Social threat perception and the evolution of paranoia

Abstract

Rapid and efficient judgments about the significance of social threat are important for species survival and may recruit specialized neurocognitive systems, consistent with biological models of threat processing [1]. We review cognitive, psychophysiological, neuropsychological, and neuroimaging evidence in support of specialized neural networks subserving the processing of facial displays of threat. Cognitive research suggests that faces depicting anger are detected quickly when presented amongst other facial expressions, on the basis of distinguishing facial features. Psychophysiological investigations using visual scanpath techniques provide evidence for increased foveal attention to facial features of threat-related expressions (anger, fear), which may facilitate rapid detection and subsequent appraisal of the significance of threat (such as the direction of impending threat). Neuropsychological and neuroimaging studies implicate a primary role for the amygdale and pre-frontal cortices in interpreting signs of danger from facial expressions and other social stimuli. Subtle disturbances in these neurocognitive systems underlying efficient threat detection (manifesting in attentional biases and aberrant neural activity) may result in abnormally heightened perception of social threat, as seen in clinical levels of social anxiety and/or persecutory delusions in schizophrenia. Clinical states of paranoia may therefore reflect normal variation (i.e. biases) in the adaptive mechanisms which have evolved, in the Darwinian sense, to facilitate efficient threat detection in humans. As such, clinical levels of paranoia may represent the inevitable cost of efficient threat perception—or ‘justified’ suspicion—that is necessary for survival of the human species.

Introduction

Accurate interpretation of emotions conveyed through facial expressions is central to effective social communication, and may encompass distinct perceptual and neurocognitive processes according to the emotion [2]. Considerable physiological [3] and cross-cultural [4] research has established the existence of basic facial expressions of emotion (see Ref. [5] for a review), and recent neuroimaging evidence has provided support for the existence of discrete neural systems subserving the perception of negative emotions such as fear, anger, and disgust [6], [7]. The influence of anxiety and mood upon the ability to detect and respond appropriately to emotion conveyed through facial expression has been well-documented, with increased attention to negative facial expressions (and decreased attention to positive expressions) reported in depression [8], [9], and increased pre-attentive orienting to threat-related facial expressions associated with social anxiety [10], [11], [12], [13], [14], [15] and delusional ideation [16].

Rapid detection of threat in the social environment is critical for species survival. The neurocognitive mechanisms responsible for fast and efficient threat detection may thus have survived as an adaptive advantage in accord with Darwinian evolution theory [17], and consistent with recent models of threat perception [1]. Studies in healthy subjects indeed suggest that the detection of angry facial expressions is fast and efficient [18], [19], [20], [21]. In line with these behavioural data, neurobiological evidence supports a primary role for the amygdala in initiating a rapid response to threat, while slower cortical processing serves to inhibit the limbic response following conscious appraisal of the significance of threat [1].

Here, we review studies which have examined the neurocognitive basis of threat perception, including cognitive research, psychophysiological studies employing visual oculography as an overt index of attention to threat, and studies which have revealed a distinct pattern of neural activity when viewing threat-related social information (e.g. faces depicting emotional expressions of fear and anger). We have focused upon studies examining the neurocognitive basis of threat perception not only in healthy populations, but also within psychiatric patients who experience inappropriately increased anxiety and threat perception in everyday social contexts: in particular, schizophrenia patients with abnormal beliefs (delusions) and hallucinations of a persecutory nature.

Cognitive studies in healthy individuals suggest that faces depicting emotional expressions of anger are detected more rapidly than faces depicting other (non-threatening) expressions. For example, angry faces are detected faster in homogeneous crowds of discrepant faces in visual search tasks employing schematic face stimuli [18], [19], [20]. Furthermore, when displays contain the same faces (congruent affect), detection of the absence of an incongruent face is slower if all the faces display angry (or sad/angry) rather than happy expressions [18], suggesting that angry faces engage attention. However, discrepant findings in visual search tasks have been noted, which suggest that the faster detection of angry schematic faces may be due to low level features of the stimulus (e.g. contrast artifact), rather than threat detection per se [22].¹ Consistent with the majority of visual search studies, a recent investigation of the time taken to identify basic emotional expressions revealed a processing advantage for facial expressions of anger in healthy individuals; that is, angry faces were identified more quickly than other non-threat facial expressions (happy, sad, neutral faces) [21].

