In accordance with theories suggesting a critical function of the amygdala in the processing of threat signals and the mediation of fear responses 12, several studies found increased amygdala activation to threatening vs neutral stimuli in individuals with anxiety disorders 345678 and in healthy subjects 91011121314. Furthermore, there are strong theoretical accounts proposing an automatic response of the amygdala to threat signals even when target stimuli are presented during attentional distraction 1214. Whereas some studies indeed suggest an automaticity of amygdala activation to threat-related stimuli under conditions of attentional distraction 91314, several recent studies in healthy subjects, however, indicated a complete inhibition of differential activation to threat vs neutral stimuli within the amygdala, given sufficiently strong perceptual load by a main task 15161718. Thus, it seems that, at least in healthy subjects, automatic activation of the amygdala to emotional stimuli does not occur when demanding cognitive tasks exhaust the available processing resources.

Bishop et al., for example, used a perceptual load task, while subjects were exposed to fearful and neutral faces. Perceptual load was induced by varying the number of task-relevant items 1920 within a letter string presented along with the facial expression. When perceptual identification was easy (low load), elevated state anxiety was associated with a heightened response to threat distractors in the amygdala and superior temporal sulcus, whereas individuals scoring high in trait anxiety showed a reduced prefrontal response to these stimuli. The latter finding was interpreted as weakened recruitment of control mechanisms when confronted with salient distractors. This finding is in accordance with theories assuming an imbalance between the stimulus-driven processing of salient threat-related stimuli, associated with automatic orienting, and goal-directed attentional control (for example, 21). This would lead to a relatively stronger role of the posterior attentional systems in the brain involved in bottom-up attention as compared to the more anterior top-down control system 22. However, in the study of Bishop et al., neither high-anxious nor low-anxious subjects showed an increased amygdala response to threat distractors when perceptual identification was more attention demanding (high perceptual load). Furthermore, high attentional load in previous studies also prevented differential activation to threat vs neutral stimuli in areas of the extrastriate visual cortex, suggesting the absence of differential processing of threat and neutral stimuli also in areas beyond the amygdala 151723.

Thus, in line with a recent model of selective attention 1920, processing of task-unrelated stimuli is prevented when task-related demands exhaust perceptual capacity limits. Even though this model does not predict that also the processing of salient emotional stimuli is impaired by high perceptual load 24, it has been extended to the domain of threat processing 1523. Furthermore, based on the findings of Bishop et al., effects of subjects' anxiety on the neural processing of threat-related stimuli seem to appear only during relatively low-load tasks. Thus, high load should prevent neural responses to threat and also the attentional processing of these stimuli, that is automatic orienting 22. This position is in contrast to cognitive models of anxiety 25 predicting a mandatory processing of threat stimuli in anxious subjects or models predicting that anxiety increases the processing of threat-related signals under high demands on the central executive 21.

Even though it has been shown that high perceptual load prevents the processing of threat stimuli in anxious healthy subjects, it is unknown whether similar findings will be observed in individuals suffering from an anxiety disorder. Automatic processing of disorder-related stimuli seems to be a main feature of anxiety disorders and this might be represented in attention-independent activation of the amygdala 128. An example is specific phobia, which is among the most common anxiety disorders 26. Neuroimaging research implicates the amygdala in the processing of phobia-related stimuli, specifically in the initial detection of such stimuli and perhaps in the lowering of thresholds for the induction of rapid fear responses, rather than in the sustained processing of phobia-relevant information 82728. For example, activation of the amygdala in spider-phobic subjects has been demonstrated regardless of whether attention was focused on the stimuli or distracted by an unrelated foreground task 8, supporting the hypothesis that the amygdala is automatically activated by phobogenic stimuli 1229. Furthermore, this attention independent response in the amygdala was associated with increased activation in the extrastriate visual cortex 8, which is typically coactivated with the amygdala in spider phobia in response to phobia-related stimuli (for example, 3303132).

