Attentional bias is a central feature of many cognitive theories of psychopathology. One of the most frequent methods of investigating such bias has been an emotional analog of the Stroop task. In this task, participants name the colors in which words are printed, and the words vary in their relevance to each theme of psychopathology. The authors review research showing that patients are often slower to name the color of a word associated with concerns relevant to their clinical condition. They address the causes and mechanisms underlying the phenomenon, focusing on J.D. Cohen, K. Dunbar, and J.L. McClelland's (1990) parallel distributed processing model.

A review of recent research on cognitive processing indicates that biases in attention, memory, and interpretation, as well as repetitive negative thoughts, are common across emotional disorders, although they vary in form according to type of disorder. Current cognitive models emphasize specific forms of biased processing, such as variations in the focus of attention or habitual interpretative styles that contribute to the risk of developing particular disorders. As well as predicting risk of emotional disorders, new studies haveprovided evidence of a causal relationship between processing bias and vulnerability. Beyond merely demonstrating the existence of biased processing, research is thus beginning to explore the cognitive causes of emotional vulnerability, and their modification.

Using a probe detection technique we have recently demonstrated that anxious subjects consistently deploy attention towards threat-related stimuli, whereas non-anxious controls tend to move attention away from such material (MacLeod, Mathews, & Tata, 1986). The current study employed the same paradigm but attempted to distinguish the role of trait and state anxiety by testing high- and low-trait students when state anxiety was relatively low (12 weeks before a major examination) and again when it was relatively high (one week before this examination). High-trait subjects alone tended to shift attention towards generally threatening material on both test occasions. Results for examination-related stimuli were more complex. Increased proximity to the examination was associated with an increase in attentional bias towards such threat stimuli in high-trait subjects, but with increased attentional avoidance of such stimuli in low-trait subjects. It is suggested that the attentional response to currently relevant stress-related stimuli may be associated with neither trait nor state anxiety alone, but with an interactive function involving both these variables. These results are discussed in relation to existing models of emotion and cognition, and alternative interpretations of the findings are considered.

Clinical anxiety disorders and elevated levels of anxiety vulnerability are characterized by cognitive biases, and this processing selectivity has been implicated in theoretical accounts of these conditions. We review research that has sought to evaluate the causal contributions such biases make to anxiety dysfunction and to therapeutically alleviate anxiety using cognitive-bias modification (CBM) procedures. After considering the purpose and nature of CBM methodologies, we show that variants designed to modify selective attention (CBM-A) or interpretation (CBM-I) have proven capable of reducing anxiety vulnerability and ameliorating dysfunctional anxiety. In addition to supporting the causal role of cognitive bias in anxiety vulnerability and dysfunction and illuminating the mechanisms that underpin such bias, the findings suggest that CBM procedures may have therapeutic promise within clinical settings. We discuss key issues within this burgeoning field of research and suggest future directions CBM research should take to maximize its theoretical and applied value.

After briefly describing the nature of emotional processing biases associated with vulnerability to anxiety, and a model of how they may be produced, we review new data on the experimental induction of attentional and interpretative biases. We show that these biases can be readily induced in the laboratory, in the absence of mood changes. However, induced biases exert effects on the processing of new information and cause congruent changes in state anxiety when they influence how emotionally significant information is encoded. We can therefore conclude that biases have causal effects on vulnerability to anxiety via their influence on how significant events are processed. Finally, we discuss how our model might account for the acquisition of processing bias and for when they can influence anxiety.