Reconstructions

Note

This base is exceptionally rich in affix potential. At the same time it has caused considerable confusion to some comparativists. Dempwolff (1938) posited *ka ‘to eat’, *ka-en ‘what is to be eaten, food, nourishment’, on the assumption that *kaen contained a base *ka plus the patient suffix *-en. However, this is a misanalysis, as the comparative evidence supports *kaen ‘to eat’ and *kaen-en ‘be eaten by someone; what is to be eaten, food, nourishment’. Wolff (2010:855-856), on the other hand, posits *kan ‘eat’, but has no explanation for the second vowel (or for vowel length) in widely separated languages such as Bunun, Amis, Tagalog, Hanunóo, various Bisayan and Manobo languages, Karo Batak, Sangir, Tondano, Mongondow, Totoli, Boano, Banggai and Proto-Bungku-Tolaki. Moreover, following a prior suggestion by Dahl he proposes *isekan ‘fish’ as a morphological derivative, although the comparative evidence points clearly to PAn *Sikan, with last-syllable *a, not *e (cp. Bunun iskan ‘fish’, vs. kaʔun ‘eat’).

The simplest explanation for these various confusions is that the base *kaen frequently contracted to a monosyllable because schwa next to another vowel is particularly weak, and tends to be absorbed. This is especially the case with affixed forms, and an examination of almost any dictionary of a contemporary language shows that reflexes of the base *kaen rarely occur unaffixed, and that when they do they are more likely to retain the medial vowel sequence, as with Aklanon káon ‘to eat’, but kán-on ‘rice, staple food’ (not **káon-on). Where *kaen has contracted to a monosyllable several languages independently have restored the favored disyllabism of content words through epenthesis, as in Puyuma əkan, Yami (a)kan, Bontok ek-ekan, KADAD akan, or Maloh aŋkan ‘eat’ (the latter with subsequent sporadic medial prenasalization).

The rich morphology of this base is ubiquitous in Taiwan, the Philippines, northern Borneo and northern Sulawesi, but contracts radically in languages further to the south, where only paradigmatically dissociated elements remain, as in the Uma Juman dialect of Kayan (Blust 1977:63-64), where kani, kanən, kinan, kuman, makan, pakan, pəŋanan and pəŋuman are no longer anchored to a common base, and where most of the morphology in this fragmented paradigm is fossilized (-ən, -in-, ma-, pa-, pəŋ- and –an are otherwise unknown in the language, and although a reflex of *-um- continues to function as the marker of active voice, its form is –əm-, not –um-). Similarly, for Ngaju Dayak Hardeland (1859:276) cites kuman as a dictionary entry, and notes that “Das Passiv von kuman is unregelmässig, heisst: kinan, gegessen, getrogen werden, (sind essbar).”