In two studies, we investigated the capacity of compositional structure and performance expression to communicate the affective dimensions of valence and arousal. We expected that performance and composition would differ in their expressive capacities, with composition better suited to expressing changes in valence and performance better suited to expressing changes in arousal. Eight highly trained musicians attempted to communicate distinct emotions in three ways: composition alone (pitches and durations composed by musicians and recorded as MIDI files), performed versions of those melodic fragments, and performance alone (e.g., changes in intensity, tempo, timing) applied to neutral or ambiguous melodic fragments. In Experiment 1, we compared valence and arousal scores for the composition and combined conditions. In Experiment 2, we compared valence and arousal scores for the performance condition. Mean scores for both dimensions varied significantly as a function of the intended emotion. Regression analyses indicated that both composition and performance contributed to valence and arousal scores. However, compositional cues had a greater influence on valence scores and performance cues had a greater influence on arousal scores. The findings underscore the collaborative and complementary nature of emotional communication by composers and performers.