Three experiments tested the proposal that music–color associations are mediated by emotional connotations as suggested by Palmer and collaborators (2013, 2016). Experiment 1 asked participants to choose 1 of 8 saturated colors for excerpts of 24 Preludes from Bach’s Well-tempered Clavier. Half of these excerpts were presented again for color choices, together with excerpts from another 12 Preludes. Participants were also asked to judge whether or not they had heard the excerpts before in the experiment. The color choices grouped together Preludes according to tempo, mode, pitch height, and attack rate. Participants included nonmusicians, musicians, absolute pitch (AP) possessors, and music–color synesthetes. Color choices followed the same general pattern across groups, and were more similar than chance when the excerpts were repeated and, to a lesser extent, when the excerpt was different but in the same key. No recognition advantage was found for synesthetes or possessors of AP. Experiment 2 asked participants to rate the colors on a number of emotion scales. Experiment 3 asked them to rate the excerpts from the Preludes on the same emotion scales, and found that the emotion ratings grouped together Preludes according to tempo, mode, pitch height, and attack rate as in Experiment 1. Finally, the color choices in Experiment 1 could be predicted by the color–emotion ratings in Experiment 2 and the music–emotion ratings in Experiment 3. These results support the proposal that music–color associations can be accounted for by the correlations between music and emotion, and color and emotion.