CV Coarticulation Research Articles

Impact of prominence type on the coarticulation of vowels following palatalized consonants

The quantitative impact of prominence (neutral vs emphatic stress) on the coarticulation of the Russian low [æ], mid [e] and high [i] front vowels is evaluated in CV sequences with palatalized consonants of various places of obstruction. The research is conditioned by the aims of speech synthesis and recognition of the Russian language, the current data in the field being insufficient. First order linear equations (locus equations) are used to evaluate the degree of CV coarticulation. The experiment is performed on CjVCjV nonsense words imbedded in the middle position of a carrier phrase «Vyros CVCV sjyljnym» («CjVCjV has grown up strong»). The first syllable of the target words was uttered both with neutral and emphatic stress. The Praat software was used to measure F2 formant values at vowel onsets and central parts. It was established that linear regression slopes are similar both in the Y-intercepts and slope coefficients for the palatalized consonants of various places of obstruction uttered with the same prominence. The degree of coarticulation was different depending upon the type of prominence (neutral or emphatic): the coefficients were lower in case of the emphatic stress compared to those of the neutral one. ANOVA analysis and Student’s t-test were performed to reveal the difference in F2 values of vowel onsets and middle parts as a function of the vowel openness and type of prominence. The results obtained were interpreted based on the degree of the openness of the vowels under study: an open vowel under emphatic stress tends to expand its vowel space by a backward displacement while close vowels tend to move forward in the mouth cavity when pronounced with emphatic stress.

Зависимость степени коартикуляции согласного и гласного от типа ударения

The paper evaluates the quantitative impact of prominence (neutral vs emphatic stress) in CV sequences. The study serves the practical needs of speech synthesis and recognition and attempts to fill an important gap as there is insufficient research in the field with regard to the Russian language. We used the locus equation method to assess the degree of CV coarticulation. The experiment was performed on CVCV nonsense words imbedded in the middle position of a carrier phrase. The first syllable of the target words was made prominent both with neutral and emphatic stress. The consonants and vowels were presented by the same consonant and vowel sounds. Russian voiced occlusive bilabial, forelingual and backlingual consonants [b], [d] and [ɡ] were combined with Russian non-palatal vowels [ɛ], [ɑ] and [u] occupying cardinal positions in the Fant triangle. Six male speakers were recorded in a sound attenuated studio using high-quality recording equipment. The total number of tokens N = 486. Praat software was used to measure formant frequencies by means of fast Fourier transformation and linear predictive coding. The experiment showed that linear regression slopes differed both in the Y-intercepts and slope coefficients for the vowels uttered with emphatic and neutral stress. We compared locus equation slopes for CV transitions of the consonants of similar place of obstruction combined with the vowels pronounced with neutral and emphatic stress, which revealed a quantitative difference in the degree of coarticulation of the bilabial and forelingual consonants both in the Y-intercept values and slope coefficients. In case of the backlingual consonant, locus equations differed only in the Y-intercept value, while the slope coefficients were practically the same. The results obtained were interpreted on the basis of existing experimental data for the Russian, English, and Arabic languages.

Alçamento sem motivação aparente em uma variedade de português brasileiro: efeitos de coarticulação de sequências CV e dispersão vocálica

Este artigo trata da motivação fonética do alçamento sem motivação aparente das vogais médias pretônicas /e, o/ (desconto ~ d[i]sconto, c[o]stela ~ c[u]stela) no português brasileiro. Realiza-se análise quantitativa (LABOV, 1972) em dados de fala da variedade de Porto Alegre, com que se confirmam os efeitos favorecedores de consoantes palatais, velares, labiais precedentes ou seguintes no alçamento das médias, registrados na literatura (KLUNCK, 2007; CRUZ, 2010; MONARETTO, 2013). A inspeção acústica de dados e o exame de literatura sobre coarticulação (GOLDSTEIN; BYRD; SALTZMAN, 2006)sugerem que o alçamento seja processo variável condicionado por múltiplos fatores, entre eles a coarticulação de sequências CV e a dispersão vocálica (MADDIESON, 1984; LINDBLOM, 1986).

Neighborhood-conditioned coarticulation effects in French listener-directed speech

Words from dense phonological neighborhoods (Hi-ND words) are realized with greater vowel hyperarticulation and increased coarticulation in English, relative to words from sparser neighborhoods (Lo-ND words) [e.g., Wright 2004]. Here, the relation between coarticulation and neighborhood density is investigated for French by looking at patterns for anticipatory and carryover nasal coarticulation, coronal and uvular CV coarticulation, and V-to-C rounding coarticulation. Twelve native French speakers (northern France) produced 82 disyllabic words of French containing a context for one of these 5 types of coarticulation in phonetically similar high-low ND pairs. Test words were produced to a real listener who had to transcribe them in the instructed positions on a grid. F1, F2, and A1-P0 (a spectral measure of nasality) were measured at 5 timepoints across each test vowel. In addition to having more hyperarticulated (peripheral) vowel midpoints, Hi-ND words showed greater anticipatory nasality and stronger coronal transitions in appropriate contexts than Lo-ND words. Carryover nasality and uvular and rounding coarticulation did not differ across NDs. These results can be interpreted in light of the relation between coarticulation and contrast and the role these effects may serve in making Hi-ND words, which are subject to greater lexical competition, easier to perceive.

