Вы находитесь на странице: 1из 35

Taba as a 'split-O' language: applicatives in a split-S system

John Bowden Linguistics Department Research School of Pacific and Asian Studies Australian National University John.Bowden@anu.edu.au 1. Introduction Taba is a mixed accusative / split-S Austronesian language spoken in eastern Indonesia which has very productive applicativisation. The major concern of this paper is to explore the characteristics of applicativisation in a language with relatively semantically transparent marking of core arguments. In accusative and ergative constructions, the differences in the semantic roles of the single arguments of intransitive clauses are neutralised insofar as their morphosyntactic expression is concerned. In an accusative construction intransitive S patterns with with transitive A whether it is actor or undergoer; in an ergative construction S patterns with transitive O whether actor or undergoer.[1] In split-S constructions, no semantic neutralisation occurs: actors occur as Actors and undergoers occur as Undergoers whether they appear with transitive or intransitive verbs. Dryer (1986) discusses the distinction between direct/indirect and primary/secondary object constructions in ditransitive clause and argues that the direct - primary object distinction is parallel to that between ergativity and accusativity. In a primary object construction, beneficiaries and recipients (primary objects) are marked in the same way as simple transitive O while patients and themes (secondary objects) are treated differently. In a direct object construction the patient or theme is marked like simple transitive O while recipients and beneficiaries are marked differently. In transitive clauses from languages of each type, semantic differences between the different object types found in ditransitives are neutralised, in much the same way that semantic differences between the arguments of transitive clauses are neutralised for the intransitive subjects of ergative and accusative constructions. It is appparent that there is a logical gap in Dryer's (1986) framework: is it possible for a language to show consistent differences in the way that object-like arguments with different semantic roles are expressed in all multivalent clause types? In this paper I argue that Taba does exhibit consistent treatment of many undergoers according to the exact nature of their semantic role, no matter what kind of clause they occur in. The Taba data suggests that the logical gap in Dryer's framework has been filled in at least this language and that a third kind of patterning needs to be added to Dryers' typology. This third type parallels split-S patterning in transitive and intransitive clauses and might usefully be referred to as 'split-O' patterning. 1.1. The Taba language and its speakers

Taba (also known as East Makian or Makian Dalam) is a South Halmahera Austronesian language spoken in North Maluku province, Indonesia by about 30,000 people. Contemporary Taba, especially as it is spoken by younger people, is increasingly affected by the local lingua franca, North Moluccan Malay, a variety of the national language, Bahasa Indonesia. The long term prognosis for the survival of the language is not good. Taba is predominantly a headmarking language and has some typological word order peculiarities such as the existence of both prepositions and a postposition. Such peculiarities have presumably come about through extensive contact with non-Austronesian languages which are also spoken in the region. The language also has speech levels in some areas of vocabulary which speakers characterise as alus 'refined', biasa 'normal' and kasar 'course'. Like many Austronesian languages, Taba has a number of 'precategorial roots' which do not belong to any word class until a word form has been derived from them in some way. Taba also has very productive applicative affixation. The major source on the Taba language to have appeared so far is Bowden (1997). 1.2. Structure of the paper Section 2 provides some background on Taba morphosyntax, looking at the nature of underived bivalent and monovalent clauses in Taba, and focussing on the phenomenon of split intransitivity. We will show that Taba is a mixed accusative / split-S language, and give an account of how the semantics of an argument predicts quite clearly whether or not it will be realised as a morphosyntactic Actor in Taba. Section three gives a detailed description of applicativisation in Taba. In section 4 we review Dryer's (1986) discussion of ditransitive argument realisation before developing a detailed typology of non-Actor arguments in Taba, arguing that Taba can be characterised not just as a mixed accusative / split-S language, but also as a mixed 'primary object' / 'split object' language. Section 5 provides a conclusion and makes some observations about implications for linguistic theory that the Taba data suggest. 2. Split intransitivity in Taba In this section, we begin by reviewing the morphosyntax of argument realisation in transitive and intransitive clauses and then discuss the semantics of Actors and Undergoers before contrasting split-S languages of the Taba sort with ergative and accusative languages. 2.1. Morphosyntactic realisation of arguments In Taba, the single arguments of some intransitive verbs are morphosyntactically realised in the same way as the more agentive arguments of transitive verbs while the single arguments of other intransitive verbs are realised in the same way as the more patient-like arguments of transitive verbs. Taba allows free ellipsis of arguments that are readily retrievable from context, so obligatoriness of arguments cannot be used as diagnostic of a verb's valence. Transitive verbs are defined as verbs which can take two noun phrases as direct arguments, neither of these noun phrases being able to occur with an adposition. Transitive verbs include the sorts of verbs which generally show a high degree of transitivity cross-linguistically such as 'hit' and 'kill' etc. Taba also has a

class of what might be called 'semi-transitive' verbs, which optionally allow an argument to be marked adpositionally. 2.1.1 Morphological marking of arguments in underived mono- and bivalent clauses Transitive verbs may appear with two arguments: an Actor, which is obligatorily crossreferenced on the verb by a proclitic in Taba, and an Undergoer which is not cross-referenced. Example (1) shows a simple transitive clause.[2]
(1) Basic transitive clause: Actor cross-referenced by proclitic; Undergoer not cross-referenced Oci nwet yak. Oci n=wet yak Oci 3sg=hit 1sg 'Oci hit me.'

Since ellipsis of readily retrievable referents is always possible, neither Oci (person's name) nor yak 'me' needs to be overtly mentioned, but whether or not these elements occur crossreferencing of the Actor is always obligatory. The single arguments of intransitive clauses occur in different ways, depending on the semantics of the verb with which they occur, and also on the animacy of the argument involved. In (2) below, an intransitive clause, the single argument is cross-referenced on the verb in just the same way as the Actor in (1).
(2) Actor intransitive ('unergative) clause: Actor cross-referenced by proclitic Iswan Iswan Iswan 'Iswan nalhod. n=alhod 3sg=run is running.'

In (3), another intransitive clause, the single argument niwi ni sapo da 'that coconut fruit', just like the Undergoer in (1), is not cross-referenced on the verb.
(3) Undergoer intransitive ('unaccusative') clause: Undergoer not cross-referenced on verb Niwi ni sapo da mtat do. coconut 3sg.POSS fruit DIST fall REAL 'The coconut has fallen.'

