Professional Documents
Culture Documents
Articuladores - Consequencias Acústicas
Articuladores - Consequencias Acústicas
Citation: The Journal of the Acoustical Society of America 50, 1166 (1971); doi: 10.1121/1.1912750
View online: https://doi.org/10.1121/1.1912750
View Table of Contents: http://asa.scitation.org/toc/jas/50/4B
Published by the Acoustical Society of America
Trading relations between tongue-body raising and lip rounding in production of the vowel /u/: A pilot
‘‘motor equivalence’’ study
The Journal of the Acoustical Society of America 93, 2948 (1993); 10.1121/1.405814
Departmentof SpeechCommunication,
Royal Institute of Technology(KTH), S-100 44 Stockholm70, Sweden
LIP
LABIAL
fHEIGHT WIDTH SHAPE
> A• L•
TIP MOUTH
AND
TONGUE >A(x) > F - pottern
BODY PHARYNX
SHAPE
LARYNX HEIGHT
1. Lip Opening
First we note that the lips resemblethe wedge-like
openingthat would be obtainedby making two slant
cuts in the wall of a cylinder-shapedobjectsothat they
Fro. 5. Empirical tongueshapefor [-o• (solid line) compared
intersect.The lip contours when projectedon a vertical
with a contourderivedby rule (dottedline). The calculatedpoints frontal planecan then be approximated by the follow-
are described in terms of two numbers: d=0.7 and c= 1.0. ing expression:
3. A pical Component
At this point we shall sketch only briefly how the w
model can be extendedto includean apical component.
The tonguetip may be positionedto form a turbulence- Fro. 6. Definition of
free constrictionas in coronalvowels or a partial or lip opening parameters.
completeobstructionas in dental, alveolar,or retroflex
obstruents.Sinceon the undersidethe apexis attached
w
to the floor of the mouth and thus cannot be stretched
beyonda certainlength,the cooperationof the mandible
and the tongue body is required in attaining a given
E
E of w, the distancebetween the innermostpoints of
contact of the upper and lower lip contours,we find
slightly more complicatedconditions.Consider the
Z30 outer pair of lip contoursindicated in Fig. 6. These
Z Fro. 7. Prediction
contours are specifiedin terms of W, the distance
of dependence of
O width (w) of labial between the mouth corners,and //, the midsagittal
20
openingon jaw open- separation.In shape,they are identicalwith the inner
ing (j). The data contoursbut are displacedvertically from these by
o points and the com-
•' lO puted line pertain a constant=l=(H--h)/2, roughly interpretableas the
o
to neutral lip con-
ditions.
thicknessof the lips. For passivelips and variable
mandiblepositionwe assumethat W remainsconstant.
.-r 0
' I I I The assumptions made so far permit us to infer that
0 10 2o 3o
the dependence of w on h is
JAW OPENING (j) mm
w= w, (4)
was reportedby Fromkin in a study of American To checkthe realismof Eq. 4, photographicdata were
6 Thesefindingsthus indicatethat p obtained for a male Swedish talker who also served as
Englishvowels.
tendsto beconstantfor anygiventalker.To predictA, subjectfor the x-ray material referred to earlier. The
sufficientto know h, w, and the positionsof the lips and the mandiblewererecordedon
it is consequently
talker-dependent value of p. In the material that we 35-mm film for a set of Swedishvowels and a range of
haveinvestigated so far, p tendsto fall in the range mandiblepositionsand passivelips. A thin and light
between2 and 3. A valueof p--2 canbe inferredfrom cap-splintwas fitted to the lower incisorsof the subject
Fromkin's data (seeRef. 6). and a firm thin metal wire was attached to the cap-
It is clear,however,that both the heightand width splint. The purposeof this wire was to indicate the
of mouth openingare controlledpartly throughthe position of the mandible.It was adjustedso as to inter-
actionof themandible,partly by meansof labiomuscular fere minimally with lip articulation. This procedure
activity proper.We have assumedthat, for no labio- was evaluatedin previouswork7 and was found to give
muscularactivity, the jaw dependence of the vertical satisfactoryresults.A specialhead-restwasconstructed
to eliminate irrelevant head movements. All measure-
separationof the lips is simplyof the form
ments were made in relation to dots painted on the
ho=j--k, (3) subject'sface in placesthat couldbe assumedto move
wherej is the separationbetweenthe jaws and k-the not as a result of lip and jaw activity but only owing
distance that the mandible must move from a closed to inevitableminor displacements of the head.The data
positionbeforethe lipsbeginto open.This assumption were taken usingconventionalenlargementequipment.
