ENG 528: Language Change
Research Seminar
Sociophonetics: An Introduction
Chapter 4: Consonants
Vowel synthesis exercise
• 1. Record yourself saying a short sentence and digitize it.
• 2. Apply some sort of filtering (lowpass, highpass,
bandpass, or band zeroing) to the signal in a way that
would be useful in a perception experiment.
• 3. Using the “To Manipulation” function in Praat, change
the F0 and timing of the utterance. You may change
different parts of the utterance in different ways.
• 4. Using the “To KlattGrid” function, change some of the
formant values enough so that the affected vowels sound
like different phonemes.
• 5. Send me your original digitized utterance and each of
the three modified soundfiles. Due October 3.
Field marks for stops
Stops are characterized by “stop gaps.” Voiced stops (right)
show a murmur; voiceless stops (left) don’t. As a rule, stops
are followed by bursts unless the stop is unreleased.
Field marks for nasals
Nasals show formants in their stop occlusions. They
may or may not show a burst. Shown: [m].
Field marks for fricatives
Fricatives show frication noise, though its robustness
varies. Shown: [f] (left), [v] (right).
Field marks for affricates
Affricates show a stop gap followed by frication, with a
burst in between. Shown: [pf] (left), [bv] (right).
Field marks for approximants
Lack of a stop gap or frication; otherwise, there aren’t any
consistent commonalities, as [w] (left) and trilled [r] (right)
show.
Featured variables involving manner of
articulation
• Interdental fricatives vs. stops
• Weakening of voiced stops in Spanish after
vowels
• Tapping/flapping in American & Australian
English
• Affrication of stops in Scouse and Québec
French
• Affricate/fricative confusion in Mexican
American English
Stopping of word-initial /ð/
after a consonant in Pearsall, Texas
70
Mexican Americans
Anglos
60
50
Percentage of Realization as [d]
• E.g., keep that pronounced
[khip dæt]
• Cases such as had that
pronounced [hæd dæt] that
are ambiguous between
stopping and assimilation
are not included in the
stopping tally
• Ordinary Least Squares
Regression used
• Ethnicity was a significant
predictor (p=.007) and
education was close
(p=.051), but year of birth
and sex were not significant.
40
30
20
10
0
1920
1940
1960
Year of Birth
1980
2000
Place of Articulation (1)
What—tube models again??!
6000
F6
F6
5000
F6
F5
F5
F5
F6
F5
F5
F6
F6
F5
F5
F5
F4
F4
F4
3000
F3
F3
palatal
velar
uvular
14
12
10
8
pharyngeal
alveolar
F2
1000
F4
F3
F3
2000
F4
F2
glottal
F4
bilabial
labiodental
dental
frequency in Hz
4000
0
16
6
4
distance constriction is from glottis in cm
2
0
Place of Articulation (2)
Place of
Articulation
General effect on:
F1
F2
F3
Bilabial
lowered
Lowered
lowered
Labiodental
lowered
lowered
lowered
Dental
lowered
raised next to back
rounded vowels,
lowered next to front
vowels
slightly raised except
next to high front
vowels
Alveolar
lowered
slightly raised except
next to high front
vowels
Retroflex
lowered
Palato-alveolar
lowered
raised next to central
and back vowels,
lowered next to mid
and high front vowels
raised next to back
vowels, lowered next
to front vowels
raised
Palatal
lowered
strongly raised
slightly raised
Velar
lowered
raised
lowered
Uvular
lowered?
lowered
slightly lowered
Pharyngeal
raised
strongly lowered
strongly lowered
strongly lowered
raised
Locus Equations
}
Stop Gap
}
Visible Transition
F2 of [ti]
Locus
F2 of [t]
3000
F2 of [tu]
2000
time
F2 of onset
2500
F2 of onset
frequency
F2 of [t
]
1500
2000
1000
1500
/b/
/d/
/
/
/b/
/d/
/
/
1000
500
500
1000
1500
F2 of vowel
2000
1000
1500
2000
2500
F2 of vowel
3000
Featured variables involving place of
articulation
• Interdental fricatives vs. labiodental fricatives
in English (especially British dialects)
• Dental, alveolar, and retroflex consonants
• Alveolar (or dental) vs. velar nasals
Frication: Peak location
Comparison of adult male and girl
smoothed spectra for sibilants
Unsmoothed spectrum of [x]
25
peak
20
adult male []
adult male [s]
10-yr.-old girl []
10-yr.-old girl [s]
20
15
peak
peak
0
10
peak
Amplitude in dB
Amplitude in dB
10
-10
-20
5
0
-5
-10
-30
-15
-40
-20
-50
-25
0
2000
4000
6000
Frequency in Hz
8000
10000
-30
0
2000
4000
6000
Frequency in Hz
8000
10000
Frication: Spectral moments
• Moment 1=mean frequency; whether the
energy is relatively high- or low-frequency
• Moment 2=variance; range of energy, i.e., how
concentrated the energy is
• Moment 3=skewness; gets at spectral tilt
• Moment 4=kurtosis; how much of a peak
there is in the spectrum
Frication: Ad, Sp, and S´p
• Jesus and Shadel (2002), applied to Portuguese
F
30
_
F
Amplitude in dB
20
Ad S'p
10
0
Sp
-10
-20
-30
0
5000
10000
Frequency in Hz
15000
20000
Featured variables involving fricative
spectra
• Laminal and apical [s]
• Dorsal fricatives: [ç], [x], [ ], and their voiced
counterparts
