Background: Behavioral studies show that attention training can alter threat bias, influence vulnerability to stress and reduce clinical anxiety symptoms. The aim of this study was to examine which cognitive functions of attention processing are modulated by attention training, and how a priori anxiety interacts with the attention training procedure. Specifically, we expected modulation in the P1/N1 event-related potential (ERP) complex if early spatial attention was to be affected by training and modulation in later ERP components (P2, N2, P3) had training affected top-down attentional processes.Method Thirty anxious and 30 non-anxious adults performed a modified probe detection task. Electroencephalograms (EEGs) were recorded throughout for later ERP analyses. Half the participants in each anxiety group were randomly assigned to undergo a training procedure designed to divert their attention away from threat and the other half received placebo training.Results Anxious participants who were trained to avoid threat showed a linear reduction in response time (RT) to targets replacing neutral faces with the progression of training. This change in RT was not observed among non-anxious participants or among anxious participants who were exposed to placebo training. Following training, the anxious participants who were trained to avoid threat showed a reduction in P2 and P3 mean amplitudes and an enhancement in N2 mean amplitude.Conclusions Attention training affects anxious participants whereas non-anxious participants seem not to respond to it. The ERP data suggest that attention training modulates top-down processes of attention control rather than processes of early attention orienting.