hasbeenborneout by photographic data. In the case In Fig. 7, measurements
of w for passivelip conditions
were plotted as a function of j, the degree of jaw
opening,whichis definedasthe verticaldisplacement of
the cap split wire from its position for completely
'-'
'•'
._.
5
I I I I I I. Iß • ,,V'•'•'ør•r•/
closedjaws. The solid line representsthe relationship
predictedwith the aid of Eqs. 3 and 4 and the assump-
tion that W=38 mm, H-h=2 mm, and k=4 mm.
These values were used in all calculations to be re-
Z / / / / Wm- -10mm ported. As can be seen,the theoreticalcurve is quali-
0 e i/
tatively in goodagreementwith the data points.
Since our purposeis to distinguishjaw-dependent
' i from labiomuscularcomponentsof labial shape, we
shouldinterpret W as equal to Woq-Wmand H as
0 1- / / - Ho-kHm where the indicesrefer to the jaw-dependent
/ 0
and labiomuscularterms, respectively.By combining
•: O- -
theseexpressions and Eqs. 2, 3, and 4 we obtain the
• I I I I I I
0 5 10 15 20 25 30 equationthat relatesthe area of lip opening,A, to its
determiningfactors: the positionof the mandible,j,
JAW OPENING (j) mrn and the state of the lip muscleswhichis representedin
Fro. 8. Predictionof dependenceof areaof labial opening(A)terms of H,• and Wm. Figure 8 demonstrateshow A
on jaw opening(j). Passivelip conditions.Also shownare aver-dependson j for H,•=0 and W,• =- 10 mm ("rounded"
ageddata for Swedishvowels.The dashedlinespertainto con-
ditionssimulatinglabiomuscularly controlledvariationof lip conditions),for Hm=0 and W•=0 (neutralconditions),
openingwidth ("rounding
..... spreading"). and for H,•=0 and Win=q-10 mm ("spread" con-
1170 Volume50 Number4 (Part 2) 1971
ACOUSTIC CONSEQUENCES OF ARTICULATOR MOVEMENT
:z
tangent to the lips, an alternative choicewhich may o
[u], according
to x-ray observations
of the subject.A
similar range of larynx heightswas noted by Perkell.1ø
Unlessotherwisestated, in the followingpresentation
the ranges of the other parametric values were as
follows:
i i i i
30 - ß - 3.0
..' '• ....... ß ..... . ...... ß
ß" ß.... O ' '
-• 2,5-ß....
ß o.ø ...o
.............. 2,5
•>- 2,0
...-' F2 - 2,0
z
O"51
Ill C : 1.0 I IßWm: 10 ,.5
m 1.0/d:-l'0]
•- Iøwm:-1ø
1.0
z
0 0.5 0.5
:
O. 1.5 -
I.U
o 0,5
-- _ _
I I I I i I I I
5 10 15 20 25 5 10 15 20 25
observed.Therefore, we used two different functions (Ref. 3, p. 100). (For further detailssee Ref. 3, pp.
for converting crossdistancesinto area values in the 100-101.)
pharynx region:one from the uvula down to the upper
point of the epiglottis,and anotherfor the bottom part III. ACOUSTIC PROPERTIES OF THE MODEL
of the pharynx. The relation between crossdistance,
o, and cross-sectionalarea, A couldbe approximatedby A. Effect of Jaw Movement
conditions can be inferred from the dashed and solid strict the vocal tract radically. For larger valuesof j,
linesof Figs.8 and 10.In thebottomrightpart of Fig. 12 a pharyngealconstrictionis produced.F• decreases as
the neutral tonguecontour(c-0) wasused.The remain- a functionof jaw opening.For c-0 andc= 1.0,d = -- 1.0,
ing graphs have the tongue shapesassociatedwith F• is relatively unaffectedby the jaw.
