Breathing and speech planning in turntaking Francisco Torreira
Breathing and speech planning in turn-taking Francisco Torreira Sara Bögels Stephen Levinson Max Planck Institute for Psycholinguistics Nijmegen, The Netherlands
A psycholinguistic puzzle A’s turn 100 -300 ms B’s turn In conversation, the most frequent transition between speakers takes only a few hundred ms (e. g. Stivers et al. , 2009; Heldner & Edlund, 2010)
A psycholinguistic puzzle A’s turn 100 -300 ms B’s turn B’s production planning > 600 ms Planning and producing language takes time: - word-picture naming: 600 ms (Levelt et al. , 1999) - simple sentence production: 1500 ms (Griffin & Bock, 2000)
A psycholinguistic puzzle A’s turn B’s production planning Speakers often plan their turns in overlap with their interlocutors’ turns (Levinson, 2013)
A psycholinguistic puzzle Direct evidence for overlapping production and comprehension during conversation is scarce Can the breathing behavior of interlocutors provide such evidence?
A psycholinguistic puzzle Direct evidence for overlapping production and comprehension during conversation is scarce Can the breathing behavior of interlocutors provide such evidence?
Research questions In read speech, deeper and longer inbreaths before longer utterances Whalen & Kinsella-Shaw, 1997; Fuchs et al. 2013 What about spontaneous conversation? What is the timing of speakers’ inbreaths relative to their interlocutors’ turns?
Conversational corpus with Respitrace inductive plethysmography
Initial observations As in controlled experiments (e. g. Mc. Farland 2001): – Vital cycles – Speech cycles But also (as in Bailly et al. 2013 for collaborative reading): – Speech-adapted vital cycles? – Apneas: listeners often stop breathing for several seconds!
Materials Conversational context in which a turn transition is relevant: Q & A Assistant identified Q & A sequences in 6 dyadic conversations (~ 5 h) We restricted the dataset following these criteria: – Answer is relevant to the question – Syntactically marked (wh-word, SV inversion) or intonationally marked (L* H-H%, H* H-H% or H*L-H%)
Breathing in Q&A sequences A’s question B’s answer Time
Measurements A’s question B’s inbreath B’s answer Time Asnwerers’ inbreaths that occurred after the beginning of the question
Measurements Acoustic signs in the speech signal attributable to either a lexical item or particle A’s question B’s inbreath Time B’s answer
Measurements Acoustic signs in the speech signal attributable to either a lexical item or particle A’s question B’s inbreath B’s answer Time First point of silence, syntactic completion, and prosodic completion
Breathing behavior and answer length A’s question Presence vs absence Depth Duration B’s inbreath Time B’s answer
Presence of an inbreath NO INBREATH 38% 62% INBREATH n=145 Not all answers are preceded by an inbreath
Answer duration & inbreaths β = 949, t = 3. 95, p <. 0005
Answer duration (ms) Inbreath depth and answer duration β = -0. 03, t = -0. 19, p = 0. 85 Speaker-normalized Inbreath depth
Timing relative to question end A’s question B’s inbreath Time B’s answer
Question question Inbreath timing to question end Inbreath Answer
Question question Inbreath timing to question end Inbreath Answer
Question question Inbreath timing to question end Inbreath Answer
Question question Inbreath timing to question end Inbreath Answer answer < 2. 5 s answer > 2. 5 s
Question question Inbreath timing to question end Inbreath Answer answer < 2. 5 s answer > 2. 5 s Partly vital? Speech inbreaths?
Timing relative to answer start A’s question B’s inbreath Time B’s answer
Question question Inbreath timing to answer start Inbreath Answer
Question question Inbreath timing to answer start Answer Inbreath -650 ms
Is the timing of answerers’ inbreaths sensitive to where questions end?
Is the timing of answerers’ inbreaths sensitive to where questions end? We examined the relationship between: - Gap duration Question question Inbreath Answer - Inbreath timing to answer start Question question Inbreath Answer
Distance to answer start (ms) Are answerer’s inbreaths anchored to question ends or answer starts? β = 0. 48, t = 10. 4, p < 0. 0001 Gap duration (ms)
Conclusions Inbreaths are more likely to occur before long answers > breathing behavior can be informative about speech planning in conversation too The timing of inbreaths before answers is sensitive to the timing of question ends, and is very often aligned with it. > evidence of interlocutors’ orientation to turn ends > speech planning often starts early during the interlocutor’s turn: A’s question Decision to take an inbreath contingent on answer length Inbreath preparation 140 -320 ms Draper et al. , 1960 B’s inbreath B’s answer
References Bailly, G. , Rochet-Capellan, A. , and Vilain, C. (2013). Adaptation of respiratory patterns in collaborative reading. Proceedings of Interspeech 2013. Draper, M. H. , Ladefoged, P. , and Whitteridge, D. (1960) Expiratory pressures and airflow during speech. British Medical Journal, 1(5189): 1837– 1842. Fuchs, S. , Petrone, C. , Krivokapic, J. , and Hoole, P. (2013). Acoustic and respiratory evidence for utterance planning in German. Journal of Phonetics, 41(1): 29– 47. Griffin, Z. M. , and Bock, K. (2000). What the eyes say about speaking. Psychological Science, 11: 274– 279 Heldner, M. and Edlund, J. (2010). Pauses, gaps and overlaps in conversations. Journal of Phonetics, 38: 555 — 568. Levelt, W. , Roelofs, A. , and Meyer, A. (1999). A theory of lexical access in speech production. Behavioral and Brain Sciences, 22(1): 1– 37. Mc. Farland, D. H. (2001). Respiratory markers of conversational interaction. Journal of Speech, Language, and Hearing Research, 44: 128– 143. Stivers, T. , Enfield, N. J. , Brown, P. , Englert, C. , Hayashi, M. , Heinemann, T. , Hoymann, G. , Rossano, F. , de Ruiter, J. P. , Yoon, K. -E. , and Levinson, S. C. (2009). Universals and cultural variation in turn-taking in conversation. PNAS, 106(26): 10587– 10592. Whalen, D. H. and Kinsella-Shaw, J. M. (1997). Exploring the relationship of inspiration duration to utterance duration. Phonetica, 54: 138– 152.
- Slides: 34