VELOPHARYNGEAL FUNCTION:
A SPATIAL-TEMPORAL MODEL-
Fredericka Bell-Berti
sounds
the vocal tract (Fant
the vocal tract
and oral
additional
abil
to
cavities
air stream
For oral
of the
the
index
the
usefulness
41
and therefore, employing the velar coarticulation data reported in the
literature, as well as data from an experiment to be reported here, we propose
to offer a model of velar function that may prove to be a useful subject for
further comparisons with the actions of the human articulatory system.
II.
MECHANISMS OF VELAR CONTROL
A
Introduction
The role of the
mechanism in speech has been of interest
for many years, but the hi
of this interest will only be surveyed briefly
in this chapter.
(See Dickson & Maue-Dickson, 1980, for a comprehensive
historical perspective.) Thus, Fritzell (1969) reports studies by Czermak
(1857, 1858, 1869) and Passavant (1863) involving both indirect and direct
measures of velopharyngeal closure during speech.[1] The conclusion of these
experiments was that velar
decreases through the vowel series [i], [u],
[0], [e], [a].
Passavant also placed tubes of varying diameters in the
velopharyngeal port region to determine how small the port must be to prevent
nasalization of oral
sounds and found that a cross-sectional area of
12.6om2 had lit tIe effect on
qual, but that a cross-sectional area
of 28.3mm 2 resulted in the nasalization of most consonants. He also reported
a bUlging in the posterior
wall, above the level of velopharyngeal
closure during the speech of a cleft palate speaker. He assumed that this
bulging 9 which has come to be known as Passavant! s ridge 9 occurs in all
9
It is possible to
two lines of investigation leading from these
early studies. The first line concerns the dimensions and mechanisms of oral
and nasal articulation.
, is oral articulation achieved
(a)
and
movement of the velum; (
a combination of velar movement and
directed movement of the posterior
pharyngeal wall (Passavant i s ) ; or (c) a combination of velar movement
and medially directed movement of the lateral pharyngeal wall? Which muscles
are responsible for
the velar port? Need the port be completely
closed for all "oral" articulations? And, is nasal articulation achieved by
the contraction of some muscle or muscle group, or
by decreasing
activity in those muscles
for oral articulation? The second line
of
concerns
of variations in
activ
both as
function of
of
and as a function of
interactions among prox
(coarticulation)
Mechanisms
It
42
is
is the muscle
Bell-Berti, 1
This
and
peoples
Calnan (1
) has di
the presence of Passavant's
ridge in
and claimed that such a mechanism would be far too
sl
sh and
be
reliable
mechanism for
wi th inad
muscul ature"
(
w
Hill,
Pettitt & Kane, 1
& Hill t
that Passavant's r
is not a mechanism used
most normal
e cleft
ate
ects tend to use more
wall movement in
ng than do normal
) concluded that,
when Passav ant's r
clefts t i t
may be used as
them
In
lel
and
(1
1)
w
directed
and that
movement
r
was
velar
Passavant'
closure
constricthe 1
also been
Lubker 1
Me Call, & Barnes
anatomithe
atal
above thi
constrictor
movements
cour
Skolnick
43
showing
of the
constrictor, for
(cf& Minifie, Hixon, Kelsey &. Woodhouse, 1970;
seems
reasonable in 1
of EMG data
i
the
constrictor musc
, a t the level
at
superior
of the
(Bell-Berti, 1
first, and simplest, is that the muscles used in closing it relax and the
elastic tissue forces open the port
The second possibil i ty is that the
contraction of some muscle or group of muscles (possibly palatopharyngeus or
palatoglossus)
s downward on the velum while the muscles involved in
the port are relax
e
In an EMG study~ Fri
I
across
ects~ but in
[i] and [u]
Bell-Berti (1
works
stical
wi th the
for open than for close
isthmus for these articulations.