[i-], [u-] and[-•-]asindicatedby thevaluesfor c andd. The largestvariationin Fa associated with mandibular
In all four cases it can be seen:that F1 rises as the movement is found in the case of c= 1.0 and d-- 1.0.
jaw openingbecomeslarger.This is true alsofor values Here the constrictionis in the palatalregionfor smallj
of d and c not exemplifiedin Fig. 12. This riseis quite values(Fig. 12, top left). It seemsreasonableto assume
substantial--severalhundredhertz--as the jaw moves that the Fa values in the rangej=5-9 mm can be ex-
froma 5- to a 23-mmopening. In the topleft graphwe plained as half-wave resonancesof the front constric-
find an I-i-I-likeformant pattern for j-9 mm and tion. For the lowestvaluesof j and both "spread" and
"spread"lips. For a configuration of this type, it is "rounded" conditions,this constrictioncomprisesnot
customary to regard F1 as associatedwith a Helmholtz only the tongue passagebut also the lip section.
resonatorwhosevolume is that of the back cavity and Spreadingby no more than W,•= 10 mm is insufficient
whose neck dimensions are those of the constricted in counteractingthe effectof closingand prolongingthe
front part. (For a furtherdiscussion of the formant- tube that the smallj valueshave (cf. Figs. 8 and 10);
cavity relationsreviewedin this section,seeRef. 3.) As hencethe low Fa valuesnot only for roundedbut also
the mandibleis lowereddrasticallythe tongue,although for spreadconditions.Note that 5c/4l providesreason-
humpedas for [i-], no longerconstrictsthe vocal tract. able approximationsof Fa at large j values.
Thus, for j-23 mm, the vocal tract approachesa
B. Effect of Tongue Parameters
straightuniform tube in shape.Here F1 comescloseto
500Hz and isbetter interpretedasa quarter-wavelength 1. TongueBody Position
resonanceof the entire tube. Also, in the two graphson Figure 13 was prepared to illustrate the effect of
the right-handside,the highF1 positionsat largevalues varying the positionof the tonguebody.When studying
of j shouldprobablybe interpretedin a similarfashion. Fig. 13, it shouldbe borne in mind that the formant-
Under these conditions no radical constrictionsexist, pattern data do not pertain to conditionsof equal
the tongueshapes b•ing thosefor [-u-]andthe neutral cross-sectionalarea, as in the nomograms of Fant
contour,respectively. In the lowerleft plot,the tongue (Ref. 3). Thus the top left plot of Fig. 13 (j= 5) shows
shapeis that for [-a•. This shapeis such that small how formant frequenciesvary under "spread" and
cross-sectional areas are observedin the pharynx for "rounded" conditionsas the tongue mass slidesfrom
large valuesof j. For area functionsfor [-a• and [-a•-] a palatal (d=- 1.0) to a pharyngeal(d= 1.0) location.
which normally exhibit-pharyngeal constrictionsof At d- -- 1.0, a minimum cross-sectionalarea of 0.16
varying extent, F1 can be approximatedas the lowest cm2 is producedin the regionof the hard palate, and at
quarter-wavelengthresonanceof the front cavity. The d-1.0 the minimum area is found in the pharynx but
low F1 values that we note in the caseof the I-u-] and is larger, 1.1 cm•. The value of 0.16 cm• is obtainedat
[a-I, and the neutral tongueshapes(right and bottom d-l.0 only with the cooperation of the mandible,
left plots)for smalljaw separations are to a largeextent which has to be lowered to 23 mm.
determinedby the smallA and fairly largel valuesthat Qualitatively, the anterior-posteriordisplacementof
the presentmodelassignsto rounded,neutral, and even the tongue body results in similar formant frequency
spreadconditionsin the lower range of j (cf. Figs. 8 changesfor all mandibular conditions.F1 is lower for
and 10). Consequently,we must concludethat, with palatal positionsof the tongue than for pharyngeal
all other parameters constant, mandible movement positions.The magnitudeof this variation is somewhat
alone causes considerable shifts in Fl. smaller than that associatedwith jaw movement (cf.
