Noting four major types of glottalization in English (the first three of which are habitual, rhetorical, low F0-equivalent [produced at prosodic sites where a low F0 might be expected and apparently perceptually very similar to low modal F0]), Pierrehumbert focuses o the fourth type, which she identifies mostly in two locations: 1) at the beginning of a vowel-initial word, especially if the word is phrase initial or follows a vowel, and 2) in coda position instead of and prior to a voiceless stop (that is, in e.g. cat the /t/ may be replaced by /ʔ/, the preceding vowel may exhibit creak or F0/amplitude drop consistent with glottalization/laryngealization, or both). In an attempt to isolate prosodic effects on glottal allophones, she controls for other sources by selecting speakers with even modal voices and garnering tokens that fall in the middle of the pitch range according to English prosody (items in an open-ended list [Ladd 1980]).


Pierrehumbert’s project was to evaluate the extent and consistency of glottalization in the two contexts given above (V_V hiatus and coda voiceless stops). Specific contexts examined were 1) postvocalic /t/ preceding sonorant consonant (e.g. scoutmaster), 2) postnasal /t/ preceding sonorant consonant (mint leaves), 3) postvocalic /p/ preceding sonorant consonant (mapmaker), V_V hiatus across word boundary (Calgary airfield), and combination transword V-V hiatus and presonorant coda /t/ (watery oatmeal). Vocal fold adduction intervocalically was expected to give rise to irregular pitch periods (creak), which Pierrehumbert and Talkin (1992) identified as the most spectrally visible feature of creak. No attempt was made to control for other possible sources of irregularity such as low airflow.


Pierrehumbert found that phrasal stress caused exaggerated articulation at phrase onsets and pitch accents, and led more often to lengthened than exaggerated articulations in other positions. She notes that word-initial glottal stops behave more like consonants than suprasegmental overlays on the vowel, becoming more strongly articulated under stress—although as Pierrehumbert notes in a critique of De Jong’s (1991) hyperarticulation model, the realization of this articulation varies dialectally (in Southern American English, for example, often manifesting as an implosive). Evidently, though, word-initial glottal stops pattern with non-allophonic (phonemic) consonants in this regard, meaning that words and contexts likely to produce this effect can and should be excluded from my study of creak. Coda /t/ allophonic glottalization, on the other hand, does not appear to be affected by these stress patterns, and can probably be safely treated as glottalization rather than consonant insertion. Article appeared in Fujimura, O. and M. Hirano (eds.), Vocal Fold Physiology 8: Voice Quality Control. ( )