How does one piece of music affect another? Listeners' affective responses to music have been reliably linked to it's predictability, with this relationship often following an inverted U-shaped pattern. Music of intermediate predictability is often liked more than overly simple or unpredictable music. But how does the context in which a piece of music is heard affect this relationship? Here, we present participant liking and predictability ratings between a set of orignal, composed melodies, and short excerpts from Western music. We compare participant responses when hearing these stimuli in a combined context versus hearing them alone (i.e. only Western music or only artificial stimuli). We find that the simpler, composed stimuli were less liked when presented alongside the more ecologically valid excerpts from Western compositions compared to when they were presented alone. Conversely, we find that Western music was more liked in the combined context. Additionally, the composed melodies were rated as more predictable in the combined context, and vice versa for the excerpts of real music. Melody predictability was also estimated using an information theoretic model of music (IDyOM). Despite having a similar range of objective predictability, there are clear differences in both subjective predictability and liking ratings, suggesting that people may use more than expectations drawn from statistical probability when evaluating a piece of music.