1
Word boundary phenomena of interest - Enchaînement without liaison (E): faite en (‘made in’) - Enchaînement with liaison (EL): fait en (‘did in’) - No enchaînement (NE): fais t ant (‘do so much’) Simultaneous perceptual challenges: detect position of pivotal consonant & disambiguate E, EL, NE phrases. Poor phoneme monitoring & sequence discrimination (Laeufer 1985; Dejean de la Bâtie & Bradley 1995; Wauquier- Gravelines 1995; Yersin-Besson & Grosjean 1996) Research questions: 1) Can listeners disambiguate sandhi phrases (E & EL) and non-sandhi phrases (NE) on the basis of durational variation between E, EL and NE consonants? 2) Is the sensitivity to durational differences across word boundary consonants a condition for the automaticity of word onset identification in L2 French? Aims: 1) Document temporal variation of fast speech, word boundary [t]s, which listeners might use to detect blurred word onsets / disambiguate E, EL, NE phrases, 2) Readdress the issue of temporal indifferentiability of E, EL and NE (Pernot 1937, Grammont 1960, Encrevé 1988) vs. temporal differentiability (Laeufer 1985, Rialland 1985, Wauquier-Gravelines 1995, Yersin-Besson & Grosjean 1996). Methodological innovations: 1) Larger pool of speakers (N=8), 2) Focus on one consonant to control for variability across consonant-inherent durations, 3) Comparison of E consonants with and without prevocalic glottal stops/pauses, 4) Separate analysis of boundary geminates for conservative & controlled comparison of word-initialism vs. sandhi: - No enchaîn t but gemination (NEG): faite Conclusion: The insignificant temporal differences between E & EL consonants (Exp. 1), the confusability of word boundary singletons, whether or not their lengths were edited (Exp. 2 & 3), and the similar interpretations of strong duration cues by beginning, advanced and native listeners (Exp. 3) suggest that the sensitivity to consonant duration changes at word boundary is not a necessary condition for the automaticity of word onset detection in L2 French. References Encrevé, P. (1988). La liaison avec et sans enchaînement . Paris: Le Seuil. Grammont, M. (1960). Traité pratique de phonétique française . Paris: Delagrave. Pernot, N., 1937: Quelques notes sur la liaison en français, liaison et enchaînement. Publications of the Modern Language Association 22 , 333-8. Laeufer, C. (1985). Language-specific and universal aspects of syllable structure and syllabification: Evidence from French and German . PhD dissertation, Cornell U. Rialland, A. (1985). Schwa et syllabe en français. In Wetzels, L. & E. Sezer, Studies in Compensatory Lengthening , pp.187-226. Dordrecht: Foris. Wauquier-Gravelines, S. (1995). Detecting ghost phonemes: Aims: 1) Examine native performance in word boundary identification. 2) Test the perceptual relevance of the E, EL < NE < NEG durational ranking obtained in Experiment 1. Method: Transcription task whereby L1 listeners (N=8) indicate position of a pivotal [t] by using the orthographic conventions –te, t, t, as in faite en, fait en, fais t ant and faite t ant. Lengths of pivotal consonants unedited from speaker’s production data. Methodological innovation: Word boundary geminates treated as separate NEG items rather than expectedly prominent NE items (cf. Yersin-Besson & Grosjean 1996). Result: E, EL < NE < NEG heard as (E, EL, NE) < NEG Perception of NEG consonants as geminates Perception of Es, ELs & NEs as singletons : although NE consonants are temporally closer to the geminates, there were more one-consonant responses (E, EL, NE) to NE stimuli than to NEG stimuli, t1(28) = 8.11, p<.0005, t2(44) = 8.58, p<.0005 (one-tailed, Bonferroni- adjusted). Confusability of Es, ELs and NEs : non- significant effects of stimulus type on proportions of E & EL consonant responses. Fig. 1. Input-to-output mappings for unedited consonant durations Method: Same orthography-based transcription task, but with consonant durations (D1, D2, D3, D4) artificially enhanced into potentially “strong” cues of word boundary along a scale of 35 to 165 ms with a constant increment of 43 ms / Subjects: N=24 (8 natives, 8 advanced, 8 beginners) Preliminary Results (6 beginners, 6 advanced, 6 natives): Similarities with Exp. 2 & between 3 proficiency groups Hypotheses: 1) at least two word boundary conditions remain confusable, 2) L1 listeners, presumably better at detecting word onsets, do not differ from L2 listeners in using strong temporal cues. Yes to either Hyp. would undermine the role of these cues. Durational cues for the identification of word onsets by L2 learners of French Durational cues for the identification of word onsets by L2 learners of French Manuel Sinor – Department of Linguistics, University of Alberta – Manuel Sinor – Department of Linguistics, University of Alberta – [email protected] [email protected] Fig. 2. Input-to-output mappings for edited consonant durations 0% 20% 40% 60% 80% 100% BEG ADV NAT D 1 as E D 1 as EL D 1 as N E D 1 as N EG D 2 as E D 2 as EL D 2 as N E D 2 as N EG D 3 as E D 3 as EL D 3 as N E D 3 as N EG D 4 as E D 4 as EL D 4 as N E D 4 as N EG Perception of D4s as geminates (top yellow stacks in chart) Perception of D1-2-3s as singletons (small “as NEG” stacks) Confusability of D1-2-3s : D1s and D2s heard as E or NE, D3s as NE (biggest stacks in blue, red, & green-tainted areas) 0% 20% 40% 60% 80% 100% P roduced E P roduced EL P roduced NE P roduced NEG Perceived N EG Perceived N E Perceived EL Perceived E