• Aspiration vs. frication: note that formants
(including low formants) are visible in
aspiration, while only high formants or none
at all are visible in frication
Direct measurement of articulation
• X-ray microbeams: phoneticians have used
them for a long time
• Electropalatograph (EPG): useful for contact
between tongue and roof of mouth
• Ultrasound: mostly for tongue because
ultrasound can’t handle air spaces
Voice Onset Time (VOT)
• Well-known term in phonetics
• Relevant only for syllable-onset or ambisyl. stops
• Lead, short-lag, and long-lag VOT
Measuring VOT
• Length of time between the burst and the onset of
vocal fold vibration
• VOT will be negative for lead VOT (pre-voicing) and
positive for lag VOT (voiceless, especially aspirated)
• Judge onset of vocal fold vibration by onset of F2
• Go with the last burst if there are more than one
• For intervocalic stops, measure from offset of F2 for
preceding vowel
• Aspiration tends to be shortest for labials and longest
for dorsals, with coronals in between
VOT in Pearsall (labials shown)
35
Number of Tokens
30
25
20
/p/ after voiced, Mex. Am.
/p/ after pause, Mex. Am.
/b/ after voiced, Mex. Am.
/b/ after pause, Mex. Am.
/p/ after voiced, Anglo
/p/ after pause, Anglo
/b/ after voiced, Anglo
/b/ after pause, Anglo
15
10
5
0
<-170
<-120
<-70
<-20
VOT
<30
<80
Glottalization
• Important variable for British dialects
• Slowed glottal pulsing is the key
Cues for the voicing distinction
property
adjacent F0 contour
adjacent F1 contour
approach of F1 to
closure
duration of closure
duration of preceding
vowel
glottal pulsing
intensity of burst after
closure
associated phonation
“voiced” (lenis or
non-spread glottis)
depressed
strongly depressed
closer
“voiceless” (fortis or
spread glottis)
elevated
not depressed as much
less close
shorter
longer
longer
shorter
may be present
greater
absent
lesser
aspiration and
glottalization less
likely
aspiration and
glottalization more
likely
Stop releases
• Main issue is whether release is present or absent
Stop release for Pearsall (left),
Robeson County, NC (right)
Laterals: clear and dark /l/
• F2 frequency is the key
Clear and dark /l/ in Pearsall
9.0
Mexican Americans
Anglos
curve for Mexican Americans
8.5
Mean Z2-Z1 for /l/ Midpoint
8.0
7.5
7.0
6.5
6.0
5.5
5.0
4.5
4.0
1920
1940
1960
Year of Birth
1980
2000
Other lateral variation
• Velar [ ] vs. vocalized: hard to tell acoustically; F3
bandwidth is everybody’s best guess now, but in
addition F1 & F2 may be a tad higher for vocalized
• Palatal lateral [ ] vs. palatal central approx. [j]: [j]
shows a higher F3, maybe a higher F1
Rhotics
• No common features of rhotics—mainly, they’re
all spelled with r
• Lots of variables in different languages
• We’ll focus on a couple of kinds of variation here
• Since we won’t cover assibilation, here’s what it
sounds like
Uvular /r/ forms
• Important if you want to study Continental
European languages
• Uvularization is characterized by lowering of
F2 and maybe a little raising of F3
• Lots of variation in uvular /r/
Bunched vs. retroflex /r/
• Important in English
• Frequency of F4 (!) turns out to be crucial
• How do you normalize it for interspeaker
variation? (We haven’t covered normalization
yet.)
Non-rhoticity
• For English, F3 frequency is the crucial factor
• My advice: use a combination of auditory
judgment and examination of F3 on spectrograms
(compare with F3 of nearby vowels)
• Pearsall results (unstressed /r/) shown below
Percent Rhotic in Unstressed Syllables
100
90
80
70
60
Mexican Americans
Anglos
50
1920
1940
1960
Year of Birth
1980
2000
Question for discussion
• What steps would you take in figuring out how
to tell the difference, acoustically, between a
pair of sounds not covered in this chapter—for
example, velar [k] vs. uvular [q], alveolar [n]
vs. palatal nasal [ ], or lateral fricative [ ] vs.
post-alveolar [ ]?
Formant plot fun
300
i
u
400
oi

r
e
600
o
u
vo
e . #.  r
e . .. e
N ..
æ æ:
500
 ai  
ai
#
vo
r
ai
r
r
l
o

o
au

=o
 
o
ai
600

..
r
'


v
o
700
æ
au
u

v
au


æ
800
700
2000
l
u
r
F1
F1
e

400

r
'
i
oi
r
r
=o

<here>
500
i
1800
1600
1400
1200
F2
1000
800
600
2500
2000
1500
F2
1000
500
References
• The diagrams on slides 9, 21, 27, and 32 are taken from:
• Thomas, Erik R., and Janneke Van Hofwegen. 2011. Consonantal
Variation in the English of a Spanish-Substrate Community. Paper
presented at 14th International Conference on Methods in
Dialectology, London, Ontario, 5 August.
• The diagram on slide 25 is taken from:
• Miller-Newman, Sara E., and Hayley E. Heaton. 2011. Final stop
accommodation in married couples. Poster presented at the 161st
meeting of the Acoustical Society of America, Seatle, WA, 27 May.
• Other reference:
• Jesus, Luis M. T., and Christine H. Shadle. 2002. A parametric study
of the spectral characteristics of European Portuguese fricatives.
Journal of Phonetics 30:437-64.