Dysfunctional attitudes have been proposed as an important vulnerability factor in the cognitive model of depression. Yet it has often proved difficult to demonstrate their existence in non-symptomatic populations. We examine the ways in which dysfunctional attitudes have been conceptualized and assessed, from self-report methods to information-processing tasks. A s dysfunctional attitudes are typically viewed as latent in non-symptomatic groups, the importance of priming or activating such variables is emphasized, together with recommendations as to how this may best be achieved. Comparative studies of depressed, control, and at-risk groups are then considered, together with longitudinal studies that have directly testedpredictions of the cognitive model. Prospective studies of non-depressed, non-clinical samples have so far had mixed results in demonstrating that dysfunctional attitudes precede depression or that specific attitudes interact with congruent events in the way the model predicts, although more consistent results emerge from clinical samples with a past history of depression. Possible reasons for the variability in findings are presented, together with suggestions for further research and a revised cognitive model of depression.