Concerning ka-, an Overlooked Marker of Verbal Derivation in Formosan Languages

Abstract

Blust (1999) shows that in Pazeh, a moribund Formosan language, causativized dynamic verbs are morphologically marked by pa-, while stative verbs are prefixed by paka-. He argues that this language provides "critical evidence" for distinguishing the two grammatically conditioned variants *pa- and *paka- in PAN. Blust's reconstruction overshadows, however, a morphological pattern found across many Formosan languages. We demonstrate that (i) in many Formosan languages where the causative pa-/paka- alternation is found, paka- is better analyzed as a bimorphemic prefix, whereby pa- indicates causativity and ka- stativity, and that (ii) the reanalysis of *pa-ka- as paka- and/or the replacement of this form in stative verbs must have taken place at a later stage in the history of the Austronesian languages. This analysis accounts for the synchronic variation across the Austronesian languages, and the Formosan languages in particular, with respect to this matter.

1. Introduction.

In a recent paper, Blust (1999) shows that, when causativized, dynamic verbs in Pazeh are morphologically marked by pa-, while stative verbs are prefixed by paka-.² This contrast is illustrated in (1a) and (1b) respectively.

(1) Pazeh (Blust 1999:347-348)³

1. a. mu-dader  'choke on something'  pa-dader     'cause to choke'

2. b. ma-ngesen 'afraid'           paka-ngesen  'to frighten'

Blust argues that Pazeh provides "critical evidence" for distinguishing the two grammatically conditioned variants *pa- and *paka- in PAN:⁴ "*pa- formed the causative of dynamic verbs and *paka- the causative of stative verbs" (Blust 1999:356-357). He further supports his claim by showing that these two prefixes are widely distributed across the Austronesian languages. Table 1, taken from Blust (1999:357), gives examples that reflect the *pa-/paka- alternation.

In our view, Blust's (1999) reconstruction overshadows a morphological pattern found across many Formosan and extra-Formosan languages, where dynamic and stative verbs differ from one another morphologically, depending on whether they occur in their "finite" forms or "nonfinite" forms. It is the contention of this paper that in most Formosan languages where the causative pa-/paka- alternation is found, there is reason to believe that paka- should be analyzed as a bimorphemic prefix whereby pa- indicates causativity and ka- stativity.⁵ It is shown that ka-

Table 1.

Sample Reflexes of PAN *Pa- and *Paka- 'Causative' in Austronesian Languages*

represents the counterpart of (stative) ma- (~Ø). Both are stem-forming affixes that appear on stative verbs: ka- occurs exclusively in nonfinite verb stems, ma- (~Ø) in finite verb stems. It is further suggested that (i) *pa- and *ka-,⁶ along with (stative) *ma-, should be reconstructed as distinct morphemes in PAN, and (ii) the reanalysis of *pa-ka- as paka- and/or the replacement of this form in stative verbs must have taken place at a later stage of the history of the Austronesian languages. This analysis accounts for the synchronic variations across the Austronesian and the Formosan languages in particular.

In section 2, we first review the evidence put forward by Blust (1999) for distinguishing two causative forms (cf. pa- vs. paka-) in Pazeh. In section 3, we argue that paka- should be treated as a bimorphemic prefix in Pazeh and provide external evidence to support this claim. The languages⁷ on which our analysis is founded include Atayal (Mayrinax), Seediq (Paran and Truku), Rukai (Mantauran) and (Southern) Paiwan.⁸ In the last section, we present data on (extra-) Formosan languages in which pa-ka- has been reanalyzed (Saisiyat), replaced (Bunun and Thao), or neutralized (Yami) or where the dynamic/stative distinction is not overtly marked on the morphological level (Tsou) and we hypothesize that in many such languages, pa-ka- must have later been reanalyzed as paka-.

2. Review of Blust's (1999) Analysis.

Blust shows that dynamic verbs are prefixed with different focus AF affixes, mu-, me-, mi-, ma₁- and m₁-, but that the occurrence of these affixes is conditioned by two factors. The first is phonological; the second, semantic. Phonologically, me- can be treated as an allomorph...