In terms of surface morphology in intransitive and canonical transitive clauses then, there are two distinct categories of intransitive arguments. In Taba, Actors are defined

morphosyntactically as those arguments which are obligatorily cross-referenced on verbs and Undergoers as arguments which are not cross-referenced on verbs. While the terms 'actor' and 'undergoer' have been borrowed from Foley and Van Valin's (1984) Role and Reference Grammar (RRG), the terms as used here are not strictly equivalent to the terms as used by Foley and Van Valin. In RRG, actor and undergoer are universal semantic 'macro-roles' which may be realised in a variety of different ways in different languages and/or constructions within languages. In this paper, I use the capitalised terms 'Actor' and 'Undergoer' to refer to morphosyntactic categories in Taba grammar, and the terms 'actor' and 'undergoer' (all lower case) to refer to macro-roles. Even if Taba's Actor and Undergoer do not strictly correspond to the RRG categories of actor and undergoer, there are, nevertheless, quite semantically transparent ways of assigning arguments to one or other of the categories. Intransitive verbs in Taba, then, are subcategorised as belonging to one of two distinct types: 'Actor intransitives' or unergatives (which take an Actor argument) and 'Undergoer intransitives' or unaccusatives (which take an Undergoer argument). A further bivalent clause type, that of 'semi-transitive' verbs can also be identified. These are distinguished by taking an Actor argument (which is cross-referenced by a proclitic) and a further argument which, like the Undergoers we have seen so far is never cross-referenced on the verb, but which, unlike the Undergoers we have so far seen, may or may not be marked adpositionally. Underived 'semi-transitives' are all motion verbs such as sung 'enter' which, in addition to an Actor, also allow locative arguments to occur, either with or without the locative postposition li. Semi-transitives derived by applicativisation also include non-motion verbs but discussion of these is left until 3.3. The motion verb sung 'enter' is illustrated in (4) below. In (4a) the locative goal um 'house' occurs with no postposition, while in (4b) it occurs marked with li 'LOC'. Although use of adpositional marking is generally optional with these kinds of arguments, it becomes obligatory when the locative goal serves as the head of a relative clause as illustrated in (4c).
(4) Semi-transitive verb: Actor cross-referenced on verb; Undergoer not cross-referenced. a. Undergoer receives no adpositional marking Yanti ncung um yanti n=sung um Yanti 3sg=enter house 'Yanti entered the house.' b. Locative Undergoer marked by posposition li Yanti ncung um li yanti n=sung um li Yanti 3sg=enter house LOC 'Yanti entered the house.' c. Obligatory adposition with relativised semi-transitive Undergoer Rauf nani um ya Yanti ncung li

Rauf n=hani um ya Yanti n=sung li Rauf 3sg=own house DEM Yanti 3sg=enter LOC 'Rauf owns the house that Yanti entered.'

The locative goals of constructions like that illustrated in (4) above are somewhat difficult to categorise in terms of the 'core' / 'argument' vs. 'oblique' / 'adjunct' distinction that is often made in linguistics. It does not make much sense to see the NP um 'house' as a good 'core' constituent, since it can be marked adpositionally. On the other hand, it is difficult to see the locative goal as strictly an adjunct either, since the verb licenses it to occur without an adposition.[3] I will label constituents such as those illustrated above as 'Remote Undergoers', to capture the idea that it is neither a truly good core nor oblique constituent. 2.1.2 Word order in underived mono- and bivalent clauses A few characteristic word ordering relationships between core arguments in basic transitive and intransitive clauses will be mentioned here. If core arguments are expressed pronominally, Actors always precede the verbs they are subcategorised by and Undergoers always follow, no matter whether the verb be transitive or intransitive.
(5) Transitive verb: pronominal Actor precedes verb, pronominal Undergoer follows I i 3sg 'He nwet yak n=wet yak 3sg=hit 1sg hit me.'

(6) Actor intransitive verb: pronominal Actor precedes verb I nalhod i n=alhod 3sg 3sg=run 'He's running.' (7) Undergoer intransitive verb: pronominal Undergoer follows verb Mot i die 3sg 'It died.' (of an animal)

If only one argument is represented by a full NP, that argument precedes the verb, whether the verb be transitive, Actor intransitive or Undergoer intransitive, and also whether the argument involved is Actor or Undergoer.
(8) Transitive verb: full NP Actor precedes verb, pronominal Undergoer follows

Oci nwet yak Oci n=wet yak Oci 3sg=hit 1sg 'Oci hit me.' (9) Transitive verb: full NP Undergoer precedes verb, pronominal Actor occurs between full NP Undergoer and verb Mon da mon da man DIST 'You hit meu hwet meu h=wet 2pl 2pl=hit that man.'

(10) Actor intransitive: full NP Actor precedes verb Iswan nayok Iswan n=ayok Iswan 3sg=cry 'Iswan is crying.' (11) Undergoer intransitive: full NP precedes verb Ubang da mlongan fence DIST be.long 'The fence is long.'

If two full NP's are found within a clause, the Actor precedes and the Undergoer follows the verb. Such clauses are quite rarely encountered in natural discourse.
(12) Transitive verb: full NP Actor precedes verb, full NP Undergoer follows verb Oci Oci Oci 'Oci nwet Iswan n=wet Iswan 3sg=hit Iswan hit Iswan.'

2.2. Semantics of Actors and Undergoers Whenever the single argument of an intransitive verb is the instigator of an action described by the verb it will be treated as morphosyntactic Actor. The term 'instigator' is used here in the sense of Van Valin & Wilkins (1996: 317) who characterise instigators as 'effectors which are capable of independent motion and action, and ... not subject to the control of another effector, animate or inanimate'. Such instigators may include true agents, but instruments are excluded from the category because they are subject to the control of a higher effector. Forces such as earthquakes, storms, etc. are properly seen as instigators since they are not subject to external control as instruments are. This is illustrated in (13) where the instigator taplod ni laylu 'tidal wave' is cross-referenced as Actor.

(13) Instigator = Actor Taplod ni laylu nwom lawe taplod ni laylu n=wom la-we eruption POSS wave 3sg=come sea-at 'The tidal wave came seawards.'

The single argument of an intransitive verb which refers to a human will occur as morphosyntactic Actor, whether that argument is an instigator or not.[4] In (14) below, the quality of tallness, when ascribed to a human being, is expressed by a verb which has the Actor cross-referenced on it by the 3sg proclitic n=. The Actor intransitive here has been derived from the Undergoer intransitive root mlongan 'to be tall/long'.
(14) Human = Actor (whether instigator or not) Namlongan. n=ha-mlongan 3sg=CAUS-be.long 'He is tall.'

In (15) the underived Undergoer intransitive mlongan is used to ascribe the quality of length to a fence.
(15) Non-human non-instigator = Undergoer Ubang da mlongan. fence DIST be.long 'That fence is long.'

