Optimizing Display Other Formats:      PDF - 1696k

Volume 10 Issue 2 (2012)        DOI:10.1349/PS1.1537-0852.A.410

Note: Linguistic Discovery uses Unicode characters to represent phonetic symbols. Please see Optimizing Display for requirements to accurately reproduce this page.

Phonology Sketch and Classification of Lawu, an Undocumented Ngwi Language of Yunnan

Cathryn Yang

SIL International

Lawu is a severely endangered, undocumented Ngwi (Loloish) language spoken in Yunnan, China. This paper presents a preliminary sketch of Lawu phonology based on lexico-phonetic data recorded from two speakers in 2008, with special attention to the tone splits and mergers that distinguish Lawu from other Ngwi languages. All tone categories except Proto-Ngwi Tone *3, a mid level pitch, have split, conditioned by the voicing of the initial segment. In the conditioning and effect of these tone splits, Lawu shows affinity with other Central Ngwi languages such as Lisu and Lahu and is provisionally classified as a Central Ngwi language.

1. Introduction

This report presents a preliminary phonological sketch of Lawu, an undocumented Ngwi (Loloish) language spoken in Shuitang District, Xinping County, Yuxi Municipality, and Jiujia District, Zhenyuan County, Pu'er Municipality, in Yunnan, China. Comparative phonological data is used to hypothesize Lawu's placement in the Central Ngwi (CN) cluster within the Ngwi sub-branch of Tibeto-Burman. Like other Central Ngwi languages, such as Lolo, Lahu and Lisu, Lawu shows splits in Proto-Ngwi Tones *1, *2, and *L, conditioned by voicing or glottal prefixation of the initial. However, Lawu has diachronically diverged from other CN languages, through the devoicing of Proto-Ngwi voiced stops and affricates, the merger of Proto-Ngwi palatoalveolar affricates and fricatives to alveolar, and the merger of Proto-Ngwi consonant clusters with plain stops, all of which result in a comparatively reduced inventory of initials. Lawu also shows later micro-splits in Tones *2 and *L, conditioned by voicing of the initial and vowel height. There is no linguistic documentation of Lawu, only passing mention of a distinct group in Chinese demographic sources (XPXZ 1993) and Pelkey's (1999) compilation of information on the Yi ethnicity. Although now ethnically categorized as Lahu, Lawu does not belong to the Lahoid language cluster (which includes Black Lahu, Yellow Lahu and Kucong). A distinct ISO 639-3 code for Lawu was applied for in July 2011. Hopefully this preliminary information will open the door for further investigation of this endangered language.

The Lawu data presented here was recorded in Jiuha Village, Shuitang District, Xinping County, Yuxi Municipality. Two native speakers recorded a list of 1,001 lexical items in 2008. These items were compared with Proto-Ngwi and lexical data from other Ngwi languages such as Hani (Southern Ngwi), Lisu (Central Ngwi), Lahu (Central Ngwi) (Sun 1991, Huang & Dai 1992), and Nisu (Southeastern Ngwi, data from author's fieldnotes). Based on shared tonal innovations, I propose that Lawu is a member of the Central Ngwi subgroup, although further documentation and comparative analysis is needed to confirm this hypothesis.

2. Background

Lawu speaker population is difficult to estimate with accuracy. Lawu are now officially classified as members of the Lahu nationality, but were previously classifed as Yi (XPDZ 1986; XPXZ 1993). A government official in Jiuha Village reported approximately 300 'Lahu' in the village. Only the grandparental generation still speaks Lawu in Jiuha, so the Lawu speaker population there is probably 50 or less. There may be additional Lawu speakers in neighboring Zhenyuan County. According to the Gazetteer of Zhenyuan County (ZYXZ 1995: 74), there were 6,455 Lawu in Zhedong and Jiujia districts in 1995. However, a more recent demographic website (YSX 2008) identifies the Lahu in Zhedong as speakers of Kucong, a different language. Jiuha villagers reported the presence of Lawu speakers in Jiujia District, but not Zhedong. The Lahu population in Jiujia District, reported online as 3,288 (BDBK 2011) may contain both Kucong and Lawu speakers. Further fieldwork is needed for a more precise estimate of how many ethnic Lawu still speak the language.

