A 2 (good vs. bad mood) × 2 (important vs. unimportant) × 2 (success vs. failure) experimental design was used to investigate whether importance could moderate mood effects on students' performance attributions. Attributions were analyzed in terms of their underlying dimensions (locus, stability, controllability) as specified by Weiner (1985). Undergraduate business students (31 men, 49 women) were randomly assigned to 1 of the 8 experimental conditions. Analysis revealed a significant 3-way interaction of mood, importance, and performance outcome (p < .01). Outcome importance significantly reduced mood biases only in perceptions of the primary dimension of locus. Perceptions of causal stability reflected a mood bias: Students in a good mood viewed success as more stable than failure regardless of outcome importance. Perceptions of causal controllability were unaffected by students' mood state or outcome importance, perhaps reflecting a basic need to view events as controllable.