A recent study with spider-phobic subjects reported attention-dependent activation of the amygdala to spider pictures 33. However, in this study the number of phobia-related stimuli and attention focus were confounded, making a clear interpretation of the results difficult. Thus, the findings might even be interpreted to support the hypothesis of automatic amygdala activation to task-irrelevant (background) spider pictures. To date, there has been no functional imaging study that employed a parametric variation of attentional load in individuals with specific phobia or any other anxiety disorder.

In the present study, we used event-related functional magnetic resonance imaging (fMRI) to explore the question whether amygdala activation to phobia-relevant stimuli is modulated by a parametric variation of attentional distraction in patients with specific phobia. We used a perceptual load task that has been previously shown to inhibit amygdala activation to threat-related stimuli in high-anxious healthy subjects 15. An absence of attentional modulation of amygdala activation in the present experiment would indicate a role of the amygdala in threat processing even under high attentional load in individuals with anxiety disorder. Additionally, we examined the neural activation in the visual cortex and several brain areas proposed to be involved in the processing of threat-related stimuli.

The results show a pattern of simultaneously increased amygdala and visual cortical activation to threat vs neutral pictures in phobic individuals, compared with controls, occurring regardless of attentional load. These findings suggest that amygdala activation to clinically relevant threat stimuli is resistant to attentional load.

The present study provides evidence for a critical involvement of the amygdala in threat processing under attentional load in subjects suffering from an anxiety disorder. Thus, amygdala activation to disorder-related vs neutral stimuli was observed regardless of attentional load. A similar finding was evident for the left fusiform gyrus.

This finding contrasts with prior studies in healthy subjects 15161718, especially with a study where the same task resulted in a strong modulation of amygdala activation to fearful vs neutral faces, with the differential amygdala activation completely inhibited under the high-load condition 15. The study of Bishop et al. found that, during a low-load condition, when perceptual distractor identification was less demanding, elevated state anxiety was associated with a heightened response to fearful faces in the amygdala and superior temporal sulcus, whereas individuals scoring high in trait anxiety showed a reduced prefrontal response to these stimuli. The latter finding was interpreted to indicate a weakened recruitment of cortical control mechanisms in anxious individuals when confronted with threat distractors. However, neither high-anxious nor low-anxious volunteers showed an increased amygdala response to threat distractors when the perceptual identification task was more attention demanding (high perceptual load).

Thus, it seems that the findings in subjects showing high, but subclinical, levels of state and trait anxiety may not necessarily be comparable to those of a subject sample meeting the diagnostic criteria for a clinically relevant anxiety disorder such as specific phobia. This difference suggests an increased responsiveness of the amygdala to threat signals in anxiety disorder patients. This increased responsiveness might be mainly associated with differences in the threat relevance of stimuli used in the different studies. Thus, while fearful facial expressions are associated with rather low anxiety ratings, disorder-related stimuli evoke strong fear responses in subjects suffering from an anxiety disorder. Furthermore, our findings are also not associated with trait or state anxiety scores of subjects, since there was no difference between groups. Thus, the use of increased trait or state scores as analogue to clinical disorders might be often of limited relevance. Rather, it may be the disorder-related importance of the stimuli that determines differential brain responses, at least in phobias.

Our results support previous findings of amygdala activation to threat under conditions of attentional distraction in specific phobia 833. However, no previous study in individuals with anxiety disorders employed a parametric manipulation of attentional load as yet. Also, prior work may differ in that the distraction conditions might not have been very demanding 8, or that they were confounded with other factors 33. The present results suggest that, at least in specific phobia, the salience of stimuli evokes differential amygdala activation to threat vs neutral stimuli independent of attentional load, even though the amygdala and other areas were found to be modulated by attentional load in general. Thus, high load led to decreased activation of the amygdala and several other brain areas. Conversely, regions implicated in attentional control and dealing with task difficulty showed increased activation under high as compared to low load. This general effect of load or attentional distraction is in line with prior work 151723.

We did not detect any evidence for a decreased prefrontal control of threat distractors as suggested by Bishop et al. However, one has to keep in mind that the results in the Bishop et al. study are based on a correlation with trait anxiety scores and a comparable significance of the facial expressions for all subjects (high and low anxious). Here, we compared subjects with anxiety disorder to healthy controls. That is, for spider-phobic subjects the spider pictures were disorder related, while for the control group the (attentional control of these) pictures had no relevance. This prevents a meaningful comparison of differential control mechanisms between groups.

