Consonants

Lizu has a complex consonant system. There is a general three-way contrast in stops and affricates: voiceless aspirated, voiceless unaspirated, and voiced. Velar and uvular stops contrast before /o/, e.g. /^R ko/ ‘to beg’ vs. /^F qo/ ‘hole, pit’. Elsewhere, they are in complementary distribution. Uvular stops are found before back non-high vowels and velar stops are more freely distributed, e.g. / ^EP kæ-kæ/ ‘to fight’ vs. /^F qɐ/ ‘steelyard’.

Footnote ² Footnote ³ Footnote ⁴

Velar fricatives show a two-way contrast between voiceless and voiced, as illustrated in /^Rɣe/ ‘anemone tomentosa’ vs. /^F xe/ ‘to hatch’. The voiced velar fricative has a uvular allophone occurring before /ɐ/ and /wɐ/, e.g. /^Fɣɐ/ [^Fʁɐ] ‘needle’, /^Rɣwɐ/ [^Rʁuɐ] ‘to thunder’.

There are four nasals /m n ɲ ŋ/. We only found contrast between /n/ and /ɲ/ before /e æ o/. /ŋ/ is the only syllabic consonant in Lizu, but with a very restricted distribution. It occurs only after the voiceless velar plosive /k/ and in very few words, e.g. /^F k / ‘seven’.

One interesting phenomenon is the nasalization of glottal-initial words, which appears to be an areal feature in a number of Ngwi languages of Southwest China, such as Lahu (Matisoff Reference Matisoff1973: 20–21, Reference Matisoff, Ferguson, Hyman and Ohala1975; Sprigg Reference Sprigg1987), and Lisu (Bradley Reference Bradley, Bradley, Henderson and Mazaudon1989). Nasalization applies without exception, e.g. /^F he/ [^F ] ‘bamboo’, /^F hu/ [^F ] ‘language’, /^LPʔoɐ/ [^LPʔ w ] ‘goose; swan’, /^EPʔiæ/ [^EPʔ j ] ‘duck’. When preceding /ɐ/, /ɹæ/, and /ɹə/, the /h/-initial is realized as voiced, e.g. /^R hɐ/ [^R ] ‘five’, /^R hɹɐ/ [^R ] ‘to obtain’, /^R hɹə/ [^R ] ‘mushroom’. Perceptually, it has more and longer frication, and the following vowel is less nasalized.

The approximant /j/ occurs mostly before front vowels with the sole exception of the back vowel /o/ (e.g. /^F jo/ ‘self’). The phonetic realization of /j/ varies according to the height of the vowel: the higher the vowel, the stronger the frication. This surface contrast is illustrated in /^F ji/ ‘to go’ vs. /^LP jæ-he/ ‘last year’.

/w/ occurs only before /æ ɐ o u/. In most cases, /w/ is realized as a voiced bilabial continuant, much like the English /w/, e.g. /^EP wæɹæ/ ‘cloth’ (but see its realization before the vowel /u/ in the section ‘Vowels’). /ɹ/ only occurs before /æ/, /ə/, and /wæ/, e.g. /^Fɹæ/ ‘yak’, /^Rɹə/ ‘to laugh’, /^Rɹwæ/ ‘chicken’.

Consonant clusters

Consonant clusters are a prominent feature of the Lizu language. Most consonant clusters involve bilabials and laterals.

Vowels

The Lizu vowel system comprises eight monophthongs. Diphthongs are only observed in recent loanwords from the local (Southwest) Mandarin dialect (e.g. /^RPʃətsai/ ‘really’, Southwest Mandarin /ʂʅ²¹ tsai ²¹³/ 实 在). Note that the actual realization of Lizu vowels is in most cases different from what may be expected from the adopted IPA symbols (see the vowel chart plotted on the relative F1/F2 formant values).

The most frequent vowels include /e/, /o/, /u/ and /æ/. They appear after almost every initial consonant. The least frequent vowel is /y/, which was found only after the initials /l/, /h/, and alveolopalatals (e.g. /^LP de-ly/ ‘white’, /^F hy/ ‘back, following’, /^R dʑy/ ‘poison’), as well as in loanwords from Mandarin Chinese (e.g. /^RP jɐy/ ‘potato’).