for the role
e
activ
to vary
more active for the vowel [a] than for
that the palatopharyngeus
but that it is more active
to narrow the faucial
Thus~ the available EMG data do not provide
as a velar
The
situation
however, for the pal
studies have
activ
occurs when
ni activ
is
is, at times
to nasal
consonant articulation (
~ Hirose, Sawashima, &:
ima, 1977;
Fritzell, 1
; Lubker, Fritzell, &: Li
ist, 1970; Lubker, Lindqvist, &:
Fritzell, Note )
In contrast however, Bell-Berti (1973 1976; Bell-Berti &:
Hi rose, 1
) has
EMG, recorded from several
, showi ng
no difference in
activ
associated with
in the status
of the velar
data show
activ
for
back vowels and vel
segments for which levator
atini
is also
Kuenzel, 1
),
involvement in
These authors have
vowel ~
rose (
Taken
universal mechanism
muscle (Bell-Berti,
nvol ves the
and for some
downward
, at the least, that there is no
velum
increased activ
in any
Rather ~ the basic mechanism for
the velar
of
i those muscles acting to close it,
the contraction of the
to provide
no evidence that the
45
among the control
articulators (cf
Lindblom ~
Stevens
House ~ 1
of vel
control
be an '
thIn, wi th variable
reI
and
correction for mechanical constraints
number of
& Shriner
for
s of
for
vowels
here, confirm the
)
,
pulsing to continue
the period of vocal tract occlusion for obstruent
(cf Bell-Berti, 1975; Perkell, 1969; van den
1958)
Conversely, some
s maintain the transglottal pressure difference necessary for glottal pul
allowing air to 'leak' through a partially opened
port (Dixit & MacNeilage, Note 1)0 Still other
accomplish this vocal
tract
ustment in other ways, including
and depressing the
root,
the lar
,or
volume (cf0 Bell-Berti,
1
Tatsumi, &
1969)
0
$
$
These
articulatory maneuvers
effective
volume, as well as the
ustment of
cross-sectional area
for vowels (cfe Bell-Bert, 1
), are
reasons e First, and
most obvious,
that
model
must account
all of the articUlatory
of
mechanism
Second, and
more
here,
interaction with
otherwise
our
of data collected
of
sequences of
which we must collect if we are
our under stand
interaction between motor
for,
the execution
0
IV
In addition
and their interaction
also tried to
uence
~~~-~~"~
OF PHONETIC
and nasal articulation
of velar function have
, the extent of the
position for proximate
motor
po si tion for
common observation
is no
and
(1)
positions than do the
velar elevation occurs
hav
ironment
for
One general
production that has been tested wi th velar
function data is
phoneme-based modele This model assumes the
input to the
to be
of phonemes that are
fied
as sets of invariant articulatory goals, or "features." It postulates a "lookahead" procedure that allows the
s of phonemes occurring later in the
to influence the current and
vocal tract
, so
as these
s are not in conflict with any more immediate
s [3] A model
from the Moll and Daniloff data proposes two
velar
s:
for
consonants and 'open' for nasal consonants
for vowels is assumed to be
fied, and determined
fied posl tion
The
lctions of thi
essentially binary
those of Henke's model of
tion, and a substantial
of the data
in
lctions of such
model
e
e
There are,
•
of the look-ahead model
The first of these
three instances in which blind
ication
account for observations of human
effect of a marked junctural
in
downstream
(McClean 1
Ushijima & Hirose,
in nasal
may result from
command to the vel um,
that this ex anamodel
and the look-ahead model concerns
NC sequences,
movement toward
for
consonant
model
issues
in
the
model
utterances whose
sequences velar
vO\rlel
) and
et 131",
A second, and more
that velar position for
velar
, limitation is
between oral
consonants,
that velar
the tacit
will be the
substantial
cav
----
FUNCTION
12.
11.0
10.
W
......
9.
8.
7.
-600
300
msec
at
x
~)(
xxxxxxxx)OOOO(xXx
)(Xxx
~
Ensemble~average
from the
carrier
contains
contains
is
50
the
velar elevation functions for two V1C V? phrases
utterance set described in Section V, B 1 spBken in the
The upper figure
sentence nIt's a
the function for
the lower
the function for the
Velar elevation
units time in msec
duration of the
on
9
e
extent of interaction effects among vowels and consonants, in entirely oral
utterances, and are described below.
B0
The
_-'l.......