Note that the stipulution that human arguments occur as Actors only applies to intransitive, and not to transitive verbs. Taba transitive verbs always have an instigator, and selection of an instigator as Actor of a transitive verb precludes any non-instigating human from occupying the Actor slot, whether the instigator argument refers to a human or not, as illustrated in (16) below.
(16) Human = Undergoer of transitive verb when another instigator exists Taplod da taplod da eruption DIST 'The eruption npun manusia te n=pun manusia te 3sg=kill people NEG didn't kill people.'

To briefly recapitulate what we have discussed so far then, two distinct categories of arguments are found in canonical transitive and intransitive clauses. These have been labelled the Actor and the Undergoer. Although these labels have been defined on morphosyntactic grounds, a clear semantic basis for assignment to the Actor category has been shown. We have also seen that there are at least two distinct kinds of Undergoers: simple Undergoers as illustrated in most of

our examples so far, and 'remote Undergoers' which were exemplified in (4) above. Finer details about the entailments of 'Undergoer-hood' will be addressed in section 4, after an overview of applicative derivation has been given in section 3. 2.3. Split-S patterning in typological perspective In this section (following Dixon, 1994) I adopt the labels S, A, and O to refer to the sole argument of an intransitive clause, to the more agentive argument of a transitive clause, and to the more patient-like argument of a transitive verb respectively. The Taba system for marking of core arguments just outlined above can be characterised as a mixed accusative / split-S systen. When the sole argument of an intransitive verb refers to a human the resultant pattern is an accusative one: S patterns with A no matter what the actual semantic role of the human referring single argument of the intransitive. When the sole argument refers to anything other than a human, however, the resultant patterning is of a split-S type: intransitive Actors (those labelled SA by Dixon) pattern with transitive Actors (Dixon's A) and intransitive Undergoers (Dixon's SO) pattern with transitive Undergoers (his O). Accusative, ergative, and split-S patterning is represented schematically in figure 1.

In accusative and ergative constructions, semantic differences between the different types of arguments found in intransitive clauses are neutralised insofar as the morphosyntactic expression of these arguments is concerned. In an accusative construction, the intransitive argument patterns morphosyntactically with transitive A and occurs in the nominative case whatever its semantic role. In an ergative construction, the intransitive argument patterns morphosyntactically with transitive O in the absolutive case, again whatever its semantic role. In split-S constructions, no such neutralisation occurs: Actors gain morphosyntactic expression as Actors whether in transitive or intransitive clauses and Undergoers are expressed as Undergoers in both transitive and intransitive clauses as well. 3. Applicative constructions in Taba We define applicatives in Taba as constructions in which a derived verb has a valence one higher than the stem from which it was derived, and in which the added argument is not crossreferenced on the verb as an Actor by a proclitic.[5]

Many of the applied arguments introduced by applicativisation have similar characteristics to the non-Actor arguments found with semi-transitive verbs, since they may optionally be marked by an adposition. While this section of the paper is simpy concerned with describing applicativebased constructions, explicit parallels between these non-Actor arguments will be drawn in section 4. Taba has two applicative affixes which derive verbs with added non-Actor arguments. Applied arguments can have a variety of different semantic roles. The affixes, and a rough characterisation of the semantic roles of the applied arguments they can introduce are:[6] -Vk[7] instrument, companion, recipient, theme, stimulus of emotion -o location, stimulus of emotion

One semantic role that is conspicuous by its absence from this listing, given its very common occurrence in similar lists of potentially applicativised roles in other languages, is that of beneficiary. In fact, there are no purely benefactive arguments in indigenous Taba whatsoever, whether in applicative derived clauses or in any other kind of clause. Many younger speakers, however, use the borrowed North Moluccan Malay preposition untuk to license beneficiaries. Using purely indigenous Taba linguistic resources, the exact nature of the benefit bestowed on someone must be spelled out in some detail in a subordinate resultative clause. To translate something such as 'I sang a song for her', one would have to use something like the equivalent of 'I sang a song so she could listen to it'; to translate 'I cooked a meal for her' one would have to say 'I cooked a meal so that she could eat it' or, using a possessive construction, 'I cooked her meal'. Note however, that in distinction to many other eastern Austronesian languages such as Saliba (Margetts, 1998) where possessive morphology is also used for marking benefactive participants, in Taba possessive marking can only be used when the putative beneficiary really does end up with something in his / her possession. In section 2 we encountered four different subcategories of Taba verbs: Actor intransitives, Undergoer intransitives, caninical transitives and semi-transitives. Any of these kinds of verbs may act as the stems for applicative formation, deriving forms given the labels sketched out in figure 2.

stem transitive

derived form 'close' ditransitive 'remote' ditransitive semi-transitive 'remote' ditransitive Actor intransitive transitive semi-transitive Undergoer intransitive non-Actor bivalent verb

Figure 2. Input and output verb classes with applicativisation

It can be seen that figure 2 lists two different kinds of ditransitive verbs derived by applicativisation. The differences between the two types of ditransitives will become apparent in the following sections where we discuss applicative derivation as it occurs with each kind of stem. 3.1. Applicative derivation with transitive stems This is no doubt the most common and most productive of the Taba applicative formation processes. All Taba ditransitives are formed through applicativisation: there are neither root ditransitive verbs nor causative derived ditransitives. Ditransitives may be formed with either the -Vk or the -o suffix. Each suffix will be discussed in turn. In ditransitive derivation, the -Vk suffix derives verbs with extra arguments having a variety of semantic roles, depending on the meaning of the transitive stem to which they are attached: instrument, recipient, or theme.[8] Translation equivalents of English 'give' are derived by applicative suffixation. There are two such verbs, and each is derived in a distinct way. Otik, illustrated in (17) below, is derived by adding an applied recipient to a transitive root having a patient or theme as its Undergoer. In this case, the applied argument yak 'me' occurs immediately post-verbally, and the underived Undergoer, yan 'fish', occurs after that.
(17) Derivation of ditransitive otik 'give' by applicativisation of transitive root with Actor and theme arguments a. Banda Banda Banda 'Banda not yan bakan. n=ot yan bakan 3sg=catch fish be.big caught a big fish.'

b. Banda notik yak yan. Banda n=ot-ik yak yan Banda 3sg=catch-APPL 1sg fish 'Banda gave me some fish.'

The other Taba verb which can be translated into English as 'to give', hatadak, illustrated in (18) below, works differently. In example (18a) the transitive stem hatada, roughly glossable as 'to present', has a recipient as Undergoer. The derived ditransitive form shown in (18b) has an applied theme. Here, just as in (17b), the theme follows the recipient, but this time, the applied argument follows the underived Undergoer.
(18) Derivation of ditransitive hatadak 'give' by applicativisation of transitive root with Actor and recipient arguments a. Latada yak. l=ha-tada yak 3pl=CAUS-presentation 1sg 'He presented me (with something)'

b. Latadak l=ha-tada-Vk 3pl=CAUS-presentation-APPL 'He gave you lots of money

meu pipis lloci pa ne? meu pipis lloci pa ne 2pl money much or PROX didn't he?'

