ENG 528: Language Change
Research Seminar
Sociophonetics: An Introduction
Chapter 5: Vowels (Continued)
Lindblom (1963), Undershoot
Third Laboratory Assignment
• From any interview consisting at least partly of conversation,
choose a diphthong and measure at least 25 tokens of it. (You’ll
need a diphthong with plenty of tokens, so don’t pick /oi/.)
• Measure the duration of the entire diphthong and formant
values for the nucleus and/or glide.
• Using Origin, plot the duration of the whole diphthong on the xaxis and F1, F2, and, if you like, F3 of either the nucleus or the
glide on the y-axis. Include regression lines. You may make
different plots or one plot showing all the formants (you may
need a logarithmic scale for that). This has to be in Origin
because I want everybody to learn how to use it.
• Print out the plot(s) and turn them in on October 24.
Undershoot: Definitions
• Target: the point that the speaker “intends” to
reach (controversy: is it an articulatory or
acoustic target?)
• Undershoot: failure to reach the target
 Because of short duration
 Because of weak stress
 Other reasons? (I.e., can some sounds be
specified as being given less effort to reach
their targets? Do some languages or dialects
show more undershoot than others?)
More Definitions
• Coarticulation: overlap of segments
• Assimilation: a segment becomes more like a neighboring
segment
• Truncation: part of a segment, such as a steady state or a
glide, is lopped off when the duration is shorter
• Compression: no lopping off, but all the components are
squeezed into a shorter time frame
• Reduction: weaker articulation; for vowels, it means
becoming more schwa-like. Two senses:
 Phonological: predictable alternation between a full
vowel and schwa; no intermediate realizations
 Phonetic: vowel is on a gradient between full realization
and schwa, with any intermediate realization possible
Plotting Undershoot
• Generally, you plot the duration of something (generally
the whole vowel) against a formant measurement
• Example for /a/ in K’iche’:
800
.5
2
42
x+
2
.26
F1 of vowel in Hz
700
1
y=
600
500
400
0
50
100
150
200
duration of vowel in ms
250
300
Undershoot plots to tease out
phonological distinctions
• Here’s what figure 5.28 is supposed to look like:
formant value of glide in Hz
2000
F1 before voiced obstruent
F2 before voiced obstruent
F1 before voiceless obstruent
F2 before voiceless obstruent
1000
900
800
700
600
500
400
300
100
150
200
250
duration of entire diphthong in ms
300
Degree of Undershoot can Differ:
/a/ vowels in three languages
900
900
F1 of vowel in Hz
8. 9
46
x+
700
5
.65
1
y=
600
Am. English
Slope=-0.04101
500
400
0
50
100
150
200
250
800
F1 of vowel in Hz
Turkish
Slope=1.655
800
700
y=-0.04101x+719.0
600
500
400
300
0
50
100
duration of vowel in ms
K’iche’
Slope=1.262
.5
422
+
x
2
.26
y=1
600
500
400
50
100
150
200
duration of vowel in ms
250
300
350
400
450
250
300
Am. English, long
outliers removed
Slope =0.08185
900
800
F1 of vowel in Hz
F1 of vowel in Hz
800
0
200
duration of vowel in ms
900
700
150
y=0.08185x+702.2
700
600
500
400
0
50
100
150
200
duration of vowel in ms
250
300
Factoring Out Coarticulation
• Measure formants at
vowel onset or offset
• Measure formants at
some point within the
vowel
• Subtract within-vowel
formant measurement
from onset~offset
formant measurement
• Use that difference to
normalize for
coarticulation
Lindblom (1963)
“Spectrographic Study of Vowel Reduction”
• this is the paper that launched experimental
phonetics as a science—it provided a testable
hypothesis
• looked at the effects of duration on Swedish short
vowels
• had subjects say vowels in b_b, d_d, and g_g
contexts in the carrier phrases så är det ___ and
__ är det så
• subjects had metronome played into their ears so
that they said the words at different speeds
Lindblom (1963)
“Spectrographic Study of Vowel Reduction”
• duration did have an
effect—vowels became
closer to their offset values
at short durations
• the results suggested that
each vowel phoneme has a
target that doesn’t vary—
the consonantal context
doesn’t affect the target
• p. 1780: He says that the
undershoot pattern isn’t
corrected by speakers and
“it presupposes that the
listener is able to correct
for coarticulation effects”
Lindblom (1963)
“Spectrographic Study of Vowel Reduction”
b_b
d_d
g_g
Reactions to Lindblom (1963)
• In subsequent years, several other researchers found that a
number of other factors besides simple duration influenced the
degree of undershoot:
 Stress (possibly a stronger effect than duration—see Harris
1978, Language and Speech 21:354-61)
 Speaking style—research opportunities for sociolinguists?
 The particular vowel phoneme
 The particular language (presumably, that could mean
dialects vary, too—opportunities for sociolinguists)
 The particular speaker (more opportunities for
sociolinguists)
• In response, Lindblom (1990) developed his “H&H” theory,
which says that speakers vary their clarity of speech in response
to the needs of their addressees
Perception and Undershoot
•
•
•
•
Lindblom & Studdert-Kennedy
(1967) found that listeners’
perception compensated for
coarticulation
Perceptual boundaries between / /
and / / shifted depending on the
context, in a compensatory pattern
This is the same process that Ohala
(1989, 1993) later called “corrective
rules”
They also noted that diphthongal
glides don’t have to reach their
targets because listeners expect to
hear them undershot (i.e.,
truncated)
Conditioned Vowel Shifts
• Occur as a result of coarticulation that, at some point, isn’t corrected for
perceptually and thus is reanalyzed by language learners
• In American English, conditioned vowel shifts are most common before /r/, /l/,
nasals, and palatals (from anticipatory coarticulation)
• Perseveratory shifts are less common, though note what happened to short
/æ/ after /w/ in Early Modern English
• What coarticulatory effects on vowel formants would you expect from each of
those kinds of consonants?
400
i
K
r
i
T
u
500
N
æ
F1
Here’s a vowel plot for a female
speaker from southeastern Ohio.
What conditioned vowel shifts
appear in her speech? What
secondary developments did the
conditioned shifts allow?
600
700
800
900
e
u
l
l=u
..
oi .
l
o
r
e 
o
r=ur2
r
r
r
'
=o =u
   1
=
=