Word boundary phenomena of interest - Enchaînement without liaison (E): faite en (‘made in’) - Enchaînement with liaison (EL): fait en (‘did in’) - No

Embed Size (px)

Citation preview

Page 1: Word boundary phenomena of interest - Enchaînement without liaison (E): faite en (‘made in’) - Enchaînement with liaison (EL): fait en (‘did in’) - No

Word boundary phenomena of interest

- Enchaînement without liaison (E): faite en (‘made in’) - Enchaînement with liaison (EL): fait en (‘did in’)- No enchaînement (NE): fais tant (‘do so much’)

Simultaneous perceptual challenges: detect position of pivotal consonant & disambiguate E, EL, NE phrases. Poor phoneme monitoring & sequence discrimination (Laeufer 1985; Dejean de la Bâtie & Bradley 1995; Wauquier-Gravelines 1995; Yersin-Besson & Grosjean 1996)

Research questions: 1) Can listeners disambiguate sandhi phrases (E & EL) and non-sandhi phrases (NE) on the basis of durational variation between E, EL and NE consonants? 2) Is the sensitivity to durational differences across word boundary consonants a condition for the automaticity of word onset identification in L2 French?

Aims: 1) Document temporal variation of fast speech, word boundary [t]s, which listeners might use to detect blurred word onsets / disambiguate E, EL, NE phrases, 2) Readdress the issue of temporal indifferentiability of E, EL and NE (Pernot 1937, Grammont 1960, Encrevé 1988) vs. temporal differentiability (Laeufer 1985, Rialland 1985, Wauquier-Gravelines 1995, Yersin-Besson & Grosjean 1996).

Methodological innovations: 1) Larger pool of speakers (N=8), 2) Focus on one consonant to control for variability across consonant-inherent durations, 3) Comparison of E consonants with and without prevocalic glottal stops/pauses, 4) Separate analysis of boundary geminates for conservative & controlled comparison of word-initialism vs. sandhi:

- No enchaînt but gemination (NEG): faite tant (‘done so’)

Observed durational ranking: E, EL < NE < NEGNE vs. E, EL: t1(7) = 4.44, p<.01; t2(92) = 3.23, p<.001. E vs. EL: t1(7) = 1.23, n.s.; t2(92) = 1.13, n.s.

