Recent timing control research has shown that when a smaller linguistic segment within a repeated utterance is produced either shorter or longer than its average duration, the resulting deviation constitutes a timing error that is temporally compensated by the remaining segments in the utterance. In the present study, timing errors and the extent to which they are compensated define different levels of temporal interaction. Temporal interactions were studied within a phrase spoken in isolation and within the same phrase spoken in a sentence context. The data, presented as significant negative Pearson correlations (p<0.05) between various adjacent segment durations, support several tentative conclusions on the timing control of repeated utterances: (1) different‐sized utterances have different levels of temporal interaction; (2) temporal interaction effects extend across phoneme, syllable, word, and phrase boundaries; (3) there appears to be no basic unit of timing for English utterances, but a hierarchy of units ranging in size from a CV or VC syllable to a sentence; and (4) identical durational errors may be compensated for by a variety of adjacent segment lengths.