Compared symptomatically depressed, clinically remitted, and normal controls using cognitive measures designed to be traitlike and statelike in cross-sectional and longitudinal designs, respectively. Remitted depressives and normal subjects did not differ in their attributional biases, endorsement of dysfunctional attitudes, or interpretation of schema-relevant ambiguous events, but both groups differed from symptomatic depressives. Depressive episodes thus affect cognition, but cognitions measured by self-reports are more statelike than traitlike.