_
ect in thi
York Ci
the
of standard Greater
utterances were 27 twoand
were [i] and [a],
the
the remaining 12
ases
combinations of
[ w i th word-boundary
positions
varied in each
vowel-order setse
This
such contrasts as 11 for ex
e, [itll
[ s t i J and [atslltiJ
Nine minimal contrasts were possible between vowel-order sets, in addition to
the
contrasts wi thin each vowel-order set among
hav
consonant
of different duration (and number of
Each phrase
and ended wi th an obstruent consonant, al
different consonants
and ended the two
The
were embedded in the carrier
sentence nIt'
ain 9t and
aced in lists in random order.. The
1 ists
had
from
to eight tokens
of each ..
0) was inserted into the
rested on the floor of the nasal
border of the hard
ate,
wall
from the level
A
thin
astic
nostril and
nasal sur face of the
supravelar surface and
VF
at 60
velum was then
The
each utterance
were
between the end of the
vowel, and
were
each
can be made
urn continues
many
51
s for
may sum cumulatively t and even the most extreme
exceeded $
another alternative~
one assuming positional
s, is that the velar
may not be achieved even during the production
of a
of five obstruent segments hav
a duration of 360 msec.
icit in this last
is a velar position goal that far exceeds the
velar
necessary to
nasal coupl
The second observation, al
mentioned
above and which admitted
cannot be
from the first~ is that velar
for vowels
differs from velar
si tion
oral consonants.
The obvious conclusion,
therefore ~ i
that the vel
s for vowels differ from those for consonants$ Furthermore the
for open and close vowel s, at the I east may
very well differ from each other
9
observations are al so po ssible • One
in velar
for different vowels
between vowel environment and maximum velar
Still other observations are concerned with
in relation to
ar position
the vowel s [
and [a],
for [
in each of the 18
(nine
sons (t
)$ These differthan in the first
refl
ar elevation for
infl uence of
3).
§
this difference
of the
elevation occurs
msec
and the average duration
ases.
of
turn, be
on
ar
extreme that
52
a
a
a
a
a
a
a
a
a
a
a
a
a
a
7.0
6.
a
LLnUL.L.
2.
U1
Lv
1
Velar position "
second
elevation
contrast
consonant strings:
the vocalic
of the first and
described in Section V,B
Velar
ordinate in
unitse Minimalthe abscissa
their
syllable 2 is at the
$
\..J1
-I>-
,"
,
a
_,'
/
x~~ ~~
~
. ~,
~~
"
..X
,
/
,,
},
,
X
.D-O=aCni
x--x -: iCno
3
Peak velar elevation, from the ensemble averages t for minimal
contrasts indicated
the abscissa. The smallest and largest
standard deviation values are shown
their respective
means.
o
11.
o
o
o
0
0
0
0
0
0
x
0
x
x
x
x
x
x
o
xX
X
x
X
x
x
8.5
x
200
300
.400
DURATION (MSEC)
Figure
lJ1
lJ1
0
Scatter-plot of peak velar elevation (along the
consonant-string duration (along the abscissa).
ordinate)
vs.
F~nally, to estimate the time at which V2 exerts more influence on peak
elevat10n than does V1 , velar elevation was compared in the nine minimal pairs
at several times before peak elevation was achieved: at 100, 150, 200, and
250 msec before the beginning of V2a The mean difference in velar position
was determined for each time point: by subtracting the value obtained for
[iC na ] strings from that obtained for [aC i] strings a So long as V.2 exerts
the greater influence, this difference si\ould be positive, and it"" should
~ecrease as the influence of V
2 diminishes, becoming negative when the
1nfluence of V1 exceeds that of V2 a These data are summarized in Table 1a
Clearly, at even 100 msec before V2 , the influence of that vowel is small
(t 8 ::1 a J9), and. at ~Oo. mse? before V,? .the mean difference across comparison
pa1rs 1S negat1ve, 1nd1cat1ng that th~ 1nfluence of V predominates.