CV coarticulation, locus and locus equation perspective on the invariance issue involving Algerian Arabic consonants

The aim of the present investigation is to strengthen the link between coarticulation and locus theory of speech perception through the analysis and discrimination of six Algerian Arabic consonants [b], [d], [ɡ], [q], [ʕ], [h] representing respectively six places of articulation: labial, alveolar, velar, uvular, pharyngeal, laryngeal. When plotting the error area of the categories formed by (F2onset, F3onset) samples within a CV syllable across all speakers against the F2 transition extent mean(abs(F2onset–F2vowel)), a linear correlation is observed. Then a formula for the degree of coarticulation is derived from this straight line and it is shown that it is more reliable than the slope of the locus line. When combined with the locus equation's linearity, another equation is obtained and appears to be well fitted with locus theory. Another observation in our data concerns the deviation of syllables with respect to the locus line, as in the previously noted case of [ɡ] with back vowels. Moreover, it is shown that these shifts are speaker dependent and a consequence of interactions between the categories cited above. Finally, it is shown how better place separation is obtained when these categories are mapped in the auditor's perception space using the modified locus equation.

A Duration-Dependent Account of Coarticulation for Hyper- and Hypoarticulation

Previous studies investigating anticipatory coarticulation in emphatically stressed CV sequences and during fast speaking rates reported that three factors contributed to the overall extent of the documented coarticulation. These factors were: (1) vowel identity, (2) vowel space expansion (emphasis) or reduction (fast rate), and a hypothesized (3) ‘deeper’ and ‘shallower’ stop closure contact in emphatic and faster speech, respectively. The objective of the current research was to conceptually and quantitatively unify these two studies. This was accomplished by showing that the opposite changes to frequency onsets of F2 transitions due to emphatic and rapid speech systematically vary as a function of the durational changes in the stop closure interval. Specifically, the decrease in coarticulation in emphatic speech is characterized by increases in F2 onsets and longer stop closures (relative to a normal baseline); the increase in coarticulation due to rapid speech shows concomitant decreases in F2 onsets coinciding with shorter stop closure intervals. Vocal tract area function simulations corresponding to emphatic and reduced speech implicitly support ‘deeper’ and ‘shallower’ closure contacts as a third factor contributing to the overall extent of anticipatory CV coarticulation.

Effect of tone on consonant + vowel interaction.

Examining consonant + vowel coarticulation for American English [Agwuele et al., 16th ICPhS, 1, 617–620 (2007) and Lindblom et al., J. Acoust. Soc. Am. 121, 3802–3813 (2007)], the effect of speech tempo and emphatic stress on F2onset and F2mid of the following vowel were quantified and dissociated from the underlying traditional coupling of consonant and vowel. Thus, these two prosodic factors were shown to pose additional confound to CV coarticulation. Adopting the approach of these works, this present study examines VCV sequences in Yoruba, a tonal language, in order to determine whether lexical tone produces as additional confounding effect to the interaction between consonant and tone bearing vowel that could be shown to be separate from the traditionally expected CV bonding. The study shows that the changing pattern of tone configuration in VCV sequences, e.g., from VMCVM [M=midtone] to VLCVL [L=low tone] produces differing consonant + vowel interaction. The study provides acoustic evidence for a possible independence of the “source” from the “filter,” i.e., the resonances of the vocal tract [Fant, G. Acoustic Theory of Speech Production (1960)].

Lexical tone and segmental coarticulation.

In studying anticipatory coarticulation in emphatic CV sequences [Lindblom et al., J. Acoust. Soc. Am. 121, 3802–3813 (2007)] and in fast speech [Agwuele et al., Phonetica 68, 194–209 (2008)], it was shown that prosodically induced modulations of speech introduce additional sources of phonetic variation separate from those generated by vowel context. Both studies were successful in dissociating these additional prosodic effects from the underlying contextual vowel variation and quantifying them. The focus of this current investigation was to study CV coarticulation, using a similar set of regression metrics, but now applied to a tonal language (Yoruba). It is to be determined whether lexical tone impacts the coarticulation of consonant and vowel. Specifically, words, do tones, produce additional effects on anticipatory CV coarticulation separate from those generated by vowel‐context similar to emphatic stress and increased tempo as documented by the aforementioned studies? Finally, of interest also is to examine whether different lexical tones impact C+V coarticulation differently.