Lawu is severely endangered, according to UNESCO's (2003, 2009) criteria for assessing language vitality and endangerment. Village leaders and wordlist participants in Jiuha Village were interviewed about the ethnolinguistic vitality of Lawu. Interview questions dealt with vitality indicators such as domains of language use, proficiency of children in Lawu, language attitudes, and contact with other languages. Only grandparents and older generations still speak the language; two generations ago, parents began speaking Chinese to their children in the hopes of giving them an advantage in school. Members of the current parental generation have some passive understanding of Lawu, but do not speak it to their children. Lawu is rarely spoken in the home domain, being only spoken between elderly spouses or friends. As most speakers are elderly and have trouble with mobility, they have few opportunities to leave their household to speak Lawu with their peers. Lawu speakers' attitudes towards their own language is laissez-faire; they do not mind if their grandchildren do not learn to speak Lawu.

There is no orthography for Lawu, and thus no educational materials in Lawu and no use of Lawu in any new print domains such as the Internet. There is no governmental or institutional support for Lawu; though the Chinese constitution gives minorities the right to develop their language, there is no monetary or administrative support for the Lawu to do so. Government encourages assimilation to the dominant language through the use of Mandarin Chinese only in the school. Chinese is used in all public domains, including village meetings. Besides this sketch, there is no documentation of Lawu whatsoever. Hopefully, more extensive documentation can begin before Lawu becomes extinct.

3. Methodology

Fieldwork was conducted in March 2008 as part of a research project on geographical variation in lesser-known Central Ngwi languages of western Yunnan, including Lalo, Lolo, Limi, and Lawu. Further details regarding this project are given in Yang 2010a. In Jiuha Village, one male and one female native speaker, both elderly, were asked to give Lawu equivalents to the Chinese prompts of a 1,001 item wordlist adapted from Pelkey 2008. The male speaker was recorded uttering lexical items in citation form, and the female speaker was recorded giving the lexical item in utterance medial form. The utterance medial form was embedded in one of several carrier phrases chosen for semantic plausibility and designed so that the elicited item was usually preceded by a syllable with mid level pitch, from Pelkey 2008. Both citation and utterance medial forms were repeated three times. An Edirol R-09 digital recorder was used to record in uncompressed .wav format.

Contrastive tone categories were determined during elicitation by asking speakers if lexical items that differed only in pitch and/or phonation were in fact minimal pairs. In the carrier phrase, the syllable preceding the elicited lexical item had a mid level pitch; by comparing the elicited item's pitch with the mid level pitch, I identified the tone categories of specific syllables. Comparison with lexical data from other Central Ngwi languages (Lolo, Lisu and Lahu (Huang & Dai 1992)) and with Bradley's (1979) reconstruction of Proto-Ngwi also helped clarify which tone category a syllable belonged to. All Proto-Ngwi forms cited in this paper are from Bradley 1979.

For the acoustic tonal analysis, monosyllabic target words were used in both citation and utterance medial form. For each tone, approximately 30 tokens were used (on average 10 lexical items, with three repetitions), for an average of 150 tokens per speaker. Fundamental frequency (F0) measured in hertz (Hz) was extracted using Praat language software (Boersma & Weenick 2009) and analyzed using procedures developed by Stanford (2008, 2011). Stanford's script in R (a free statistical software package) converts Hertz values to semitones, a logarithmic transformation that has been shown to better model speakers' intuitions about pitch difference than the Hertz scale (Nolan 2003). The speaker's mean pitch value of Tone 3, the mid-level tone, was defined as the zero level pitch value and thus served as the benchmark for normalized pitch (Stanford 2008). Stanford's script divides the pitch token into 600 equally spaced locations throughout the vowels' duration, and cuts off the first 25% of the vowel onset and the last 10% of vowel offset to avoid syllable edge effects. Similar to Stanford's (2008) study of Sui, the tone tokens in this study had a mix of onsets in terms of voicing and sonority; therefore, a larger portion of the onset was omitted, as different onsets have been shown to locally affect F0 (Hombert 1978, Hombert et al. 1979). For each tone category, pitch values were averaged for all tokens at each normalized time point. The tonal inventories in Figure 3 and Figure 4 are presented in semitones and normalized for duration and mean Tone 3 F0. When transcribing tone, Chao's (1930) system of pitch numbers is used to represent the relative pitch height of the tone (1 is low and 5 is high).

4. Phonology sketch

Lawu's syllable template is (C)V(ŋ) with suprasegmental tone. There are 22 initial consonants, six monophthongs that may be followed by -ŋ, five diphthongs, and four tones. Like other Central Ngwi languages (Matisoff 2003), Lawu has lost Proto-Ngwi syllable-final consonants. Only a few instances of Proto-Ngwi *-ŋ are preserved, as in kəŋ³³ 'bracelet' from Proto-Ngwi *ɡoŋ¹.

