Emotion-laden stimuli include those involving emotional expressions or affective scenes, in addition to originally neutral items that might have acquired affective significance by previous pairing with aversive events (e.g. pairing with mild shock). It is hypothesized that affective significance impacts both perceptual competition and executive control (). Perceptual competition, which takes place in visual cortex, is affected because emotional content enhances sensory representations of emotional items (, arrow 1), which is well documented in human visual cortex [1
]. Such enhancement depends, at least in part, on output connections from the amygdala, which is known to project to multiple levels of visual cortex, including the primary visual cortex [15
Figure 1 Dual competition model framework. Affective significance impacts the flow of information processing both in a (a) `stimulus-driven' (note fearful face paired with shock as input) and a (b) `state-dependent' fashion based on motivational manipulations (more ...)
Executive control is affected by emotional content because; firstly, strengthened sensory representations will receive prioritized attention (, arrow 2). For example, items with increased visual responses can direct spatial attention towards those locations – this will occur as long as sufficient processing resources are available [16
]. Secondly, executive control is modulated because affective information might be directly conveyed to control structures (, arrow 3). For instance, amygdala outputs might convey the significance of an item via connections with anterior cingulate cortex (ACC) territories, which might help direct attention towards the location of the emotional item via connections with the dorsolateral prefrontal cortex (PFC) (see later). In this manner, the modulation of executive control eventually affects visual processing [17
] (, arrow 4).
The impact of an emotion-laden stimulus on behavior crucially depends on how it affects the flow of executive functions. It is hypothesized that this will depend on the level of threat, which, accordingly, will determine if emotional content enhances or impairs behavioral performance. When emotional content is low in threat, processing is biased in favor of the emotional item () – this situation also extends to positive stimuli [18
] (Box 1). In particular, the spatial locus of the emotional item is privileged, possibly because items that are low in threat are somewhat ambiguous and so might attract further attention as part of additional information gathering [19
]. In this manner, emotional content enhances target processing with relatively minor effects on irrelevant stimuli and other executive functions that might be needed (e.g. if task switching is involved). Thus, in the low-threat case, although emotional items are prioritized, the impact on behavior is modest – in this sense, it can be said that a `soft' prioritization occurs. Because the effect on performance is relatively weak, behavioral findings can be difficult to replicate and might be observed only in high-anxious individuals (e.g. Ref. 
). Furthermore, whereas low-threat emotional stimuli comprise a privileged stimulus category, their processing is highly dynamic and depends on the interplay of a host of factors that sculpt the associated neural responses, including attention, task context, awareness and perceptual interpretation [21
Figure 2 Executive control and competition are viewed as involving multiple mechanisms, or resources. Larger unfilled ellipses represent executive control; smaller shapes represent processing resources. (a) When threat level is low, affective significance enhances (more ...)
When emotional content is high in threat, resources are diverted towards the processing of the item. The mobilization of the resources is more extreme, and the effects on behavior considerably more dramatic [22
]. In this case, the main impact on behavior comes from the recruitment of attentional/effortful control that is required to prioritize the processing of high-threat information () – thus, `hard' prioritization occurs. In particular, attentional/effortful control is envisaged as involving processing resources that are strongly shared by several executive functions (see also Refs. [24
]). Because high-threat is expected to recruit such `common-pool resources', it will impair other executive functions that are reliant on them, including inhibition, shifting and updating. For instance, in a recent study, performance during response inhibition was compromised when participants viewed high-versus low-arousing pictures [27
]. Specifically, emotional scenes preceding both go and stop stimuli increased the stop-signal reaction time, a measure of the temporal evolution of inhibitory processes (see also Ref. 
). The processing of threat typically will require further actions and, in addition to the consumption of common resources, could involve the triggering of multiple mechanisms that are specific to the task at hand ().
Although the notion of resources has at times been viewed as vague [29
] (but see Ref. 
), one approach to understanding resource consumption could be to probe the correspondence of brain sites that are sensitive to specific experimental conditions. It is particularly instructive, for instance, to observe the overlap between attentional manipulations and those that are sensitive to higher levels of threat. The `attentional network' has been extensively researched and is believed to involve fronto-parietal regions, including the middle frontal gyrus (MFG), ACC, inferior frontal gyrus (IFG) and anterior insula [31
]. To assess brain regions that are sensitive to high levels of threat, the activation sites of the contrast of CS+ (i.e. stimuli paired with an unconditioned stimulus) versus CS− (i.e. stimuli never followed by an unconditioned stimulus) of 34 aversive conditioning studies were reviewed here. In addition to the amygdala, several frontal activation sites were consistently reported, including MFG, ACC, IFG and anterior insula (). Thus, it seems that high-threat processing engages key nodes of the attentional network, consistent with the notion that it is linked to resource consumption.
