The effect of boundary recoverability on geminate length in English.

Abstract

[Payne (2005)] argues that fake geminates are longer than true geminates, and [Ridouane (2007)] shows that their preceding vowel duration is also significantly longer. These effects are likely due to the boundary that distinguishes fake from true geminates. If this is correct, then we might expect that boundary strength will also affect geminate realization. We investigated this possibility by comparing the absolute and relative durations of assimilated and concatenated word-internal and cross-word boundary fake geminates in English (e.g., immoral versus unnamed versus fun name). Eight speakers produced five repetitions of 32 stimuli (four types: three geminate, one singleton) in two speech styles. The results showed that although cross-boundary geminates were longer than word-internal concatenated and assimilated geminates in absolute terms, both types of concatenated geminates were shorter than assimilated geminates in relative terms. A follow up experiment, comparing geminates in compound words to those emerging across words showed no differences in relative duration between the two. These results suggest that boundary strength may be less important than boundary recoverability in production. Overall, we argue that boundary recoverability impacts the phonetic implementation of geminates and likely does so whether these are true or fake.