Conclusion: The insignificant temporal differences between E & EL consonants (Exp. 1), the confusability of word boundary singletons, whether or not their lengths were edited (Exp. 2 & 3), and the similar interpretations of strong duration cues by beginning, advanced and native listeners (Exp. 3) suggest that the sensitivity to consonant duration changes at word boundary is not a necessary condition for the automaticity of word onset detection in L2 French. ReferencesEncrevé, P. (1988). La liaison avec et sans enchaînement. Paris: Le Seuil.Grammont, M. (1960). Traité pratique de phonétique française. Paris: Delagrave.Pernot, N., 1937: Quelques notes sur la liaison en français, liaison et enchaînement. Publications of the Modern Language Association 22, 333-8.Laeufer, C. (1985). Language-specific and universal aspects of syllable structure and syllabification: Evidence from French and German. PhD dissertation, Cornell U.

Rialland, A. (1985). Schwa et syllabe en français. In Wetzels, L. & E. Sezer, Studies in Compensatory Lengthening, pp.187-226. Dordrecht: Foris.

Wauquier-Gravelines, S. (1995). Detecting ghost phonemes: the “liaison enchaînée” in French. Proceedings of the 13th International Congress of Phonetic Sciences.

Yersin-Besson, C., & Grosjean, F. (1996). L’effet de l’enchaînement sur la reconnaissance des mots dans la parole continue. L’Année Psychologique 96, 9-30.

Aims: 1) Examine native performance in word boundary identification. 2) Test the perceptual relevance of the E, EL < NE < NEG durational ranking obtained in Experiment 1.

Method: Transcription task whereby L1 listeners (N=8)

indicateposition of a pivotal [t] by using the orthographic conventions–te, –t, –t, as in faite en, fait en, fais tant and faite tant. Lengthsof pivotal consonants unedited from speaker’s production data.

Methodological innovation: Word boundary geminates treatedas separate NEG items rather than expectedly prominent NEitems (cf. Yersin-Besson & Grosjean 1996).

Result: E, EL < NE < NEG heard as (E, EL, NE) < NEGPerception of NEG consonants as geminates Perception of Es, ELs & NEs as singletons: although NE

consonants are temporally closer to the geminates, there were more one-consonant responses (E, EL, NE) to NE stimuli than to NEG stimuli, t1(28) = 8.11, p<.0005, t2(44) = 8.58, p<.0005 (one-tailed, Bonferroni-adjusted).

Confusability of Es, ELs and NEs: non-significant effects of stimulus type on proportions of E & EL consonant responses. Fig. 1. Input-to-output mappings for unedited consonant durations

Method: Same orthography-based transcription task, but with consonant durations (D1, D2, D3, D4) artificially enhanced into potentially “strong” cues of word boundary along a scale of 35 to 165 ms with a constant increment of 43 ms / Subjects: N=24 (8 natives, 8 advanced, 8 beginners)

Preliminary Results (6 beginners, 6 advanced, 6 natives): Similarities with Exp. 2 & between 3 proficiency groups

Hypotheses: 1) at least two word boundary conditions remain confusable, 2) L1 listeners, presumably better at detecting word onsets, do not differ from L2 listeners in using strong temporal cues. Yes to either Hyp. would undermine the role of these cues.

Durational cues for the identification of word onsets by L2 learners of French Durational cues for the identification of word onsets by L2 learners of French

Manuel Sinor – Department of Linguistics, University of Alberta – [email protected] Sinor – Department of Linguistics, University of Alberta – [email protected]

Fig. 2. Input-to-output mappings for edited consonant durations

0%

20%

40%

60%

80%

100%

BEG ADV NAT

D1 as E D1 as EL D1 as NE D1 as NEGD2 as E D2 as EL D2 as NE D2 as NEG

D3 as E D3 as EL D3 as NE D3 as NEGD4 as E D4 as EL D4 as NE D4 as NEG

Perception of D4s as geminates (top yellow stacks in chart) Perception of D1-2-3s as singletons (small “as NEG” stacks)Confusability of D1-2-3s: D1s and D2s heard as E or NE, D3s as NE (biggest stacks in blue, red, & green-tainted areas)

0%

20%

40%

60%

80%

100%

ProducedE

ProducedEL

ProducedNE

ProducedNEG

Perceived NEG

Perceived NE

Perceived EL

Perceived E