The effect of jaw openingon F2 is most pronounced Fig. 12).
in the case-ofc= 1.0,d =0 (Fig. 12, top right). To under- It is important to note that all calculationswere
stand this drastic rise in F2 as a function of mandible undertakenon the assumptionthat the impedanceof
lowering,we shouldrememberthat for j= 5 mm there the vocal-tract walls is infinite. As Fant and Sonesson 15
are two constrictionspresent' one at the lips (alsofor point out, this is true only in a first approximation.The
"spread"conditions)., and one near the velum.The mass of the walls plays a role in tuning F1, and its
formant pattern is in suchcasesapproximatelythat of effect increaseswith an increasingpharyngeal-cavity
a doubleHelmholtz resonatorconfiguration;hencethe wall area. Consequently, the F1 valuesassociatedwith
loweredF•. At j=23 mm, no constrictionis present configurationsfor which d<0 are somewhattoo low.
(in spiteof c--1.0), and we might regardF• as a three- If the wall effect had been considered,the F1 curvesin
quarter-wavelength resonanceof the entire vocal tract. Fig. 13 would have started at somewhathighervalues
In the bottom left graph of Fig. 12, the tonguehump is (sinced=- 1.0 and the wall area is large) but would
displacedin a pharyngealdirection.At small j values, have terminated at about the same frequencies as
this pharyngealdisplacementis not sufficientto con- shown(d= 1.0 and the participatingareais small).
3.0 -
, , , , Ij =5 ' i j__11
2.5
2.0
1.5
1.0 I'Wmm:_1
o_
'r 0.5
• 0
z I I li , I I i I I I I
-1.0 -0,5 0 0,5 1.0 -1.0 -0,5 0 0,5 1.0
0 2.o
1.5
1.0
0.5
i i i i i • I I I I
-1.0 -0.5 0 0,5 1.0 -1,0 -0.5 0 0,5 1,0
3.0 X NEUTRALSHAPE
O d=0
- 3.0 about the geometry of rounding and spreading in
b d=10 normalhuman speech.It is likely that the lip muscles
ß d=-l.0
2.5 2.5 contract to adjust not only the width but also the
height dimensionsof lip opening. Nevertheless,the
2.0•- 2.0
measurementsperformed should give fairly direct
information on the acousticconsequences of a labio-
muscular jaw-independentgesture. Its most marked
1.5 1.5 effectsare found in Fa and F•., as is apparentfrom Figs.
12 and 13 (filledand opencircles).In the top left graph
1.0 1.0 of Fig. 12, the dotted pair of curvesindicatesa situ-
=
ation where rounding-spreadingmodifiesFa consider-
0.5• -
ably. The tonguebody is here in a palatal position.It is
0.5
probably correctto associateFa with a half-wavelength
resonance of the palatal constriction."Spreading"tends
0 - - 0
I I I , I i I to shorten this constrictionwhereas"rounding" makes
0 0.5 1.0 o 0.5 1.o
it longer; hence the shift in Fa. For other parameter
TONGUE BODY SHAPE (c) combinations(--1.0>d>l.0 in Fig. 13 and d--0 in
Fro. 14. The effectof tonguebody shape. Fig. 12), it is F•. that exhibitsa largerangeof variation.
We concludethat "rounding" lowers all formant fre-
quenciesunder all conditions.This loweringis particu-
width parameter,by 10 mm. "Spreading"refersto an
larly pronouncedfor the Fa of vocal-tract shapes
increaseof this parameter by the same amount. It
with palatal constrictionsand for F•. associatedwith
appearspossibleto interpret this parameter as closely
(palato-)velarand velopharyngeal
constrictions.
