Reconstructions

Note

Also Bontok ákew ‘steal from a house when the residents are out’, Kankanaey ákew ‘steal’, Ifugaw áko, ákaw ‘steal’, Ifugaw (Batad) ákaw ‘steal, esp. of valuable items; do something surreptitiously’, Maranao panekeo ‘steal’, Melanau (Mukah) tikaw ‘theft’, menikaw ‘steal’, t-en-ikaw ‘was stolen by’, Maori whánako, whēnako ‘steal; theft; thief; thievish’. This exceptionally complicated comparison requires extensive comment. The following numbered points concerning first the shape, then the morphology and finally the meaning of *Cakaw and its derivatives should cover the major issues:

(1) All Formosan reflexes of *Cakaw which distinguish *C from *t point unambiguously to *C in this form. Outside Taiwan, PMP *t- (the outcome of the merger of PAn *C and *t) is widely reflected as the stem-initial consonant, but there are a number of irregularities, including the unexplained loss of *t- in some of the languages of northern Luzon and the widespread nasal replacement of *t in other languages. The latter irregularities will be treated below under morphology. Although both Samal (southern Philippines) on the one hand, and the Batak languages and Nias (northern Sumatra) on the other hand, reflect a form with medial prenasalization, I take the prenasalization of *k in this word to be a product of two independent changes.

(2) Milke (1968:149) proposed a doublet *nakaw next to Dempwolff's *ta(ŋ)kaw ‘steal’. Although forms compatible with *nakaw occur in forms such as Kayan nakau ‘steal’, Singhi Land Dayak noku ‘steal’, and Karo Batak naŋko ‘steal’, these items are synchronically derived form stems with initial t. The same cannot, however, be said of Kapampangan náko, Tagalog nákaw ‘steal’ and of widespread reflexes of *nakaw in the Lesser Sundas and central Moluccas. A doublet *nakaw thus seems unavoidable, despite the fact that in some languages forms of the same shape are regual morphological derivatives of a reflex of *takaw.

(3) Although Aklanon panákaw ‘theft’, Ngaju Dayak panakau ‘thievish, stealing inveterately’ and Toba Batak panaŋko ‘thief’ are derived synchronically from underlying disyllables with initial t, Itbayaten panakaw, Manobo (Kalamansig Cotabato) penakaw, Tiruray fenakwa, Kadazan Dusun panakau and reflexes of POc *panako ‘steal’ are independent lexical entries. We might, therefore, posit a third PMP doublet *panakaw next to *takaw and *nakaw, an interpretation which appears to be supported by the reconstructions *panakaw-an and *panakaw-en. However, the evidence for a doublet *panakaw differs from the evidence for a doublet *nakaw in one crucial respect: PMP clearly had a prefix *paŋ- which, among other things, derived agentive nouns, whereas there is little reason to believe that active verbs were derived by simple homorganic nasal substitution (as opposed to prefixation with *maŋ-). The patterning of PMP morphology thus suggests that *panakaw is a morphological derivative of *takaw. Despite this conclusion two serious problems remain: 1. *panakaw presumably meant ‘thief’, yet of the available reflexes only Toba Batak panaŋko supports this interpretation, most languages reflecting *panakaw as a verb or as a non-agentive noun which itself can be affixed to form a verb, 2. reflexes of *panakaw have become independent lexical items in a number of widely distributed languages, yet this has not happened to other morphological derivatives of PMP *takaw.

(4) The morphology of some reflexes of *Cakaw appears superficially to be cognate, but does not stand up to closer inspection. Thus, Tetun hanaʔo ‘rob, steal’ appears at first to reflect *panakaw, but on closer examination appears instead to contain the Tetun causative prefex ha- plus a reflex of *nakaw. Similarly, Paiwan ma-tsakaw ‘be, become stolen’ and Pazeh ma-sakaw ‘steal, rob’ appear at first to be cognate with Sangir ma-tako ‘thievish’, Mongondow mo-takow ‘thievish in nature’, but the Sulawesian forms clearly contain a reflex of *ma- ‘stative’, whereas the Formosan forms do not. Likewise, the similarity of Rotinese manako (ma-nako) and of Kapampangan, Bare'e menako ‘steal’ to the Paiwan and Pazeh forms, or to Isneg mag-tákaw ‘steal’, appears to be a product of convergent innovations in languages that in any case share a common ancestral morphology.

(5) Most challenging of all is the task of interrelating the affixed forms of PAn *Cakaw and PMP *takaw. Wolff (1973) has proposed a convincing system of verbal affixation in PAn which distinguishes independent from future-general action dependent subjunctive modes. The independent mode of verbs includes active, direct passive (DP), local passive (LP) and instrumental passive (IP) voices in both nonpast (-um-, -en, -an, i- respectively), and past tense forms (-inum-, -in-, -in-...-an, i-...-an). At the PAn level the direct available evidence permits us to reconstruct *Cakaw and *C-in-akaw. The former was the stem for verb forms meaning ‘steal’, but when unaffixed probably meant ‘theft’. The latter clearly meant ‘what is stolen, stolen goods’. However, more generally PAn *-in- had two functions: (1) it was a common nominalizer (as in *C-in-akaw); (2) it marked the past tense of all four verbal voices, for the active, LP, and IP in combination with other affixes, but for the DP it occurred with a zero allomorph of the DP suffix *-en, thus acquiring a ‘portmanteau’ function. The forms which have been reconstructed above are only a fragment of what must have been a fuller paradigm, pieces of which are preserved either in the available Formosan evidence or the available MP evidence, but not in both. Taking into account the morphological evidence form other cognate sets we can tentatively posit the following PAn paradigm for *Cakaw:

The active nonpast form *C-um-akaw is reflected in Paiwan ts-m-akaw ‘steal; do secretly’ and the irregular Ifugaw um-áko-áko ‘steal frequently, be a thief’. No reflexes of *C-inum-akaw (Wolff's past active) are known. The nonpast direct passive *Cakaw-en is recorded only from Philippine languages, but must have been present in PAn if *C-in-akaw functioned both as a noun and as the past tense form of the direct passive. Reflexes of *Cakaw-an are known only in WMP languages, where they do not clearly reflect a local passive meaning. This meaning is, however, clearly reflected in Kadazan Dusun panaka-an ‘the place where one steals’ and Mongondow ponakaw-an ‘place where something was stolen or will be stolen’. No past tense form of the local passive, nor any form of the instrumental voice or the future-general action dependent subjunctive mode is known for this form.

(6) The principal changes which the foregoing paradigm underwent in PMP were the innovation of *manakaw and *panakaw. Both *t-um-akaw and *manakaw must have coexisted for some time as markers of the nonpast active voice in PMP, but the latter appears to have dominated the former. It is unclear whether PWMP *mananakaw derives from a form prefixed with *manaŋ-, or from a form prefixed with *maŋ- followed by reduplication of the first stem syllable.

(7) The focal meaning of PAn *Cakaw, PMP *takaw was ‘steal’. In addition to this gloss, however, *Cakaw, *takaw evidently included the meaning ‘do secretly’, as such a sense is reflected in Paiwan, Ngaju Dayak, Rotuman, and perhaps Karo Batak and Motu (in both of which the notion of clandestine or illicit sexual relations is also expressed).