This article presents cognitive models of generalized anxiety disorder (GAD) and a critical review of the empiric literature in which GAD has been associated with affect-related information-processing biases, particularly in the domains of attention, memory, and problem solving. Not included in this article are studies of cognitive features of nonclinical anxiety and worry, although this body of research may be relevant to GAD. The core feature of GAD, worry, is regarded as maladaptive information processing, biased in the direction of threat. The centrality of “threat” to everyday events and possibilities clearly is mirrored in the concerns commonly expressed by individuals with GAD—judging another's look of disapproval as a sign of personal social failure or judging lateness as a sign of personal incompetence, but despite the flood of research in this area and the robust evidence for certain types of cognitive biases in GAD, the empiric investigation of threat-related information processing is in its infancy. Most paradigms used to date fail to test key assumptions of the cognitive models of GAD. Also, testing of the extent to which affect-related information-processing biases are unique to GAD or are common to all of the anxiety disorders generally is absent. After reviewing the extant literature, the authors suggest several novel tests of affective information processing and highlight the benefits of using multiple comparison groups and prospective studies to fully test the hypothesis that cognitive biases contribute to the development and maintenance of GAD.