Phonetic status of Cj combinations in Korean and Spanish.

Korean allows combinations of consonant and glide (CG) at onset position but forbids obstruent+liquid (OL) clusters. This is a potential problem for the universal sonority dispersion principle, which can be avoided if the Korean CG combinations are secondary‐articulated consonants rather than clusters. To see if this hypothesis is phonetically supported, we compared Korean with Spanish, which allows both CG and OL onsets. F2 at the vocoid onset of (C)jV syllables varied as a function of the backness of the following vowel in Korean, whereas it stayed constant in Spanish. The vocoid duration increase from CV to (C)jV was also smaller in Korean. This shows that the Korean /j/ is weak in that it lacks a target frequency and its tongue position is decided by the following vowel. However, this property is not confined to the Cj combination, as CjV and jV syllables behaved in the same way. Our results thus support the idea that the Korean CG is close to secondary‐articulated consonants, but the weak realization of /j/ is not a contextual variation after a consonant. One possible explanation for this pattern may come from the inherent property of CV coarticulation in Korean.

Tonal coarticulation in Yoruba: Locus equation analysis

This study examines the acoustic coarticulatory effects of tone on VCV sequences of Yoruba, when each vowel of the VCV sequences bears a different tone. The first aim is to understand how the presence of tone affects CV coarticulation in the traditional locus equations parlance as used by Krull (1987). This marks the first application of the locus equation metric to a tonal language in the study of CV coarticulation. The second aim is to determine how the stop consonant in VCV sequences is affected by prosodic overlay, independently of the vowels altered location in articulatory/acoustic space. In order to achieve the second objective, the modified locus equation regression metric that was used by [Lindblom et al. J. Acoust. Soc. Am. (2007)] to dissociate vowel context effects from rate-induced effects on consonantal F2 onsets are applied. Similar to the findings of Lindblom et al. (2007), the analyses document separate effect for F2 locus relative to F2 nucleus.

The effect of emphatic stress on CV coarticulation

The effect of emphatic stress on CV coarticulation was investigated in the speech of one male and one female native speaker of American English using locus equation slope as a measure of CV coarticulation. Stressed real word C1V2C2 tokens where C1=/b,d,g/ and V2=/i, I, e, ε, æ, u, o, ɔ, a/ were put in carrier sentences with <ub;-.2>the, <ub;-.2>thirty<ubx>, or <ub;-.2>two<ubx> preceding the test word. Each sentence was read three times in a normal manner and three times with emphasis on the test token. This resulted in a total of 486 tokens per speaker (3 stop consonants * 3 V1 contexts * 9 V2 contexts *2 emphasis patterns *3 repetitions). Locus equation slopes were derived by plotting F2 onset of C1 against V2 F2 mid-vowel frequency and fitting a regression line to data points. Consonant closure duration, V2 duration, F0, and amplitude were also measured. Despite a significant increase in the acoustic correlates of emphasis, locus equation slopes remained constant as a function of emphasis and varied as a function of place of articulation. This study provides further evidence of the stability of locus equation slopes as phonetic descriptors of stop place of articulation. [Work supported by NIH.]

The effect of syllabification and gemination on F2 onsets in Swedish

The quest to explain the continuousness of speech on the physical level has been dominated by the co‐production theory [Hman, J. Acoust. Soc. Am. 41 (1966)]. According to this view, the production of a VCV sequence involves a diphthongal movement from V1‐to‐V2, with a superimposed consonantal gesture. However, data from recent studies are at variance with this position—Modaressi [Ph.D. dissertation, UT Austin, 2002], Perkell [Coarticulation Strategies Speech Commun. 5, 47–68 (1986)]. These studies document a trough‐phenomena, which suggests a discontinuity in muscular activity during the production of the consonant. This paper provides additional acoustic evidence in support of sequential programming of consonant‐vowel events as advocated by Joos [Acoustic Phonetics Lang. 24 (1948)]. It examines symmetrical VCV sequences in Swedish natural speech with syllable boundaries and duration of consonant gemination altered to produce different types of temporal interval between V1 and V2; i.e., V♯CV, VCC♯V, VC♯CV, VCC♯CV. F2 of V1mid, V1offset, V2onset, and V2mid were measured. Locus equations were plotted for all VC contacts. Statistical analysis of these data show: (1) de‐activation of tongue movement at the CV boundary, (2) a reduction of the influence of V1 on V2 as a function of increasing consonant duration, and (3) a weak degree of CV coarticulation.

The effects of changes in hearing status in cochlear implant users on the acoustic vowel space and CV coarticulation.

