Modern Risiangku Tamang (Tibeto-Burman, Nepal) has a system of four tones, two high tones and two low tones, which results from the two-way split of an earlier two-tone system linked to the disappearance of an older voicing contrast on initial consonants (Ci). The voicing contrast has fully disappeared in sonorant and fricative Ci, thus phonologizing the tone system, and partially disappeared in plosive Ci. In the latter case, three cues are used in production : F0, voice quality, and some residual voicing of Ci (Mazaudon & Michaud 2008). To summarize, (1) low tones are produced with breathier voice than high tones ; (2) 20% of the low tone syllables (vs. 0% of the high tone syllables) are produced with prevoiced Ci.
Variation in cue weighting has been reported for a number of languages where tone is still incipient like Korean (Silva 2006) or Afrikaans (Beddor 2015 ; Coetzee, p.c.). In Tamang, tone is phonologized ; however, old features/cues, i.e., Ci prevoicing and breathy voice, are still present in production. This study explores how old and new cues are used in the perception of tones in Risiangku Tamang.
We conducted a four-alternative forced-choice identification test on Risiangku listeners living in Nepal with a quasi-quadruplet. All stimuli were synthesized with equalized segmental duration and intensity contour. Five parameters were manipulated : degree of breathiness (modal, breathy, or super breathy) ; Ci prevoicing (present or absent) ; F0 of V1 onset (115, 130, 145, or 160 Hz) ; F0 slope of V1 (rising or falling) ; F0 slope of V2 (rising or falling). The resulting 96 stimuli were presented in a carrier sentence to 28 listeners individually.
Concerning the difference between high and low tones, results show that F0 onset, Ci prevoicing, and voice quality are all in play. The identification rate of high tones (T1-2) increases with the F0 onset. However, the slopes of all the identification curves are less steep than what we would expect of a sharp categorical perception. Listeners also rely heavily on both Ci prevoicing and voice quality : prevoiced Ci stimuli yield low tone identification in 50% of cases even with the highest F0 onset ; and the high tone identification rate is lower for breathy than for modal stimuli, and for super breathy than for breathy stimuli. Statistical (GLM) and classification (J48 tree, SMO) analyses confirmed a preponderance of the F0 onset cue in high vs. low tone identification, followed by reliance on the prevoicing of the Ci when present. The use of voice quality comes in to resolve conflicting cues : i.e., a high F0 with Ci prevoicing or a low F0 without Ci prevoicing. Moreover, when cues are conflicting, response times increase.
Our study shows a predominance of F0 as an identification cue, which is consistent with the historical evolution. However, the old features/cues of [voice] and [breathiness] continue to be used in perception as in production. The substantial role of [voice] in perception in spite of its reduced role in production might suggest that its progressive disappearance is production-driven.
Beddor, P. S. (2015). The relation between language users’ perception and production repertoires. Proc. of the 18th ICPhS, Glasgow, UK
Mazaudon, M., & Michaud, A. (2008). Tonal contrasts and initial consonants : a case study of Tamang, a ’missing link’ in tonogenesis. Phonetica, 65, 231-256.
Silva, D. J. (2006). Acoustic evidence for the emergence of tonal contrast in contemporary Korean. Phonology, 23, 287-308