Two auditory priming experiments tested whether the effect of final phonological overlap relies on syllabic representations. Amount of shared phonemic information and syllabic status of the overlap between nonword primes and targets were varied orthogonally. In the related conditions, CV.CCVC items shared the last syllable (e.g., vi.klyd-p[image omitted].klyd) or the last syllable minus one phoneme (e.g., vi.flyd-p[image omitted].klyd); conversely, CVC.CVC items shared the last syllable (e.g., py[image omitted].tib-g[image omitted]l.tib) or the last syllable plus one phoneme (e.g., pyl.tib-g[image omitted]l.tib). Both experiments required participants to repeat back the targets, with Experiment 2 including foils (e.g., vygl[epsilon]t-bugl[epsilon]f). Foils made shadowers adopt a more conservative mode, but had no systematic influence on the magnitude of the final facilitation. More importantly, neither set of data fully aligned itself with the syllabic hypothesis. These results therefore argue against the idea that syllables serve as coding units in speech perception. (Contains 4 figures.)