Constraints underlying bimanual coordination have traditionally been explained by dynamic interactions between the effectors. However, the present experiments demonstrate that a fundamental constraint on bimanual performance is the manner in which task goals are represented. In Experiment 1, participants vocalized during in-phase and anti-phase bimanual movements. As expected, most participants spontaneously exhibited temporal coupling between the manual and vocal responses. However, the form of coupling differed for the in-phase and anti-phase conditions. For anti-phase movements, there was a strong bias to produce two vocalizations per cycle; for in-phase movements, participants were equally likely to produce one or two vocalizations per cycle. We hypothesized that the spontaneous vocalizations probed the cognitive representation of the task, and the results indicated that anti-phase movements did entail a more complex event structure than in-phase movements did. In Experiment 2, we manipulated the event structure by having participants vocalize either once or twice per hand cycle. As predicted, coordination stability was reduced when the event structure was more complex.