1. Anticausative alternation
About 150,000 people speak Gyalrongic languages (Trans-Himalayan/Sino-Tibetan) in north-western Sichuan Province, China.Footnote 1 One valency-reducing operation in Gyalrongic languages is anticausativization, characterized by a detransitivizing P → S shift and completely removing a causal agent.Footnote 2 Anticausative derivations are uniformly characterized across languages as transforming a causal action into a non-causal one by entirely omitting the causal agent. Conversely, passives, which are also P → S derivations, may still include an agent, typically represented as an oblique argument in various languages (Kulikov Reference Kulikov and Son2010; Zúñiga and Kittilä Reference Zúñiga and Kittilä2019; Bahrt Reference Bahrt2021). An example of anticausative derivation can be seen in the Stau verb pair sili “to cause to fall” → zili “fall over (without an agent)”. Anticausativization in Gyalrongic languages is unproductive and fossilized in a regular morphophonological derivational process found in a limited set of verb pairs. In each verb pair, the transitive stem has a voiceless stem initial, and the intransitive stem has a voiced initial. In many of these intransitive stems, there are also nasal preinitials. It has been argued strongly elsewhere that the direction of derivation is from transitive to intransitive due to an ancient *N- prefix originally marking spontaneous/non-volitional derivation (Gates et al. Reference Gates, Honkasalo and Lai2022). The evidence for this directionality claim includes aspiration neutralization (e.g. pʰ and p → b), the lack of a devoicing mechanism (e.g. the sigmatic causative becoming voiced before voiced initials), and recent transitive verb loanwords from Tibetan incorporating a nasal preinitial into their stems to derive anticausatives. The *N- prefix explains why stem initials are voiced, even with stems that no longer have a nasal preinitial reflex.
2. Comparing cognates for “melt”
Many Gyalrongic languages share cognates for the anticausative verb pair “melt”, as displayed in Table 1. At first glance, one might suppose that the Gyalrongic words for “melt” are borrowed from Tibetan <bzhu> or <bzhus> “melt”. However, “melt” in Gyalrongic languages was not borrowed from Tibetan <bzhu> or <bzhus> “melt” for the following reasons. Firstly, and most importantly, when the onset <bzh> has been borrowed into Gyalrongic languages it never yields mtʂ/mdʐ, mtɕ/mdʑ, or mtʃ/mdʒ, but regularly yields vʑ, βʑ, or ʑ, among other reflexes depending on the modern lect; e.g. <bzhugs> “stay, sit” > Stau vʑu “sit (honorific)”, <bzhar> “shave” > Stau vʑɛr “shave, cut”, <bzhogs> “cut” > Japhug βʑoʁ “cut, peel”, <bzhugs> “stay, sit” > Japhug ʑɯɣ “say goodnight”. Secondly, if “melt” is a loanword from <bzhu> or <bzhus>, there is no explanation for a voiceless transitive counterpart in the derivational pair. In other words, there are no identifiable mechanisms for the cause of devoicing and all evidence favours an alternation from voiceless to voiced (see Gates et al. Reference Gates, Honkasalo and Lai2022).
Examples of “melt”
Most transitive forms for “melt” have a labial preinitial: p-, f-, m. Tangut, Bawang, and Puxi Stosde do not have a labial preinitial, but do have a labial medial -w-/-v-. In Pre-Tangut, -w- can be reconstructed to the preinitial *p- (Jacques Reference Jacques2014: 29–30), a labial causative prefix which will be discussed later. It can be generalized that for the word “melt”, languages with a labial preinitial do not have a labial medial, and those without a labial preinitial do have a labial medial. Zongke Stosde is the only language with neither labial preinitial nor medial. Transitive forms have the expected voiceless stem initials, which are voiceless palatalized affricates (palato-alveolar or aveolo-palatal) in West Gyalrongic languages (Tangut, Bawang, Puxi, Zongke, Geshiza, Stau, Bragmda’ rTa’u, and Khroskyabs) and voiceless retroflex affricates in East Gyalrongic (Japhug and Bragbar). The nasal preinitial in the transitive form for Bawang is explained by Yang (Reference Yang2021: 158) as an autive prefix.