Whilst this evidence suggests that healthy individuals detect and identify facial expressions of anger with rapid efficiency, several studies of facial affect recognition in schizophrenia have revealed greater differential impairment for the recognition of negative emotions of fear [25], [26], anger [27] and sadness [25], [28]. Of course, this finding has been revealed in the context of a long history of evidence for a broader impairment in facial affect recognition in schizophrenia [26], [29], [30], [31], [32], [33], and support for a differential deficit that is specific to negative emotion recognition remains equivocal [30], [34] (indeed, some studies report no evidence for any deficit in facial affect recognition in schizophrenia [35], [36]). Specific impairments in the identification of negative facial affect may therefore not be robust, or expressed in all schizophrenia patients. Indeed, one study examining schizophrenia patients on a case-by-case basis demonstrated that the recognition of threat-related emotions thought to rely upon intact amygdala function was impaired overall, yet no universal emotional identification deficits were demonstrated, and not all patients demonstrated specific deficits in the recognition of threat-related expressions [37].

The specificity of emotion recognition deficits in schizophrenia therefore requires further investigation in studies which examine the processing of particular emotional states in association with symptom variations in schizophrenia. For example, it may be that only sub-groups of schizophrenia patients with negative symptoms (i.e. anhedonia, social withdrawal) demonstrate impaired recognition of negative facial affect such as sadness, while other symptom profiles may be associated with aberrant processing of other emotions in various ways. For example, in contrast to the evidence for impaired recognition of negative facial affect in heterogeneous schizophrenia groups, there is robust evidence for superior emotion perception in paranoid schizophrenia patients (i.e. those experiencing predominantly persecutory delusions and/or hallucinations, compared with non-paranoid patients) [38]. This finding has been reported specifically with regard to negative emotions [38], [39], [40], [41], [42], and suggests that heightened perception of certain negative emotions may be relevant to the genesis of persecutory delusions.

Investigations of cognitive biases in schizophrenia have accumulated to implicate biased processing of threat-related information in relation to persecutory delusions. Some of the first investigations of this type employed a variation of the original Stroop [43] procedure, to investigate pre-attentive processing of words referring to delusional themes (rather than colour words). Using this paradigm, schizophrenia patients with persecutory delusions are slower in naming the ink-colour of ‘paranoid’ or threat-related words, versus meaningless strings of O's, neutral words, or words signifying negative affect, compared with depressed and normal controls [44]. These findings have been interpreted as evidence for heightened pre-attentive processing of threatening information in schizophrenia patients with persecutory delusions.

In attempting to demonstrate a similar bias in selective attention, two studies examined memory for threatening information in patients with persecutory delusions [45], [46]. In one study [46], subjects listen to stories that differed in terms of threatening content; when asked to recall as many propositions from the stories as possible, persecutory deluded subjects recalled more of the threatening propositions from the stories than the healthy control group. Similarly, Bentall et al. [45] asked schizophrenia patients with persecutory delusions, depressed patients, and healthy participants to recall items from a list of threat-related, depression-related and neutral words. The deluded participants displayed better memory for threat- and depression-related words than the normal controls, and demonstrated a significant tendency to repeat threat-related words during recall. In a related study [47], groups of patients suffering from persecutory delusions, depression, and healthy controls were asked to rate the frequency with which selected positive, negative, and neutral events had actually happened to them, as well as the predicted likelihood of these events happening to them in the future. All patient groups rated negative events as occuring relatively more frequently in comparison to the control subjects, and levels of magical ideation (e.g. paranormal beliefs) were associated with estimates of the frequency of negative events for both self and others. Collectively, the studies by Bentall and colleagues suggest that persecutory delusions are associated with biased attention and memory for material associated with personal threat; memory biases may be a consequence of deeper encoding of threatening material due to increased pre-attentive processing of threat, thus facilitating greater recall and expectation of threatening information and personal experiences.

