Say again? Individual acoustic strategies for producing a clearly-spoken minimal pair wordlist

Abstract

People make their speech clearer in difficult conversational contexts using global mechanisms (e.g. “Lombard Speech”) and by targeted enhancements of linguistic constituents (“hyperspeech”). We describe production changes observed in four speakers of Scottish English who produced three repetitions of twelve CVC words: V was one of six monophthongs and C_C was either /p_p/ or /m_m/. Thus each word differed (near-) minimally from six others. In a “neutral” condition each participant read aloud from a randomised wordlist. A “clear” condition was an interactive task in which an interlocutor had to repeat back every word correctly, despite their hearing being impaired by headphone-delivered noise. If the speaker was mis-perceived by the interlocutor, the speaker tried again, until the word was correctly repeated. We describe the surprisingly speaker-specific acoustic hyperspeech effects (in vowel F1, vowel space area, and acoustic segment durations) in the clear speech. A companion paper describes the associated articulatory changes.