/i/ after /b/, /p/ and /l/ is realized as /ʝi/ (i.e. perceptually with considerable frication) such as in /^F bi/ [^F bʝi] ‘bee’, /^R li/ [^R lʝi] ‘ashes’. This is in contrast to its realization in /^Fɲi/ ‘gold’ or /^F mi/ ‘monkey’.

/u/ is a high back but slightly centralized vowel. It is produced with greater lip compression than the cardinal /u/. The pronounced lip compression leads to some phonetic variation in the realization of the vowel and sometimes even the initial consonant onset. /u/ trills after bilabial and alveolar stop initials and is realized in this environment close to [ ]. This is a common, conceivably areal phenomenon in many neighboring languages, such as Nuosu (Li & Ma Reference Li and Ming1983: 52–53, 77), Yongning Na (Yang Reference Yang2009: 3), or Namuzi (Lama Reference Lama1994: 52). In contrast to Nuosu, where the trill is produced on only one side of the lips (Bradley Reference Bradley2008), the Lizu trill is produced with both lips vibrating against one another (see Figure 1). The bilabial trill is particularly evident in the minimal contrastive pair /^F tu/ [^F t ] ‘bean’ vs. /^F to/ [^F to] ‘to look’. Compare also the realization of /u/ after a dental stop, /^F tu/ [^F t ] ‘bean’, with that after an alveolar affricate, /^LP(de-)tsu/ ‘to put on (a hat)’.

Figure 1 Pronunciation of the word /^F tu/ [^F t ] ‘bean’ photographed at the beginning and end of the vowel production.

/u/ also introduces the allophonic variation of /x/ with [f] (e.g. /^F xu/ [^F fu] ‘garlic’), and of /w/ with [v] (e.g. /^LP wu-li/ [^LP vu-li] ‘head’, /^RP kʰe-wu/ [^RP kʰe-vu] ‘to buy’).

The difference between /æ/ and /ɐ/ is, for many non-native ears, rather subtle. Minimal pairs include /^R ntsʰæ/ ‘to do, to make’ vs. /^R ntsʰɐ/ ‘liver’, and /^RPʃæ-ʃæ/ ‘to search, to look for’ vs. /^RPʃɐ-ʃɐ/ ‘long’.

/ə/ co-occurs with a restricted number of consonant onsets (/s z t s t sʰ dz ʃ ʒ tʃ tʃʰ dʒ k kʰ ɡ/). After an onset, it is realized in two different ways. When following /s z t s t sʰ dz ʃ ʒ tʃ tʃʰ dʒ/, /ə/ is homorganic to the preceding consonant onset. For example, /^EP sə-sə/ [^EP s -s ] ‘to wage war’, /^F zə/ [^F z ] ‘son’, /^F tsʰə/ [^F tsʰ ] ‘salt’, /^F dzə/ [^F dz ] ‘to give birth’, /^Fʃə/ [^Fʃ ] ‘meat’, /^R tʃə/ [^R tʃ ] ‘star’, /^F tʃʰə/ [^F tʃʰ ] ‘sedan chair’, /^F dʒə/ [^F dʒ ] ‘pan’. When following velar stops, /ə/ is realized close to [ɘ], as in /^F kə/ ‘eagle’, /^RP kə-kə/ ‘big’.

Lizu has an additional set of nasalized vowels, which are restricted to (mostly recent) loanwords from Tibetan and Mandarin Chinese, where the donor language has or had a nasal coda (-m, -n, -ng) or a nasalized vowel (see also above on prenasalized clusters). For example, /^EP s mu/ [^EP s ] ‘to make smoke offering’ (WT bsang). This is different from the cases of nasalization conditioned by the glottal initial consonants /h/ and /ʔ/, or the effect of prenasalized cluster (see above). Overall, vowel nasalization in Lizu must be regarded as subphonemic, and only needs to be marked in those cases where it is unpredictable (i.e. in recent loanwords).