Two recent studies have also investigated the time taken to recognize facial emotions in delusion-prone individuals and schizophrenia patients with persecutory delusions. Both delusion-prone and clinically deluded individuals were significantly slower to identify angry faces, reflecting a lack of the processing advantage shown by healthy control subjects [21], [48]. The delayed processing of angry faces in relation to delusional ideation may reflect a prolonged, conscious evaluation of threat that is not undertaken during the normal efficient and automatic processing of directly threatening social material. Persecutory delusions may therefore be associated with an extended early processing of threat, or problems disengaging threat from conscious attention. Notably, recent studies have shown delayed disengagement of threat-related faces in association with high trait-anxiety [49] and delusion-proneness in healthy participants [16]. Again, it is plausible to suggest that this extended processing of threat may result in deeper encoding of threat-related information in long-term memory stores, coinciding with heightened recall of threatening material and higher expectation of negative events in paranoid schizophrenia as reported in previous studies (reviewed above).

Visual scanpaths comprise voluntary eye movements (saccades) and foveal fixations, and represent a psychophysiological measure of visual attention [50]. During examination of a visual stimulus, rapid eye movements direct the fovea to areas of interest, where ‘fixations’ occur. Fixations involve the fovea remaining relatively stationary on a particular region of the stimulus, usually for a few hundred milliseconds, depending upon task requirements [51]. Foveal input during fixations provides the visual system with detailed information about the stimulus, while cognitive processes operate in parallel, using parafoveal and peripheral retinal information to determine the location of the subsequent fixation, a process termed visual scanning [52].

Studies using brief (less than 200 ms) tachistoscopic presentations of objects have shown that coarse discrimination of visual stimuli occurs very quickly, but that the perception of finer details about the stimuli are governed by later, controlled processes of selective attention [53]. The distinction between controlled and automatic processing [54] has been prominent in the study of cognitive processes in both normal and schizophrenic cognition [55]. Controlled processes involve a slow sequence of temporary mental operations that are under voluntary control of the individual, placing demands upon limited attentional resources. Automatic processes, on the other hand, involve a fixed sequence of mental operations occurring relatively quickly in response to a particular stimulus, which occur outside of conscious awareness and involve access to long-term memory [55]. The study of visual scanpaths thus enables the investigation of information processing during both controlled stages of selective attention, and automatic processing of sensory information [56]. In terms of face processing, visual scanning paradigms can therefore be used to assess automatic sequences involved in early perceptual processing (e.g. recognition of an object as a face), as well as providing a real time measure of controlled attention to stimulus features (i.e. particular facial regions) during later stages of face processing that may be necessary for the discrimination of particular facial expressions (such as those depicting direct threat).

The first study of visual scanpaths to facial expressions depicting threatening emotions was conducted with primates [57]. In this study, macaque monkeys spent more time attending to the eye regions of con-specific and human faces exhibiting obvious displays of social threat (i.e. furrowed brow, bared teeth). These authors concluded that increased attention to specific facial features (e.g. eyes) may be useful to detect the significance (i.e. directed at self versus others) of impending social threat.

Following from this study, a recent investigation has revealed the existence of distinct visual scanpath strategies in human subjects for threat-related (i.e. anger, fear) facial expressions [58]. Visual scanning of threat-related faces was characterized by increased fixations to feature areas (i.e. eyes, nose, and mouth) and extended lengths between fixations, when compared with non-threat facial expressions that included both emotion-laden (happy, sad) and non-emotional (neutral) faces. The increased attention to feature areas of threat-related faces may thus facilitate the detection of threat from facial expressions, consistent with previous behavioural research that attests to the role of feature-based detection of threat from facial expressions of anger and fear [23], [24].

