension of eye movements was negatively correlated with gender discrimination accuracy. Hence, the combination of SSRI-induced changes in ocular exploration over fearful expressions and over all faces could lead to a more functional attention deployment toward socially salient stimuli, contributing to the long-term anxiolytic effect of antidepressants. If this were true, antidepressant action would appear to occur via an ‘exposure' mechanism similar to cognitive-behavioral treatments, as we have previously suggested (Di Simplicio et al, 2013).

Summary
In summary, our results demonstrate that short-term antidepressant administration can modify the pattern of attention for facial stimuli by changing ocular movements exploring faces in subjects at risk for depression and anxiety, before any changes in affect. Our data illustrate how antidepressant medication could reduce the difficulties in extracting information from facial stimuli by increasing evaluative capacity. This can have different clinical impact depending on preexisting attentional biases. It might temporarily increase anxiety in subjects hypervigilant for threat, but it can also increase positive affect in subjects generally avoidant of social contacts and thus deprived of positive interpersonal feedback. This is consistent with the variable responses to SSRI treatment in clinical practice, including the so-called early ‘anxiogenic effects' (Sinclair et al, 2009; Yeragani et al, 1992). Further studies will need to verify whether our findings can be extended to clinical populations and to investigate how modification of ocular exploration of socially salient stimuli influences remission of anxious and depressive symptoms over time.