Shifts
related to the major axis of the elliptical slingsof
in F1 are of smaller magnitude and are slightly larger
musclefibersthat enclosethe lip opening.[Note that
in the case of pharyngeal constrictions(F1 front-
the other width parameter,w, (Fig. 6) cannotbe inter-
cavity dependent).
preted in terms of underlying muscular activity as
easily,sinceit is dependenton the vertical dimension
of lip openingas well.] ChangingWmby q-10 mm has
D. Effect of Larynx Lowering
consequences for the area of lip openingA, as demon-
stratedin Fig. 8. The effecton the point of vocal-tract The simulated larynx lowering correspondsto an
termination of this changecan be read from Fig. 10, increaseof the pharynxcavity length,and thus alsoof
in which it is seenthat Lm=q-5 mm, an automatic the total vocal-tractlength.Consequently,it lowersall
resultof the variationin W•. By this particularchoice formant frequencies,especiallythose that can be re-
of parametricvalueswe do not want to imply anything garded as a back-cavity resonance.The percentual
decreasesobservedfor different vowel qualities similar
TAB•.•.II. Effect of larynx lowering.The upperset of numbers to thoseof the subjectare givenin Table II. The follow-
refersto formantpatternscloselysimilarto thoseof the subject, ing observationscan be made. For most vowels, the
and the lowerset refersto the loweredlarynx conditions.
effecton F1 keepscloseto 5%-6%. For I-u-]and I-a-I,
F• F• Fa F4
the effectis strongerand for [a• weaker.The effecton
F•. in the vowelsarticulatedwith a tongueshapesimilar
250 2255 2702 3534
[-i• 235 2038 2663 3424 to that of [i-] is quitelarge(around8%). This appearsto
be a consequence of the fact that F•. tendsto be a half-
[-e• 315295
1940
1810
2520
2490
2955
2786 wave resonanceof the back cavity for those vowels.
460 1840 2597 3165
•e• 436 1704 2520 3040 F•. for the back vowels[u•, [o-], and [a• is muchless
199 1935 2060 2883 affected,but is largefor [a•. For all vowelsexcept[u•,
[-Y• 188 1830 2040 2713 Fa is rather insensitive,which might be explainedby
410 1720 2480 2965
[-½• 389 1600 2418 2870 the front-cavity affiliation of this formant. On the
246 775 2340 2964 average,F4 dropsabout5%--somewhatlessfor [-o-]
[-u• 224 766 2224 2813 and [-i-I,somewhatmorefor [-a-]and I-a-].
452 818 2423 2770
[o• 435 800 2410 2670 Summarizing,we may say that F3 is least sensitive
650 920 2663 3510 to an expansion of the pharynxcavity in the epiglottal
[-a• 600 882 2580 3260
683 1130 2550 3576 regionand, in termsof percent,F •.is the mostsensitive
•a• 662 1034 2470 3327 formant. The absoluteformant frequencychangeis
570 1563 2498 3109
[-ae• 548 1407 2445 2945 largestfor F4. The net effectof the larynx loweringon
282 1612 2300 2911 Fa andF4 is to decrease the frequencydistancebetween
[-u• 271 1495 2247 2744 them,whichhasbeenclaimedto be a characteristic of
sungvowelsarticulatedwith a loweredlarynx.•6
in the pharynx that is sufficientlynarrow for the [(•]- and their articulatoryattributesis unique; that is,
quality. The reasonis that the jaw movesalsoforward only small variations in parametricvalues preserve
as it is elevated(cf. Sec.I-A-l). The tonguecompen- the auditoryimpressions
intact, especiallyin the pe-
sationnecessaryfor this "close"[a• is thus alsoa form ripheralseries.
