Perception & production of [voice] contrasts
in Dutch word-initial plosives
Alexandra M. Pfiffner | LSA 2021
roadmap
background on acoustic cues & Dutch [voice] contrasts
research questions
experiment methodology
results
what next?
acoustic cues
phonological contrast is signaled by multiple phonetic cues (Lisker, 1986)
[voice]
glottal pulsing
VOT
f0
transition
F1
transition
acoustic cues
phonological contrast is signaled by multiple phonetic cues (Lisker, 1986)
[voice]
glottal pulsing
VOT
f0
transition
F1
transition
cue weights
glottal pulsing
f0
transition
f0
transition
glottal pulsing
Dutch [voice]
[+voice] in Dutch is traditionally realized with prevoicing
/t/elg 'descendant'
/d/elg
'paying off debt'
/p/erk
'limit'
/b/erk
'birch tree'
van Alphen & Smits (2004)
25% devoicing rate
Pinget (2015)
22-30% devoicing rate
*differed by regional dialect
following vowel f0
/p, t, k/
/b, d/
6 hz
16 hz
Slis & Cohen (1969)
van Alphen & Smits (2004)
Dutch
Afrikaans
Afrikaans
Coetzee et al. (2018)
-
younger women (20s): 83% devoicing rate
-
older women (40s-60s): 44% devoicing rate
-
robust difference in following vowel's f0
-
listened to a young woman's voice in 2AFC
-
artificially manipulated continua varying in VOT and f0
-
all participants relied on both VOT and f0 as cues to underlying voicing
-
when in conflict, VOT > f0
Afrikaans
Pfiffner (in progress)
-
men and women, ages 20s and 60s-80s
-
devoicing rates: women > men of all ages
-
robust difference in following vowel's f0, but larger difference in older participants
-
listened to four different voices in 2AFC
-
artificially manipulated continua varying in VOT and f0
-
when prevoicing is present, it's the dominant cue
-
in the absence of prevoicing, f0 is a cue for all listener/speaker combos but less for the perception of male speakers
6 hz
16 hz
25% devoicing
Slis & Cohen (1969)
van Alphen & Smits (2004)
Pinget (2015)
22-30% devoicing
(2020)
?
research questions
What is the current state of [voice] contrasts word-initially in Dutch?
-
devoicing rate
-
differences in following vowel's f0
-
differences by age and gender
-
which cues signal a [voice] contrast
-
relative cue weighting
-
differences by age and gender - both speaker and listener
experiment
25 native speakers of Dutch
Speech Lab, University of Amsterdam
ages 22-29; 61-71
experiment
pad
telg
Randomized wordlist
20 minimal pairs
-
10 /p,b/
-
10 /t,d/
+ fillers
experiment
pad bad
tal dal
2AFC
artificially maniuplated continua
-
/pad/-/bad/ 'road', 'bath'
-
/tal/-/dal/ 'quantity', 'valley'
experiment
experiment
prevoicing
f0
none
100%
changing
changing
changing
changing
stable; ambiguous
changing
4 different continua
experiment
female
male
younger
older
4 speaker voices
separated by block
results
younger females: 25%
younger males: 17%
older females: 0%
older males: 9%
mean devoicing rates
individual means
results
following vowel's f0
normalized
Brunelle et al. (2020)
z-scores --> converted for readability with group means & SDs
p < 2e-16 ***
results
following vowel's f0
non- normalized
35 Hz
24 Hz
59 Hz
23 Hz
results
younger females
younger males
older females
older males
devoicing
25%
17%
0%
9%
Hz difference
35
24
59
23
results
continua types
no prevoicing - f0 changing
100% prevoicing - f0 changing
prevoicing changing - f0 constant
prevoicing changing - f0 changing
results
continua type
no prevoicing - f0 changing
100% prevoicing - f0 changing
prevoicing changing - f0 constant
*collapsed across all listeners and speakers
results
continua type
no prevoicing - f0 changing
100% prevoicing - f0 changing
prevoicing changing - f0 constant
*sorted by speaker voice
results
continua type
100% prevoicing - f0 changing
*sorted by speaker voice
no prevoicing - f0 changing
prevoicing changing - f0 constant
continua type
100% prevoicing - f0 changing
listener groups hearing a 60-year-old woman
no prevoicing - f0 changing
prevoicing changing - f0 constant
results
continua type
prevoicing changing - f0 constant
*sorted by speaker voice
no prevoicing - f0 changing
100% prevoicing - f0 changing
results
both prevoicing and f0 can cue phonological voicing
f0 effect: alveolar > bilabial
in the absence of prevoicing:
-
f0 cues voicing more when listening to a younger male speaker and older female speaker
when prevoicing is present:
-
prevoicing is the dominant cue
-
high f0 can outweigh full prevoicing when listening to an older female speaker
research questions
devoicing rate: 0-25% (lot of individual variation!)
differences in following vowel's f0: significant differences, 23-59 Hz
differences by age and gender: Yes!
-
younger speakers devoice more
-
female speakers have larger f0 differences
research questions
which cues signal a [voice] contrast: both prevoicing and f0; though there are place of articulation effects
relative cue weighting: prevoicing > f0
differences by age and gender:
speaker: f0 is stronger with older female voice
listener: no large effects
implications
Dutch may be in the beginning stages of tonogenesis, but too early to be certain
right now: we can see cue re-weighting happening
devoicing in progress
f0 differences widening
f0 can cue contrast with or without prevoicing
what next?
-
examine task effects: word list versus reading passages
-
look closer at results on an individual basis
-
connect perception & production in individuals
-
make larger connections with Afrikaans
This research is supported by NSF Doctoral Dissertation Research award 1918306 as well as two grants from Georgetown University.
Many thanks to Dirk Vet and Paul Boersma for welcoming me into the Speech Lab at the University of Amsterdam.