The two 'give' verbs illustrated above can be thought of as appropriate for different speech levels: otik 'give-1' as biasa 'normal' and hatadak 'give-2' as alus 'refined'. Taba speakers consistently explain the difference between the two forms with respect to politeness. [9] The two verbs are typically used to refer to different kinds of giving events which have differently conceived sociocultural consequences. Hatadak is more appropriately used when whatever is transferred is viewed as an important gift, an account of which must be kept and repaid sometime. On the other hand, otik is more appropriate for everyday transfers of things for which a close accounting of is unnecessary. As far as example (17) is concerned, for example, it is normally expected in Taba culture that anyone with a reasonable quantity of fish or surplus vegetables etc. simply gives them to others who are not so well endowed. The gift of large sums of money, however, as in (18) is not an everyday occurence in the same way as the passing out of surplus fish.[10] The derived ditransitive verbs just illustrated are of the type which were labelled 'close' ditransitives in figure 2. In close ditransitives, neither of the non-Actor arguments can be marked adpositionally, but instruments and locations may be marked by adpositions in 'remote ditransitives'. We can note here that close ditransitives are used in constructions somewhat reminiscent of the 'double object' constructions discussed by Dryer (1986), in which recipients are treated as 'primary objects' and themes are treated as 'secondary objects'. The Taba equivalents of these grammatical relations are provisionally labelled 'primary' and 'secondary' Undergoers. While the verbs of transfer of possession just discussed are certainly quite common tokens in Taba disourse, applicativisation with the -Vk suffix is at its most productive in licensing applied instruments. Some examples illustrating the licensing of an instrument are given in (19) and (20). In all of these examples the applied argument occurs after the underived Undergoer. The underived Undergoer retains its immediately post-verbal position.
(19) Applicativisation of transitive verb with -Vk licenses applied instrument in ditransitive derived form a. Oci nliko Oci n=liko Oci 3sg=tread.on 'Oci trod on the manik. manik chicken chicken.' sapatu. sapatu shoe his shoe.'

b. Oci nlikok manik Oci n=liko-Vk manik Oci 3sg=tread.on-APPL chicken 'Oci trod on the chicken with

(20) Applicativisation of transitive verb with -Vk licenses applied instrument in ditransitive derived form a. Yak kgoras nik kumu.

yak k=goras ni-k kumu 1sg 1sg=shave POSS-1sg beard 'I'm shaving (my beard).' b. Nggorcak kapaya kobit. n=goras-ak kapaya kobit 3sg=shave(take.seeds.out)-APPL pawpaw knife 'He shaved (i.e. took the seeds out of) the pawpaw with a knife.'

Example (21) further illustrates the addition of an applied instrument with the verb pun 'to kill'.
(21) Applicativisation of transitive verb with -Vk licenses applied instrument in ditransitive derived form a. Npun babang da. n=pun babang da 3sg=kill moth DIST 'She's killing the moth.' b. Ahmad npunak kolay peda. Ahmad n=pun-Vk kolay peda Ahmad 3sg=kill-APPL snake machete 'Ahmad killed the snake with a machete.'

Example (22) illustrates the verb pun 'to kill' being used metaphorically. The applied argument catur 'chess' is again an instrument, this time the instrument by means of which the metaphorical killing was brought about.
(22) Instrumental applicativisation with metaphorical use of transitive root pun 'kill' I i 3sg 'He npunak yak catur. n=pun-ak yak catur 3sg=kill-APPL 1sg chess killed me at chess.' [lit. 'He (metaphorically) killed me by means of chess.']

The suffix -o may also be used to derive ditransitive verbs from simple transitives but its use is not as common as that of -Vk. The -o suffix licenses applied locative arguments. In (23) below, the applied argument is a locative goal.
(23) Applicativisation of transitive verb with -o licenses applied locative argument a. Yak kgoras kapaya yak k=goras kapaya 1sg 1sg=scrape papaya 'I'm scraping out the b. Yak kgorco ni kowo. ni kowo 3sg.POSS seed papaya seeds.' kowo bbuk.

kapaya ni

yak k=goras-o kapaya ni kowo bbuk 1sg 1sg=scrape-APPL papaya 3sg.POSS seed book 'I'm scraping the papaya seeds onto the book.'

In (24), the applied argument is a locative source.


(24) Applicativisation of transitive verb with -o licenses applied locative argument a. Mina nyol woya. mina n=yol woya Mina 3sg=take water 'Mina is taking water.' b. Rauf nyolo wola ai coatco. rauf n=yol-o wola ai coat=so Rauf 3sg=take-APPL rope wood CLASS=one 'Rauf untied the rope from the fire-wood.'

In this last example, the root Undergoer is wola 'rope' and the applied argument is ai coat co 'a bundle of firewood'. The literal meaning of the sentence is thus 'Rauf takes the rope from the bundle of firewood'. Cross-linguistically, it is often the case that arguments licensed by applicatives in one language may be licensed in different ways in another language. Sometimes adpositions may be used; in other cases the twin objects of a ditransitive verb in one language may be licensed individually by the separate verbs of a serial verb construction. Sometimes, such different ways of licensing arguments may be found within one language, as is the case in Taba. The instrumental arguments licensed by -Vk and the locatives licensed by -o may also be licensed adpositionally. In addition, instruments can be licensed by means of serial verb constructions. (The applied arguments in transfer ditransitives, on the other hand, can only be licensed by applicativisation.) Examples (25) and (26), with adpositionally licensed NP's can be used to refer to the same real world events as (18b) and (23b) respectively.
(25) Transitive verb with preposition licensing instrumental adjunct Nggoras n=goras 3sg=shave(take.seeds.out) 'He took the seeds out of kapaya ada kobit. kapaya ada kobit pawpaw with knife the pawpaw with a knife.'

(26) Transitive verb with postposition licensing locative adjunct Yak kgoras yak k=goras 1sg 1sg=shave 'I'm scraping kapaya ni kowo bbuk kapaya ni kowo bbuk papaya 3sg.POSS seed book the papaya seeds onto the li. li LOC book.'

