- The Numeral Confix *i- -(e)n
In his 1981 paper on Austronesian numerals, Otto Dahl adduced evidence from Paiwan, Ma'anyan, and Malagasy for positing a numeral confix meaning 'so and so many days', which he reconstructed as *(ma)ka- -(a)N. In this squib, I discuss the evidence for reconstructing another numeral confix of the same meaning but having the form *i- -(e)n. As far as I know, reflexes of this other numeral confix are limited to the Bali-Sasak languages and five language groups of Sulawesi. This distribution suggests that *i- -(e)n was a local innovation within Western Malayo-Polynesian.
1. *itelun AND HIGHER FORMS.
I begin with the evidence for reconstructing a temporal adverb of the form *itelun. This form referred to a period of time, 'in three days', or—if one believes the ancients were more likely to have counted the number of nights passing by—perhaps rather 'after three nights'. Both possibilities are reflected in the glosses for present-day forms supplied by my sources, which I have not attempted to harmonize. As one may note, in many cases another morpheme is present that disambiguates time in the past from time in the future. Owing to the considerable variability of this additional element, however, no such morpheme can be reconstructed at the same level as *itelun.
Balantak: r-itulun 'three days ago', itulun 'three days hence'
Dondo/Tialo: itoun-aa 'three days ago', me-itoun 'three days hence'
Totoli: ko-itolom 'three days ago', itolom 'three days hence'
Napu: na-italu 'three days hence'
Moma: na-itolu 'three days hence'
Pamona (Sigi dialect): itolu 'three days ago', ne-itolu 'three days hence'
Wolio: na-italu 'third day from now' [End Page 167]
Tolaki: wingi itolu 'memorial evening three days after death' (wingi = 'night')
Padoe: itolu 'three nights'
Mori Bawah: itolu 'after three nights'
Moronene: n-itolu 'three days ago', itolu 'in three days'
Muna: n-etolu-mo 'three days ago', na-etolu 'in three days'
Balinese: itelun 'three days ago', buin telun 'in three days' (buin = 'again, yet to come')
Present-day Kaili-Pamona, Wotu-Wolio, and Southeastern Celebic languages have for the most part lost final consonants, including all nasals; however, the reconstruction of final *-n is solidly based on the tomini-tolitoli, Balantak, and Balinese forms, in which languages final consonants were retained. Were this all, *itelun would make for a run-of-the-mill reconstruction, with perhaps its only notable feature being the distribution of its reflexes, which is highly skewed toward the island of Sulawesi. Other facets, however, make this an interesting case in historical reconstruction.
First, there is clearly both a semantic and formal connection between *itelun 'period of three days' and the well-known Proto-Austronesian reconstruction *telu 'three', by which we may consider the temporal adverb to be a complex form to be parsed as *i-telu-n. Although the *-n suffix is reminiscent of some kind of genitive marker or ligature, all present-day languages suggest that *itelun was an independent form, and was not required to be followed by a morpheme meaning 'day' or 'night'. Balinese presents an interesting case in this respect, having preserved the form itelun 'three days ago' next to paraphrastic expressions like di petang dina 'four days ago' and the such. For that matter, in Balinese, itelun also contrasts formally with telung, the regular form of the numeral 'three' used before nouns or classifiers. In present-day Tolaki, itolu is preserved only in the expression wingi itolu 'memorial evening three days after death', where not only does wingi 'night' precede itolu, but it furthermore appears that the phrase did not originally mean 'three nights' but rather 'the night of the third (day)'. Although paraphrastic constuctions are used in Mori Bawah to express 'after one night' and 'after two nights', respectively, asa wongi and rua wongi, the equivalent expression for 'after three nights' is simply itolu, not *itolu wongi. Similar observations can be made, as far as I know, for all the other languages given above that reflect *itelun.
Second, as mentioned at the beginning, *itelun did...