The recognition of intervals is likely to underlie skilled sight-reading, since auditory coding and fingering selection are normally dependent on the relative pitches of notes in the score. Therefore, the purpose of this experiment is to investigate the processing of visually presented intervals. Participants were required to perform a matching task, indicating whether two consecutively presented intervals were the "same" or "different". These pairs of intervals were visually similar, visually dissimilar or spatially dissimilar. Sight-readers were most accurate at judging "same" intervals when pairs were visually similar and least accurate at judging "same" intervals when pairs were spatially dissimilar. Conversely, sight-readers were most accurate at judging "different" intervals when pairs were spatially dissimilar and least accurate at judging "different" intervals when pairs were visually similar. This interaction between actual similarity and visual similarity was greater for poor sight-readers compared with good sight-readers.