- Email: [email protected]
Vowel length and ‘voicing’ in a following consonant
Journal of Phonetics (1981) 9, 305-308
Vowel length and 'voicing# in a following consonant Thomas Walsh and Frank Parker Linguistics Program, Louisiana State University, Baton Rouge, Louisiana 70803, U.S.A. Received 27th June 1980
Abstract:
Most attempts to explain vowel lengthening before voiced consonants have failed because they have assumed that lengthening is either 'learned' (without explicit motivation) or is caused by physical factors alone (e.g. physical limitations of the vocal tract). These assumptions are the result of the failure to distinguish between physiological (physical) and phonological (abstract) definitions of 'voicing'. (Physiologically, 'voicing' is the presence of vocal cord vibration during closure; phonologically, it is the abstract property that, for example, divides English segments into those that take an fsf plural and those that take a fzf plural.) Spectrographic analysis, however, indicates that vowel lengthening in English is triggered by phonological rather than physiological 'voicing'. The acquisition of the lengthening rule is in turn motivated by perceptual factors . Speakers perceive vowels before phonologically 'voiced' consonants as longer than those before phonologically 'voiceless' consonants (Javkin , 1976).
It is well known that vowels are typically lengthened before voiced consonants among the languages of the world and that this effect is particularly apparent in English (see Javkin, 1976, for references). However, the explanation for this phenomenon has been a particularly elusive goal. Raphael (1975), for example, conducted two electromyographic experiments to determine what sort of muscular activity underlies the durational differences in vowels before voiced and voiceless consonants. From the results of his study, Raphael concludes that muscular activity is maintained longer during vowels before voiced consonants than in those before the corresponding voiceless consonants. Although Raphael's account may be accurate physiologically, it is tautological linguistically: it essentially states that vowels before voiced consonants are longer because speakers sustain the articulatory gesture for them longer. In short, Raphael's study answers the question of what happens but not why it happens. 1 Raphael's analysis is noteworthy, however, not because it is incorrect (for it apparently is not), but because it is particularly reductive. It appears to be a logical extension of many 'It is only fair to emphasize that Raphael makes it clear that he is interested only in the physiological activity which underlines durational differences. However, his conclusions may be misleading, since he attributes explanatory power to his findings. For example, he states, "Results indicate that the most reasonable hypothesis explaining the durational differences is the one which posits a sustention of muscular activity in the articulatory gesture of the vowel preceding voiced consonants" (1975, p. 25; emphasis added). Raphael may have determined the physiological mechanism used to implement vowel lengthening, but in no sense has he explained the lengthening.
0095-4470/81/030305+04 $02.00/0
© 1981 Academic Press Inc. (London) Ltd.
306
T. Walsh and F. Parker
earlier attempts to account for this same phenomenon, in which a direct , physical cause and effect relationship is proposed between vowel lengthening and vocal cord vibration (or , more generally, laryngeal activity) in a following consonant (see Javkin, 1976, for references) . We believe , however, that a major stumbling block to accounting for the effect of a 'voiced' consonant on a preceding vowel has been the often overlooked fact that the term 'voicing' can be understood in two totally different ways. It can be defined physiologically as the presence or absence of vocal cord vibration during the articulation of the consonant, or it can be defined phonologically as the abstract distinctive feature [±voice], which has a number of acoustic and articulatory correlates, only one of which is the presence or absence of vocal cord vibration (see Parker, 1977, for a discussion of this distinction). In order to investigate the relation between vowel length and 'voicing', it is absolutely essential to determine first if lengthening is correlated with physiological or phonological 'voicing'. Consider the hypothesis that vowel lengthening is caused by vocal cord vibration in the following consonant. Such a relationship would require that vowel lengthening occur only before a consonant exhibiting vocal cord vibration. Counter examples would be provided by cases in which [+voice] (i.e . phonologically 'voiced') consonants (/b, d, gf etc. in English) exhibit no vocal cord vibration following a relatively long vowel, and , conversely, by cases in which [-voice] consonants (/p , t, k/ etc.) exhibit vocal cord vibration following a relatively short vowel. Counter examples such as these are , in fact, not hard to find. Figures 1 and 2 display spectrograms made from recordings of the speech of two native speakers of English , one of whom consistently produces post-vocalic [+ voice] consonants without vocal cord vibration, and both of whom sporadically produce post-vocalic [-voice J consonants with vocal cord vibration. What is significant in each case, however, is that vowel length is predicted by the phonological feature [± voice J and not by the presence or absence of vocal cord vibration? First, consider bopped /bapt/ and bobbed /babd/ produced by speaker A (Fig. 1). Note