4.1 Initials

Table 1 gives the inventory of initials in Lawu, with allophones in brackets. Alveolar and alveopalatal obstruents are contrastive before all vowels except /e/; before /e/, only the alveopalatal occurs. The alveolar series does not occur before [i], as the allophone of /i/ after alveolar affricates and fricatives is [ɿ], an apical anterior vowel (Lee 2005). The female speaker distinguished between alveolars and alveopalatals before /i/ (i.e. [tɕi] and [tsɿ] were distinct in her pronunciation), but the male speaker had merged both syllables in favor of [tɕi]. There are no examples of [tɕa] in the wordlist, but /tɕ/ contrasts with /ts/ before all other vowels, e.g. tso⁵⁵ 'drill a hole' contrasts with tɕo⁵⁵ 'lean (adj.),' and tsu³³ 'teach/train' contrasts with tɕu³³ 'cold (adj.)'.

/f/ has a restricted distribution, only appearing before close vowels /i/ and /u/. It is contrastive with /x/ before /i/: compare a⁵⁵ fi³³ 'left' with xi³³ 'situated crooked,' and with /s/ before both /i/ and /u/: compare fu⁵⁵ 'swell,' su⁵⁵ 'iron' and a³³sɿ³³ 'who?'. Southwestern Mandarin likewise has no opposition between [xu] and [fu]; /x/ before [u] is pronounced as [f] (Ho 2003:130). There is a gap in the inventory, with no voiced alveopalatal fricative [ʑ]. Proto-Ngwi *ʒ, the most likely source for [ʑ], instead is usually reflected as palatal glide [j] (e.g., *ʒakᶫ 'descend'> ja³³), thus merging with some reflexes of *y. /n/ is realized as [ɲ] before close front vowel /i/ and /e/. Table 2 gives lexical examples of Lawu initial consonants.

Table 1: Inventory of initial consonants in Lawu

Table 2: Examples of initial consonants in Lawu

Table 3 summarizes the development of manner of articulation of Lawu stops and affricates, compared with other CN languages (Matisoff 2003:17). Bilabial stops (P, P^h, B) are used in the table to represent all places of articulation for stops and affricates. *C is a cover symbol for the Proto-Tibeto-Burman voiced consonantal prefixes *b, *d, *g, *r, *l (Bradley 1979). *m- represents the Proto-Ngwi homorganic nasal prefix. Proto-Ngwi voiced and nasal prefixed stops and affricates have unconditionally devoiced in Lawu, so that Lawu no longer has any voiced stops and affricates, unlike most other CN languages. This development may be due to the influence of Southwestern Mandarin, whose voiced stops and affricates have also devoiced. Other languages Lawu is in contact with, i.e. Nisu and Lolo, have retained voiced onsets. The devoicing of voiced stops and affricates was the trigger for the phonologization of the tone splits described in Section 4.3.

Table 3: Development of manner of stops and affricates in Central Ngwi

Proto-Lawu had preglottalized sonorants *ʔm, *ʔn, *ʔl, and *ʔw, a coalescence of the Proto-Ngwi *ʔ or *s- prefixes before sonorants. Preglottalization has since been lost in Lawu, but at the time of the Lawu tone splits (see Section 4.3), preglottalization blocked the lowering of the pitch seen in plain voiced sonorants. Modern pitch reflexes in syllables with Proto-Lawu preglottalized sonorants match those of voicless initials rather than voiced, e.g. Proto-Lawu *ʔla¹ 'spirit' > la⁵⁵ and *ʔpa¹ 'exchange'> pa⁵⁵ versus *la¹ 'come' > la³³.

Table 4 summarizes the Proto-Ngwi sources of Lawu initials. The majority of Proto-Ngwi consonant clusters of stop plus *-l, *-r, *-y have merged with the simple stops. Only *ky and *gy palatalized to the alveopalatal affricate before close vowels (e.g., *ky > tɕ/_*i, e, o) and merged with the alveolar affricates before rhymes with *a (e.g., *gyaw² 'waist' > [tso²¹]; *kyaŋ² 'friend' > [ts^hu²¹]). Palatoalveolar affricates and fricatives merge with the alveolar series (*c > /ts/ and *ʃ > /s/). Before rhymes *a and *aŋ, *my merges with *n to /n/. Proto-Ngwi resonants *r and *y become fricatives before back vowels, with the place of articulation conditioned by the vowel, e.g., *r and *y > v/_*u and > z/_*-a-. Elsewhere, *r becomes /ɣ/, and *y remains a palatal glide /j/. *w also becomes a fricative, usually /v/. *h merges with *x, and becomes /ɕ/ before high front vowels. Labial prefixes preempt initial liquids *l and *r, with both *b and *m becoming a labial stop, e.g., *b-le² 'four' > [pi²¹], *m-le² 'grandchild'> [pi²¹], *m-rwe¹ 'snake' > [pe³³]. This suggests an early merger in Proto-Lawu of the *m- and *b- prefixes to an initial *b. Labial stops before *w became labiodental fricatives, as in * b-way² 'left' > *vi³³ > [fi³³] and *C-pwam² 'swell' > [fu⁵⁵]. Similarly, glottalized *ʔ-pl becomes /f/ in *ʔ-plu¹ 'porcupine,' but plain *plu¹ 'silver/white' merges with /p^h/, becoming [p^hu⁵⁵].

