Abstract When a movement sequence is repeated, the second execution is faster than the first. This demonstrates that the brain retains some trace of the just executed sequence, the earliest form of sequence memory. Currently it is unclear whether this memory trace is represented at the level (a) transitions between movements, (b) chunks of multiple movements, or (c) the entire sequence. To answer this question, we instructed human participants to generate sequences of 11 finger presses in a delayed response paradigm. From one trial to the next, segments of variable length (1, 2, 4, 6, 11) could be repeated from the previous trial. We observed that repetition benefits appeared when a segment of 4 consecutive finger presses or longer was repeated from the previous trial. This suggests that the benefit of repetition is not merely the sum of improvements in individual transitions, nor does it require the entire sequence to be repeated. The repetition benefit was small for the first transition of a repeated segment and increased with additional repetitions. This suggests that the memory supporting the repetition effect is mainly activated when a series of past movements matches the memory trace. Planned future movements had less of an effect on the repetition effect. Our results provide insight into the structure of the earliest memory traces for motor sequences. Significance Statement Many motor skills involve concatenating single movements into complex sequences. Already after a single execution of a sequence, humans retain a memory trace that accelerates the execution of the sequence if it is repeated in the next trial. Our results show a repetition benefit even when only 4 movements from the previous trial are repeated, suggesting that full sequence repetition is not necessary. Detailed analyses indicate that the memory trace is activated when the last 2-3 movements match the current execution. Our work, therefore, sheds light on the structure of the earliest sequence memory and will interest researchers studying motor skill learning and the neuronal basis of movement control.