Can intrinsicvowel F0 be explained by source/tract ß ,3 coupl,ng. William G. Ewan

Department of OralBiology,University of Connecticut HealthCenter,Farmington, Connecticut 06032 (Received7 September1978;revised26 March 1979) There is extensiveevidencethat in the samephoneticenvironmentthe voicefundamentalfrequency(F0) of vowelsvariesdirectly with vowel "height." This F0 differencebetweenvowelscould be causedby acoustic interactionbetweenthe first vowel formant and the vibratingvocal folds. Sincehigher vowelshave lower first formants than low vowels the acousticinteractionshould be greatestfor high vowels whose first formant frequenciesare closerin frequencyto F0. Ten speakerswere usedto seeif acousticinteraction could causethe F0 differences.The consonant[m] was recordedin the utterances[umu] and [ama]. Althoughthe formant structureof [m] in [umu] and [ama] shouldnot differ significantly,the F0 of each [m] allophonewas significantlydifferent. However, the F0 of each [m] allophonedid not differ

significantlyfrom the F0 qf the followingvowel.Theseresultsdid not supportacousticinteraction. However, it is quite reasonableto concludethat the F0 variation of [m] was causedby coarticulatory anticipationof the tongueand jaw for the followingvowel.Another experimentis offeredin order to help explain the physical causesof intrinsic vowel F0. In this experimentF0 lowering was found at the beginningof vowelsfollowing Arabic pharyngealapproximants.This finding indicatesthat the F0 of pharyngealconstrictingvowels,e.g., [•e] and [a], might be loweredas a resultof similar articulatory movements,viz. tonguecompressionand activepharyngealconstriction. PACS numbers:43.70.Bk, 43.70.Ve

INTRODUCTION

intrinsic F0. Third, an F0 versus F1 plot using data from Peterson and Barney (1952) and Atkinson (1972)

The fundamentalfrequency of voice (F0) of vowels varies directly with vowel "height." A "high" vowel

showed a disturbing mount of scatter. The scatter observed by Ohala indicates that acoustic coupling, at least acoustic coupling alone, probably does not cause intrinsic vowel F0.

suchas [u] may have an F0 as much as 25 Hz higher than a "low" vowel such as [a] in the same phonetic environment[see reviews,of this subjectby Atkinson (1973), Ohala (1973), Hombert (1975), Petersen (1978),

and Ewan (1976)]. Slight but statistically significant differences in F0 between vowels have even been observed when vowels were sung at the same pitch (C.K. Chuang, personal communication). If the cause of this phenomenon were known, it might help us to better understand the mechanism available to speakers for

voluntary F0 regulation.

A hypothesis has been offered by Atkinson (1973) to account for the intrinsic variation in F0 between

vowels. In referring to Flanagan and Landgraf (1968), he suggestedthat the closer the first resonance (F1) of the vocal tract is to F0, the greater effect that re-

Another possibility is that intrinsic F0 is caused by a physical change in the vocal tract which alters the length or tension of the vibrating vocal folds. This

possibility was first consideredby Ladefoged(1964). Ladefoged suggested that the tongue pulls increasingly on the vocal folds for progressively higher vowels.

I infer from Lindau, Jacobsen, and Ladefoged (1972) that the pull affecting the vocal folds was expected primarily through the hyoid bone. For example bunching of the tongue in order to articulate high vowels would presumably pull the hyoid bone toward the genoid process and effectively pull up or forward on the vocal folds. However, MacNeilage (1969); Menon and Shearer (1971); Lindau, Jacobsen, and

sonancewill have in raising F0. This source/tract

Ladefoged (1972); and Atkinson (1973) found that the

coupling would be greater for higher vowels since

hyoid bone position is generally inversely corre-

the F1 is lower (and therefore closer to F0) than it is

lated with vowel height,i.e., highfor the vowel [a] and low for the vowel [i]. If the positionof the hyoidbonereflects tonguepull, the vowel [a] wouldbe

for

low

vowels.

