We examined some of the factors that influence evaluations of improvised music. Using the Consensual Assessment Technique (CAT), 10 expert judges evaluated 16 pieces of music, improvised on an electronic keyboard. Overall evaluations of improvisations were associated with their perceived complexity, creativity, and technical goodness. These predictors accounted for 76% of the variance in preference judgments. The intrinsic motivation of improvisers was also assessed, and was correlated with overall evaluations. No support was found for a nonlinear relation between the perceived complexity of the music and its aesthetic appeal, as suggested by Berlyne's optimal-complexity model. We outline a model linking characteristics of improvisers and their improvisations to judges' overall liking for the music. The benefits of using the CAT to assess music are also discussed.