Table 4: Proto-Ngwi sources of Lawu initial consonants

4.2 Rhymes

Lawu has six monophthongs and five diphthongs, given in Table 5. All monophthongs can be followed by the velar nasal -ŋ except -u-. Diphthongs may be followed by the velar nasal in Chinese loanwords (e.g., p^haŋ²¹kuaŋ³³ 'bladder'), but not in native vocabulary. Lawu is unlike most other Central Ngwi languages in having a contrastive nasal final. In a few items, Proto-Ngwi *-ŋ has been retained, as in pəŋ³³ 'full' from Proto-Ngwi *m-bliŋ' and tsəŋ⁵⁵ 'tree' from Proto-Ngwi *dzin¹. The nasal final is also seen in Chinese loanwords, e.g. ɕaŋ²¹fu²¹ 'blessed, happy,' and in some words whose diachronic origin is unclear (e.g. p^hiŋ³³ 'knife'). In originally open syllables beginning with nasal stops, a nasal final has been added after back vowel *o, e.g. *s-mo¹ 'mushroom' > məŋ⁵⁵ , and *C-no² 'awake' > nəŋ²¹. The close vowel [i] is realized as apical vowel [ɿ] after alveolar affricates and sibilants (i.e. /z, s, ts and ts^h/). Table 6 gives examples showing minimal or near minimal pairs.

Table 5: Inventory of rhymes in Lawu

Table 6: Examples of rhymes in Lawu

The following near minimal pairs further distinguish phonetically similar rhymes:

1. Front vowels:

a²¹pi³³ 'duck'

pie³³ 'explode'

pe³³ 'snake'

piŋ³³kaŋ⁵⁵ 'rainbow'

2. Back vowels:

tu²¹ 'dig a hole'

tue⁵⁵ 'exit'

Table 7 gives the Proto-Ngwi sources for Lawu rhymes, arranged in order of decreasing frequency in the 1,001-item wordlist. I compared the Lawu lexical data to Bradley's (1979) Proto-Ngwi reconstructed vocabulary of 866 items to determine the Lawu reflexes of Proto-Ngwi rhymes. The monophthongs [e] and [o] have fewer diachronic sources than [i] and [u] and are less frequent than other monophthongs. Monophthongs are more frequent than diphthongs or nasal final rhymes. Some nasal final rhymes and diphthongs occur only rarely, such as [eŋ] and [ua], which have only one example each in the wordlist.

Table 7: Proto-Ngwi sources of Lawu rhymes

4.3 Tones

This section gives an acoustic analysis of the Lawu tonal system and describes Lawu's complex tone splits and mergers. Lawu has four contrastive tones: three level pitches (high, mid, low) in modal phonation and one mid falling pitch in harsh phonation. Compared to other Central Ngwi languages, this tonal inventory is rather small: Lahu has seven (Matisoff 1982 [1973]), Lisu six (Bradley 2003), Central Lalo five (Yang 2010a). The reduced number of tones is due to the unconditioned merger of Proto-Ngwi Tones *1 and *H, and the multiple mergers with Tone *3 that occurred after the splits in Tones *2, *L, and *1/H.

The Proto-Ngwi tonal system had a three-way pitch height contrast in syllables ending in vowels or nasals: Tones *1 (*T1, high), *2 (*T2, low), and *3 (*T3, mid) (Bradley 1977). In syllables ending in *-p, *-t, and *-k, Proto-Ngwi had two distinct tone categories, *Low-stopped (*TL) and *High-stopped (*TH) (Matisoff 1972). In most Ngwi languages, syllable-final *stops merged to a glottal stop and then to laryngealized vocal register on the vowel, often seen as harsh (or tense) phonation. Harsh phonation is marked with an underscore under the vowel, e.g., [ a̠].

