Perception of Yoruba Word-Initial [gb]and [b]

Labial-velar obstruents /g͡b/, /k͡p/, /ŋ͡m/ occur in hundreds of African languages in sub-Saharan Africa, dozens of languages of the Pacific (mostly in Papua New Guinea and Vanuatu), and a handful of languages in South America (Cahill 1999 and references therein). To date 615 languages in my personal database have phonemic labial-velars, approximately 8.9% of the 6912 languages in the most recent Ethnologue (Gordon 2005), and I expect a more thorough count will increase this figure to approximately 10% of the world’s languages. This paper is a preliminary investigation into what cues are used by native Yoruba speakers to distinguish phonetic [g͡b] and [b]. More specifically, I focus on word-initial occurrences of these sounds, not on intervocalic position (e.g. [ag ͡ba]) or post-nasal position (e.g. [ŋg͡ba]). One might assume that word-initial occurrences (e.g. [g͡ba] vs. [ba]) would be inherently more difficult to distinguish than intervocalic occurrences (e.g. [ag ͡ba] vs. [aba]), for at least two reasons. First, both CV and VC cues are present in [ag ͡ba], while only CV transition cues are present in [g͡ba]. Second, though labial-velars are commonly described as having “simultaneous” labial and velar articulations, this is not quite accurate. They have overlapping but not coterminous articulations. The start of the velar articulation precedes the labial and ends sooner as well. Thus, the release of a labialvelar is labial (Ladefoged and Maddieson 1996 and references therein). This is shown in formant transitions in spectrograms from several languages in Connell 1994 and Ladefoged and Maddieson 1996, as well as below (Note especially the difference in F2 in transitions into and out of the [gb].)