The clustering of ‘fluencemes’ in French and English

Degand, Liesbeth [UCL] Gilquin, Gaëtanelle [UCL]

Different features can contribute to the fluency (or disfluency) of discourse, among which speech rate, (filled and empty) pauses or discourse markers. These ‘fluencemes’ (the term is taken from Götz 2011) seem to be present in all languages. However, the relatively few studies that have performed a contrastive analysis of fluencemes reveal cross-linguistic differences. Thus, it is well-known that filled pauses are language-specific, from English uh or uhm and French euh to Spanish pues or Japanese eeto, for instance (see Clark & Fox Tree 2002: 92). Their use and selection may also differ significantly from one language to the other, as demonstrated by Zhao & Jurafsky (2005) for English and Mandarin Chinese. Similarly, discourse markers appear to have several equivalents cross-linguistically, which points towards language-specific functions (see, e.g., Aijmer & Simon-Vandenbergen 2006). In this presentation, we adopt a French-English contrastive approach to test the hypothesis that fluencemes do not occur in isolation, but tend to cluster together (cf. Stenström 1990: 222, Aijmer 1997: 27). Using corpus data representing spontaneous conversations in French and English (from the VALIBEL database and ICE-GB, respectively), we start from filled pauses (French euh and English uh/uhm) and examine what other fluencemes occur in their immediate environment, with special emphasis on discourse markers, defined here very broadly as “any element of language that plays a role in the organization of discourse” (Vincent 2005: 189). We investigate how French and English compare in this respect, both quantitatively (e.g. what proportion of filled pauses cluster with other fluencemes in each language?) and qualitatively (e.g. what types of discourse markers tend to cluster with filled pauses?). Preliminary results reveal interesting differences between French and English. While filled pauses are more frequent in English than in French (260 occurrences per 10,000 words in English, as against 181 in French), French filled pauses are proportionally more likely to be immediately preceded or followed by a discourse marker than English filled pauses (42% for French vs 28% for English). The most frequent discourse markers accompanying French filled pauses are connectives (et, mais, donc), whereas there seems to be more variation in English (and, but and so, but also well, I mean, you know, etc.). A qualitative analysis will have to reveal whether these markers are used in this context according to their ‘usual’ profile of use, or whether specific features are at play. A word cluster analysis furthermore shows that the most frequent clusters with filled pauses involve empty pauses, viz. euh immediately followed by a short pause or a long pause in French, and uhm immediately followed or preceded by a short pause in English. It thus appears that, both in French and in English, fluent or disfluent stretches of discourse result from a combination of fluencemes, as illustrated by the following examples: (1) ben oui mais donc euh / on / on a / on travaille jusqu'à mardi (2) I mean I think uhm space is you know I mean <,> you know just the obstacles that you have in a room This suggests that fluency (or disfluency) should be viewed as a multifaceted phenomenon, which should be investigated by means of an integrated approach (cf. Götz 2011). Further contrastive research will have to establish whether such an approach can prove equally rewarding for other languages besides French and English.