Harsh phonation is an aperiodic phonatory quality produced when the laryngeal sphincter (i.e., the thyroarytenoid muscle complex) is constricted and the ventricular folds incur over the vocal folds (Edmondson & Esling 2006: 162). High levels of jitter (irregularity in the duration of the glottal pulse cycle), shimmer (irregularity in amplitude), and high-frequency energy are acoustic correlates of harsh (Scherer 1986). In some Ngwi languages, such as Lalo (Yang 2010b), harsh phonation also correlates with increased pitch and shorter duration, compared to modal phonation. Figure 1 below gives the spectrogram of the male speaker's utterance of [na̠³¹] 'sticky', in harsh phonation. Note the blacker regions at high-frequency levels, reflecting high-frequency energy. Harsh [na̠³¹] has 2.7% local jitter and 12.8% local shimmer, measured with Praat. In comparison, modal [na²¹] 'rich' (spectrogram seen in Figure 2) has only 1.2% jitter and 8.3% shimmer.

Figure 1: Spectrogram of [na̠³³] 'sticky' by male speaker

Figure 2: Spectrogram of [na³³] 'rich' by male speaker

Figure 3 and Figure 4 show the acoustic tonal analysis of the male and female speaker, respectively, using the methodology described in Section 3. N is the number of tokens used in the analysis. Lawu's synchronic tone categories are abbreviated without asterisks as T1/TH (a merger of *T1 and *TH), T2, T3, and TL. The male speaker's tokens were all in citation form, and the female speaker's were in utterance medial form. In Figure 3, the merged T1/TH category is a high level pitch, and T3 a mid level pitch. T2 starts low and gradually falls to the bottom of the pitch range, while TL starts at the mid pitch and falls rapidly. TL is pronounced with slight harsh phonation, with a higher pitch onset and shorter duration than modal T2. The average duration for the male's TL tokens is roughly 75% of the duration of the average T2 token, and the duration difference between TL and T2 is statistically significant (p < 0.01). In the figures below, durations are normalized to depict relative differences, so TL is depicted in Figure 3 as 25% shorter in duration than the other tones.

Figure 4 below, based on the female speaker's utterance medial forms, shows some differences from Figure 3. Acoustic analysis of additional speakers from both genders uttering citation and utterance medial forms is needed to investigate the cause of the differences. With the current limited data, I cannot conclude whether the differences are related to gender or to the phonetic environment of the token used. Unlike the male speaker, the female speaker's TL and T2 do not differ in their duration. Although the female's mean TL pitch is still higher than the mean T2 pitch at a 0.05 level of significance, TL's pitch onset in Figure 4 is lower than the mid pitch onset seen in Figure 3. One possible explanation for the differences is that harsh phonation may be neutralized in utterance-medial speech. This neutralization also occurs in some Lalo languages (Yang 2010a: 180) and Phula languages (Pelkey 2011). The eventual loss of harsh phonation would not be surprising, as SW Mandarin, the language that is currently replacing Lawu, does not have harsh phonation.

Figure 3: Acoustic tonal analysis of Lawu male speaker in citation form

Figure 4: Acoustic tonal analysis of Lawu female speaker in utterance medial form

Table 8 below gives the Lawu tone box, showing the splits and mergers that were conditioned by voicing of the initial consonant. A tone box presents the historic interaction between initial and tone, originally developed by Gedney (1972) for Tai languages. Proto-Ngwi *T1 and *TH merged unconditionally to a high level pitch, and then underwent a split wherein voiced initials conditioned the lowering of the high level pitch to a mid level pitch. This resulted in a partial merger with *T3. Syllables with voiceless initials (including preglottalized and voiceless aspirated) remained high. In *T2 (low) and *TL (low and harsh), voiceless initials conditioned a raising of the pitch to mid level, resulting in another partial merger with *T3. Syllables with voiced initials remained low in T2 or became mid falling in TL. After the tone splits and mergers described here, voiced obstruent initials became devoiced, making the mid falling and low pitches contrastive. Lawu tone splits go up or down a prosodic staircase: voiced initials condition the pitch to either step down one level (high > mid) or remain low if the original pitch was low, and voiceless initials condition the pitch to either go up one level (low > mid) or stay high if the original pitch was high.