Syllable structure

The canonical Lizu syllable minimally consists of an obligatory nucleus and a tone. It may also contain up to three optional elements in the following linear structure: (C1)(C2)(C3)V, where C1 can be nasal, C2 can be any consonant, and C3 can only be one of the following set: -j-, -w-, -ɹ-, -ʑ-, -z-, -ɕ-, -ʃ-; V stands for vowel, and parantheses indicate optional constituents.

1. (1) V /^Fæ/ ‘I, first person singular pronoun’

2. (2) CV /^RP be-be/ ‘to crawl’

3. (3) CCV /^RP bje-bje/ ‘thick, coarse’

4. (4) CCCV /^RNbje/ ‘mountain’

Similar to its linguistic neighbors, Lizu is phonologically monosyllabic with a strong tendency towards disyllabicity in its lexicon. Tri-syllabic and tetra-syllabic words are mostly composite, e.g. /^LP toNbu m u/ ‘nose hair’ (< /^LP toNbu/ ‘nose’, /^F mu/ ‘feather, fur’), /^RP sɐlɐ mutsə/ ‘wild cat’ (< /^RP sɐlɐ/ ‘forest’, /^RP mutsə/ ‘cat’). There is also a handful of trisyllabic monomorphemic words (both native and loanwords), e.g. /^EPʃeNbelje/ ‘buttocks’, /^EPɬoNbotʃʰe/ ‘elephant’ (WT glang po che).

In disyllabic composite forms, where the second syllable has zero initial, the two adjacent vowels merge into one vowel or a diphthong, a process that typically results in a long vowel or diphthong. This change characteristically occurs when the perfective marker is added to a verb stem, and when the recipient marker is added to a nominal form. The two markers are homophonous (/æ/) and assimilate to the preceding vowel. For example, compare the realization of the verb /^RP de-pʰɐ/ ‘to smash’ in isolation and when followed by the perfective marker /æ/, i.e. /^RP de-pʰɐ=æ/ ‘have smashed’. Compare also the realization of the second person singular pronoun ‘you, thou’ in isolation, /^F ne/, and when followed by the recipient marker, i.e. /^F ne=æ/ ‘(e.g. to give) to you’.

Prosodic organization

Similar to many of its linguistic neighbors in the provinces of Sichuan and Yunnan (western Tibeto-Burman languages), Lizu has a sparse or ‘culminative’ tonal system, in which no more than one pitch pattern is pronounced per word (or longer phonological unit) (see Evans Reference Evans2008, Reference Evans2009, for discussion). For multisyllabic words, we also observe prominence difference among the syllables, resembling stress difference in Germanic languages (see below for details). Above the word (i.e. compounds), the tone of the initial word remains and affects the pitch realization of the rest of the constituent. Such a culminative tonal ‘spreading’ pattern of Lizu bears some resemblance to that described for Tamang (e.g. Mazaudon Reference Mazaudon and Kaji2005) and for some Wu dialects of Chinese such as Shanghai Chinese (e.g. Duanmu Reference Duanmu1995, Reference Duanmu1999; but see Chen Reference Chen2008). Culminative tonal systems in the languages of Southwest China have been little researched to date (but see J. T.-S. Sun Reference Sun2005, Ding Reference Ding2007, Chirkova & Michaud Reference Chirkova and Michaud2009, Greif Reference Greif2010 for recent studies), and available data allow considerable room for interpretation. For this reason, the following discussion of Lizu prosody is limited to the presentation of observable surface phenomena, as found in a corpus of vocabulary items, sentences (including a number of semi-controlled production studies targeting tones and tone sandhi in different domains), and traditional stories.

Word-level tonal patterns

Lizu is a tone language. For monosyllabic words, there are two contrastive tones: low rising (R) vs. high falling (F). Contrastive pairs abound: /^R ne/ ‘two’ vs. /^F ne/ ‘you, thou’; /^Rŋu/ ‘silver’ vs. /^Fŋu/ ‘cow’; /^R tʃə/ ‘star’ vs. /^F tʃə/ ‘frost’. See Figure 2 for the contrastive pair /^Rŋu/ ‘silver’ vs. /^Fŋu/ ‘cow’. In the five-scale pitch system developed by Chao (Reference Chao1930), these tones may be annotated as 23 for the rising tone and 51 for the falling tone. We will, however, refrain ourselves from using this notation as variation abounds in the actual realization of the two lexical tones, and in particular, in the falling tone where the actual f0 fall of the falling tone is not obligatory. This lack of dynamic f0 change has also been observed over the rising tone although not as often as for a falling tone. The extreme cases can be found in a few words, including /^R xwɐ/ ‘bird’ and /^F te/ ‘one’. Here, both renditions (i.e. one with a clear f0 fall/rise and one without any perceptually salient f0 fall/rise) are equally acceptable for our native language consultants. We think that the rich latitude for variation is probably due to the fact that there are only two tonal contrasts to be made in this language over monosyllabic words.