Allowinga somewhat
largerrangefor W•
of "superpharyngealization." Clearly, the model as wouldfor examplecontributeto decreasing the j and
describedabove fails to generatethesesupershapes.c valuesfor [e] and [•]. As remarkedearlier,the man-
Further researchand examinationof x-ray data will be dible positionsof Table II are in good qualitative
necessaryin order to extend the inventory of tongue agreementwith traditional descriptionsin terms of
shapesthat the model is capable of generatingat degreesof openingand with previously reportedjaw-
present. However, a preliminary extension suggests positiondata (seeRef. 7).
itself as we attempt to interpretthe resultsof the jaw From theseresults,we canconclude that the model
experiment.It is likely that the compensatory tongue explainsthe originof the articulatoryopen-close di-
shapespreserveacousticallyrelevant aspectsof the mensionin vowels. 19It makesan explanationpossible
area functions.Let theseshapes,therefore,be defined because it is jaw-basedand derives"tongueheight"
as earlierwith respectto the mandibleand as the con- indirectlyand as a secondary featurecharacteristic
tours that for any given vowel and mandibleposition of the finalvocal-tractconfiguration. The explanation is
leave the area function invariant. Since all super- basedon a principleof "minimal articulatoryantag-
shapesrepresenta deformationof the tongue shape onism"which enablesus to understandwhy qualities
from the neutral shapethat goesbeyondthe rangeof suchas [i, y, u, i, u• appearin normalpronunciation as
c-valuesexploredso far (0_<c_<1.0), it is natural to as- "close"vowelswith a smalljaw opening and [a,a,•e•
signto the supershapes a value of c that variesbetween appearas"open"vowels withlargerjaw openings. With
1.0 and 2.0, dependingon the degreeof deformation. referenceto area functions,it would not make senseto
(Note that mostbut not all of the compensatory con- talk aboutopenand closevowels.Usingthe minimum
toursderivedfor the vowelsof Table II aresupershapes. constrictionarea as a criterionof opening,we would
For example,[e• and [•, whenproducedwith smaller haveto group[i, i, u, o,•, a, a• together in a categoryof
jaw openingsthan normal, presuppose compensatory "close"vowels.In the "open" classwe would find
shapesfor which c< 1.0.) [e, e, •e•. This is obviouslyat variancewith the phono-
Characteristicof the supershapes is the antagonism logical behaviorof these segments.The open-close
betweenthe tonguemusclesand the jaw muscles.In a feature correlatesbetter with jaw opening.
compensatoryvariant of [i-[ pronouncedwith wide Normally, children acquire the ability to pro-
jaws, the depressormusclesof the mandible and the duce vowels and consonantswith the characteristic,
genioglossusassistedby variousother muscles,suchas "natural"jaw positions.
They do so withoutexplicit
the mylohyoid, opposeeach other. Similarly, in an instructionasto thesearticulatoryhabits.How do they
"open" [-u• the labial musculature and the stylo- do it? In view of the precedingargumentit might
glossusamongothermusclesparticipatein overcoming tentativelybe suggested that the principlereferredto
the pull of the jaw muscles.And in a "close"[a-[, abovecorresponds to an innatemechanism.
increasedactivity in the hyoglossusand pharyngeal The precedingremarksnecessitate a few commentson
constrictorsis likely to occur.It is obviousthat these the questionof whetherthe mandibleshouldbe in-
supershapesare not physiologicallyoptimal, and we corporated in articulatorymodelsof speech production
shouldexpectthe systemto strive for a minimization or not.Theanswer depends onthephilosophy andgoal
of suchantagonism.The problemcan be formulatedas of ourresearch. If weconstruct modelsof speech produc-
follows' For any given formant pattern or area func- tionin orderto contribute to phonetictheory--thatis,
tion, what is the optimal positionof the mandible, to our generalunderstanding and knowledgeof the
optimal position being defined as minimum displace- speechsignaland the mechanisms that generateit--
ment of lips and tongue from their neutral positions? the answer is in the affirmative. 2ø If we construct
Severalproblemsmake the numericalsimulationof this modelsof speechproductionprimarily in order to
principledifficult' We can only guessthat the "physio- solvespecific technological problems that arisein the
logicalcost" risesas a function of c and Win, but we applicationof this theory,e.g., to producepeaking
know nothingabout the form of theserelationshipsor machines for practicaland commercial purposes, the
how to combine and weight the labial and lingual answermay not be as self-evident.However, these
gestures.However, if as a first approximationwe re- aspectsmust not be confused.