While (26) is allowable in Taba, the construction is somewhat marked. As we noted earlier, there is a strong preference in Taba discourse for locative adjuncts to be marked first by an independent locative such as a directional and for the postpositional locative phrase to occur as a complement of the independent locative as in (27).
(27) Transitive verb with directional and postposition licensing locative adjunct Yak kgoras kapaya ni yak k=goras kapaya ni 1sg 1sg=shave papaya 3sg.POSS 'I'm scraping the papaya seeds onto kowo appo bbuk li. kowo ap-po bbuk li seed ALL-down book LOC the book.'

Example (28), where serial verb pake 'use' licenses an instrument is also referentially equivalent to (18b) and (25).
(28) Serial verb construction licensing instrument Nggoras n=goras 3sg=shave(take.seeds.out) 'He shaved (i.e. took the kapaya npake kobit. kapaya n=pake kobit pawpaw 3sg=use knife seeds out of) the pawpaw with a knife.'

What is probably most unusual about Taba from a cross-linguistic perspective is that instruments and location can be doubly licensed with both applicative and adpositional marking. Example (29) can be used to refer to the same real world situations as (18b), (25), and (28), while (30) could refer to the same event as (23b) and (26).
(29) Double licensing of instrument with applicative suffix and preposition Nggorcak kapaya ada kobit. n=goras-ak kapaya ada kobit 3sg=shave(take.seeds.out)-APPL pawpaw with knife 'He took the seeds out of the pawpaw with a knife.' (30) Double licensing of location with applicative suffix and postposition Yak kgorco kapaya ni yak k=goras-o kapaya ni 1sg 1sg=scrape-APPL papaya 3sg.POSS 'I'm scraping the papaya seeds onto kowo bbuk li. kowo bbuk li seed book LOC the book.'

One further point needs to be made with respect to the double licensing of locative and instrumental arguments with both an applicative and an adposition before we turn our attention to other kinds of applicative derivation. This concerns the behaviour of applied arguments in

relative clauses. When these arguments are used as the head of a relative clause, they must be doubly licensed. Example (31a) illustrates relativisation of the instrumental argument in a ditransitive clause. Note the clause final occurrence of the preposition ada, which along with the -Vk applicative suffix on the verb, licenses the applied object. (31b) is ungrammatical because the relativised instrument is not doubly licensed.
(31) Obligatory double-licensing of applied instrument when it serves as head of a relative clause a. Banda nyat peda npunak kolay ada. Banda n=yat peda n=pun-Vk kolay ada Banda 3sg=carry machete 3sg=kill-APPL snake with 'Banda is carrying the machete he killed the snake with.' b. * Banda nyat peda npunak kolay. Banda n=yat peda n=pun-Vk kolay Banda 3sg=carry machete 3sg=kill-APPL snake

Relativised locations from ditransitive clauses must also be doubly licensed.


(32) Obligatory double-licensing of applied location when it serves as head of a relative clause a. Iswan nbaca bbuk Iswan n=baca bbuk Iswan n=baca bbuk 'Iswan is reading kgorco kapaya ni kowo li. k=goras-o kapaya ni kowo li 1sg=scrape-APPL papaya 3sg.POSS seed LOC the book I'm scraping the papaya seeds onto.'

b. * Iswan nbaca bbuk kgorco kapaya ni kowo. Iswan n=baca bbuk k=goras-o kapaya ni kowo Iswan n=baca bbuk 1sg=scrape-APPL papaya 3sg.POSS seed

Note that the second non-Actor in a transfer ditransitive cannot be licensed adpositionally and is not affected by this rule.[11]
(33) No double licensing for theme of transfer ditransitive when it serves as head of a relative clause Lahon yan mon ya l=ha-hon yan mon ya 3pl=CAUS-eat fish man up 'They're eating the fish lwagik lai mo ya. l=wag-Vk lai mo ya 3pl=sell-APPL just come up the man just sold them.'

The ditransitives with derived instruments and locations which we have just discussed are all examples of what was labelled 'remote' ditransitive verbs in figure 2 above. While close ditransitives (the verbs of transfer of possession) are used in clauses that are basically reminiscent of the double object constructions found in many languages, 'remote' ditransitives

(with applied instruments and locations) are distinguished from close ditransitives by the following characteristics:

The applied arguments are optionally marked by adpositions The applied arguments are obligatorily marked by adpositions when they occur as the head of a relative clause.

3.2. Applicative derivation with Actor intransitive stems Applicativisation of Actor intransitive stems results in derived verbs which are either transitive or semi-transitive. The classification of the derived verb depends on the semantic role of the applied argument. When -Vk applicativisation adds an instrument, and when -o applicativisation adds a locative argument, the derived forms are semi-transitive verbs. When -Vk applicativisation adds a theme or a companion, and when either -Vk or -o applicativisation adds a source of emotion, the derived forms are basic transitive verbs. 3.2.1. Deriving semi-transitive verbs from Actor intransitives The derived forms discussed in this section are categorised as semi-transitive verbs because the non-Actor arguments found with these verbs share the same basic characteristics as the remote Undergoers of the motion semi-transitives discussed in 2.1, and of remote ditransitives discussed in 3.1: they can be optionally marked by an adposition as well as by the applicative suffix. Example (34) below illustrates the addition of an applied instrument. In (34a) the noun phrase wangsi 'children' is the Actor cross-referenced on the intransitive verb agawil 'swim'. In (34b) the applicative derived verb agawlik is shown. Here, an applied instrument has been added, in this case the NP hopan 'planks'. When -Vk applicativisation licenses an instrument, the derived form is a semi-transitive verb, and the applied argument functions just like the remote Undergoers of underived semi-transitives (2.1) and remote ditransitives (3.1).
(34) Derivation of semi-transitive verb from Actor intransitive with -Vk suffixation a. Underived Actor intransitive Wangsi lagawil turus. wang=si l=agawil turus child=PL 3pl=swim all.the.time 'Children are always swimming.' b. Derived semi-transitive with instrumental argument Jul nagawlik hopan. Jul n=agawil-Vk hopan Jul 3sg=swim-APPL plank 'Jul is swimming with a plank.' [i.e. holding onto a plank as he swims and using it as an aid for floatation]

In (35) we illustrate applicative derivation from an Actor intransitive stem with the -o suffix. (35a) shows the Actor intransitive verb battaln 'to sit', with the pronominal Actor i '3sg' and the locative adjunct yase um li 'up in the house'. In (35b), -o has licensed the locative argument kurusi 'chair' to appear as an argument of the derived semi-transitive verb battalono 'sit on'.
(35) Derivation of semi-transitive verb from Actor intransitive with -o suffixation a. Underived Actor intransitive i nbattalon yase um li. i n=battalon ya-se um li 3sg 3sg=sit up-ESS house LOC 'He's sitting up in the house.' b. Derived semi-transitive with locative argument Nbattalono kurusi. n=battalon-o kurusi 3sg=sit-APPL chair 'He's sitting on the chair.'

