A Quantitative Study of Voice in Malagasy

Abstract

This paper is a quantitative study of the voice system in Malagasy (a Western Austronesian language spoken on Madagascar). We show that nonactive verbs in Malagasy have a very different distribution in texts than nonactive verbs in English, German, and Dutch: they occur far more frequently and they typically present Agent phrases. This we claim reflects the very different role of the voicing system in the grammars of Western Austronesian and Western European languages. Part 1 reviews the voice system of Malagasy, classifying the various voice forms into active vs. nonactive, with the latter divided into passive and circumstantial; part 2 presents the results of our text study; and part 3 draws some conclusions regarding the nature of the voice system in Malagasy.

1. The Voice System of Malagasy.

We first exemplify the Malagasy voice forms. Examples are given in the standard orthography, augmented when helpful by hyphens to indicate morpheme boundaries, "´" to mark main stress, and block parentheses to indicate constituency. Pronounced forms are noted in round parentheses next to (or beneath) their morphemic decompositions (when not simply the concatenation of the latter).

(1)

1. a. [N-i-vídy     akanjo  hoan'ny zaza] i     Vao.
   PAST-ACT-buy    clothes   for'the    child   ART  Vao
   'Vao bought clothes for the child.'

2. b. [No-vidy-ina    i-Vao (novidín'i Vao) hoan'ny zaza]  ny  akanjo.
   PAST-buy-ACT      ART-Vao               for'the    child   the  clothes
   'The clothes were bought by Vao for the child.'

3. c. [N-i-vidy-ananai-Vao (nividiánan'i Vao)     akanjo] ny  zaza.
   PAST-ACT-buy-CIRC ART-Vao                      clothes   the  child
   'The child was bought clothes for by Vao.'

There is massive evidence (Keenan 1972, 1995; Pearson 2000) that the bracketed strings in (1a,b,c) are constituents, called here PREDICATE PHRASES (PredPh). For example, to form the yes-no questions corresponding to (1a,b,c), it suffices to insert the particle ve at the right edge of the PredPh. Insertion internal to the PredPh is ungrammatical.

Another example: the only relativizable position in (1a,b,c) is that of the NP external to the PredPh (henceforth the EXTERNAL ARGUMENT, EA, a usage we take from Pearson 2000). Thus (2a,b,c) are grammatical; relativizing any other NP, as in (3), is not.

(2)

1. a. ny olona (izay) nividy akanjo hoan'ny zaza
   'the person (that) bought clothes for the child'

2. b. ny akanjo (izay) novidin'i Vao hoan'ny zaza
   'the clothes (that) were bought by Vao for the child'

3. c. ny zaza (izay) nividianan'i Vao akanjo
   'the child (that) were bought-for by Vao clothes'

(3) *ny akanjo (izay) nividy hoan'ny zaza i Vao
'the clothes (that) bought for the child Vao'

Reading (3) through active nividy we understand that it is the clothes that bought something. T he r est of ( 3) lacks a Theme a nd p resents an NP, Vao, which lacks a semantic role. Similar ungrammaticality results when attempting to relativize NPs other than the EA from (1b and c). The EA in (1a,b,c) is replaceable with pronominal forms drawn from the following series, which we call NOMINATIVE:

(4) aho 'I'; izahay 'we EXCL'; isika 'we INCL'; ianao 'you SG'; ianareo 'you
PL'; izy '3 SG or PL'

This series contrasts with those in (5a), called here accusative, and (5b), called GENITIVE. (Items are given in the same order as in [4].)

(5)

1. a. ahy; anay; antsika; anao; anareo; azy

2. b. -ko; -nay; -(n)tsika; -nao; -nareo; -ny

The accusative forms replace Patient and Theme arguments internal to the PredPh, as in (1a,c). The genitive forms are used for the Agent phrases in (1b,c) and for possessors in general.

(6)

	ny trano	'the house'	ny trano-ko	'my house'
	ny trano-ny	'his house'	ny trano-nao	'your sg house'
	ny tranon'i Vao	'Vao's house'	ny tranon-dRabe	'Rabe's house'

Note the nasals in the last line, and see Paul (1996b) for the complicated morpho-phonemics involved in binding genitives to their hosts.

Definitions.

Verbs that take...