Figure 2 Pitch contours of the rising (R) and falling (F) tones illustrated with /^Rŋu/ ‘silver’ and /^Fŋu/ ‘cow’, respectively.

In disyllabic words, three pitch patterns are observed. No minimal three-way contrast, however, has been attested. The attested binary tonal contrasts yield the following three patterns:

1. (i) Both syllables sound equally prominent and no special lengthening is observed on either syllable (despite the presence of final lengthening, due to the fact that these words were elicited in isolation as an utterance on its own). There is also no salient rise or fall over either syllable. Rather, it seems to be two mid-level pitch contours with the first slightly higher than the second, which might also show a slight falling pattern. Hereafter, we will refer to this pattern as Equally-Prominent Contour (EP). Examples include /^EP midzə/ ‘hare’, /^EP səNɡe/ ‘lion’ (WT seng ge).

2. (ii) The duration of the second syllable, despite final lengthening, is relatively shorter and the syllable also sounds less prominent. The high f0 peak is typically realized before the end of the first syllable, where the pitch starts to fall already and continues to fall in the second syllable. Hereafter, we will refer to this pattern as Left-Prominent Contour (LP). Examples include /^LP midzə/ ‘pepper’, /^LP melje/ ‘place’.

3. (iii) The duration of the second syllable is relatively longer than that of the first one. Within the first syllable, there is a slightly rising pitch contour. The high f0 peak is realized within the second syllable where there is also a clear fall. The second syllable sounds more prominent. Hereafter, we will refer to this pattern as Right-Prominent Contour (RP). Examples include /^RP mutsə/ ‘cat’, /^RP melje/ ‘wind’.

The relative difference in syllable length is particularly striking when compared across minimal pairs, such as /^EP midzə/ ‘hare’ vs. /^LP midzə/ ‘pepper’, and /^LP melje/ ‘place’ vs. /^RP melje/ ‘wind’. A three-way contrast can be observed over the comparison of the words /^EP midzə/ ‘hare’, /^LP midzə/ ‘pepper’, and /^RP mutsə/ ‘cat’ in Figure 3, noting that these three words have comparable segmental composition.

Figure 3 Illustration of the three-way prominence contrast in disyllabic items with /^EP midzə/ ‘hare’, /^LP midzə/ ‘pepper’, and /^RP mutsə/ ‘cat’. Dotted lines indicate syllable edges. Within a word, the length contrast between the two syllables is indicated via their percentage of duration.

Native speakers of Lizu seem to be very sensitive to the three different prominence patterns, as suggested by a pilot perception experiment. We substituted the segmental composition of 10 triplets (of comparable segments but different prominence patterns) with the same syllable /a/ which was resynthesized with the same acoustic parameters (i.e. intensity, duration, and f0) of each corresponding word. Listeners were asked to identify the synthesized word they heard as one out of the three choices. Listeners showed a high degree of consistency in their judgements (Chen & Chirkova Reference Chen and Chirkova2012).

The observed differences are reminiscent of the effect of a stress system in languages such as English (e.g. the contrast in apex, apple, and above). It is worth noting that unlike English, Lizu does not exhibit vowel reduction in non-stressed positions.

In tri- and tetra-syllabic words (both monomorphemic and polymorphemic), we observe again three common patterns, comparable to the ones in disyllabic domains:

1. (i) The Equally-Prominent Contour. All syllables within a domain are relatively equal in terms of duration and prominence. The pitch contour over the tonal domain remains mid or high level until the last syllable where there is a clear falling pattern. Examples include /^EPʃeNbelje/ ‘buttocks’, /^EP tɐNtsʰə pimæ/ ‘lizard’.

