Presentation is loading. Please wait.

Presentation is loading. Please wait.

Project 4 Oliver Niebuhr University of Kiel Meghan Clayards McGill University, Montreal April 5, 2011.

Similar presentations


Presentation on theme: "Project 4 Oliver Niebuhr University of Kiel Meghan Clayards McGill University, Montreal April 5, 2011."— Presentation transcript:

1 Project 4 Oliver Niebuhr University of Kiel Meghan Clayards McGill University, Montreal April 5, 2011

2 Project 4 Research subject of P4 was assimilation of place of articulation… …contrastively investigated for English and French. Involved Institutions: York, Aix, Geneva. People: Meghan Clayards, Oliver Niebuhr, Gareth Gaskell, Christine Meunier, Noel Nguyen, Uli Frauenfelder, Leonardo Lancia, Sophie Dufour. 2 Steps: – (1) Identification of the PD of place assimilation in E … and F? – (2) Cross-linguistic comparison of the listeners’ sensitivity/ability to use these details in the decoding of words.

3 Project 4 Part I Why place assimilation and why French and English? Perception-oriented step (2) was the actual starting point. – French and English are supposed to differ clearly in terms of place assimilation. – English is well known for having regressive place assimilation across word boundaries (e.g., “swis{s/h} shop”, “ba{d/g} girl”) –  Place assimilation is claimed to not occur in French at all. – So, what do French listeners do with potential instances of place assimilation, and does it make a difference whether assimilation is partial is complete? – Ultimately: What kind of cognitive strategy do listeners use in coping with assimilations? Is there is single mechanism? Is it language-specific or language-universal?

4 Project 4 Part I Two parallel acoustic analyses of single-sentence read-speech corpora of 4 English and 4 French speakers. Elicited with a fast speaking rate similar to the spontaneous speech of the individual speakers. Based on place assimilation in sibilant sequences (/s/, /  /, /z/, /  /) across word boundaries… …and additional reference sequences of sibilants and labial consonants across word boundaries. Acoustic analyses included: – Durations of reference sibilants and sibilant sequences. – Centre-of-gravity (1,5-12kHz) means and ranges of each sibilant and sibilant sequence.

5 Project 4 Part I Main result: English and French are more similar than commonly thought! – Like English, French does show place assimilation in sibilant sequences. – Like English, the French assimilations are overall gradual, ranging from non-assimilations to (temporally & spectrally) complete assimilations. – But the English assimilation is direction-guided  strictly regressive (in some cases /  s/  [ss], e.g., “British sea”). –  French assimilation is quality-guided  towards post-alveolar into either direction, regressive and progressive. Regressive assimilation is stronger, but still less consistent than in English. French assimilation occurs independently of simultaneous voice assimilation, which can go into the opposite direction.

6 Project 4 Part I Results of duration measurements: = sibilants in a sequence have equal durations French: If there are two separable sibilant sections in a sequence, then the postalv. section is almost always clearly longer. English: There are less sequences with separable sibilant sections; but if they are separable, the second sibilant is longer, i.e. no order effect like in French.

7 Project 4 Part I Results of centre-of-gravity measurements: French: single av. and pav. sibilants clearly differ in mean CoG, both av-pav and pav-av sequences overlap more or less strongly with the pav. reference English: single av. and pav. sibilants clearly differ in mean CoG, av-pav sequences are almost all identical with CoGs of pav. reference  pav-av differ from pav. and av. in both CoG means and ranges.

8 Project 4 part II Perception experiments – Phonetic context can radically change pronunciation – how does the listener recognize “dresh” as “dress” in “dresh shop”? – At what level is the problem solved? Crosses word boundaries so can’t be purely lexical does it require top-down knowledge?

9 Visual context 9

10 Visual context 10

11 Accounts Contrast enhancement low level increases dissimilarity Phonological Inference 11 language dependant high level ‘undo’ learned patterns works on partial assimilation not language dependant works on partial and complete

12 Perception Experiment Compare languages that differ in assimilation patterns Equate stimuli and control acoustics for two language groups -Artificial lexicon Use a range of assimilation strengths -Continuum Examine the time course of compensation 12

13 Task English listeners (York, UK) 26 French listeners (Geneva, Switzerland) Rendez le cavisse pagune s’il vous plait Render the cavees pagoon please pagoon cavees

14 Task English listeners (York, UK) 26 French listeners (Geneva, Switzerland) pagoon Rendez le cavisse pagune s’il vous plait Render the cavees pagoon please cavees

15 Word 1Word 2 sival caveesshinnow caveeshpagoon pentuf Word 1Word 2 pidas tamashsamal nalipshamal remop Task from Pirog Revill et. al 2008 Word 1 = objects Word 2 = textures Materials 15

16 Predictions Contrast Enhancement Phonological Inference 16 EnglishFrench Right context Yes Left context Yes EnglishFrench Right context Yes? Left context no?

17 English listeners 17 “sh”“s” s sh control Left Context “Render the...amal” control tamash pidas /ʃ/ /s/

18 French listeners 18 Left context “sh”“s” s sh control “Render the...amal” control tamash pidas /ʃ/ /s/

19 Right context English listeners 19 “sh”“s” s sh control “Render the cavee... control shinnow” sival” /ʃ/ /s/

20 “sh”“s” s sh control French listeners 20 Right context “Render the cavee... control shinnow” sival” /ʃ/ /s/

21 Summary Contrast Enhancement Phonological Inference 21 EnglishFrench Right context Yes Left context Yes EnglishFrench Right context Yes? Left context no?

22 Eye-movements 22 If compensation is an auditory process, it should happen quickly Does it?

23 Eye-movements “cavees” “caveesh” “pagoon” 200 ms Trials …cavees pagoon… Time /s/ - /sh/ 1 0 Time % fixations 1 0

24 Eye-movements 24 Coronal bias (looks to /s/ - looks to / ʃ /) Average RT Render the cavee… Looks to button start 1 7 cavees caveesh Occulo-motor delay (200 msec)

25 Eye-movements 25

26 Eye-movements 26

27 Summary 27 Compensation for complete assimilation requires language specific experience Some other language general mechanism may also be at play for moderate assimilation Doesn’t seem to be rapid but might require some top down knowledge as well

28 Thanks to... Sound to Sense: Uli Frauenfelder, Université de Genève Sarah Hawkins, Cambridge University Christine Meunier, Université de Provence Noel Nguyen, Université de Provence Eyetrackers: Gerry Altmann, University of York Dirk Kerzel, Université de Genève Gareth Gaskell, University of York Oliver Niebuhr, University of Kiel

29 Modeling Predictions Contrast Enhancement CoG (Hz) “sh” Proportion “s” response “s” 29

30 Modeling Predictions Phonological Inference CoG (Hz) “s” “sh” Proportion “s” response 30


Download ppt "Project 4 Oliver Niebuhr University of Kiel Meghan Clayards McGill University, Montreal April 5, 2011."

Similar presentations


Ads by Google