Flapping and Reduction of Stops: Speech Style, Phonological Environment, and Variability

Natasha Warner ([email protected]) University of Arizona, Tucson, U.S.A., and Max Planck Institute for Psycholinguistics, Nijmegen, the Netherlands

1.0 2.0 I.Introduction III.Results .9 1.8

h • Figure 1: Conversational speech style has the most reduction There is tremendous variability in how sounds are realized in t 1.6 g o .8 i

n (significant for all four measures; style interacts with stop for two). t Segmental env.

connected speech, and many variants sound natural to listeners. Flaps a STOP e 1.4 r r

t Conversation does not elicit the full set of conditions, so other .7 pre-schwa y in English can be reduced to an approximant or deleted: s t b 1.2 i t analyses examine only wordlist reading.

s pre-l

.6 n n ‘computer’ ‘matters’ ‘yesterday’ deleted d a 1.0 R R e phrasal • Figure 2: Segmental environment affects reduction inconsistently. t m

.5 r

burst n

I k .8 Stops before a full vowel /i/ are anomalous: they are unlikely to have

o pre-r F .4 p .6 pre-i clear formants, even though they do remain fully voiced. Thus, stops

post-r before /i/ may reduce in a different way than others. These are .3 t .4 V2 max V1 max F2, F3 throughout conversation story list d t excluded from other analyses. min. inten. Fig. 1: Comparison of speech styles, by Fig. 2: Effect of segmental environment • Figure 3: Stress and stop significantly affect reduction as measured on formant strength (0=F2 and F3 cease by intensity ratio of stop to surrounding vowels, and they also

s stop. Ratio of minimum intensity of e • How often are flaps reduced? r stop to averaged maximum intensity of during stop, 1=weak F2 and F3, 2= interact. /b, g/ are less reduced than /t, d/ in post-stress environment, u

g strong F2 and F3 throughout stop). • Is flap reduction influenced by segmental environment, stress i preceding and following vowel. but approximately as reduced as /d/ when between unstressed f

l environment, and/or formality of speech style? l syllables. /t/ is significantly more reduced than /d/.

a .9 2.0

• Do flaps differ from other stops in how they are reduced? Do flaps r • Figures 4 & 5: Effects on formant strength and on proportion with a o f .8 h derived from /t/ vs. /d/ differ? n voiceless period during the stop are similar to intensity (except that t

o Stop 1.5 STOP i o g i • Does a categorical phonological rule apply to /t, d/ in flapping t /t/ and /d/ do not differ on voicelessness). /p, k/ are also slightly t n

c .7 a e

u b b r r

environment, or is flapping phonetically variable reduction? reduced between unstressed syllables. d t y e S t

r d d i .6 1.0

• Much past research focuses on whether /t, d/ become flaps (Kahn 1976, t • Figure 6: Effects on presence/absence of a burst are less clear than s r n e n a t g g

Patterson & Connine 2001). This project examines further reduction in e other measures. Because some bursts have extremely low amplitude, a

t .5 m e n r r environments where they should flap. I k .5 k this may not be a useful measure. o g

F

e .4

t p p a c II.Methods i .3 t 0.0 t IV.Conclusions d

n post-stress inter-unstressed post-stress inter-unstressed • 22 native English speakers were recorded, data from 1 is reported. i • Reduction of all intervocalic stops is common, as indicated by several

s

• 3 speech styles: casual conversation, story reading, word list reading. e acoustic measures. Even /p, k/ in careful wordlist reading can be u Fig. 3: Effect of stop and stress Fig. 4: Effect of stop and stress l • Words containing /p, t, k, b, d, g/ in six segmental environments and a environment on intensity ratio. environment on formant strength. realized as fully voiced. /b, d, g, t/ often have F2 and F3 continuing v

two stress environments were used for the reading conditions. 10 items r throughout the constriction, frequently lack bursts, and have only a e h

per condition for list reading where possible, 2 items for story. g 1.0 1.0 small dip in intensity. i t H

u • /t, d/ are more reduced than /b, g/ in post-stress environment on most t o s

h .8 .8

Sample stimulus words by stop and stress r measures, suggesting that there is something categorically different

g STOP STOP u u

Stop Post-stress Inter-unstressed b about alveolars. This could be evidence that a categorical

o t r .6 b .6 b u /t/ status limited h phonological rule applies to alveolar intervocalic stops, and/or that t o d d h d

/d/ credit prejudice t tongue tip gestures are more susceptible to reduction (de Jong 1998, i e .4 .4 c

g w g i

/p/ appetite precipice Fukaya & Byrd 2005). . o v p

/b/ inhibit halibut k k • /t/ is more reduced than /d/ on most measures. Surprisingly, this is the . .2 o .2 r p P

/k/ recognize applicable o p p opposite direction from an incomplete neutralization effect (Dinnsen r P /g/ magazine esophagus 0.0 t 0.0 t 1985). This could reflect frequency. /t/ being at least as reduced as post-stress inter-unstressed post-stress inter-unstressed /d/ suggests a categorical phonological rule applies at least to /t/. Sample stimulus words by segmental environment Fig. 5: Effect of stop and stress Fig. 6: Effect of stop and stress • Stress environment seems to affect /b, g/ and /p, k/, but not the Before schwa status environment on whether voicing environment on presence/absence of a alveolars. This indicates a difference in how phonological rules and Before syllabic /l/ cattle continues throughout the stop. visible burst. phonetic reduction apply to alveolars vs. other stops, but it is not clear Before /‘/ butter why /t, d/ are immune to reduction based on stress environment. Before full vowel /i/ pretty References • All styles show reduction, but reduction is greater in conversation After /r/ forty Connine, C., & Patterson, D. 2001. Variant frequency in flap production. Phonetica 58:254-275. than reading. However, the effect of speech style is relatively small. de Jong, K. 1998. Stress-related variation in the articulation of coda al veolar flaps: Flapping revisited. Phrasal (Across word boundary, before schwa) write a letter Journal of Phonetics, 26, 283-310. Dinnsen, D. 1985. A re-examination of phonol ogical neutralization. Journal of Linguistics, 21, 265-279. V.Future research Fukaya, T., & Byrd, D. 2005. An articulatory ex am ination of word-final flapping at phrase edges and • Measurements: ratio of minimum intensity (21 msec window) during interiors. Journal of the International Phonetic Association, 35, 45-58. • Large number of speakers to examine the degree of inter-speaker and consonant to average peak intensity of preceding and following vowel; Kahn, D. 1976. Syllable -based generalizations in Engl ish phonol ogy. Indiana University Ling uistics Club. intra-speaker variability. Mitterer, H., & Ernestus, M. In press. Listeners recover /t/s that speakers reduce: Evidence from /t/- presence/absence of burst; approximate formant strength during lenition in Dutch. Journal of Phonetics. • Analysis of word frequency. Pluymaekers, M., Ernestus, M., & Baayen, R.H. Submitted. Articulatory planning is continuous and consonant (none, weak, clear); presence of voicelessness during sensitiv e to informational redundancy. • Analysis of intonation and discourse prominence in conversation and consonant; duration measures. Port, R. 1977. The influence of tempo on st op closure duration as a cue for voicing and place. Haskins story reading (cf. Pluymaekers et al, submitted). Laboratories: Status Report on Speech Research SR-51/52, 59-73. • Perception and processing studies: how do listeners process reduced forms? (cf. Port 1977, Mitterer & Ernestus in press.)