In order to examine the role of hearing status in controlling coarticulation, eight English vowels in /bVt/ and /dVt/ syllables, embedded in a carrier phrase, were elicited from 7 postlingually deafened adults and 2 speakers with normal hearing. The deaf adults served in repeated recording sessions both before and up to a year after they received cochlear implants and their speech processors were turned on. Each of the two hearing control speakers served in two recording sessions, separated by about 3 months. Measures were made of second formant frequency at obstruent release and at 25 ms intervals until the final obstruent. An index of coarticulation, based on the ratio of F2 at vowel onset to F2 at midvowel target, was computed. Changes in the amount of coarticulation after the change in hearing status were small and nonsystematic for the /bVt/ syllables; those for the /dVt/ syllables averaged a 3% increase--within the range of reliability measures for the 2 hearing control speakers. Locus equations (F2 at vowel onset vs. F2 at vowel midpoint) and ratios of F2 onsets in point vowels were also calculated. Like the index of coarticulation, these measures tended to confirm that hearing status had little if any effect on coarticulation in the deaf speakers, consistent with the hypothesis that hearing does not play a direct role in regulating anticipatory coarticulation in adulthood. With the restoration of some hearing, 2 implant users significantly increased the average spacing between vowels in the formant plane, whereas the remaining 5 decreased that measure. All speakers but one also reduced vowel duration significantly. Four of the speakers reduced dispersion of vowel formant values around vowel midpoint means, but the other 3 did not show this effect.

Locus equation parameters in non-native speech

The principled relation between the magnitude of CV coarticulation and the language-specific system of phonological contrasts was investigated by comparing slope values of the locus equation [Lindblom, J. Acoust. Soc. Am. 35, 1773–1781 (1963)] for three stop places of articulation in English and Russian. The slopes of the locus equation were hypothesized to be shallower in Russian, indicating less vocalic influences on adjacent stop consonants, than in English, due to the requirement maintaining the phonological distinctions between plain and palatalized stops imposed only on Russian [Cohn, UCLA WPP 69, 51–59 (1987)]. More importantly, this study compares the slope, y-intercept, and R-squared values of native speakers with those of English learners of Russian. Preliminary analyses on bilabial stops indicated the following. (i) The locus lines of Russian were characterized by substantially shallower slopes and larger y-intercepts than those of English. (ii) The locus lines of relatively experienced English speakers of Russian showed shallower slopes and larger y-intercepts than those of native English speakers. An advantage of investigating the locus equation parameters as indices of acquisition effects of CV coarticulation is that the locus equation parameters characterize the acquisition of coarticulatory regularities of a whole distribution of CV tokens, not just the acquisition of a single CV token.

Stress Effects on CV Coarticulation and their Implications for Phonology

Proceedings of the Twenty-Fourth Annual Meeting of the Berkeley Linguistics Society: General Session and Parasession on Phonetics and Phonological Universals (1998)

A cross-linguistic validation of locus equations: Are there ‘‘phonetic hot spots’’ in CV acoustic space?

A previous study [Sussman etal., J. Acoust. Soc. Am. 90, 1309–1325 (1991)] of American English CV coarticulation showed a remarkably linear relationship between onset frequencies of F2 transitions (plotted along the ordinate) in relation to midvowel target frequencies (plotted along the abscissa) in CVC tokens with initial /bdg/ followed by ten medial vowels. Slope and y intercept of regression functions (‘‘locus equations’’) fitted to these coordinates systematically varied as a function of stop place. Discriminant analyses, using as predictor variables, slope and y intercept, yielded 100% correct classification of stop place categories. Locus equations provide a systematic lawfulness to coarticulatory variation and a relationally invariant phonetic index for stop place classification. The present study extends the metric to three additional languages−Thai, Cairean Arabic, and Urdu. Resulting scatterplots were extremely linear and varied as a function of stop place. Plotting slope Xy intercept yields a derived map of CV phonetic space with which to relationally compare 2, 3, and 4-stop place languages. Within-language stop place contrasts were consistly divergent in CV space. Variability of labial, alveolar/dental, and velar coordinates across five languages showed fairly broad clustering of stop place categories, rather than narrowly focused ‘‘phonetic hot spots.’’ The data are discussed in relation to quantal and adaptive dispersion theories. [Work supported by NSF.]

Spectral and perceptual evidence for CV coarticulation in devoiced /si/ and /syu/ in Japanese.

We examined the spectral and perceptual characteristics of devoiced /si/ and /syu/ syllables in Japanese. Measurements showed that the deleted vowel can color the spectrum of the [integral of] in these syllables, although the mean influence is small. Perception tests showed that Japanese listeners can use this spectral coloring as a cue to the identity of the underlying vowel, although identification was substantially worse than in syllables where the /i/ or /u/ is maintained. These results suggest that a supposedly lower-level coarticulation between the fricative and the vowel can occur before a higher-level process deletes the vowel, contradicting the order implied by traditional accounts of speech as a translation of discrete phonological units.