Just as the applied instruments and locations of ditransitives may be doubly licensed with applicatives and adpositions, double licensing of the applied arguments in the semi-transitive verbs derived from Actor intransitive stems is also possible. Examples (36) and (37) can refer to the same real world events as (34b) and (35b).
(36) Double licensing of instrument with applicative and preposition Jul nagawlik ada hopan. Jul n=agawil-Vk ada hopan Jul 3sg=swim-APPL with plank 'Jul is swimming with a plank.' (37) Double licensing of location with applicative and postposition Nbattalono kurusi n=battalon-o kurusi 3sg=sit-APPL chair '(S)he's sitting on li. li LOC the chair.'

We saw in 3.1 that when applied instruments and locations of ditransitive clauses serve as the heads of relative clauses they are obligatorily licensed by both an applicative suffix and an adposition. The same is true of the instruments and locations which are arguments of applicativised Actor intransitives. This is illustrated in (38) and (39).
(38) Obligatory double licensing of instrument as head of relative clause

Kam hopan Jul nagawlik ada. k=am hopan Jul n=agawil-Vk ada 1sg=see plank Jul 3sg=swim-APPL with 'I can see the plank Jul swam with.' (39) Obligatory double licensing of locative as head of relative clause Nik babasi lpe kurusi nbattalono li. nik baba=si l=pe kurusi n=battalon-o li 1sg.POSS father=PL 3pl=make chair 3sg=sit-APPL LOC 'My father made the chair (s)he's sitting on.'

3.2.2. Deriving transitive verbs from Actor intransitives The semantic roles of the added arguments which -Vk applicativisation allows in transitive derivation are those of theme, companion, and stimulus of emotion. Transitive verbs are also derived by -o applicativisation, and the added arguments in these cases are always stimulii of emotion. The derived forms with all of these types are real transitives since adpositional marking of the added Undergoer is not allowed, and the applied arguments behave like the unambiguously core Undergoers of canonical transitive clauses. Example (40) shows the addition of a theme by -Vk applicativisation. In (40a) we have the activity verb poas 'to row', while (40b) shows its derived transitive applicativised counterpart posak 'to row something'. Note that (40b) also shows a relativised Undergoer, and that the relativised Undergoer bears no adpositional marking.
(40) Derivation of transitive verb from Actor intransitive with -Vk suffixation a. Underived Actor intransitive Banda npoas kwat Banda n=poas kwat Banda 3sg-row be.strong 'Banda rows hard.' b. Derived transitive with theme as applied argument Ltala wog Banda nposak do. l=tala wog Banda n=poas-Vk do 3pl=find canoe Banda 3sg=row-APPL REAL 'They found the canoe Banda was rowing.'

Example (41) illustrates the addition of an applied companion. Example (41a) shows intransitive wom 'to come' with Bib, which is a man's name, the Actor cross-referenced on the verb. Example (41b) shows the addition of an applied companion argument.
(41) Derivation of transitive verb from

Actor intransitive with -Vk suffixation a. Underived Actor intransitive Bib nwom do Bib n=wom do Bib 3sg=enter REAL 'Bib has come.' b. Derived transitive with theme as applied argument Bib nwomak Bib n=wom-Vk Bib 3sg=come-APPL 'Bib has come with a gina lloci. gina lloci stuff much lot of stuff.'

Note that (41b) is not an example of causativisation, although it may appear at first glance to be so. For (41b) to be true, Bib must have come at the same time as the stuff he was carrying. A true causative can be used whether the Actor him or herself comes or not, as in (42).
(42) Bib nawom si nidi wangsi. Bib n=ha-wom si nidi wang=si Bib 3sg=CAUS-come 3pl 3pl.POSS child=PL 'Bib made his children come.'

A few examples of both suffixes licensing a stimulus of emotion to occur as Undergoer are also found. The use of -Vk applicativisation to license a stimulus of emotion is illustrated in (43).
(43) Derivation of transitive verb from Actor intransitive with -Vk suffixation; applied argument = stimulus of emotion a. Nkiu. n=kiu 3sg=be.scared 'He's scared.' b. Wangsi lkiuak wang=si l=kiu-ak child=PL 3pl=be.scared-APPL 'The children are scared of baratci. barat-si westerner=PL westerners.'

The -o applicative licenses a stimulus of emotion in (44).


(44) Derivation of transitive verb from Actor intransitive with -o suffixation; applied argument = stimulus of emotion a. Acan namara kwat. Acan n=ha-mara kwat Acan 3sg=CAUS-be.angry EMPH 'Acan is very angry.'

b. Oci namaro Iswan. Oci n=ha-mara-o Iswan Oci 3sg=CAUS-be.angry-APPL Iswan 'Oci is angry at Iswan.'

Adpositional marking of the applied argument is not possible when the applied argument is a stimulus of emotion. 3.2.3 Applicative derivation with semi-transitive stems This process is rather rare in Taba. It always involves -Vk suffixation, and the added argument is always a companion which the Actor takes with him / her to the goal specified by the locative goal of the semi-transitive verb. Example (45a) shows the same semi-transitive verb we have already encountered, sung 'enter', occurring with the goal um (li) 'house (LOC)', and the Actor Bib (man's name). Its derived counterpart in (45b) shows the companion argument Nou (child's name). Note that as with example (41) above, the following example is not a causative for the same reasons mentioned with respect to (41): for the construction in (45b) to be true, Bib must have entered the house at the same time as Nou, by picking him up and carrying him in.
(45) Derivation of ditransitive verb from semi-transitive with -Vk suffixation a. Underived Actor intransitive Bib ncung um Bib n=sung um Bib 3sg=enter house 'Bib entered the house.' (li) (li) (LOC)

b. Derived transitive with theme as applied argument Bib ncungak Nou Bib n=sung-Vk Nou Bib 3sg=enter-APPL Nou 'Bib entered the house um (li) um (li) house (LOC) with Nou.'

