Paths to tone in the Tamang branch of Tibeto-Burman (Nepal)

We examine a phonological change in progress in Tamang-Gurung-Thakali-Manangke (TGTM), a group of Tibeto-Burman dialects or languages of Nepal. Data from eight language varieties, five of them studied first-hand in the field, are presented. The phonological change studied is a modern-day instance of the tonal split which swept through the whole of Asia in the Middle Ages: Chinese, Vietnamese, Thai and many less known languages underwent a merger of two of their series of initials (most commonly voiced/voiceless), resulting in a split of their tonal systems. Hypotheses about the modalities of implementation of this change have been offered, but modern day traces of intermediate stages are very limited. The languages of the Himalayas are situated at the geographical and chonological end of this wave so that the change is still in progress. In all the TGTM dialects studied here, the tonal split is phonologically completed, but traces of previous distinctions in manner of articulation and in phonation type survive, offering possible models for previously unobserved intermediate stages in tonogenesis. From the similarities and differences observed between the dialects, some conclusions can be drawn. In diachrony, the common passage by a breathy stage between consonant-borne voice contrasts and tone, which has been proposed for the pan-Asian tonal split, is corroborated for all TGTM languages. But after the phonologization of tone, the degree, modality and factors of retention of the old features of voice and breathiness differ from dialect to dialect. Building on the repetition of distinct but similar changes, a tentative “law” is proposed for the evolution of breathiness, emphasizing the interplay of phonetic and phonological constraints in historical development: in a language where breathiness is used as a cue to a given tone and is not an independent feature orthogonal to tone, it is retained only as long as the phonetic pitch of the tone remains low. For the synchronic analysis of linguistic states (which may last indefinitely) during which different cues contribute to the identification of a given “tone”, we propose that conceptualizing a toneme as a bundle of cues, some of them non-pitch features, rather than as defined by a single distinctive feature accompanied by “redundant” features may better account for the variability observed in tonal realization.