For the intransitive forms in Table 1, we find the expected voicing of the stem initial (except for Bragmda’ rTa’u, which will be explained anon). The East Gyalrongic languages have retained the anticausativizing nasal preinitial that descends from the autive *NV- prefix (Jacques Reference Jacques2015, Reference Jacques2021: 18), while the West Gyalrongic languages have lost the nasal preinitial.
3. Jacques’ explanation for n- and not m- in East Gyalrongic
Jacques claims that if ftʂi was the original transitive stem for “melt” in Japhug, we would expect the anticausative alternation to be †mdʐi instead of what we find, which is the anticausative stem ndʐi (Jacques Reference Jacques2021: 920).Footnote 3 However, Jacques does not explicitly mention why we would expect †mdʐi. Although there is an assimilation rule that states that w- becomes m- before nasal (*wN- → mN-, see Jacques Reference Jacques2021: 73, 85), since †mw- is not permitted phonologically in Japhug, it is plausible that there is also a phonological rule that causes w- to become m- after a nasal, which results not only in assimilation but also a merging of the nasal and w, e.g. *Nw- → m-. This later rule, which is perfectly plausible phonologically, would be necessary for the expectation of an anticausative alternation as †mdʐi in Japhug and †mdʐî in Bragbar.
However, the problem remains that we find ndʐi and not †mdʐi. Jacques provides two possible reasons why we find ndʐi as the anticausative form. One solution is that the original root was tʂi, perhaps a labile verb, which became transitive through *wɐ- resulting in ftʂi and intransitive through *N- resulting in ndʐi (Jacques Reference Jacques2021: 851, 920). The f- (/w-/) preinitial may be a descendent of a Proto-Gyalrongic causative prefix *wɐ- (Jacques Reference Jacques2021: 851), which may have descended from an even older Trans-Himalayan/Sino-Tibetan labial causative prefix (Jacques Reference Jacques2019). This hypothesis – that tʂi was originally a labile verb – finds support in the Bragmda’ rTa’u form tɕə, which has an unvoiced onset and an anticausative function. Another possibility, proposed by Jacques (Reference Jacques2021: 920), is that the preinitial f- is part of the original transitive stem, perhaps as *ptʂi (as in Bragbar), and the anticausative form underwent cluster simplification, e.g. *N-ptri → *N-tri → ndʐi.
It is impossible to determine which of the above two hypotheses is correct or if these hypotheses are even falsifiable. Evidence in favour of the first hypothesis would include a Gyalrongic language with a triplet containing a labile verb (e.g. †tʂi, †tɕə, etc.), a causative verb (e.g. †ptʂi, †ftɕə, etc.), and an anticausative verb (e.g. †ndʐi, †dʑə, etc.). Such a triplet has not yet been found.
4. Explaining the variation in West Gyalrongic
Regardless of which hypothesis we adopt, there is still much variation to be explained in West Gyalrongic. There is a bilabial nasal m- preinitial in the Mazur Stau and Geshiza causative form mtɕʰə “melt”, and yet Khanggsar Stau and Bragmda’ rTa’u have the form ftɕʰə. There is a labial medial in Tangut, Bawang, and Puxi Stosde, and there is neither preinitial nor medial in Zongke Stosde.Footnote 4
4.1. Stau, Bragmda’ rTa’u, and Geshiza
The m- in Mazur Stau and Geshiza can be explained as coming from a fossilized nasal autive prefix (as in Bawang ᶰtʃʰwə “melt”), which went before the labial preinitial (causative prefix or not), creating the change *N-*p-*tɕʰə > mtɕʰə (the expected yet unattested phonological change for Japhug). West Gyalrongic languages like Khangsar Stau, Bragmda’ rTa’u, and Wobzi Khroskyabs did not fossilize the autive prefix. The fact that both Geshiza and Mazur Stau have this innovation, while languages much closer to Mazur Stau, like Khanggsar Stau and Bragmda’ rTa’u, do not, suggests that this innovation either happened independently in both languages or that it came about through contact between the two languages.
