An experiment is reported on the effects of extensive practice in a task in which subjects (N = 36) pressed a succession of nine keys with nine separate fingers, each key press in response to a corresponding stimulus. The order of the key presses remained constant over practice. Key-pressing cycles followed each other without interruption. A stimulus was usually presented immediately upon depressing the previous key, but in the structured conditions a stimulus was preceded at either two (the 45 group) or three (the 333 group) positions by a 750-ms response-stimulus interval (RSI). This partitioned the sequence into three response groups for subjects in the 333 condition and into two response groups for subjects in the 45 condition. On occasion, all subjects performed in the unstructured condition, in which RSIs were zero. Interkey times in this condition clearly reflected the position of the 750-ms RSIs in the structured conditions. This suggests that motor chunks developed in the structured conditions were also used in the unstructured condition. Rhythm-based control, as proposed by Summers (1975), was rejected as an explanation for this effect because group-start/within-group ratios exceeded 2:1 and because the theory of rhythm-based control could not predict intervals in the unstructured 45 condition. Unstructured within-group intervals were slower than structured within-group intervals, which effect was more pronounced in the 333 than in the 45 condition. Also, the initial element of the unstructured four-key group was faster than the initial element of the unstructured three-key group. These and other findings (a) accord with the notion that in unstructured sequences preparing a forthcoming response group concurs with the execution of the preceding group and (b) suggest that shorter group are more slowed by concurrent preparation for the forthcoming group than are longer groups.