1
Table 1
Mean difference in velar position between laC il and liC al utterances, taken
at 50 msec intervals before the (acoustic) ~nd of the Honsonant string (t::O
msec) a The difference is
at t::50 msec, where
exerts the
influence, and smallest at
wnere the infl
of V is
1
comparison time
(msec before V )
2
( 50)
100
Mean
Difference
200
150
08
-11 e 7
250
7
19
a
C
This n-ary
fication of
tion both of the
fication of
consonants,
Additional
consonants; it
po
nasal
the appropriate
close vowel is
1
'Model
the
already in the literature,
This model
res
or movement
s, one each for
vowel, and obstruent consonants
hal f-close vowel sand sonorant
other hand, be
to
velar
interaction between the nasal consonant and
s
for nasal zed
that for
open vowelsa
0
some
which it
rather than
56
abruptly, and to be
some fixed time after the (acoustic) end of the
segment for which it
fied
The model assumes that the velum is
programmed to achieve its maximum excursion for a
before the (acoustic) end of the
Once the velum has achieved this maximum
ment, it moves ei ther towards its rest position or po
, some neutral,
y
(cf@
& Halle, 1
p@ 300)@
(It should be
to determine whether
not this movement away from the maximum
is toward the rest position
the 'neutral'
on
vel
movement
before marked
al boundaries,
where the neutral po
ex
and in utterance-final positions, where the
be ex
@) The
fication may
take the form
mov
some
position,
or, alternatively,
or
tento The
model is not able
two al ternatives.
In ei ther
however,
successive
fications
described.
the
to the second
the data
/
will have
vowel because
; that is, closer
indeed,
in
velar
si tion
values
Bell-Berti
57
It is important to note that this
that it is the result both of temporal
component
comprising a particular
overlap, or co-occurrence, of gestures for successiv
not permit determination of the effect of
in lex
ng rate on the
of the
and end of
relation to the acoustic onset of the
for which
Nor does this model contain
the
among the
of
it claims that a vowel
infl uence an
about 150 to 200 msec before the acoustic onset of the
ins in relation
no claims about when the velar
movements for
to
is constant
of lexical
rate
Obviously, a
However, after the values
established
it should
infl uences on
model may be used to
determine the model'
Dixit,
58
velar
function
and
tend
Once
impl ies
among the
and of temporal
These data do
stress and in
velar
in
are
fied.
relationThus while
consonant
vowel, it makes
the
of
this
ng
available.
have been
account for
done the
utterances,
Bell-Berti, F.
An electromyographic study of velopharyngeal function in
speech. Journal of Speech and Hearing Research, 1976, 11, 225-240.
Bell-Berti, F. The extent of a vowel's influence: Evidence from a study of
velar function. In J. J. Wolf & D. H. Klatt (Edse) t Speech communication
papers presented at
97th Meeting of
Acoustical Society of
America. New York: Acoustical Society of America, 1979, 79-82.
Bell-Berti, F., Baer, T., Harris, Ko S", & Niiroi t So Coarticulatory effects
of vowel quality on velar elevation
Phonetica, 1979, 36, 187-193
Bell-Berti, F., & Hirose, He Patterns of palatoglossus activity and their
implications for
organization •
1973,
activity in vOlclng distinctions
A
rose, He
of
simul taneous
and electromyographic study.
Phonetics, 1975,
Benguerel, A.-Pe, Hirose, He, Sawashima, M , & Ushijima, T. Velar coarticulation in French: An electromyographic study
of
, 1977,
159-167.
Bj~rk, L.
Velopharyngeal function in connected
1961 ,
study of the
Bosma, J. F. A
upper pharynx by cadaver dissection and
after maxillo-facial
1953,
New York: AppletonThe
.......&.
,
,1
~------
1•
Modern view on Passav ant's
1957, 0 89-113
A., &
H. Lo
of Passav ant's Pad
New York
, N.,
Row, 1
Czermak, Je Ne Ueber das Verhalten des weichen Gaumens beim
reinen Vocale.
1857,
Fri tzell, 1969) e
reine
und
nasalierte
Vocale
Czermak,
Je N.
Ueber
( ci ted in Fr i t ze 11, 1969).
~~~~:.:::.:::~~~, 1
und
der Stimm- und
Wilhelm
~~~~~ _~
, Vol.
Dickson, D. R.