.
..
   
N
N
o
v
=
ai ai
=


au
g
d
1000
æ
2500
2000
æF
æ
P
æ
1500
F2
1000
Discussion Question
• How might difference in the degree of
undershoot take on social meanings? What
social meanings could they index, both for
inter-speaker variation and for stylistic
variation?
Fun with Vowel Plots
300
i
T
K
u
u
r
r
'
e 
F1
e
600
F
ai 
N
d
æ
au
P
æ
800
2500
2000


v
F2
1000
oi

o =
'
l
l
r
r
r

=o =u 1

600



N
d
æ
700
500
o
l
r=ur2
N

N
<result>
ai
1500

e
æ



u
500
o
æ
æ
e
l
=


oi




u
r
<bulletins>
...
K
i
<poor>
o
u
r
400
o
r
500
700
u
l
i
400
T
i
F1
300
g





=
=
o
ai
au
r
ai
æ
ai
P


=
æ æF
v
800
2400 2200 2000 1800 1600 1400 1200 1000 800
F2
600
Today’s lab demo
• Use of Origin for plotting
• If you know what you’re doing with it, it’s
versatile and can make esthetically better
plots than programs whose primary purpose is
something else
References
• Harris, Katherine S. 1978. Vowel duration change and its underlying
physiological mechanisms. Language and Speech 21:354-61.
• Lindblom, Björn. 1963. Spectrographic study of vowel reduction. Journal
of the Acoustical Society of America 35:1773-81.
• Lindblom, Björn. 1990. Explaining phonetic variation: A sketch of the H&H
theory. In William J. Hardcastle and Alain Marchal (eds.), Speech
Production and Speech Modelling, 403-39. Dordrecht: Kluwer.
• Lindblom, Björn E. F., and Michael Studdert-Kennedy. 1966. On the rôle
of formant transitions in vowel recognition. Journal of the Acoustical
Society of America 42:830-43.
• Ohala, John J. 1989. Sound change is drawn from a pool of synchronic
variation. In Leiv Egil Breivik and Ernst Håkon Jahr (eds.), Language
Change: Contributions to the Study of its Causes, 173-98. Berlin/New York:
Mouton de Gruyter.
• Ohala, John J. 1993. Coarticulation and phonology. Language and Speech
36:155-70.