Anxious and depressed humans typically view circumstances more pessimistically than non-depressed individuals. Here, we explore the proposal that such cognitive biases also exist in non-human animals, and could be used as novel measures of animal welfare. Specifically, we test the hypothesis that wild-caught captiveEuropeanstarlings (Sturnus vulgaris) are more optimistic in their interpretation of ambiguous stimuli when they are housed in cages designed to promote greater welfare compared with when they are housed in standard laboratory cages. Starlings were trained using a choice procedure to discriminate between two temporal stimuli (2 s versus 10 s duration light stimulus) associated with outcomes of a different value (instant or delayed food). Next, the birds’ responses to ambiguous, unreinforced stimuli of intermediate duration ranging from 2 to 10 s were examined under two housing regimes designed to manipulate the birds’ welfare: big enrichedcages versus standard cages (smaller and unenriched). The birds’ probability of classifying an intermediate stimulus as that associated with the instant food outcome was significantly higher in the enrichedcage compared with the standard cage. Thus, the birds displayed greater optimism in the face of uncertainty under housing conditions in which other measures indicate better welfare. These findings support the use of cognitive bias-based tasks as a novel, non-invasive technique for assessing affective state in non-human animals.