that both the post-vocalic /b/ in /babd/ and the /p/ in /bapt/ exhibit little vocal cord vibration at most, yet the vowel preceding the /b/ is almost twice as long as the vowel preceding /p/. The same phenomenon is apparent when the post-vocalic stops are replaced with fricatives, as in pushed / p u st/ and buzzed [bAzd/ (Fig. 1). Whlle vocal cord vibration continues well into the /s/ in /pust/, it stops shortly after the beginning of the /z/ in /bAzd/, even though the vowel in /bAzd/ is more than twice as long as the vowel in /p ust/ . In both pairs the length of the vowel is predicted by the distinctive feature specification (i.e. [± voice]) of the post-vocalic consonant and not by the presence or absence of vocal cord vibration during closure. Second, compare speaker B's productions (Fig. 2) with those of speaker A (Fig. 1). Speaker B exhibits more vocal fold activity in the production of post-vocalic /t/ in midst /mrtst/ than speaker A does in the post-vocalic /b/ in bobbed /babd/; yet comparison of speaker B's vowels in midst /mrtst/ (1 02 ms) and bids /brdz/ (150 ms) shows clearly that 2 The discussion of these spectrograms is obviously meant to be suggestive rather than exhaustive. Part of what they are meant to illustrate is, in fact, well documented . Raphael, for example, states, "Examples of voiced consonants with little or no voicing during consonant closure are not uncommon .. . . Similarly, voiceless consonants which evidence vocal pulsing throughout most or all of their closure period are not uncommon" (1972, p. 1301). ('Voiced' and 'voiceless' as used in this citation obviously refer to phonological voicing; whereas the term 'voicing' must refer to physiological voicing.)
Voicing in a following consonant
/bapt ' -t
-.--
/babd/
/pust/
(b)
(c)
307
• /bl\zd/
~
--~
··-\ '''"''''~~' ~ (a)
Figure 1
Spectrograms of bopped-bobbed and pushed-buzzed for speaker A.
/mrtstl
(a)
Figure 2
(d)
/brdz/ ·
(b)
Spectrograms of midst and bids for speaker B.
/mrtst/ contains a short vowel. In short, vowel lengthening (in English at least) must be seen as a function of the abstract feature specification of the following consonant. There is apparently no cause and effect relationship between vocal cord vibration (or laryngeal activity) in a post-vocalic consonant and lengthening of the preceding vowel. Still, if there is no direct physiological correlation between vocal cord vibration and vowel lengthening, the question remains as to how and why speakers of English (and perhaps other languages) should have internalized a phonological rule which is triggered by an abstract distinctive feature. The answer may well be that proposed by J avkin (1976). In his study, listeners were asked to adjust a tone to agree in length with synthetic and selfgenerated tokens of hiss, his, and has. His findings indicate that speakers tend to perceive
308
T Walsh and F. Parker
vowels preceding [+ voice J consonants as longer than those before [-voice] consonants, even when the lengths of the vowels are the same. He concludes "that the continuation of voicing into the con sonant causes the perception of greater vowel length (eventually leading to the production of greater vowel length)" (1976, p. 81). Thus, if we accept Javkin's theory, the phonological rule of vowel lengthening before [+ voice] consonants is motivated by the perceptual mechanism rather than by physiological constraints on production. That is, the rule is the result of a sort of self-fulfilling prophecy : speakers make vowels before [+ voice] consonants longer because they hear them as longer. Note that this account does fulfill the requirement of a (partial) explanation. Vowel lengthening (a physical phenomenon) is triggered by the feature specifications of the following consonant (an abstraction). This process can be characterized by an (abstract) phonological rule in the grammars of speakers of English:
c
.
V -+ [ + long] / - [+ voice] The acquisition of this rule , in turn, is motivated by speakers' perceptual apparatus (a (partially) physical phenomenon). In conclusion, we have tried to make four points. First, any explanation of a phonological phenomenon (or of any other kind for that matter) must account not only for what happens but why it happens. Second, a central problem in explaining the relationship between vowel lengthening and 'voicing' in a following consonant stems from the failure to distinguish between physiological and phonological 'voicing'. Third, spectrographic analysis indicates that vowel lengthening in English is a function of phonological 'voicing'. Finally, the inclusion of the vowel lengthening rule in English is motivated by perceptual factors. Speakers make vowels before [+voice J consonants longer because they perceive them as longer; and they perceive them as longer because of the continuation, in some cases, of physiological voicing into the consonant. References Javkin, H. R. (1976). The perceptual basis of vowel duration differences associated with the voiced/ voiceless distinction. Report of the Berkeley Phonology Laboratory, 78-92. Parker; F. (1977). Distinctive features and acoustic cues. Journal of the Acoustical Society of America, 62, 1051-1054. Raphael, L. J. (1972). Preceding vowel duration as a cue to the perception of the voicing characteristic of word-final consonants in American English. Journal of the Acoustical Society of America, 51, 1296-1303. Raphael, L. J. (1975). The physiological control of durational differences between vowels preceding voiced and voiceless consonants in English. Journal of Phonetics, 3, 25-33.
