If people could tap 2 rhythms independently, 1 rhythm with each hand, train
ing people to tap the rhythms separately should enable them to tap the rhyt
hms concurrently. However, nonmusician participants in the present experime
nts were unable to produce accurate intervals when tapping bimanually after
they had mastered the rhythms individually. That finding implies that tapp
ing concurrent rhythms requires an integrated sensory-motor representation
incorporating the actions of both hands into a single pattern. Although tra
ining the rhythms separately cannot specify an integrated code, such specif
ication is possible with bimanual training. Bimanual training quickly led t
o accurate tapping, but most participants then did not tap correctly either
of the separate rhythms that had been mastered in the context of concurren
t tapping. That finding suggests that the integrated representation does no
t code either rhythm independently.