Table 8: Lawu tone box

In an additional twist, the Proto-Lawu vowel had to match the place of articulation of the initial in order to trigger the split: after anterior initials (labial and alveolar), the vowel had to be both high and front (i.e., *i) for the split to occur, and after non-anterior initials (e.g., velar), the vowel had to be high and back (i.e., *u). For example, as seen in Table 9, Proto-Lawu *pi² 'old,' *ti² 'to pound,' and *ku² 'head' (from Proto-Ngwi *ʔ-koŋ²) all show the high level pitch, but not *pay² 'duck' (the split happened before *ay merged to /i/), *ʔmu² 'to brood,' or *ʔna² 'ear.' Unfortunately, the wordlist has no examples of *T2 preglottalized sonorants with the correct conditioning for the tone split, but since the split in *T2 and *TL parallel each other in other aspects, I assume that *T2 glottalized sonorants also took part. Likewise, in *TL *ʔli̠ᶫ 'take off' (Proto-Ngwi reconstruction unknown, compare Proto-Lalo *ʔli̠ᶫ ) and *ku̠ᶫ 'to bend' became li⁵⁵ and ku⁵⁵, respectively. For Proto-Ngwi *k-lokᶫ 'stone' (probably with an epenthetic schwa after the *k- prefix, *kəlok ᶫ ), the vowels in the two syllables likely underwent metathesis while the initial prefix triggered glottalization, i.e., Proto-Ngwi *kəlokᶫ > *kə'ʔlu̠ᶫ > *ku'ʔli̠ᶫ > li⁵⁵ku⁵⁵. As in *T2, the *TL vowel had to match the place feature of the initial, so neither Proto-Lawu *ʔlu̠ᶫ 'graze' or *tu̠ᶫ 'to light a fire' took part in the split. Vowel quality conditioned tone splits are rare, but have been shown to occur in Proto-Lalo (Yang 2010a: 182-186), another Central Ngwi language, and in Lugbara (Andersen 1986).

Table 9: Examples of vowel-conditioned tone split in T2 and TL

Table 10: Examples of conditions that did not incur tone split

5. Lawu's position within Ngwi

Lawu is clearly a Ngwi language within the Tibeto-Burman language family, with transparent correspondences to Proto-Ngwi tone categories and many cognates shared with other Ngwi languages. Ngwi has been subgrouped into four main branches: Northern, Southern, Central, and Southeastern (Bradley 2002, Pelkey 2011). Lawu does not share any of the distinctive innovations that characterize Northern, Southern, or Southeastern Ngwi languages, but does show affinity with Central Ngwi through its tone splits in *T1, *T2 and *TL. In Northern Ngwi languages such as Nasu, *L became the highest pitch in the tone system (Bradley 1979), a change not seen in Lawu. In Southern Ngwi languages such as Hani and Akha, phonation caused a split in the manner of obstruents, also not seen in Lawu (Thurgood 1982). Other Southern Ngwi languages such as Bisu or Sangkong are very different from Lawu in their development of obstruents, with Proto-Ngwi glottal-prefixed stops becoming aspirated and nasals becoming voiced or prenasalized stops (Matisoff 2003: 17). Nor does Lawu show the development of velar- or alveolar-lateral clusters that distinguish Southeastern Ngwi languages (Pelkey 2011).

Instead, Lawu shares the Central Ngwi tendency to split *T1, *T2, and *TL, conditioned by voicing or prefixation of the initial. Table 11 shows a comparative tone box for some Central (C), Southern (S), and Southeastern (SE) Ngwi languages, organized by manner of Proto-Ngwi initial. An underscore below the Chao pitch number represents harsh phonation. Lawu's conditioning of the *T2 split (*voiceless and *ʔ /*s- prefixed initials) overlaps with Lisu's ( *ʔ /*s- prefixed) and Lahu's (*voiceless sibilants and *ʔ /*s- prefixed); Lawu and Lahu may have expanded the original Central Ngwi conditioning environment of *glottal stop and *s prefixation. In the conditioning of the *TL split, Lawu, Lisu and Lahu again share *glottal stop/*s- prefixation, with Lisu and Lahu expanding to include *C- prefixation (*C represents Proto-Tibeto-Burman voiced consonantal prefixes *b, *d, *g, *r, *l). In contrast, Hani (S) shows no such splits in any tone category. Nisu (SE) shows a glottal-conditioned split in *T1 but not elsewhere. Sani (SE) shows tone splits similar to Lisu, but Pelkey (2011) groups Sani as a Southeastern Ngwi language based on other evidence. Although the tone splits suggest affiliation with Central Ngwi, Lawu does not show typical Central Ngwi lexical items such as reflexes of *a na² for 'dog' and *a ʔdukᶫ for 'fire,' described in Bradley2004a. However, such lexical items are not a requirement for membership in Central Ngwi, as Lahu does not show them either.