&
der
in
Normal and
mechanism.
in
and function
A
),
and
59
Fujimura, 00' & Lindqvist, J 0 Sweep-tone measurements of vocal-tract characteristics e Journal of the Acoustical Society
, 1971,
541
5580
Fujimura, Oe, Tatsumi, 10 F., & Kagaya, Ro
Computational
of
palatographic patterns e Journal of
, 1973, 1, 47-54
Hagerty, R. Fe, & Hill, Me Jo
Pharyngeal wall and
movement
postoperati ve cleft
ates and normal
ates.
_ _~
, 1960, 1, 59-660
, Ho
, M" J", Pe t tit, H. S.,
J J
al wall movement in normals
203-210
1958,
ton, R. A study of the
closure"
of
HarriS-:- ~-production
0
0
Henke,
Massachusetts Institute
To
Kent
60
Role of
wallo
Mermelstein, P. Calculations of the vocal-tract transfer function for speech
synthesis applications.
In Proceedings of the Seventh International
Congress on Acoustics, Vol. 3. Budapest: Akademiai Kiado, 1971, 173176.
Minifie, F. D., Abbs, J. H., Tarlow, A., & Kwaterski, M. EMG activity within
the pharynx during speech production.
ournal of Speech and
Research, 1974, 17, 497-504.
Minifie, F. D., Hixon-:- T. J., Kels~ey, A. A., & Woodhouse, R. J.
Lateral
pharyngeal wall movement during speech production.
of ~ - Hearing Research, 1970, 13, 584-594.
Moll, K. L. Velopharyngeal closure on vowels. J
of .......::.._Research, 1962, 5, 30-77.
Moll, K. L., & Daniloff, R. G. Investigation of the timing of velar movements
during speech. Journal of the
o f , 1971,
678-684.
Moll, K. L., & Shriner, T. H Preliminary investigation of a new concept of
velar activity during speech.
Palate Journal, 1967, 4, 58-69.
Niimi, S., Bell-Berti, F., & Harris, K. S. Dynamic aspects of velopharyngeal
closure. Haskins
, 1978,
45-62.
Nylen, B. O. Cleft palate and speech.
1961,
Ohala, J. J.
Monitoring soft palate movements in
Linguistic Analysis Reports (Phonology Laboratory,
tics, University of California, Berkeley), 1971, 1
Passavant, G.
Ueber
Verschliessung des
Frankfurt
J D. Sauerl~nder, 1863 cited
Perkell, J. S.
of
......&_-
a
1
RUbin, P., Baer, T., & Mermelstein,
perceptual research.
Research, 1979, SR-57 ,
Sc ull y, C., & Shi r t, M. A.
contexts. Proceeding
Sciences. Vol. 1.
Copenhagen, 1979, 212
Shprintzen, R. J., Lencione, R. M., McCall, G. N.,
dimensional cinefl
anal
of
speech and nonspeech activities in normals
ll, 412-428.
Skolnick, M. L. Video
emphasis on lateral
for
&
Skolnick, M L. A three
closure during
, 1974
1
61
Stevens, K. N., & House, A. S. Perturbation of vowel articulation by consonantal
contex t:
An acoustical stud y • Journal of Speech and
Research, 1963, 6, 111-128.
Subtelny, J. D., Koepp-Baker, H., & Subtelny, J. D.
function and
cleft palate speech. Journal of Speech and
-"",
, 1961,
213-224.
Ushijima, T., & Hirose, H.
Electromyographic study of the velum dur
speech. Journal of Phonetics, 1974, 2, 315-326.
Ushij ima, T., & Sawashima, M,.
FiberscoP'ic observation of velar movements
during speech.
Annual Bulletin (Research Insti tute of Logopedics and
Phoniatrics, University of Tokyo), 1972,
25-38.
van den Berg, J. Myoelastic-aerodynamic theory of voice production.
Speech and Hearing
, 1958, 1, 227-244
Warren, D. Nasal emission
and velopharyngeal function"
Journal, 1967, 4, 148-1560
Zagzebski, J. A" Ultrasonic measurement of lateral pharyngeal wall motion at
two levels in the vocal tract. Journal of Speech and
1975, ~, 308-318.
