Plan de seguimiento para verificación de las medidas propuestas

Fundamental frequency is slightly lower in vowels following voiced stops. For stressed vowels surrounded by the same stop on either side, House and Fairbanks (1953) found just a 4 Hz difference (a ratio of 1.04), tentatively attributing it to a lower intrinsic or natural fundamental frequency for glottal pulsing during voiced stops. Jessen (1998) is a relatively recent and through examination of F0perturbation following word-initial obstruents in German. In the first five pitch periods following

stop release, or for fricatives essentially starting with the onset of formant structure, F₀after voiced obstruents was reported consistently lower on the order of 10–20 Hz for at least 75–150 ms into the vowel. And although there had been claims that the voicing difference was the difference between a falling and a rising contour, Ohde’s (1984) study of word-final stressed syllables in English found that F0 tended to decrease in the first five pitch periods in both voicing categories. — F0 simply

seems to start off higher for unvoiced consonants. (Also Edwards 1981, and Castleman and Diehl 1996 for references to other studies including perception studies.)

But when actual aspiration is considered separately from the phonological contrast, a different pattern emerges — indicating there are actually two separate F0 perturbation effects. Ohde (1984)

performed a three-way comparison of English speakers’ 1) [+voice] stops (e.g. in nonsense word

/h@"bib/), 2) [-voice] stops in a position where they are aspirated (e.g. in/h@"p(h)ip/), and 3) [-voice] stops in onset clusters following an ‘s’ where they are unaspirated (e.g./h@"spip/). The difference in

Voiced Voiceless Unaspirated 154 188

Aspirated 120 178

Table 2.3: F0perturbation in the four-way contrast of Hindi: Mean fundamental frequency in Hz in

a vowel following a stop. F0was measured following the release of voiced stops and following the

start of voicing for voiceless stops (Kagaya and Hirose 1975).

Stevens (1998, page 466), however, computed based on a mathematical model of the glottis that the effect on F0should be on the order of 5–7 percent, much less than the 10–15 percent difference

that is observed (according to Stevens). Vocal cord tension may be a contributing factor, but the rest of the effect remains unexplained. Ishihara (1998), for instance, reported that F0perturbation in

Japanese and English persisted whether or not the voiced half of the minimal pair had pre-voicing. (Jessen 1998 reported something similar for German.) Granted, there may be vocal cord tension differences even if they do not result in pre-voicing, but I think some difference would be expected. It seems then that F0 perturbation is linked to neither aspiration (as explained above) or glottal vi-

bration, which leaves few options besides it being a part of the phonological specification of [voice] in much the same manner than glottal vibration itself is.

Other aerodynamic explanations had been proposed for F0 perturbation (see citations in Ohde

1984 and Jessen 1998), but they do not fit the data. The aerodynamic effect, based on higher airflow causing a Bernoulli effect on the glottis, is expected only in the first 10–15 ms of the following vowel. This would be useful for explaining the effect due to aspiration, except that they predict higher F0 with aspiration, contrary to fact. Jessen (1998, page 109) noted that voiceless obstruents

tend to be followed by more breathy voice quality than voiced obstruents, in German, English, and other languages (see references within). This might affect F0. Ohde (1984) noted that the height of

the larynx also varies between the voicing categories and could be the source of another explanation. For a survey of explanations, see Kingston and Diehl (1994).

In fact, neither a tension or an aerodynamic explanation could be the sole explanation. Jessen (2001) noted that the same perturbation effect is found in languages that make a binary voicing distinction in initial position based primarily on aspiration including English, Cantonese, Mandarin, and Danish, or based primarily on pre-voicing such as French and Japanese. And in Tamil, in which voice is not contrastive but predictable from gemination, no perturbation effect is found (Kingston and Diehl 1994), and thus F0 perturbation could not be a physiological consequence of any other

aspect of [voice] used by Tamil.

There is also some literature on the effect of voice on the fundamental frequency preceding the consonant. For this context, more work has been done in perceptual studies than production studies. Codas are perceived more often as voiced when the fundamental frequency steady state or offset is lowered in the preceding vowel (Castleman and Diehl 1996). In CVC syllables, a mere 15 Hz increase in F0 from 95 to 110 increases the rate of voiceless perception by roughly 20 percentage

points. In VCV sequences, the effect is much smaller, and is smaller than the perceptual effect of varying F0 at the onset of the second vowel. Hawkins and Nguyen (2004) reported from acoustic

measurements of production data no F0difference at the onset and mid-point of the vowel (at least

not greater than 3 Hz, which is hardly perceivable), and that F0was approximately 20 Hz higher at

the vowel end before a voiced consonant. However, they attributed the difference at the vowel end to the difficulty of measuring fundamental frequency in the vicinity of glottalization (of voiceless

codas). Looking at /l/ onsets, Hawkins and Nguyen (2004) reported no effect on onset fundamental frequency of coda voicing, but note their earlier work that showed a perceptual effect on coda voice when onset fundamental frequency is varied (lower F0 is, again, correlated with voice).