We investigate the effects of different levels of delay (or latency) on the coordination, pace and timing regularity of musicians who are in remote locations—a situation encountered in an interactive network performance. Two pairs of musicians performed two Mozart duets while isolated visually and connected through microphones and headphones. Different levels of latency (0, 20, 40, 50, 80, 100, 120, 150, and 200 ms) were introduced into the performing environment (musicians heard themselves in real time and only the other part delayed); the musicians performed the duets under these conditions and rated their musicality and level of interactivity. Although the musicians chose different strategies to handle the latency, which resulted in different levels of success in maintaining coordination, pacing and regularity, both duets were strongly affected by latency at and above 100 ms. At these levels, the musicians rated the performances as neither musical nor interactive, and they reported that they played as individuals and listened less and less to one another.

A model is proposed of the effect of parallelism on meter. It is wellknown that repeated patterns of pitch and rhythm can affect the perception of metrical structure. However, few attempts have been made either to define parallelism precisely or to characterize its effect on metrical analysis. The basic idea of the current model is that a repeated melodic pattern favors a metrical structure in which beats are placed at parallel points in each occurrence of the pattern. By this view, parallelism affects the period of the metrical structure (the distance between beats) rather than the phase (exactly where the beats occur). This model is implemented and incorporated into the metrical program of D. Temperley and D. Sleator (1999). Several examples of the model's output are presented; we examine problems with the model and discuss possible solutions.

Persons who do not possess absolute pitch (AP) often display latent AP-the ability to produce or recognize the pitch level of familiar songs without reference to specific labels-and also use relative pitch (RP) memory to encode and identify familiar melodies. Three experiments investigated the interaction of latent AP and RP memory in listening and production tasks, using ordered lists of songs. In Experiment 1, participants listened to excerpts from pairs of songs in the list, which were either at their original pitch or transposed 1 semitone, and identified whether the second song in the pair was at its original pitch level. Experiment 2, a follow-up study, used a refined methodology and transposition levels up to 3 semitones. In Experiment 3, after hearing the end of one song-either at its original pitch level or transposed 1 semitone-participants were asked to sing the next song in the list. The results showed that participants used both latent AP and RP cues in listening and production tasks: RP memory reinforced latent AP when songs were played at their original pitch level, leading to higher levels of accuracy. When songs were transposed, with RP cues maintained, accuracy decreased in pitch-level identification and singing at the original pitch level. Also, in Experiment 2, the identification of pitch levels was more accurate for consecutive songs on the list-which retained RP cues-instead of nonconsecutive songs, and with a transposition of 3 semitones instead of 1 semitone.