Perception of VOT and First Formant Onset by Spanish and English Speakers

Many languages exhibit a voicing distinction with two series of stops, voiced and voiceless, which differ primarily in the relative timing between the consonantal release and the onset of laryngeal vibration, known as voice onset time (VOT) (Lisker and Abramson, 1964). While the exact implementation of VOT for the voicing contrast differs among languages, long lag VOTs, with a long delay between release and onset of laryngeal vibration, generally signal a voiceless stop. Short lag VOTs, with a short delay between release and onset of laryngeal vibration, signal a voiced stop. In Spanish, the voiceless stops /ptk/ are produced with a near-simultaneous release and onset of laryngeal vibration, resulting in VOT values that are approximately zero or a few milliseconds. Spanish voiced stops /bd/ in utterance-initial position are prevoiced such that the onset of laryngeal vibration precedes the release by 40 ms or more, resulting in a negative VOT of less than -40 ms. Thus, for Spanish, the short lag VOTs of voiced stops are negative, while the voiceless stops have with a long lag VOTs between zero and 10 ms (Lisker and Abramson, 1964; Cho and Ladefoged, 2000). In English, voiceless stops /ptk/ have a substantial delay between the release and the onset of laryngeal vibration, resulting in a VOT of 30 ms or longer, corresponding to the aspiration interval. English voiced stops /bd/ in utterance-initial position are generally not prevoiced but released simultaneously with the onset of voicing, for a VOT of approximately zero (Caisse, 1982; Docherty, 1992). In many other languages, As Lisker (1986) notes, the timing relation between oral release and onset of laryngeal vibration produces a number of acoustic cues, primarily manifested in differences in formant transitions. In English voiced stops, voicing begins simultaneously with the release. As a result, acoustic energy from vocal fold vibration excites the first formant (F1) during the entire consonant-vowel (CV) transition. Low frequency periodic acoustic energy from laryngeal vibration makes F1 audible during its rise from the consonantal release to the vowel steady state frequency. The long VOTs for English voiceless stops result in F1 transitions that differ radically from corresponding transitions of voiced stops. Since voicing onset occurs much later than the release in voiceless stops, F1 is not excited until very late in the CV transition, at which time the vocal tract is close to the vowel steady-state configuration. The frequency of F1 at voicing onset frequency is much higher for English voiceless stops than for voiced stops. This delay in the excitation of F1 is also known as F1 cutback (Liberman et al., 1958). Early work in speech perception has verified the role of VOT in the perception of voicing in utterance-initial position for speakers of English and other languages (Liberman et al., 1958; Lisker and Abramson, 1970). In these studies, listeners have been found to classify stops as voiced or voiceless depending on the VOT value, consistent with the observed VOTs of voiced and voiceless stops for the appropriate language. Stops with VOT values longer than some boundary value are classified as voiceless, while stops with VOTs shorter than the boundary value are classified as voiced. In addition to the effects of VOT, the effects of F1 transition and frequency at voicing onset have also been shown to be important in the perception of the voicing contrast in English (Stevens and Klatt, 1974; Lisker et al., 1977; Summerfield and Haggard, 1977; Kluender, 1991; Pind, 1999 for the Icelandic aspiration contrast). These studies show that stops with a longer F1 transition and/or lower F1 frequency at voicing onset, all other things being equal (such as VOT), will be classified as voiced relative to stops with shorter F1 transitions and/or higher F1 frequency at voicing onset. Thus, VOT and the F1 transition pattern have a trading relation (Repp, 1982) in the perception of the voicing