A cardinal characteristic of anxiety disorders is the so-called attentional bias [1, 2], which refers to alterations of attentional processes in response to emotional, especially threat-related, stimuli. The attentional bias of subjects with trait anxiety [3, 4] as well as anxiety disorders [5, 6] including PTSD  has recently been described as a 'vigilance-avoidance reaction'. This model describes the tendency to initially attend to threat before this stimulus is avoided in a subsequent reaction. It is hypothized that while aversive cues evoke a rapid response, anxious subjects subsequently initiate attentional avoidance as an attempt to alleviate the fear reaction . Thereby, this model can reconcile the apparently contradicting findings of an increased allocation of attentional resources towards aversive stimuli in PTSD that has been found in some studies [8, 9], as well as an attentional bias away from threat reported elsewhere [10, 11]. A recent magnetoencephalographic study indicated that the vigilance-avoidance pattern in PTSD consists of two distinct processes that are temporally and spatially dissociated . PTSD patients showed a hyper-reaction to threatening stimuli in the ventrolateral prefrontal cortex as early as 130 ms after stimulus onset, which was followed by an attentional avoidance reaction in occipital regions starting at 200 ms after stimulus onset. Catani et al.  found that the avoidant reaction to aversive pictures in PTSD, indicated by a reduced activity in the left occipital lobe, remained stable for at least the first four seconds after stimulus onset. This reduced activity in brain regions that are associated with elaborate visual processing, was interpreted as disengagement from a detailed visual stimulus exploration . Confronted with potential threats, subjects with anxiety disorders seem to focus on the initiation of a rapid flight reaction rather than concentrating on the attentive evaluation of the threat cue . As it is plausible to assume that such a mechanism can contribute to the maintenance of anxiety symptoms, it could be a promising target for therapeutic interventions. However, the early onset of both the transient vigilant response as well as the prolonged avoidant reaction indicates that both reactions are automatic, stimulus driven bottom-up processes, which cannot be intentionally altered by verbal instruction.
Accumulating evidence from treatment trials demonstrates that PTSD can be effectively treated with psychotherapy. A large number of randomized controlled trials have shown that variants of trauma-focused cognitive behavioral therapy (CBT) are the most successful interventions for the treatment of PTSD [13, 14]. The active mechanism of therapeutic change is thought to be the imaginal exposure to the episodic memory of the traumatic event. In this procedure, the patient is instructed to overcome avoidant strategies and to narrate the traumatic experience in detail. The repeated re-experiencing is thought to reconstruct the distorted autobiographic memory of the trauma  and to change pathological mechanisms of fear processing.
Little is known about the neurobiological foundations of psychotherapy for PTSD. Three uncontrolled, small-scale studies have examined neuronal changes over the course of psychotherapy. While no changes were found on a structural level [16, 17], Felmingham and colleagues (2007)  reported a change of activity of basic fear processing mechanisms in eight subjects who underwent CBT for PTSD. In particular, after therapy, amygdala activity declined while the activity of the anterior cingulate cortex, which is associated with a top-down influence on fear processing, increased. As this study lacked a control group however, brain changes could not be causally linked to psychotherapy. To our knowledge, there are only two randomized trials that assessed functional changes of neural correlates in PTSD. These studies reported a decrease of cortical activity in the right anterior hemisphere during exposure to trauma-related pictures  and the right middle frontal gyrus during trauma script-driven imagery .
Although the primary intention of CBT is to target trauma memories rather than to change attentional processes, the interaction of episodic memory retrieval and attention supports the assumption that trauma focused CBT alters memory as well as attention processes. A cortical structure that might play an important role in the context is the parietal cortex. Recently, Cabeza et al. (2008) formulated a theory that connects attentional and episodic memory processes. According to this theory, goal-directed memory retrieval can be regarded as an effortful process that involves voluntary top-down attention towards memory content. These goal-directed attentional processes are driven by parietal structures including the superior parietal cortex . This finding is important in interpreting the role of the parietal cortex in PTSD considering the general assumption of a deficit in top-down regulation of fear in PTSD , which may be modified by exposure therapy .
In the present randomized controlled treatment trial, we wanted to investigate the influence of trauma-focused therapy on attentional processing of aversive pictures. We compared the effects of trauma-focused CBT with a waitlist control group in a population of severely traumatized victims of war and torture. In both groups, we assessed neurophysiological indicators of attention to threat cues before and after treatment or the corresponding waiting time. We used Narrative Exposure Therapy (NET) as a variant of trauma-focused CBT. NET is a manualized short-term approach that has been adapted to meet the needs of traumatized survivors of war and torture . NET has been shown to be effective in the reduction of PTSD symptoms in various randomized controlled trials [24–27]. The advantage of NET is that rather than combining various techniques such as cognitive methods and skills training, this treatment focuses exclusively on the reconstruction of episodic trauma memory by a detailed and emotional narration of the biography, with particular emphasis on the traumatic events. Therefore, potential effects can be attributed to a specific intervention rather than to a mixture of various therapeutic methods.
We applied magnetoencephalography (MEG) to measure steady-state visual evoked fields (ssVEF) as an indicator of attention processes. Here, we were interested in the processing pattern of threatening cues, which consisted mainly of pictures with trauma-related (war and attack scenes) or generally aversive (mutilations) stimuli. The ssVEF represents an ongoing cortical oscillatory neuromagnetic response elicited by a repetitive visual stimulus that is presented at a certain frequency (e.g., 10 Hz) having the same fundamental frequency as the driving stimulus . One major advantage of the ssVEF technique is that a high signal-noise-ratio can be achieved even with a limited number of trials . This technique has proven its efficiency in several studies investigating higher order attentional mechanisms during the processing of emotional stimuli in healthy participants [29–31], and psychiatric patients [11, 32].
We expected that NET would reduce attentional disengagement towards aversive stimuli . Therefore, we expected a relative increase of activity towards threat-related pictures in brain regions associated with elaborate visual processing. Moreover, we aimed to explore potential changes in activity in the parietal cortex as a marker of enhanced top-down regulation of attention as well as episodic memory.