This post continues an earlier one (here) where a name of the rice plant, pre-reconstructible as #am, was shown to occur in languages of three distinct branches of the non-Sinitic part of Sino-Tibetan.
In addition, I cited a related pair of forms in Chepang: ʔamʰ ‘cooked rice’ and yam ‘rice plant’, noting the parallel alternation in Chepang ʔat and yat, both ‘one’. Prothesis of y- in yam and yat calls for an explanation.
I also noted related forms where #am is apparently preceded by an m- formative: Jingpo mam³³ ‘rice’ (paddy) and Thulung (Allen 1975), a Kiranti language, mam ‘grain of rice remaining unhusked after milling’.
It is possible that this m- formative is the remnant of a morpheme cognate with Chinese 米 *(C.)mˤ[e]jʔ > mejX > mǐ ‘grains of any cereal, dehusked and polished’. In the STEDT website, Matisoff gives this cognate set, with the gloss RICE/PADDY, but, as in Chinese, the same morpheme can be used of millet, e.g. Garo (Burling) mi-si-mi ‘millet’, Gyarong (Ma’erkang) sməi khri ‘millet’. Thus Jingpo mam³³ ‘rice’ (paddy) would be from #mej-am ‘grain(s) of rice/rice in grains’; and Thulung mam ‘grain of rice remaining unhusked after milling’ would also be from ‘grain of rice’, without the need for a wide-ranging semantic shift.
This idea has the additional potential of explaining the prothesis of y- in the Chepang form cited above, if we suppose an evolution of the kind of #mej-am > mej-jam > jam.
As to the morpheme #am ‘rice plant’ itself, it seems likely that it arose out of a verb ‘to eat’, for which the STEDT has this cognate set. For a parallel, the Proto-Austronesian verb *kaen ‘to eat’ has a patient noun derivative with *-en: *kaen-en ‘food’ which evolves to ‘rice’ in certain Philippine and Bornean languages where rice is the main staple.