FOOTNOTES
1It is possible to observe articulator movements associated wi th
gestures in two fundamentally different ways (cf" Bell-Berti, 1973)..
first of these,
Viewing, involves measurement of articulator
for example, measuring the elevation of the velum over time
include visual observation (using posterior rhinoscopy or
cinematography t cinerad
t
ul trasonic echo
, and
recording of reflected light.
The second group of methods,
the cause or result of articulator
specifying articulator movements, includ
acoustic, and transillumination recordings
---~
The
, involves measurements of
acement, implying but not
flow
is of some interest to note that all of these
recent data
provide general confirmation of Passavant's (1863)
that a
al port cross-sectional area
12.6 mm 2 had little effect on the
speech, while a cross-sectional area of 28 mm 2 resulted in nasal
oral speech sounds, and thus, in distorted speech
3Another fr
nikov and Chistovich
the basic uni ts of
described as
CV
model accounts for
for vel
function
consonant
unless we assume that
different
62
4Binary models are frequently proposed because of their simplicity
However, if a binary model requires a large number of reorganization instructions to account for observational data, it seems that an n-ary model may have
equal, or even greater, elegance.
50ne would expect cumulative velar position as the response of an openloop system. Such a system would obviate the need for continuous
of velar position, while guaranteeing
closure
preventing nasal coupl
dur
oral segments.
A.
Before considering the acoustic effects of
the nasal resonator to
the pharyngeal and oral resonators it seems prudent to provide
and/ or descriptions of
that will, of
, find their \-lay into
the following discussion. Thi treatment, of course, will not, and could not,
be exhaustive
9
Traditionally, in
an acoustic
we view the
vocal tract as an acoustic tube, one hav
variable
and
For
oral speech sounds, the tube is a simple one, hav ing no side branches, wi th
one end at the glottis and the other at the
[ 1] For voiced oral
sounds, the acoustic
of such a tube can be described
its
, which is the ratio of the volume
at the
to
- - -sound
- source (the
).
transfer function can be
described
its
, resonances that can be described
their
and bandwid ths, or
The resonance
and their bandltlid ths
are a function of
and
of the tube (Fant, 1 1; Stevens &
House, 1955, 1961). For voiceless speech sounds, the transfer function is the
ratio of the volume
at the 1
to the sound pressure of the source,
which, in this condition, is the
noi
or transient exci tation
at the vocal tract constriction (Bell.
• Heinz stevens,
1).[2]
B
Adding a side
acoustic
interactions of the
wi ttl those of the
branches
shunt,
the vocal tract tube
several ways
6
n
63
nasal-branch driving point admittance decrease, approach the
of
their paired zeroes, and are cancelled (Fuj imura & Lindqv ist, 1971) " The
poles of the nasal-branch driving-point admittance are the fre
at
which zeroes are observed in the transfer function of the vocal tract
therefore, the closer the pole frequencies of the nasal-branch admittance to
the resonances of the rest of the vocal tract, the more extensive will be
effects of adding the nasal branch to the
In spite of this
ity, however, it is possible
qualitatively, the results of some of the interactions among
and
resonators" (We will confine ourselves to the
on the transfer functions of vowel , and not
because of an interest in understanding observed differences
for different vowel s,,)
First, the lowest formant
will fall between the lowest nasal-branch resonance
formant of the
non-nasal zed vowel
effects of nasal
occur in the
admi t tances into the
and nasal
in the
reduction, aero
vowel,
~f F , and
is minimized (cf
Lindqvist. 1
J
been established, however
It has not
has the
acoustic effects of nasal
ThUS, while it is known that close vowels will
than will
at smaller velar
~'J;
Henderson & Marshall, 1
resul ts from
the
of nasal
FOOTNOTES
1This is an
for the
considerations
acoustic
Shirt
64
3An advance in understanding the interactions between coupled pharyngeal,
oral, nasal branches was effected by Mermelstein (Mermelstein, 1971; Rub in,
Baer, & Mermelstein, 1979), who established a method for calculating the vocal
tract transfer function, based on the independence of the driving point
admittances looking into each branch from the velopharyngeal port and of the
pressure gain across each branch. This has simplified the techniques necessary for calculating the coupled-system transfer function
e
65