2. (ii) The Left-Prominent Contour. The first syllable within the domain tends to be longer and sounds more prominent. There is a rising contour, the peak of which varies slightly depending on the length of the word. For a three-syllable word, the peak is typically realized at the beginning of the second (following) syllable, while for a four-syllable word, the peak is further delayed within the second syllable, after which there is a clear falling contour till the end of the word. Examples include /^LP tɕjɐNtʃʰokʰɐ/ ‘north’ (WT byang phyogs), /^LPɲimæ mitɕo/ ‘sunflower’.

3. (iii) The Right-Prominent Contour. The last syllable within the domain sounds more prominent, although lengthening was not consistently observed. This is probably due to the fact that utterance-final syllables usually undergo robust lengthening. The f0 peak is realized within the last syllable and the preceding syllables show a mid-level f0 contour. Examples include /^RPɲilɐɡu/ ‘daytime’, /^RP jeɕy kælæ/ ‘bat’.

Tonal patterns of compounds

In compounds, lexical tones undergo sandhi changes, where only the tonal contour of the initial word is retained and realized over the whole compound domain. Given that disyllabic compounds are by and large lexicalized, in this overview, we will essentially focus upon tri- and tetra-syllabic compounds. When a compound is composed of a monosyllabic word followed by a disyllabic word, we observe two tonal patterns, namely, the Equally-Prominent Contour and the Right-Prominent Contour. For example, a compound beginning with a monosyllabic word that has a falling tone, e.g. /^Fŋu/ ‘cow’, typically has the Equally-Prominent Pattern, as in the compound /^EPŋu m eNtʃʰo/ ‘cow's tail’ (< /^LP meNtʃʰo/ ‘tail’). On the other hand, a compound beginning with a monosyllabic word that has a rising tone, e.g. /^Rɹwæ/ ‘chicken’, typically has the Right-Prominent Contour, as in the compound /^RPɹwæ meNtʃʰo/ ‘chicken tail’. This is illustrated in Figure 4. Within each compound, the f0 contour of the word ‘tail’ is different from that of that word in isolation. In other words, the tonal contour over the whole compound is here regarded as determined by the f0 contour of the initial word in isolation. Note, however, that the durational difference observed over the lexical items does not seem to hold at the compound level.

Figure 4 Illustration of tone sandhi changes over compounds. Here, ‘tail’ with a left-prominence pattern surfaces with different pitch contours when preceded by words with different tonal patterns (‘chicken’ with a rising tone and ‘cow’ with a falling tone). Dotted lines indicate word boundaries.

If the initial word of a compound is disyllabic, we observe three contours. Again, the resulting contours are determined by the tonal contour of the initial disyllabic item, which, we know has three possible realizations. Here the tonal contours of these items are comparable to the three different contours we observed in the disyllabic lexical items (i.e. the Equally-, Left-, and Right-Prominent Contours). The following examples illustrate:

• /^EP səNɡe/ ‘lion’ + /^LP meNtʃʰo/ ‘tail’ > /^EP səNɡe m eNtʃʰo/ ‘lion's tail’

• /^LP to-Nbu/ ‘nose’ + /^LP wu-li/ ‘head’ > /^LP to-Nbu w u-li/ ‘tip of the nose’

• /^RP mutsə/ ‘cat’ + /^LPNdo-qo/ ‘eye’ > /^RP mutsə Ndo-qo/ ‘cat's eye’

Lizu function words and discourse particles (e.g. the genitive particle /i/, the focus particle /le/ in the recorded passage) are never pronounced in isolation. Their surface tone realization depends on the tone of the preceding (host) lexical word (similar to tone sandhi in compounds).

Transcription of recorded passage ‘North Wind and the Sun’

The original recording (made with a solid-state recorder Fostex FR-2 and a Beyerdynamic M88 N microphone) has been made available to JIPA along with this analysis. The acoustic analysis of the recording was made using Praat (Boersma & Weenink Reference Boersma and Weenink2009). In the transcription, only lexical items are marked for tone, whereas function words are not.