quire that c never exceed 1.0 and Wm never exceed
+10 mm, we obtain the j data of Table III. This
VI. SUMMARY OF RESULTS AND CONCLUSIONS
table representsthe best approximations
of the vowel
qualities indicated. (1) X-ray data wasinterpretedto suggestthat the
Within the constraintsof the model and the given mandiblecan be usedto controlthe acousticproperties
conditions,
the relationshipbetweenthe vowelqualities of vowels(Sec.I-A-l).
1178 Volume 50 Number 4 (Part 2) 1971
ACOUSTIC CONSEQUENCES OF ARTICULATOR MOVEMENT
(2) Model simulationsindicate that this inferenceis Stevenswith whom we have had many profitable dis-
likely to be correct. With all other parameterskept cussionsduring the courseof this work. Our thanks are
constant,jaw movementalone was found to cause due to Mrs. S. Felicetti for her editorial help in pre-
shiftsin F1 of severalhundredcyclesper second(Sec. paring the manuscript and to Dr. C. Lagergren at
Karolinska Sjukhuset, Stockholm, for providing us
(3) The tongueparametersof the model have the with the necessaryx-ray material. The research was
followingacousticcorrelates(Sec.III-B). Changingthe supportedby National Institutes of Health under a
locationof the tonguebody downwardalong the tract researchgrant and by the Bank of SwedenTercentenary
can be an extremely effective means of lowering F2 Fund under contract.
and contributes towards raising F• somewhat. An
increasein the degreeof constriction(definedwith
respectto the jaw) modifiesprimarilyF2.
(4) The lowering associatedwith decreasesin lip * Also, Institute of Linguistics,Dept. of Phonetics,Stockholm
openingarea and concomitantincreasesof vocal-tract University.
• P. F. MacNeilage and G. N. Sholes,"An Electromyographic
length(•rounding") is particularlypronouncedfor the Study of the Tongue during Vowel Production," J. Speech
Fa of palatal configurationsand for F• values associ- Hearing Res. 7, 211-232 (1964).
•'T. Smith and M. Hirano, "Experimental Investigationsof the
ated with (palato-)velarand velopharyngealconstric- Muscular Control of the Tongue in Speech," Working Papers in
tions.Shiftsin F• areof smallermagnitude(Sec.III-C). Phonetics,UCLA, No. 10, 145-155 (1968).
(5) A larynx loweringof 10 mm decreases the fre- a G. Fant, Acoustic Theory of SpeechProduction (Mouton,
The Hague, 1960).
quenciesof all formants.Percentagewise, F• is affected 4 B. Lindblom, "Studies of Labial Articulation," Z. Phonetik
most' 6.3% on the average.The largest absolutefre- Sprachwiss. Kommunikationsforsch. 21, 171-172 (1968).
quencyshiftsare observedin F4;/xF4= 160 Hz. 5 B. Lindblom and H. Soron,"Analysisof Labial Movement,"
J. Acoust. Soc. Amer. $8, 935(A) (1965).
(6) The model generatesmost of the basic vowel 6V. Fromkin, "Lip Positionsin American English Vowels,"
qualitiesobservedin the languages of the world (retro- Languageand Speech7, 215-225 (1964).