Vowel length and 'voicing# in a following consonant Thomas Walsh and Frank Parker Linguistics Program, Louisiana State University, Baton Rouge, Louisiana 70803, U.S.A. Received 27th June 1980
Abstract:
Most attempts to explain vowel lengthening before voiced consonants have failed because they have assumed that lengthening is either 'learned' (without explicit motivation) or is caused by physical factors alone (e.g. physical limitations of the vocal tract). These assumptions are the result of the failure to distinguish between physiological (physical) and phonological (abstract) definitions of 'voicing'. (Physiologically, 'voicing' is the presence of vocal cord vibration during closure; phonologically, it is the abstract property that, for example, divides English segments into those that take an fsf plural and those that take a fzf plural.) Spectrographic analysis, however, indicates that vowel lengthening in English is triggered by phonological rather than physiological 'voicing'. The acquisition of the lengthening rule is in turn motivated by perceptual factors . Speakers perceive vowels before phonologically 'voiced' consonants as longer than those before phonologically 'voiceless' consonants (Javkin , 1976).
It is well known that vowels are typically lengthened before voiced consonants among the languages of the world and that this effect is particularly apparent in English (see Javkin, 1976, for references). However, the explanation for this phenomenon has been a particularly elusive goal. Raphael (1975), for example, conducted two electromyographic experiments to determine what sort of muscular activity underlies the durational differences in vowels before voiced and voiceless consonants. From the results of his study, Raphael concludes that muscular activity is maintained longer during vowels before voiced consonants than in those before the corresponding voiceless consonants. Although Raphael's account may be accurate physiologically, it is tautological linguistically: it essentially states that vowels before voiced consonants are longer because speakers sustain the articulatory gesture for them longer. In short, Raphael's study answers the question of what happens but not why it happens. 1 Raphael's analysis is noteworthy, however, not because it is incorrect (for it apparently is not), but because it is particularly reductive. It appears to be a logical extension of many 'It is only fair to emphasize that Raphael makes it clear that he is interested only in the physiological activity which underlines durational differences. However, his conclusions may be misleading, since he attributes explanatory power to his findings. For example, he states, "Results indicate that the most reasonable hypothesis explaining the durational differences is the one which posits a sustention of muscular activity in the articulatory gesture of the vowel preceding voiced consonants" (1975, p. 25; emphasis added). Raphael may have determined the physiological mechanism used to implement vowel lengthening, but in no sense has he explained the lengthening.
0095-4470/81/030305+04 $02.00/0
© 1981 Academic Press Inc. (London) Ltd.
306
T. Walsh and F. Parker
earlier attempts to account for this same phenomenon, in which a direct , physical cause and effect relationship is proposed between vowel lengthening and vocal cord vibration (or , more generally, laryngeal activity) in a following consonant (see Javkin, 1976, for references) . We believe , however, that a major stumbling block to accounting for the effect of a 'voiced' consonant on a preceding vowel has been the often overlooked fact that the term 'voicing' can be understood in two totally different ways. It can be defined physiologically as the presence or absence of vocal cord vibration during the articulation of the consonant, or it can be defined phonologically as the abstract distinctive feature [±voice], which has a number of acoustic and articulatory correlates, only one of which is the presence or absence of vocal cord vibration (see Parker, 1977, for a discussion of this distinction). In order to investigate the relation between vowel length and 'voicing', it is absolutely essential to determine first if lengthening is correlated with physiological or phonological 'voicing'. Consider the hypothesis that vowel lengthening is caused by vocal cord vibration in the following consonant. Such a relationship would require that vowel lengthening occur only before a consonant exhibiting vocal cord vibration. Counter examples would be provided by cases in which [+voice] (i.e . phonologically 'voiced') consonants (/b, d, gf etc. in English) exhibit no vocal cord vibration following a relatively long vowel, and , conversely, by cases in which [-voice] consonants (/p , t, k/ etc.) exhibit vocal cord vibration following a relatively short vowel. Counter examples such as these are , in fact, not hard to find. Figures 1 and 2 display spectrograms made from recordings of the speech of two native speakers of English , one of whom consistently produces post-vocalic [+ voice] consonants without vocal cord vibration, and both of whom sporadically produce post-vocalic [-voice J consonants with vocal cord vibration. What is significant in each case, however, is that vowel length is predicted by the phonological feature [± voice J and not by the presence or absence of vocal cord vibration? First, consider bopped /bapt/ and bobbed /babd/ produced by speaker A (Fig. 1). Note that both the post-vocalic /b/ in /babd/ and the /p/ in /bapt/ exhibit little vocal cord vibration at most, yet the vowel preceding the /b/ is almost twice as long as the vowel preceding /p/. The same phenomenon is apparent when the post-vocalic stops are replaced with fricatives, as in pushed / p u st/ and buzzed [bAzd/ (Fig. 1). Whlle vocal cord vibration continues well into the /s/ in /pust/, it stops shortly after the beginning of the /z/ in /bAzd/, even though the vowel in /bAzd/ is more than twice as long as the vowel in /p ust/ . In both pairs the length of the vowel is predicted by the distinctive feature specification (i.e. [± voice]) of the post-vocalic consonant and not by the presence or absence of vocal cord vibration during closure. Second, compare speaker B's productions (Fig. 2) with those of speaker A (Fig. 1). Speaker B exhibits more vocal fold activity in the production of post-vocalic /t/ in midst /mrtst/ than speaker A does in the post-vocalic /b/ in bobbed /babd/; yet comparison of speaker B's vowels in midst /mrtst/ (1 02 ms) and bids /brdz/ (150 ms) shows clearly that 2 The discussion of these spectrograms is obviously meant to be suggestive rather than exhaustive. Part of what they are meant to illustrate is, in fact, well documented . Raphael, for example, states, "Examples of voiced consonants with little or no voicing during consonant closure are not uncommon .. . . Similarly, voiceless consonants which evidence vocal pulsing throughout most or all of their closure period are not uncommon" (1972, p. 1301). ('Voiced' and 'voiceless' as used in this citation obviously refer to phonological voicing; whereas the term 'voicing' must refer to physiological voicing.)