Increased attention to the features of threat-related faces may also provide information about the spatial location of impending threat [57], consistent with cognitive research showing the normal tendency to attend in the direction of another person's gaze [59]. This pattern of visual appraisal is consistent with the way that monkeys view expressions of facial threat, suggesting that increased attention to the feature areas of threat-related expressions may be have evolved to facilitate cognitive appraisal of the significance (i.e. identification of threat as direct?) of impending social threat. Furthermore, the extended scanpaths (longer distance between fixations) exhibited by human subjects might reflect normal ‘vigilance’, or excessive monitoring of salient cues for social threat.

Further investigations of visual attention to social threat in delusion-prone and clinically deluded individuals have revealed an overt avoidance of fixating upon facial threat in healthy individuals with a vulnerability to delusional thinking [58], and schizophrenia patients with persecutory delusions [60]. That is, delusion-prone individuals exhibited fewer fixations and extended scanpaths for faces depicting anger and fear, compared to age matched controls with low scores on a measure of delusional ideation [58]. Similarly, schizophrenia patients with predominant persecutory delusions spent less time viewing (i.e. showed fewer fixations of shorter duration) the feature areas of negative facial expressions (anger, sad), compared to healthy controls [60], and showed significantly fewer fixations to expressions of fear when compared to a non-deluded schizophrenia group. The pattern of viewing directly threatening facial expressions (i.e. anger) in delusion-prone and deluded schizophrenia subjects thus reflects a controlled attentional bias away from threat.

Reduced visual scanning of the feature areas of threat-related faces in deluded schizophrenia patients suggests that regions such as the eyes may be particularly threatening for paranoid individuals, consistent with an earlier study in which schizophrenia patients with persecutory delusions spent less time viewing the feature areas of neutral faces during an identity recognition task [61]. Of course, it is important to consider these findings in the context of previous studies of visual scanpath aberrations in schizophrenia, in which a robust pattern of ‘restricted’ scanning (comprising fewer fixations of longer duration and shorter distances between fixations) emerged consistently when viewing neutral faces [61], [62], [63], [64], [65] and emotional faces [66], [67]. This pattern of restricted visual scanning in schizophrenia has also been demonstrated when viewing complex geometric and scene stimuli, most commonly in association with negative symptoms [68], [69], [70], [71], [72]. By contrast, an opposing style of ‘extensive’ scanning (characterized by fewer fixations of shorter duration and longer distances between fixations) was associated with positive symptoms (e.g. delusions) of schizophrenia [73], [74] in early investigations of relatively long periods (e.g. 1 min) of viewing complex scene stimuli.

There have been two other visual scanpath studies of threat perception in deluded schizophrenia, which employed complex stimuli comprising photographs of social scenes depicting various degrees of threat [75], [76]. Freeman et al. [75] found that patients with persecutory delusions displayed reduced visual appraisal for ‘happy’ and ‘potentially threatening’ scenes, compared to scenes depicting ‘direct’ and ‘hidden’ threat (which were viewed in a similar fashion to control groups). In the study by Phillips et al. [76], participants viewed social situations rated as neutral, ambiguous, or overtly threatening in content. During an initial free-viewing task, all subjects viewed an overtly threatening scene to a greater extent than an ambiguous (potentially threatening) scene, but schizophrenia patients with persecutory delusions demonstrated reduced appraisal of threatening areas of the ambiguous scene, and a corresponding increase in the number and duration of fixations to non-threatening areas. Additionally, deluded participants repeatedly viewed (checked) threatening foreground areas of all scenes less than normal and psychiatric controls.

These findings using social scene stimuli suggest that persecutory delusions may be associated with the perception of threat in inappropriate places within ambiguous contexts, while the reduced re-appraisal of foreground areas might reflect hasty decision-making about the content of the scenes. Moreover, the results do not support a direct association between persecutory delusions and increased attention to threat, as suggested by earlier studies of pre-attentive cognitive bias toward threat-related words during emotional Stroop tasks [44]. Notably, scenes are necessarily more ambiguous and complex than the word stimuli used in emotional-stroop tasks. In summary, scanpath studies show that deluded schizophrenia patients pay comparatively less attention to the salient features of threat-related faces [58], and to the threatening areas of social scenes [75], [76].