7 B. Lindblom, "Vowel Duration and a Model of Lip Mandible
flex and coronalvowelsexcepted). Coordination," SpeechTransmissionLab. Quart. Progr. Status
(7) The modeloffersa possibilityof explainingwhy Rep. 4/1967 (Royal Inst. Tech., Stockholm),pp. 1-29.
human languageshave "open" and "close"vowels,or 8 D. Abercrombie,Ele•nentsof GeneralPhonetics(University
Press,Edinburgh, 1967).
moreprecisely,why I-a-]normallyappearsas an "open" 9LEA is the electrical line analog on which the measurements
vowel and [-i-! and [-u-]are "close"vowels.The need of formant frequencieswere made (seeRef. 3).
for an explanationbecomesapparentwhenwe examine •0j. S. Perkell, Physiologyof SpeechProduction:Resultsand
Dnplicationsof a QuantitativeCineradiographic Study (MIT Press,
the area functions of fundamental vowel qualities Cambridge,Mass., 1969).
which offer no basisfor an open-closecategorization, n j. M. Heinz and K. N. Stevens,"On the Relations between
or when we recall the "fixed-mandibleexperiment," Lateral Cineradiographs,Area Functions,and AcousticSpectra
of Speech,"in CongressReports(5th International Congresson
which seemsto indicate that tongueheight rather than Acoustics,Li[ge, 1965),Vol. Ia, paper A44.
jaw positionis essentialfor the productionof vowel • J. Sundberg,"Articulatory DifferencesbetweenSpokenand
quality. Accordingto the explanationproposed,the Sung Vowels in Singers," SpeechTransmissionLab. Quart.
Progr.StatusRep. 1/1969,33-46 (RoyalInst. Tech.,Stockholm).
degreeof openness of a vowelreflectsa positionof the •aG. Fant, "Formant and Cavities," in Proceedingsof the
jaw that is "optimized" in the sensethat it cooperates 5th International Congressof PhoneticSciences,M•nster 1964
with the tonguein producingthe desiredarea function. (Verlag S. Karger AG, Basel, 1965), pp. 120-141.
•4W. R. Zemlin, Speechand Hearing Science(Prentice Hall,
Such cooperationprevents excessivetongue shape EnglewoodCliffs, N.J., 1968).
deformation.As a partial justification of the present 15G. Fant and B. Sonesson,"Speechat High Ambient Air-
model, we regard the fact that this reasoningcan be Pressure,"SpeechTransmissionLab. Quart. Progr. Status Rep.
2/1964, 9-21 (Royal Inst. Tech., Stockholm).
formulated in numerical terms. The explanatorypro-
•6J. Sundberg,"Formant Structureand Articulationof Spoken
posalrestsupon the introductionof somenovelfeatures and SungVowels,"Folia Phoniatrica22, 28-48 (1970).
of articulatorymodeling'the mandibleand dimensions •7J. M. Stewart, "Tongue Root Position in Akan Vowel
of neutral-nonneutrallip and tongueshape. Harmony," Phonetica16, 185-204 (1967).
•8M. Halle and K. N. Stevens,"On the Feature 'Advanced
(8) Our resultsconsequently suggestthat "tongue TongueRoot,'" Quart. Progr. Rep. No. 94, Res. Lab. Electron.,
height," althoughprimary with respectto its acoustic MIT, 209-215 (1969).
consequences, had better be modeled as a derived •9Note that we do not claim that it explainsthe origin of the
phonological
featureof opening,that is, why languages
havecon-
articulatoryfeature characteristicof the final vocal- trasts such as /i/-/a/, /e/-/e/, etc. The presentclaim concerns
tract configuration.Unlesswe treat "tongueheight" in only their articulatory realization.For an attempt to give an
this fashion,our modelswill fail to reflect the under- accountof the originof the oppositions, seeJ. Liljencrantsand
B. Lindblom, "Numerical Simulationof Vowel Quality Systems:
lying principleof articulatorysynergism. The Role of PerceptualContrast"(to appearin Language).
•0After the submissionof the presentarticlefor publication,we
ACKNOWLEDGMENTS had the opportunityto readan accountof an interestingattempt
to developa similarjaw-based articulatorymodel;seeP. Mermel-
stein,S. Maeda,and O. Fujimura,"Descriptionof Tongueand
The authors are grateful to Gunnar Fant, Peter Lip Movementin a Jaw-BasedCoordinateSystem,"J. Acoust.
Ladefoged, Johan Liljencrants, and Kenneth N. Soc.Amer. 49, 104(A) (1971).