Even though Lawu is most likely a Central Ngwi language, it does not clearly belong to the Lisoid (Lisu, Lipo, Lolo), Laloid (Central, Southern, Eka), or Lahoid (Lahu, Kucong) clusters within Central Ngwi. Lawu speakers were previously classified as belonging to the Yi nationality (XPXZ 1993), but are now classified as Lahu. However, there is no linguistic evidence that the Lawu language belongs to the Lahoid cluster of languages (e.g., Kucong, Black and Yellow Lahu). Lawu does not share any distinctive Lahu phonological or lexical innovations, such as *li > [h ɔ] and others listed in Bradley 2004b. Lawu does not share the distinctive Lalo *T2 split described in Yang 2010a, in which *T2 syllables with glottalized obstruents and the rhyme *-a shifted to *TL. Neither does Lawu share the creation of a high rising tone from a split in *H, as is seen in Lisu and Lamu (Bradley 2004a). At this point, it is unclear which Central Ngwi language Lawu shares the closest relationship with. Further data on other aspects of Lawu phonology and lexicon, as well as morphology and syntax, are needed to confirm Lawu's place within Central Ngwi.

Table 11: Tone reflexes in some Ngwi languages

6. Conclusion

Based on diachronic phonology, Lawu is most likely a Central Ngwi language, but has diverged from more well known Central Ngwi languages such as Lisu or Lahu. The Lawu community is in the process of shifting to Southwestern Mandarin and Lawu phonology shows possible influence from contact with SW Mandarin. Although data was gathered from only two Lawu speakers, these preliminary observations lay a foundation for a more comprehensive investigation. Lawu's status as a severely endangered language highlights the urgent need for such investigation, while there are still speakers.

References

Andersen, Torben. 1986. Tone splitting and vowel quality: Evidence from Lugbara. Studies in African Linguistics 17(1). 55-68.

BDBK. 2011. Baidu Baike: Jiujia xiang [Baidu Encyclopedia: Jiujia District]. http://baike.baidu.com/view/1439397.htm (accessed July 19, 2011).

Boersma, Paul & David Weenick. 2009. Praat: doing phonetics by computer. Version 5.1.07. http://www.praat.org/. (accessed 17 Jun, 2009).

Bradley, David. 1977. Proto-Loloish tones. In David Bradley (ed.), Papers in Southeast Asian Linguistics No. 5, 1-22. Canberra: Pacific Linguistics.

-----. 1979. Proto-Loloish (Scandinavian Institute of Asian Studies Monograph No. 39). London: Curzon Press.

-----. 2002. The subgrouping of Tibeto-Burman. In Christopher Beckwith & Henk Blezer (eds.), Medieval Tibeto-Burman languages, vol. 2, 73-112. Leiden: Brill.

-----. 2003. Lisu. In Graham Thurgood & Randy J. LaPolla (eds.), The Sino-Tibetan languages, 222-235. London; New York: Routledge.

-----. 2004a. Endangered Central Ngwi languages of northwestern Yunnan. Paper presented at the 37th International Conference on Sino-Tibetan Languages and Linguistics (ICSTLL), Lund.

-----. 2004b. The Lamu language. Paper presented at the 37th International Conference on Sino-Tibetan Languages and Linguistics, Lund, Sweden.

Yuanren Chao. 1930 Ə ˈsɩstəm əv toʊn ˈlɛtə z. Ma'tre Phon'tique 30. 24-27.

Edmondson,Jerold & John H. Esling. 2006. The valves of the throat and their functioning in tone, vocal register and stress: Laryngoscopic case studies. Phonology 23. 157-191. doi:10.1017/s095267570600087x

Gedney, William J. 1972. A checklist for determining tones in Tai dialects. In S.M. Estellie (ed.), Studies in linguistics in honor of George L. Trager, 423-438. The Hague: Mouton.

Ho, Dah-An. 2003. The characteristics of Mandarin dialects. In Graham Thurgood & Randy J La Polla (eds.), The Sino-Tibetan languages, 126-130. New York: Routledge.

Hombert, Jean-Marie. 1978. Consonant types, vowel quality, and tone. In Victoria A. Fromkin (ed.), Tone: A Linguistic Survey, 77'112. New York: Academic Press.

-----, John J. Ohala & William G. Ewan. 1979. Phonetic explanations for the development of tones. Language 55(1). 37-58. doi:10.2307/412518

Huang, Bufan & Qingxia Dai, eds. 1992. Zangmian yuzu yuyan cihui [A Tibeto-Burman lexicon]. Beijing: Zhongyang Minzu Daxue Chubanshe.

