PIs : Christine Meunier (Aix-en-Provence, France) / Yao Yao(Hong-Kong, SAR China)
This project aims to build a connection between studies which examined two different factors on a single parameter (vowel reduction) and on a single speechs tyle(conversational speech). In the first one, a study of Gahl et al. (2012) investigates the role of phonological&neighborhood on vowel reduction in conversational speech in English. In the second one, Meunier & Espesser (2011) investigates the effects of some phonological and grammatical properties of lexicon on vowel reduction in conversational speech in French.
Speech reduction in conversational speech is currently a challenging topic, since its manifestation highlights the complex interaction and relationship between linguistic domains. It involves questions about the processes that lead speakers to produce speech in a non-canonical way and how listeners perceive it. The main question is why and how is speech reduced. Another question is to understand the process of speech reduction according to languages. Vowel reduction in casual speech is a highly robust phenomenon across languages (Gendrot & Adda-Decker, 2007) but we do not know if the factors that induce reduction are similar in all languages. This project tries to highlight different properties of lexicon and their influence on vowel reduction according to languages.
The proposed research has the following objectives:
1. Effects of phonological neighborhood structure for French and English
To investigate the effects of phonological (prosody) and grammatical (word category) on vowel reductioninEnglish. ; To compare with previous results obtained in French. To shed light on the process of phonetic reduction across languages according to phonological versus grammatical properties.
Research plan andmethodology
The current project will mainly consist of a corpus-based quantitative analysis of phonetic variation. The speech data we use will come from the CID corpus (Bertrand et & al. 2008), which contains eight hours of transcribed French conversational speech and from the Buckeye corpus (Pittetal.,2007) which contains 40 hours of transcribed conversational English speech. Following Gahl et& al.& (2012), we will examine two aspects of phonetic variation: durational shortening/lengthening and vowel reduction/dispersion. The critical factors are the neighborhood measures (neighborhood density, neighbor frequency and their interaction),which will be calculated from the CLEARPOND database (Marian et & al .2012). A separate set of lexical and contextual measures that could potentially cause phonetic variation (e.g. word frequency, phonotactic probability, contextual predictability, local speech rate, disfluency, etc.) will be calculated as control factors (Gahlet&al.2012, Meunier and Espesser 2011). Mixed-effects models will be constructed to examine the effects of neighborhood measures (if any) on phonetic variation when other factors are controlling for.
Moreover, there hasnt been any direct evidence regarding the effects of neighborhood density on French word production (in terms of reaction times and accuracy). Consequently, a production experiment (a picture-namingtask) will be conducted in a way to have a complete experimental design comparing French and English.