Previous research has established that anxiety patients demonstrate a cognitive bias that selectively favours the processing of threat related information. The current experiment employs a variant of a well established colour naming paradigm to address three issues concerning the nature of this anxiety linked pattern of selective processing. First, by considering non-clinical subjects and employing an experimental design capable of dissociating the influence of state and trait anxiety, the current study addresses the hypothesis that state anxiety elevations will elicit differential patterns of selectivity in high and low trait anxious subjects. Second, by presenting stimulus materials within or outside awareness, the study addresses the hypothesis that these processing biases occur automatically, without requiring the use of consciously mediated strategies. Third, by including both emotionally valenced stimulus materials which are related to, and which are unrelated to, the particular source of stress experienced by the subjects, the experiment addresses the hypothesis that such anxiety linked processing biases will be restricted to materials falling within the domain of current personal concern. The results of the present study clearly demonstrate that these three issues are not independent. Fundamentally different patterns of anxiety linked selective processing were observed when stimuli were presented outside awareness, permitting only automatic cognitive processing, and when stimuli were presented within awareness, permitting the use of consciously mediated strategies. The relative roles played by state and trait variables, and the degree to which these effects were influenced by the personal relevance of the stimulus materials, differed in each case. The implications of these findings for future research in this field are discussed.