Beyond its role in the rapid induction of defense behaviors 12, the amygdala might also be involved in attentional functions 242434445, for example, through the modulation of activation in visual areas by feedback connections 46. This influence of the amygdala might allow the enhanced perception of threat 47. Accordingly, it has been shown that the amygdala drives the activation of areas within the inferior temporo-occipital cortex such as the fusiform gyrus 14 and increased activations to threat even under distraction conditions or perceptual unawareness have been found in visual areas 4849. In line with these findings, our data revealed a significant activation of the fusiform gyrus to spider vs neutral stimuli in spider-phobic subjects occurring in conjunction with the amygdala activation during both attentional conditions.

The amygdala's influence on attentional functions is not specific for anxious subjects or anxiety disorders, but can be found in healthy subjects as well (for example, 434450). Animal research also implicates the amygdala in forming a crucial part of a pervasive vigilance system subserving facilitated processing of biologically relevant information 455152. Thus, the meaning of automatic amygdala activations for phobic symptomatology might be associated with such functions. Individuals suffering from specific phobias show increased vigilance for phobia-relevant stimuli 4853. Under divided attention conditions, the amygdala might be activated even by crude representations of threat stimuli requiring the brain to gather more information by potentiating subsequent sensory information processing.

It should be noted that we do not suggest that these findings are necessarily specific for the processing of phobogenic stimuli. Rather, the processing of phobogenic stimuli represents a highly aversive condition and might be a specific case where personally relevant and salient aversive stimuli are processed even during high perceptual load. Generally, we suggest that activation of the amygdala and visual cortex is due to the interplay between the saliency of stimuli and available cognitive resources. Thus, other threat stimuli might be processed in non-clinical populations as well, given that the saliency and the personal importance of these stimuli are sufficiently high. Future studies should use negative and positive affective control stimuli in order to disentangle the general role of valence and arousal for amygdala responses under high perceptual load.

Furthermore, there was a remarkable reduction of the activation of the amygdala by high load regardless of group and picture category. This is in accordance with previous work (for example, 16) and provides clear evidence that even in the absence of emotional stimuli the activation in the amygdala is affected by attentional conditions. However, in our study, the differential activation to phobia-related vs neutral pictures was stable across load conditions, indicating a dissociation between a general decrease of the amygdala responsiveness regardless of the specific stimuli and intact relative increased amygdala activation to phobia-related vs neutral stimuli during high load.

Remarkably, there were no effects on task performance in spider-phobic subjects as compared to healthy subjects. However, this finding is in accordance with previous results 8. Furthermore, impairments in task performance are not consistently observed in subjects with phobias (for example, 785455). Moreover, for the kind of task used in the present study, Bishop et al. showed differential brain activation in anxious subjects that was not accompanied by indications of behavioral impairment. Thus, effects on brain activation can be dissociated from those on behavioral measures, at least when assessed through reaction times and errors. Future studies should investigate whether amygdala responses can predict other behavioral measures. Furthermore, it would be interesting to investigate whether automatic amygdala activations can be modified by successful psychotherapy and if these responses are associated with therapeutic success in the short and long term.

Our results indicate a hyper-responsiveness of the amygdala to disorder-related stimuli in phobic subjects that proved to be independent of attentional load when using a task which induces a high load and which has been shown to prevent amygdala activation to threat in high anxious subjects. This suggests that anxiety disorder patients are characterized by a high level of automaticity of their amygdala responsiveness. Although we did not find an effect of perceptual load on differential amygdala responses, future work might aim to investigate whether a further increase of perceptual load may result in different outcomes as revealed in this study. Thus, also in anxiety patients, the amygdala response to threat might be characterized by a relative instead of an absolute automaticity.

The authors declare that they have no competing interests.

TS participated in the design and the data analysis of the study and drafted the manuscript. AS and JL carried out the experiments. AS and MML established the experimental procedures. AS performed the data preprocessing and analysis and wrote parts of the manuscript. WM participated in the development and coordination of the study. All authors read and approved the final manuscript.