The current study provides the first experimental evidence that attentional bias training can modify neural systems known to be involved in the control of attention to emotional stimuli (13,39
). Specifically, lateral PFC activity depended on the type of attentional training undertaken (attend-threat or avoid-threat) and, across all participants, was greatest when the direction of participants' attention was contrary to their training. Connectivity between the identified lateral PFC clusters and face selective sensory cortex was consistent with that predicted by the behavioral effects of training and current models of selective attention (11
). These results are in line with the prediction (22,23
) that pharmacological (16–18,20,21
) and psychological interventions that alter attentional function are mechanistically distinct.
While our main analysis showed that attentional training modulated activity in the lateral prefrontal cortex in an attentional task, it could not directly test whether these regions were actually involved in attentional control. It is conceivable, for example, that the training effect is encoded elsewhere in the brain and that the increased lPFC activity observed when the training rules were violated arise because behaving contrary to training is less practiced and thus more effortful, in essence, a form of task switching effect (40
). By this interpretation, altered lPFC activity results as a consequence of training rather than mediating its effect. We therefore sought to test our interpretation of the results by examining the pattern of connectivity between the identified lateral PFC regions and face selective visual sensory association cortex. In this analysis, we reasoned that if the lPFC was controlling attention to the emotional faces as we hypothesized, there should be evidence of a functional link between the control areas and the visual sensory association cortex (11
). The demonstrated pattern of connectivity is consistent with our hypothesis that the lPFC regions identified in the main analysis are indeed influencing attention. Clearly, our PPI analysis alone cannot prove that lPFC controls activity in the fusiform cortex; the observed pattern of connectivity could equally well be produced by the fusiform controlling activity in the lPFC. However, our interpretation is in line with both the models of attentional control (10–12
) and the more general understanding of the lPFC as providing a supervisory role in cognition (41
Although we were able to demonstrate the predicted pattern of connectivity between lPFC and sensory cortex, we did not find an effect of attentional training on the gross activity of the face selective fusiform cortex (Supplement 1
), which would have strengthened the interpretation of our results. While a single training session appears insufficient to individually demonstrate the effects of our intervention on every node of the attention circuit, future studies using longer training regimes may be able to show such an effect.
We had predicted that the rostral anterior cingulate cortex would be identified in our whole-brain analysis but did not find a significant effect. However, with a region of interest approach (Supplement 1
), a small region of the rACC was found to display the same pattern of activity as the lPFC. Thus, it seems likely that the lPFC regions identified in our main analysis are one node of a larger control circuit that incorporates the rACC. It may also include the striatum, because our whole-brain analysis revealed bilateral striatal activity with a similar pattern of activity. We had not predicted these findings, so interpretation must be cautious; however, the striatum is a component of a well described circuit that includes the lPFC (42
) and thus the striatal activity may reflect the efferent or afferent connections with the PFC.
We have suggested that attentional training may provide a model of one of the mechanisms involved in more complex psychological interventions such as CBT. Indeed, there is some evidence that CBT ameliorates the negative attentional biases found pretreatment in patients with anxiety (43
). However, our study compared avoid-threat training and attend-threat training, either or both of which may actively influence attentional function. While this design provides the most sensitive measure as to which areas of the brain are influenced by attentional training, it cannot discriminate whether the observed effects result from the attend-threat training, the avoid-threat training, or both together. As the avoid-threat training is predicted to be therapeutic in anxiety (5,6,44,45
), an interesting next step would be to assess the effects of this form of training in comparison with a control condition. Further, such a study could also incorporate an assessment of attentional function both before and after training, providing a more direct assay of the effect of the intervention on attention than the between-subject approach used in the current study.
A single session of attentional training was sufficient to tune lateral prefrontal function even when assessed using emotional stimuli of a completely different type (faces vs. words) to those employed in training. This generalization of training effect across stimulus type was also supported by the behavioral data, where a word-based training procedure influenced attention to pictures of faces. Interestingly, the effect of training was only evident when prototypical expressions of fear were used in the testing session, with no effects apparent when less intense facial expressions were employed. One interpretation of such results is that there is a threshold of emotional signal above which the training effect is manifested. Clearly, if attentional training is to be effective in clinical settings, it is important that it produces an effect on attention extending beyond the specific stimuli used in training, as demonstrated here.
The interpretation of studies that investigate treatment mechanisms in clinical groups can be confounded by factors other than exposure to the treatment. Thus, when treatments improve clinical state or significantly change behavior (e.g., [16,21,46
]), there is an inevitable confounding of the treatment effects by variation in psychopathology (e.g., mood) or behavior (e.g., time spent looking at negative pictures). The design of our study minimizes such factors, first by studying a nonclinical population who did not experience a profound change of mood or anxiety, and second by using a behaviorally insensitive task during imaging such that the performances of the groups were equivalent. This allows a more straightforward interpretation of our results as the direct effect of attentional training. While it is important that these findings are extended to clinical groups, translational studies such as ours are well suited to demonstrating the basic neural mechanics underpinning treatment effects and for proof of concept in developing novel training strategies or specific psychotherapies.
In summary, the current study demonstrates that lateral prefrontal activity to emotional stimuli may be modified by a simple cognitive intervention known to alter attentional bias. This supports the proposal that modification of PFC function contributes to the effects of psychological interventions that target attentional processes and suggests that such interventions are mechanistically distinct from pharmacological approaches.