In derived ditransitives of this sort, locative arguments are still optionally marked by the postposition li, and companion arguments are unable to be marked adpositionally. Locatives here retain the characteristics of remote Undergoers while the applied theme functions as an unambiguously core Undergoer. 3.4. Applicative derivation with Undergoer intransitive stems The addition of applicative suffixes to Undergoer intransitive stems results in derived forms which no doubt look somewhat unusual cross-linguistically. These derived forms are labelled 'non-Actor bivalent verbs', since although they have two arguments, neither of the arguments is cross-referenced on the verb as an Actor. When applicative derivation occurs with Undergoer

intransitive stems, the applicative suffix licenses much the same kind of argument as licensed by the same suffix when it is added to a transitive or Actor intransitive stem: the -Vk applicative licenses an applied instrument and the -o applicative license an applied location. First, we examine -Vk. In (46a), the Undergoer argument calana da 'those trousers' is attributed the quality of 'blackness' with the Undergoer intransitive verb kuda 'be black'. In (46b), the Undergoer from (46a) is still the entity to which blackness is being ascribed, but in this case it is asserted that the blackness of the trousers has been brought about as a result of the instrumentality of asfal 'asphalt / tar', an applied argument licensed by the applicative sufix -Vk. Neither of the arguments in this derived form has the status of Actor since neither of them is cross-referenced on the verb.
(46) Derivation of non-Actor bivalent verb with -Vk suffixation; applied instrument a. Calana da kuda do. calana da kuda do trousers DIST be.black REAL 'Those trousers are black.' b. Calana da calana da trousers DIST 'The trousers kudak asfal. kuda-Vk asfal be.black-APPL asphalt have been blackened by asphalt.'

The same kind of process is illustrated in (47), again with the -Vk applicative. In (47a), the Undergoer intransitive verb posa 'be boiled' appears with loka 'bananas' as its Undergoer. In (47b), the same stem, posa 'be boiled', and the same affected entity, loka 'banana' are found. Here, the applied argument is niwi 'coconut', the thing with which the bananas were boiled.
(47) Derivation of non-Actor bivalent verb with -Vk suffixation; applied instrument a. Loka posa do. banana be.boiled REAL 'The bananas have been boiled.' b. Loka ne posak niwi. loka ne posa-Vk niwi banana PROX be.boiled-APPL coconut 'The bananas are boiled with coconut.'

We noted in section 2 above, when addressing Van Valin and Wilkins (1996) discussion of the 'effector' semantic role, that instruments can be characterised as non-instigating effectors. Instigators (which always occur as Actors in Taba) are 'primary' effectors since they are not subject to any external control. True instruments, on the other hand, are only secondary effectors because their effects can only be achieved when they are manipulated by some higher effector. Although no higher effector is encoded in Taba non-Actor bivalent clauses, it is clear that with all the examples cited above, some higher effector must have done something that allowed the

instrument to bring about its effect. Bananas don't get boiled with coconut unless there is human intervention; trousers cannot be blackened by asphalt unless someone brings them into contact with asphalt in some way, by, for example, walking through asphalt while wearing them. Since these instruments are not true instigators, they do not occur as Actors in Taba. Although its use is rather infrequent, the -o applicative suffix can also license applied locative arguments when it is added to an Undergoer intransitive stem. The forms derived by this process typically describe the relative position of one object vis-a-vis another object. In (48a), the Undergoer intransitive verb mfati 'to be covered' is seen with the single argument faffati 'curtains'. In (48b), the same stem mfati 'to be covered' is used. The NP which would have been Undergoer of the underived verb is bbuk pso 'one book' and the applied argument is bbuk maleo 'other book', the book which is covering the first mentioned book.
(48) Derivation of non-Actor bivalent verb with -o suffixation; applied location a. Faffati mfati do pa tesu? curtain be.covered REAL or NEG-POT 'Are the curtains closed yet or not?' b. Bbuk pso mfato bbuk p-so mfati-o book CLASS-one be.covered-APPL 'One book is covered over by the bbuk maleo. bbuk maleo book other other book.'

Since books do not come into contact with other books, except through the mediation of a higher effector, the second argument here cannot occur as Actor, for the same reasons that the instruments shown in (46b) and (47b) above could not occur as Actor. We have seen that the remote Undergoers of semi-transitives (2.1), the remote Undergoers of remote ditransitives (3.1) and also the remote Undergoers of derived semi-transitives (3.2) can all optionally be marked adpositionally. The same is true of derived instruments and locations with non-Actor bivalent verbs. Examples (49) and (50) can refer to the same situations as (47b) and (48b) respectively.
(49) Double licensing of applied instrument in non-Actor bivalent derivation Loka ne posak ada niwi. loka ne posa-Vk ada niwi banana PROX be.boiled-APPL with coconut 'This banana has been boiled with coconut.' (50) Double licensing of applied location in non-Actor bivalent derivation Bbuk bbuk book 'One pso mfato p-so mfati-o CLASS-one close-APPL[abut] book is abutting the other bbuk maleo li. bbuk maleo li book other LOC book.'

Just as the applied instrumental and locative arguments of other types of verbs must be doubly licensed with both an applicative suffix and an adposition when they occur as the head of a relative clause, so too must they be doubly licensed with non-Actor bivalent clauses. This is illustrated in (51) and (52).
(51) Obligatory double-licensing of applied instrument in non-Actor bivalent clause as head of relative clause Alhoe nyat niwi loka posak ada? alho=e n=yat niwi loka posa-Vk ada who=FOC 3sg=bring coconut banana be.boiled-APPL with 'Who brought the coconut that the banana has been boiled with?' (52) Obligatory double-licensing of applied location in non-Actor bivalent clause as head of relative clause Yak ktono yak k=tono 1sg 1sg=look.at 'I'm looking at bbuk bbuk book this ne ne PROX book bbuk da mfato li. bbuk da mfati-o li book DIST close-APPL[abut] LOC which is abutting that book.'

Although this just described set of non-Actor bivalent clauses is probably somewhat unusual cross-linguistically, similar constructions are encountered in other languages. Such constructions seem to occur in languages which, like Taba, have relatively semantically transparent systems for the morphosyntactic realisation of arguments. Mithun (1991:517) discusses a set of what she calls 'double patient' constructions in Lakhota, also a split-S language, which are similar to the Taba constructions discussed here. Other formally similar sub-classes of verbs are found in inverse languages such as Oluta Popoluca for which Zavala (1998) has discussed what he calls 'non-agentive bivalent verbs'. Given the semantic basis for determining whether an intransitive verb will have an Actor or an Undergoer as its sole argument, and given the functions of the applicative, it is not surprising that non-Actor bivalent verbs should occur in Taba. None of the stems for these verbs has a human or an effector as an argument, and the applied arguments in each case are of much the same sort as the applied arguments found with other derived verb classes. The real puzzle about these forms is not so much why they should exist in Taba, but how we should best view the status of all nonActor arguments in these, and other applicative constructions. We now address this issue. 4. The status of non-Actor arguments in Taba In section 2 we saw that Taba could be characterised as a mixed split-S / accusative language with respect to the patterning of arguments found in basic transitive and intransitive clauses. In this section we review the evidence pertaining to applicative derived verbs with the aim of setting out a basic typology of arguments in all the multivalent clause types we have seen.