Voicing in a following consonant
/bapt ' -t
-.--
/babd/
/pust/
(b)
(c)
307
• /bl\zd/
~
--~
··-\ '''"''''~~' ~ (a)
Figure 1
Spectrograms of bopped-bobbed and pushed-buzzed for speaker A.
/mrtstl
(a)
Figure 2
(d)
/brdz/ ·
(b)
Spectrograms of midst and bids for speaker B.
/mrtst/ contains a short vowel. In short, vowel lengthening (in English at least) must be seen as a function of the abstract feature specification of the following consonant. There is apparently no cause and effect relationship between vocal cord vibration (or laryngeal activity) in a post-vocalic consonant and lengthening of the preceding vowel. Still, if there is no direct physiological correlation between vocal cord vibration and vowel lengthening, the question remains as to how and why speakers of English (and perhaps other languages) should have internalized a phonological rule which is triggered by an abstract distinctive feature. The answer may well be that proposed by J avkin (1976). In his study, listeners were asked to adjust a tone to agree in length with synthetic and selfgenerated tokens of hiss, his, and has. His findings indicate that speakers tend to perceive
308
T Walsh and F. Parker
vowels preceding [+ voice J consonants as longer than those before [-voice] consonants, even when the lengths of the vowels are the same. He concludes "that the continuation of voicing into the con sonant causes the perception of greater vowel length (eventually leading to the production of greater vowel length)" (1976, p. 81). Thus, if we accept Javkin's theory, the phonological rule of vowel lengthening before [+ voice] consonants is motivated by the perceptual mechanism rather than by physiological constraints on production. That is, the rule is the result of a sort of self-fulfilling prophecy : speakers make vowels before [+ voice] consonants longer because they hear them as longer. Note that this account does fulfill the requirement of a (partial) explanation. Vowel lengthening (a physical phenomenon) is triggered by the feature specifications of the following consonant (an abstraction). This process can be characterized by an (abstract) phonological rule in the grammars of speakers of English:
c
.
V -+ [ + long] / - [+ voice] The acquisition of this rule , in turn, is motivated by speakers' perceptual apparatus (a (partially) physical phenomenon). In conclusion, we have tried to make four points. First, any explanation of a phonological phenomenon (or of any other kind for that matter) must account not only for what happens but why it happens. Second, a central problem in explaining the relationship between vowel lengthening and 'voicing' in a following consonant stems from the failure to distinguish between physiological and phonological 'voicing'. Third, spectrographic analysis indicates that vowel lengthening in English is a function of phonological 'voicing'. Finally, the inclusion of the vowel lengthening rule in English is motivated by perceptual factors. Speakers make vowels before [+voice J consonants longer because they perceive them as longer; and they perceive them as longer because of the continuation, in some cases, of physiological voicing into the consonant. References Javkin, H. R. (1976). The perceptual basis of vowel duration differences associated with the voiced/ voiceless distinction. Report of the Berkeley Phonology Laboratory, 78-92. Parker; F. (1977). Distinctive features and acoustic cues. Journal of the Acoustical Society of America, 62, 1051-1054. Raphael, L. J. (1972). Preceding vowel duration as a cue to the perception of the voicing characteristic of word-final consonants in American English. Journal of the Acoustical Society of America, 51, 1296-1303. Raphael, L. J. (1975). The physiological control of durational differences between vowels preceding voiced and voiceless consonants in English. Journal of Phonetics, 3, 25-33.