Tonal patterns, associations, and alignment of peaks in regional French

Abstract

The intonation of regional varieties of French has been under-studied and remains an important empirical question. In this paper, I present an analysis of two varieties of French that are geographically distant but historically related. The main purpose of the analysis was to examine the effect of the regional factor on the tonal patterns, associations, and timing of the stress group initial and final peaks. The results showed no dialect effect on tonal pattern inventory and distribution or on tonal association and alignment. However, in both dialects women produced less utterance-medial falls and significantly delayed the peaks, compared with men. Additionally, the analysis suggested that the group initial versus final peaks in French appear to differ in secondary associations, which allows us to extend this concept to express the contrasts between different types of high targets and to detail the French language intonational grammar. This work contributes thus to the prosodic grammar of French and to the description of its dialects by providing evidence that they share the same intonational grammar but that groups of speakers demonstrate differences pertaining to its surface realization.

Keywords:

• intonation
• alignment
• tonal association
• French dialects
• variation

Acknowledgments

Thanks are extended to the Department of Statistics and Actuarial Science of the University of Waterloo and to Erin Harvey personally for helping with the statistics.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. The concepts of starredness (Grice 1995; Arvaniti et al. 2000) and edge tones are very important theoretical issues, but they are beyond the scope of this analysis.

2. See Fónagy (1980), Rossi (1980), Pasdeloup (1990, 1991), Vaissière (1991), Hirst and Di Cristo (1997), Astésano (1999, 2001), and Di Cristo (1999, 2000), among others, for a discussion of stress (and even its existence) in French.

3. To neutralize the effect of rate and vowel duration on interval values, normalized latencies were also calculated, and the corresponding analyses were performed using these values. The results for the normalized values followed the time values, with one exception specified in the note (6) below. I chose to report the results for the time values because they allow for observations pertaining to vowel duration and rate.

4. Intervals, traditionally measured in milliseconds, are reported here in seconds. This is done for the readers’ convenience to examine the boxplots, where the measurements appear in seconds.

5. The interval values in Table 5 indicate that the Hi tone appears at approximately ¾ of the syllable duration, starting its beginning, whereas the H* tone appears at ¾ of the vowel duration. It seems that the normalized values of the intervals may provide evidence supporting segmental anchoring in French because the same distance (¾ of the duration from the beginning of a landmark) describes the moment of realization of the peaks, except that the landmark is different for Hi vs. H*. Further analysis focusing on methodology would provide insight into the impact of the methods on the modeling of French intonation.

6. The normalized values exhibited exactly the same results, where gender factor was significant (F(1, 242) ≥ 3.887, p< .05), with only one difference: for the alignment of H* relative left syllabic boundaries, the interaction between the dialect and gender factors appeared marginally significant; in the VEN dataset, the difference between men and women was even greater than in the QUE sub-corpus (F(1, 242) 4.008, p = .046). In general, women reached the peak at approximately 60% (for Hi) and 80% (for H*) of the vowel duration and at approximately 75% (for Hi) and 70% (for H*) of the syllable duration. In men, the Hi tone tends to appear at approximately the 45% mark of the vowel duration and at the 70% mark of the syllable duration (these numbers are based on the averages in Table 6). At this stage, it is difficult to interpret this difference between time and normalized measurements. This methodological issue should be further evaluated, especially considering that different methodologies may lead to different observations (D’ Imperio 2000). Here, however, both methodologies revealed the same significant effect of the gender factor on the peak timing, with women realizing Hi and H* significantly later than men.

Additional information

Funding

This research was sponsored by UW/SSHRC Seed Grant (project #107964).