4.2. Tangut, Bawang, Puxi Stosde: labial causative infixation
An interesting correspondence has been observed by Jacques (Reference Jacques2014: 29–30, Reference Jacques2019) between the labio-dental fricative preinitial f- in Japhug and the labiovelar approximant medial -w- in Tangut, as can be seen in Table 2. Thus, the following sound change has been proposed by Jacques from Proto-Gyalrongic to Tangut: *pC- → *-Cw-.
Tangut -w- from *p- preinitial (adapted from Jacques Reference Jacques2019)
Furthermore, there is evidence that in Tangut there was a causative infix -w-, as displayed in Table 3. Jacques proposes that this causative infix is cognate with the Proto-Gyalrongic causative *p- > *wɐ-.
The causative -w- infix in Tangut (adapted from Jacques Reference Jacques2019)
In Tangut, Bawang, and Puxi Stosde, we find -w-/-v- correspondences, possibly attributable to a labial infix, in the transitive and intransitive forms for “melt”, as seen in Table 1. If Zongke Stosde once had the labial infix, it was later lost through syllabic simplification. The transitive forms correspond to the labial preinitials in Geshiza, Stau, Khroskyabs, Situ, and Japhug. The intransitive forms with the labial infix are unexpected. A solution for the unanticipated labial infix in Tangut, Bawang, and Puxi Stosde intransitive forms is that at an earlier point in these languages’ history, there was double marking (both causative and anticausative) on these forms, which created an indirect causative distinction. In Japhug, there can be causative and anticausative prefixes on the verb “melt” to make a direct causative (1a) and indirect causative (1b) distinction, which also contrasts with the anticausative (1c).
Japhug direct and indirect causative, and anticausative for “melt”
-
1. Direct causative
kʰru nɯ cʰɯ-ftʂi-nɯ tɕe, tɕe nɯnɯ pjɯ-lɤ-nɯ nɯ-ɕti tɕe
pig.iron DEM IPFV-melt-PL LNK LNK DEM IPFV:DOWN-release-PL SENS-be.AFF LNK They melt the pig iron and pour it into (the mould). (Jacques Reference Jacques2021: 928; 25-qraR, 33)
-
2. Indirect causative
ta-mar cʰɤ-sɯɣ-ndʐi-t-a
INDEF.POSS-butter IFR-CAUS-ACAUS:melt-PST:TR-1SG
I let the butter melt (by forgetting it next to a source of heat). (Jacques Reference Jacques2021: 928; elicited)
-
3. Anticausative
jisŋi tɕe lonba to-ndʐi .
today LNK all IFR-ACAUS:melt
Today it [the snow] has melted completely. (Jacques Reference Jacques2021: 1136; conversation, 17 November, 2023)
Thus, it is possible that an indirect causative form marked with a voiced onset and labial infix (Tangut dʑjwi¹ 3956, Bawang dʒwə, Puxi Stosde dʒvəˠ) replaced a voiced onset anticausative form without the labial infix (Tangut *dʑji¹, Bawang *dʒə, Puxi Stosde *dʒəˠ), but adopted a solely anticausative function. In other words, the present anticausative forms in Tangut, Bawang, and Puxi Stosde originally may have been indirect causative forms if a functional distinction between indirect and direct causatives previously existed and was lost.
5. Conclusion
In conclusion, this note has demonstrated evidence for the following arguments:
-
1. The verb pairs for “melt” in Table 1 are apparent cognates within Gyalrongic and not Tibetan loanwords.
-
2. There is a phonological rule that needs to be explicitly stated for Japhug: w- becomes m- after a nasal, which results not only in assimilation but also a merging of the nasal and w, e.g. * Nw- → m-.
-
3. The presence of the preinitial m- in Mazur Stau and Geshiza for the word “melt” can be explained as a fossilized and fused version of an ancient autive prefix *N- leading to the change * N-*p- *tɕʰə > mtɕʰə. This change may have happened independently in each language or through language contact.
-
4. Tangut, Bawang, and Puxi Stosde share a labial medial -w-/-v- for the word “melt”. This may be a labial causative infix corresponding to an ancient labial causative prefix.
