Rgyalrongic is a branch of Sino-Tibetan mainly spoken in culturally Tibetan areas of the Chinese province of Sichuan. With complex initial consonant clusters and an intricate verbal morphology of great antiquity, Rgyalrongic languages are increasingly appreciated for their central importance in Sino-Tibetan historical linguistics. This grammar of the Kyom·kyo dialect of Sìtǔ Rgyalrong is a welcome contribution to this group of endangered languages that only recently began to be properly described.
The author made a fortunate choice to study Kyom·kyo (腳木足 Jiǎomùzú). Sìtǔ Rgyalrong has a deep isogloss between deprenasalizing dialects and softening dialects. We can compare Cog·tse (卓克基 Zhuókèjī) Rgyalrong, a deprenasalizing dialect, with Kyom·kyo, a softening dialect. (The Cog·tse data are cited from Lin You-jing's MA thesis and Huáng Liángróng's dictionary.) Deprenasalizing dialects reduce initial prenasalized stops to plain voiced consonants: “to give”, Cog·tse kɐ-wə̂, Kyom·kyo ka-mbu; “pork fat”, Cog·tse tɐ-wōr, Kyom·kyo ta-nɡor. Softening dialects soften initial voiceless stops under certain contexts to plain voiced consonants: “to do”, Cog·tse kɑ-pɑ̄, Kyom·kyo ka-va; “needle”, Cog·tse ta-kāp, Kyom·kyo ta-waʔp. Softening is an isogloss deep enough to split also Upper Rgyalrong, dividing it into non-softening Japhug and Tshobdun and softening Zbu. Many other Burmo-Qiangic languages show reflexes akin to the softening dialects: Tangut 
·wji¹ “to do”, Burmese ap “needle”. As the bulk of scholarly attention was hitherto concentrated on deprenasalizing dialects like Cog·tse, this grammar is sure to fill an important dialectal gap in our knowledge of Sìtǔ Rgyalrong.
Kyom·kyo, as other softening dialects, show some forms with deprenasalizing or non-softening reflexes: the cognate words “clothing” and “to wear” are respectively təŋɡe and ka-waʔt (cf. Cog·tse təwā, ka-wāt); “this year” is pəjvaʔ, but “a year” is tə-pa (cf. Cog·tse pipā, tə-pā); “to help with physical labour” is ka-woʔr, while “to help” in general is ka-koʔr. Other lexical features also betray a complex linguistic history: Proto-Rgyalrongic *-a shows both non-brightened and brightened reflexes: compare “to eat” ka-ndza with “food” təndze (cf. Cog·tse ka-zā, təzā). Whether such cases reflect dialect mixture or complex internal development remains the subject of future studies.
The author chose to describe the result of “learn[ing] what [she] could from friends and colleagues” hailing from “several locations in Jiǎomùzú”. While Kyom·kyo dialects roughly show the same pattern in affixal morphology (but see for example p. 430, where past tense marker -s is used only in certain villages), the “phonetic and lexical” differences are significant: we learn that “towards” is wə-phaj in Mkho·no, but wə-mbaj in Phar·ba (pp. 15–6). The lexicon abounds with doublets, such as kə-rtok and kə-rdok for “one”, ka-mbu and ka-mbəm for “to give”, ka-rwe and ka-rwas for “to rise”, given as is without geographical, generational or register indications. The free mixing of material from different dialects and the absence of a consistent transcription implies that the data need to be rechecked for each individual dialect. The verb doublets, in particular, leave a lingering doubt about possible morphological alternations not treated in this grammar.
Words borrowed from Chinese, marked with a scarab ¤, are often not transcribed in IPA, but, confusingly, in a variant of pīnyīn-IPA where voicing indicates lack of aspiration. “Handbags”, bāobāo [paw˥˥paw˥˥] in Standard Mandarin, is transcribed bawbaw, while “beer”, píjiǔ [phi˨˥tɕiəw˨˩], is transcribed piɟo.
Kyom·kyo phonology is typically Rgyalrongic. Maximal initial cluster is CCC-: ʒɡroʔ “mouth harp”. The author analyses distinctive, free stress, which coexists with a distinction of presence or absence of glottal stop on the final vowel. Disyllables have hence four suprasegmental possibilities: σˈσ, σˈσʔ, ˈσσ, ˈσσʔ. The Kyom·kyo system is comparable to most other tonal varieties of Rgyalrong, where, however, non-final stress usually neutralizes the final glottal distinction, leaving three possibilities: σσ̂, σσ́ʔ, σ́σ. In contrast, non-final stress preserves glottality in Kyom·kyo: na-ˈa-moʔt-w “She smoked” is [ˈn̪amoʔt̪] in narrow transcription (p. 468). Final stress, default and not transcribed, contrasts with a strong final stress in imperatives (p. 543, p. 592).
The verb morphology in Kyom·kyo is as complex as that of any other Rgyalrong variety. Person marking follows the canonical Rgyalrong pattern, hierarchical with inverse marking. Unlike many Sìtǔ dialects, there is a distinction between direct and inverse in 3 > 3 (non-local) situations, making the inverse a distinct verbal category. When the inverse marker wu-/o- follows an orientational prefix in past perfective or imperative, a portmanteau prefix no-, labelled “attention flow” in this study, precedes or replaces the orientational prefix. The lengthy treatment devoted to this prefix has not completely elucidated how it fits in the verb system.
Stem alternations play a crucial role in the tense-aspect-modality-evidentiality system. However, stem forms are not indicated in the glossary, so readers are not provided with the necessary information to conjugate a verb by themselves. Alongside imperfective negator ma- and prohibitive mə- of clear Sino-Tibetan pedigree, the perfective negator is surprisingly ɟi-, by all appearances an innovation of the softening dialects.
This study provides a useful overview of a softening Sìtǔ dialect and gives ample pointers for future research. The writing is pedagogical and skimmable. The terminological choices are often pleasant (“orientation” is better than “direction”) but not immune from some quibbles (“applicative” for antipassive), especially when motivated by a desire to invent a label vague enough to extensionally cover the whole semantic territory of a category (“prominence”, “attention flow”, “viewpoint”).
References are made whenever necessary to recent literature on Rgyalrongic languages, making the work a useful vade mecum to the field. The grammar is duly appended with text samples, with much to find of Tibetological interest, as well as a glossary. The author will do the community a great service to deposit a collection of annotated audio recordings into a permanent archive, for example, the Pangloss Collection (LACITO, CNRS), as well as follow the grammar with a comprehensive dictionary, in which words are annotated with necessary morphological information.