Figure 3 Processing resources and threat. Summary of results from 34 positron emission tomography (PET) and fMRI studies of conditioning from 1995 to 2008, illustrating the coordinates provided for the contrast of threat (CS+) versus safe (CS−). (a) Activation (more ...)
It is possible to further operationalize resource consumption by linking observed evoked functional magnetic resonance imaging (fMRI) responses and behavioral performance. For instance, in a recent experiment [33
], subjects performed a search task under low and high attentional demands (), which were contrasted to determine brain sites sensitive to the availability of processing resources. Differential responses (high versus low) were observed in several fronto-parietal regions commonly associated with the attentional network, including the ones previously listed. In the same study, subjects were also shown task-irrelevant threat and safe faces (). Interestingly, increased responses to threat versus safe faces were observed in several of the same fronto-parietal regions. To further test the idea that additional processing resources were recruited during the viewing of threatening stimuli (relative to safe), in a new analysis, we correlated evoked fMRI responses in the regions modulated by attentional load with behavioral accuracy during the task. As illustrated in , the higher the ACC recruitment during the threat condition, the worse the behavioral performance (relative to the safe condition; p<0.05). Interestingly, a similar pattern of results was observed in multiple regions, including MFG, IFG and anterior insula, in addition to superior parietal lobule (although the exact spatial overlap between attentional load and threat effects varied slightly for these regions). Consistent with the increased processing of shock-paired stimuli, such stimuli exhibited increased behavioral priming and fMRI repetition effects relative to unpaired faces during a subsequent implicit-memory task [33
]. These findings indicate that consumption of processing resources engaged by task-irrelevant threat faces (as indicated via, e.g. ACC responses) impaired performance on the main task.
Overall, interactions between high threat processing and executive functions are proposed to take place via at least three types of neural mechanisms (). Firstly, it is hypothesized that threat processing engages attentional/effortful control mechanisms in the ACC and, in particular, the dorsal site observed in the previous analysis (see inset in ) – in contrast to more rostral sites [34
]. The ACC is important for integrating inputs from multiple sources, including affective and motivational inputs [35
] – and in this respect works in close cooperation with the anterior insula and OFC [37
]. The ACC has also been suggested to be involved in conflict detection, error likelihood processing and error monitoring, and helps determine the benefits and costs of acting. It is suggested here that ACC engagement during threat will impair executive function because common-pool resources that are required to prioritize threat processing are taken up. Secondly, threat also recruits multiple PFC sites that are involved in specific
executive functions (, green arrows). This recruitment is suggested to depend, at least in part, on the ACC, whose signals are known to influence activity in other brain regions and to modulate cognitive, motor and visceral responses [35
]. For instance, the ACC might engage the MFG, which is important in the manipulation of information, among other important functions. In this manner, additional specific processing resources are diverted to the processing of threat information (, orange regions). Thirdly, threat affects executive functions by inducing state
changes that are implemented via ascending systems [38
] (, red arrows).
Figure 4 Effects of threat and motivation on executive function. Key brain regions mediating the interactions between emotion and/or motivation with executive control function. Both types of interaction are hypothesized to depend on the anterior cingulate cortex (more ...)
The neural interactions described previously indicate that the effects of affective significance on behavioral performance will typically depend on multiple factors. For example, emotional content will enhance stimulus-driven processing in a way that could enhance or impair task performance. An important dimension in determining the impact of affective significance on information processing is task relevance. Specifically, an emotion-laden item that is task relevant will often improve behavioral performance because additional processing resources will typically be devoted to it (relative to neutral). At the same time, a task irrelevant emotional item will usually impair performance because resources will be taken away from the main task. As described, another important dimension of emotional information corresponds to the level of threat. On the one hand, emotional items that are relatively low in threat will benefit from sensory enhancement, which might improve, for instance, reaction time when the item is task relevant. On the other hand, emotional items that are relatively high in threat will lead to enhanced sensory enhancement but, crucially, will also divert processing resources away from other mechanisms. Thus, in many tasks, items that are high in threat will impair behavioral performance even though sensory processing is enhanced. The impairment will be typically observed when the item is task irrelevant, especially in high-anxious individuals [26