Lee, Wai-Sum. 2005. The articulatory and acoustical characteristics of the "apical vowels" in Beijing Mandarin. 150th Meeting of the Acoustical Society of America/NOISE-CON 2005. vol. 118, p. 2027. http://hdl.handle.net/10722/54214.

Matisoff, James A. 1972. The Loloish tonal split revisited (Center for South and Southeast Asia Studies, Research Monograph No. 7). Berkeley: University of California.

-----. 1982 [1973]. The grammar of Lahu (2nd ed., University of California Publications in Linguistics, No. 75). Berkeley: University of California Press.

-----. 2003. Handbook of Proto-Tibeto-Burman: System and philosophy of Sino-Tibetan reconstruction. vol. 135 (UC Publications in Linguistics). Berkeley: University of California Press.

Nolan, Francis. 2003. Intonational equivalence: an experimental evaluation of pitch scales. In Juergen Trouvain & William J. Barry (eds.), Proceedings of the 15th International Congress of Phonetic Sciences (ICPhS). 771-774. http://www.ling.cam.ac.uk/francis/Nolan%20Semitones.pdf (accessed 30 Jun 2009).

Pelkey, Jamin R. 1999. Yunnan's myriad Yi: Profiles on the peoples classified as Yi in Yunnan province. Unpublished manuscript.

-----. 2008. The Phula languages in synchronic and diachronic perspective. Melbourne: La Trobe University PhD dissertation.

-----. 2011. Dialectology as Dialectic: Interpreting Phula Variation (Trends in Linguistics: Studies and Monographs). Berlin: Mouton de Gruyter.

Scherer, Klaus R. 1986. Vocal affect expression: A review and a model for future research. Psychological Bulletin 99(2). 143-165. doi:10.1037/0033-2909.99.2.143

Stanford, James N. 2008. A sociotonetic analysis of Sui dialect contact. Language Variation and Change 20. 409-450. doi:10.1017/s0954394508000161

-----. 2011. Procedure for extracting tones from Praat to txt file, normalizing in R, exporting to Excel. http://www.dartmouth.edu/~jstanford/Procedure-for-extracting-Praat-pitches-to-Excel-and-normalizing-in-R.pdf. (accessed Apr 30, 2011).

Sun, Hongkai, ed. 1991. Zangmian yuyin he cihui [Tibeto-Burman Sound Systems and Lexicons]. Beijing: Zhongguo Shehui Kexue Chubanshe.

Thurgood, Graham. 1982. Subgrouping on the basis of shared phonological innovations: A Lolo-Burmese case study. BLS 8. 251-260.

UNESCO. 2003. Language Vitality and Endangerment. http://www.unesco.org/culture/ich/doc/src/00120-EN.pdf. (accessed Mar 26, 2009).

-----. 2009. UNESCO Interactive Atlas of the World's Languages in Danger. http://www.unesco.org/culture/ich/index.php?pg=00139. (accessed Feb 19, 2010).

XPDZ. 1986. Xinping Yizu Daizu zizhixian diming zhi [Geographic place name gazetteer of Xinping county]. Xinping: Renmin Zhengfu.

-----. 1993. Xinping xian zhi [Gazetteer of Xinping county]. 880-02 Di 1 ban. edn. Beijing: Shenghuo, Tushu, Xinzhi San Lian Shudian.

Yang, Cathryn. 2010a. Lalo regional varieties: Phylogeny, dialectometry, and sociolinguistics. Melbourne: La Trobe University PhD dissertation. http://arrow.latrobe.edu.au:8080/vital/access/HandleResolver/1959.9/153015.

-----. 2010b. Tone change in Lalo. In Yvonne Treis & Rik De Busser (eds.), Selected Papers from the 2009 Conference of the Australian Linguistic Society. http://www.als.asn.au/proceedings/als2009.html (accessed 30 Apr 2010).

YSX. 2008. Yunnan shuzi xiangcun [Digital Yunnan]. http://www.ynszxc.gov.cn/szxc/ProvincePage/default.aspx. (accessed Feb 23, 2010).

ZYXZ. 1995. Zhenyuan Yizu Hanizu Lahuzu Zizhixian zhi [Gazetteer of Zhenyuan Yi Hani Lahu autonomous county]. Kunming: Yunnan Renmin Chubanshe.

Contact information:
Cathryn Yang c/o SIL East Asia
53 Hao, Qing Nian Lu Yan Chang Xian,
18th Floor, Block B, Bin Jiang Da Sha,
Kunming, China 650051
cathryn_yang@sil.org
+86-13211695460

Appendix A: Lawu Word List