This hypothesis has been challenged (althoughnot completely rejected) by Ohala (1973) with the following counterevidence- First, Beil (1962) found that during helium speech, when all formants are higher and acoustic coupling is weaker, a similar intrinsic F0 difference

expected to have the higher intrinsic F0; just the opposite

is true.

Delos, Guerin, Mrayati, and Carre (1976) suggest that the position of the larynx should explain intrinsic

between vowels still occurs and has the same magni-

F0 dueto the mechanicaldependence of the positionof

tude as that duringnonheliumspeech.t Second,

the supraglottal articulators on larynx position. However Ewan and Krones (1974) and Ewan (1976) provide evidence that vertical larynx movement, which is

sounds

such as laterals

and nasals

have low first

for-

mants but apparently do not have the predicted high

358

J. Acoust. Soc.Am.66(2),Aug.1979

0001-4966/79/080358-05500.80

(D1979Acoustical Societyof America

358

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 130.113.111.210 On: Sat, 20 Dec 2014 22:20:58

often correlated with F0 variation (as observed at the

expectedF0 pattern occurs. The vowels[u] and[o] are

size of the laryngeal cavity may not provide counterevidence for the acoustic coupling hypothesis. Presumably the pharyngeal and laryngeal cavities change their volumes in the same way during vowel production. If the pharyngeal and laryngeal cavity are generally the primary source for F1, an increase in cavity size

produced under experimental

would lower F1 (increase coupling) and at the same

laryngeal prominence), is not related in a simple way to intrinsic F0. For example depending on the speaker

either the vowel [i] or [a] may have the highestlarynx position in a given phonetic environment,

and yet the

conditions with a lower

larynx than the vowels [i] and [a] (Ewan 1975, 1976). Thus the only consistent larynx height pattern

has a

low larynx for back rounded vowels. Ewan (1975) speculates that because the articulation of the tongue for back rounded vowels is farther back than a high

vowel such as If], it may be more difficult to ef-

time increase vocal-fold tension. Ohala (1973) also cites studies showing that the volume of the laryngeal cavity increases with an increase in •0, which is similar to what occurs with a change in vowel quality. This evidence does not support acoustic coupling, since •0 should vary without a necessary change in Fl.

fect a lateral expansion of the pharyngeal (and la-

ryngeal) cavity in order to help lower F1, especially if passive tissue pull is the primary means for expanding the pharyngeal cavity. Lowering the larynx would help to increase the vocal tract volume by stretching tissues of the laryngeal cavity in a direction basically opposite to that of the tongue pull. This maneuver to expand the pharyngeal cavity during the production of back rounded vowels would help to cause a low F1, weak resonant peaks, and wide bandwidths. A secondary effect of this cavity expansion would

be to increase

vocal-fold

tension.

In-this

case

larynx lowering (not raising) with the raising of the tongue body would presumably result in a slight in-

crease in F0. This is pure speculation and is not clearly supported by evidence reported in Ewan

(1976), which showsthat subjects produce an F0 change in which the larynx height versus F0 function

obtainedwhile pronouncingthe vowel [a] intersects that obtainedfor the vowels [i] and [u]. If intrinsic vowel F0 were caused by larynx movement, larynx position occuring with one vowel for a given F0 should not coincide with that occurring with another at the same F0o

Ohala (1973) attempted to rescue the tongue-pull hypothesis proposed by Ladefoged by suggesting that the tongue pull may be done through the soft tissue between the tongue and the vocal folds rather than via the hyoid bone, specifically, via the aryepiglottic folds. This

action

would increase

the vertical

tension

of the vocal

folds andcause an increase inFo bypossibly elevating the

I. EXPERIMENT

I

The effects of acoustic coupling can be tested phon-

etically. The •0 of a nasal consonant,for example [m], should not be altered by the F1 of a following vowel if

acoustic coupling causes intrinsic vowel •0.

A nasal

consonant should eliminate the hypothesized acoustic interaction caused by adjacent vowels, and should not

vary in its effect on •0 with various adjacent vowels. The lowest nasal resonance, which would presumably

cause an •0 perturbation, is not like the F1 of the following vowels, and, in any case, should not change

significantly whether producedbefore [i], [a], or [u] (House, 1957; Fujimura, 1962). A.

