Realization of the English postvocalic [voice] contrast in F1 and F2

Before a [-voice] coda, F-1 is higher for monophthongs but lower for /aI/ than before [+voice]. We test the hypothesis that this is due to hyperarticulation before voiceless consonants (Thomas, J. Phonetics 88 (2000) 1). Experiment 1, with 16 American English speakers, found the /aI/ pattern of more peripheral F-1 and F-2 in the offglides /I eI au/ as well, showing that it is part of the realization of [-voice]. The diphthong nuclei were less affected than the offglides. Experiment 2, with different 16 American English speakers, collected "tight"-"tide" judgments of a synthetic stimulus in which offglide F-1, offglide F-2, and nuclear duration were varied independently. "Tight" responses were facilitated by lower F-1, by higher F-2, and by shorter nuclei. Log-linear analysis showed that the three factors contributed independently, and that F-2 was a stronger cue than F-1 in terms of logits per Bark. Experiment 3, with a different 16 speakers, replicated Experiment 2 for "ate"-"aid" judgments (/eI/ having shown the smallest effects in Experiment (1)). Thus, [-voice] is correlated with, and cued by, peripheralization of diphthong offglides. (C) 2003 Elsevier Science Ltd. All rights reserved.