We investigated how the surface and structural information of pitch and time in melodies contribute to the perceived expectancy of melodic segments. The contour (pitch surface), tonality (pitch structure), rhythm (time surface) and metre (time structure) were preserved or altered in factorial fashion, either for the full length of a melody (Full condition) or only its last phrase (Last condition). Participants (N = 24) with a range of musical training received instructions to rate how expected the second portion of a melody was, having heard its first part. Additionally, instructions varied across blocks to attend selectively to pitch, time, or both. Expectancy ratings for the Last condition were lower than for the Full condition, indicating that ratings truly reflected expectancy (rather than overall goodness, which would predict the opposite). Interestingly, tonality and rhythm contributed to global expectancy ratings, but not contour or metre. Furthermore, listeners were unable to ignore entirely either dimension, but successfully attenuated their influence in accordance with instructions. These findings offer a unique insight into music perception by testing expectancies of melody segments (beyond single-note continuations), factorially varying both the surface and structure of pitch and time, and using a selective attention manipulation.