Method

A simpletestwasconducted to seeif the•0 Ofa nasal consonant does or does not change in different vowel

environments. Tenphonetically naive adult subjects (five male, five female) pronouncedthetwoutterances[ area] and[umu]in the sentenceframe "Say again." This sentence frame helped to give each VCV utterance a

level •0. tained.

Twenty-one tokens of each type were obA randomized

used as a guide.

list

of the two utterances

was

The recordings were processed using

the PDP-11/45 computer and GT-40 graphics terminal at Haskins Laboratories.

Each token was low-pass

filtered at a cutoff frequency of 600 Hz (30 dB/octave attenuation). Then the remaining'waveform was sampled by the computer at 10 kHz and displayed on the

ventricular folds and stretching, vertically, the tissue in the laryngeal pharynx, including the vocal folds. Ohala provides evidence for this change in soft tissue tension by citing x-ray studies of the laryngeal ventricle and laryngeal cavity which seem to show a greater ventricle size for high vowels than low vowels

GT-40 scope display. A manually controlled cursor was moved across the displayed waveform to the center of the nasal consonant. This cursor position marked the beginning of one fundamental frequency period. The cursor was then moved from this beginning point to the end of the fundamental frequency period. This distance on the display between the first and last cursor posi-

(van den Berg, 1955), and which show a change in ven-

tions (i.e., one fundamental frequency period) ap-

tricle size corresponding to the order generally found

peared in the corner of the scope display in ms. The utterance-final vowels were measured using the same procedure at a position roughly 100 ms following the vowel onset (i.e., approximately one-third of the way into the vowel waveform display). Presumably this point was far enough into the vowel to have occurred after the hypothesized effects of acoustic coupling were established. All 420 measurements were recorded by hand.

for intrinsic F0 (i.e., in the order [a,o,e,i,u] from a small to a large cavity) (Shimizu 1960,1961). Recent x-ray studies by Appelman (1967) show similar patterning in the positioning of the epiglottis during

singing(i.e., with the vowel [a] or [•e] havingthe most posterior positionand[i] or [u] the most anterior). These changes in the position of the epiglottis indicate that the soft aryepiglottic tissues pull on the vocal folds as a result of tongue pull.

Changes in both the position of the epiglottis and the

359

J. Acoust.Soc.Am.,Vol. 66, NO.2, August1979

The initial vowel was audibly more nasalized, and therefore acoustically more like the nasal stop than

WilliamG. Ewan: Intrinsic vowelFo

359

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 130.113.111.210 On: Sat, 20 Dec 2014 22:20:58

the final vowel. Since the effects of acoustic coupling should take a little time to be manifested following a

(1973) has made a case for tongue pull.

However

vowel nasalizatlon probably increased the chances for

other attempts to decide between an acoustic-coupling hypothesis and tongue-pull hypothesis have not been completely successful. For example, Lindblom and

the lowest resonanceof the nasal [m] to causethe F 0 of [m] to be measurablydifferent from the F0 of the

Sundberg (1971), and later Lindblom, Lubker, and Gay (1977) found that when the jaw was artifically propped

final

open subjects were able to hyperarticulate the tongue body so that vowel formants were basically unchanged.

change in phonetic environment,

the heavier initial

vowel.

B. Results

Ohala and Eukel (1976) predicted that this hyperexten-

The F0 meansfor the two nasal allophones[m], and [m]u andthe final vowelswere convertedinto Hz and are shown in Table i. The measurements (in ms) of both the female and male data were pooled. A twoway analysis of variance was used in order to find

whetherthe F0 of the nasal allophones[m]a and differed significantlyfrom [a] and [u], respectively. The F0 of [m]a and[m]uwere significantlydifferent (p

tract coupling?

Can intrinsicvowel F0 be explained by source/tract ß ,3 coupl,ng. William G. Ewan Department of OralBiology,University of Connecticut HealthCenter,Fa...
732KB Sizes 0 Downloads 0 Views