A low glide in Marphali

International audienceWe describe in the Marphali dialect of Thakali (Nepal) an unsusual phoneme which we characterize as a consonantal or semi-vocalic ‟a”. This uvular or pharyngeal approximant, which is to /ɑ/ or /ʌ/ what /j/ is to /i/, /w/ to /u/ or /ɰ/ to /ɯ/ does not have a representation in the IPA. It can appear between an initial consonant and the nuclear vowels /i/ and /e/.Understanding the relationship of a low vowel to the consonantal domain is a challenge for the classical classification of vowels, where ‘close’ is equated with ‘high’. To solve this representation problem, we resort to Catford’s analysis of vowels in terms of place of articulation and stricture type. In that analysis, the place of maximal constriction rather than the point of maximal height of the tongue is taken into consideration, and a natural class [i-ɨ-u-o-ɔ-ɑ] of ‟peripheral ‘narrow approximant’ vowels” (represented by the external arc on a polar co-ordinate diagram) emerges (Catford 1977 :186). This class has the potential to shift, historically, morpho-phonologically etc, between vocalic and consonantal status. We trace a similar ‟a-glide” in a few other languages (Gurung, Gurage, Spanish, Aghem, Middle Chinese, etc) where it occurs in limited syllabic positions, or only as a phantom phonological item, or is revealed mostly by morpho-phonological alternations. Comparative evidence is adduced to show several possible origins of this fully audible, if transient, phoneme in Marphali.Theories of syllable structure based on a sonority hierarchy are discussed in the light of the fact that the non-syllabic status of this ‟a” in the presence of more sonorous vowels like [i] or [e] cannot be computed from generally admitted sonority hierarchies where ‟low vowels” are always considered as the most sonorous sounds. We conclude that we need either to recognize [ɑ] and [ɑ̯] as potentially different phonemes, parallel to [u] and [w], and/or to revise the sonority hierarchy, or we will have to indicate the place of the nucleus in the lexical form of words. We suggest that in all events a symbol should be added to the inventory of the IPA for an ‟[ɑ]-approximant” between the (vowel) [ɑ] and the (fricative) [ʁ]