In 4.1 we begin by reviewing Dryer's (1986) work on the differences between primary and direct object constructions. An interesting feature of Dryer's typology that has so far gone largely unremarked in the literature is that while Dryer notes that the differences between direct and primary objectivity have explicit parallels in the differences between ergativity and accusativity, there is a logical gap in the range of systems he discusses, i.e. a third pattern for realising objectlike arguments which would provide a parallel for split-S patterning in intransitive clauses. Section 4.1 ends with a brief sketch of what such a pattern, which might be labelled 'splitobjectivity' or perhaps 'split transitivity', would look like. In 4.2 we turn back to the Taba data and argue that in fact Taba has a number of multivalent clause types which do fit the requirements for such a split-object pattern of alignment. It is argued that Taba multivalent clauses can be characterised as being of a mixed primary object / split object type. 4.1 Primary objectivity vs. direct objectivity Dryer (1986) shows that languages (or more properly constructions) commonly treat the 'objects' of ditransitive clauses in two distinct ways. In some constructions, patients/themes group together as 'direct objects' while recipients/beneficiaries group together as 'indirect objects'; in other constructions recipients/beneficiaries group together as 'primary objects' while patients/themes group together as 'secondary objects'. Dryer's account of the phenomena is given in a Relational Grammar framework. Thus, the diagrams in figure 3 list 'final terms' of subject, object, ergative, and absolutive across the top of figure 3a, and show how the 'intial terms' of subject and object map onto these final relations. The 'final terms' of direct object, indirect object, secondary object, and primary object are listed across the top of figure 3b, and the way that the 'initial terms' of direct object and indirect object map onto these categories is illustrated in the body of the diagram.

It is noteworthy that Dryer explicitly draws a parallel between primary / direct objectivity and ergativity / accusativity in the linguistic systems he discusses. He says:

just as some languages employ the grammatical relations ERGATIVE and ABSOLUTIVE, which are defined in terms of subject and object-as in Figure 1a [here, figure 3a, J.B.]-so too some languages employ the grammatical relations Primary Object and Secondary Object, which can be defined in terms of DO and IO-as in Figure 1b. [here, figure 3b, J.B.] (Dryer, 1986: 814) Dryer maintains that the distinction between primary objectivity and direct objectivity is independent of any distinction between ergativity and accusativity. Languages which have mostly ergative characteristics can also have either primary object or direct object constructions. Likewise, languages which have mostly accusative characteristics can also have either kind of object alignment. In addition, just as some languages exhibit split ergativity, some also exhibit what Dryer calls 'split objectivity'. Primary objectivity, then, according to Dryer, should be seen as a property of constructions rather than of languages, just as ergativity should be seen. English, for example, is a language which could be argued to have both kinds of constructions, cf. 'I gave her a book' / 'I cooked her a cake' (primary objectivity) vs. 'I gave a book to her' / 'I cooked a cake for her' (direct objectivity).[12] We argued in 2.3 above that accusative and ergative patterning involved a neutralisation of semantic differences between intransitive subjects of different types. Whatever the semantic role of an intransitive subject, it patterns with the (Actor) transitive subject in an accusative structure and with the (Undergoer) transitive object in an ergative structure. A set of parallel observations can be made about the objects of monotransitive clauses shown in figure 3b: again there is a neutralisation of semantic differences between different monotransitive object types. In primary/secondary object systems, the single objects of monotransitive clauses pattern together with the recipients and beneficiaries etc. of ditransitive clauses whatever the precise semantic role of the monotransitive object, e.g. patient in 'Someone stained Monica's dress', or beneficiary in 'Kenneth didn't help Monica'. In direct/indirect object structures, likewise, the monotransitive object patterns together with the patient and theme of a ditransitive, no matter what the role of the monotransitive object.[13] We saw in 2.3 that the split-S patterning found in Taba differs from ergative and accusative systems in that there is no neutralisation of semantic roles with respect to the morphosyntactic coding of intransitive arguments. The question we now wish to address is whether the same thing can be said with respect to the argument types that are found with applicative derived verbs in Taba. To what extent are the semantic differences between arguments neutralised in the full range of multivalent constructions discussed so far? In the next section we seek to answer that question. 4.2. A typology of Undergoers in Taba It should be clear by now that there is more than one kind of Undergoer to account for in Taba. The immediate task at hand, then, is to categorise the different kinds of Undergoers found, to explicitly list their different morphosyntactic characteristics and to see which semantic roles are found with Undergoers of different types. We will argue that Taba has three distinct kinds of Undergoers, each with distinct sets of properties. First, we review the behaviour of Undergoers

with respect to adpositional marking, and then with respect to preferred word order before presenting an overview of Undergoer properties in all clause types. 4.2.1 Adpositional marking We have already reviewed a fair amount of evidence concerning the ability of Undergoers to be marked adpositionally in previous sections. A clear division between 'remote Undergoers' and what we might call 'close Undergoers' can be made, on the basis of whether or not the arguments concerned can be marked adpositionally. A listing of the semantic roles of the NP's which occur as 'close' and 'remote' Undergoers is given in figure 4. Close Undergoer Remote Undergoer object of affect instrument non-instigating theme location recipient stimulus of emotion Figure 4. 'Close' vs 'remote' Undergoers and their semantic roles 4.2.2. Word order So far we have not explicitly addressed the question of word order as it pertains to different sorts of Undergoers. We begin here by examining word order in clauses where all of the arguments are expressed by full NPs. The relative order of Undergoers in these clauses is fixed, so we can readily label the arguments concerned as 'First Undergoers' and 'Second Undergoers' on the basis of this relative ordering. The First Undergoer in a ditransitive clause is always an object of affect or a recipient. The Second Undergoer is always an instrument, theme or location. In non-Actor bivalent clauses, the First Undergoer is always an object of affect and the Second Undergoer is either an instrument or a location. The semantic characteristics of First Undergoers and Second Undergoers are outlined in figure 5. First Undergoer Second Undergoer object of affect instrument theme recipient location Figure 5. Arguments in ditransitive and non-Actor bivalent clauses First Undergoers in both these clause types have much the same characteristics as the sole Undergoers of canonical transitive and Undergoer intransitive clauses. When no full NP Actor is expressed, and if the First Undergoer is a full NP, the First Undergoer may precede the verb. Compare (53) and (54) below with (9) above, repeated below as (55).

(53) Ditransitive verb: no full NP Actor; full NP First Undergoer precedes verb Kolai kolai snake 'They da lpunak peda da l=pun-Vk peda DIST 3pl=kill-APPL machete killed the snake with a machete.'

(54) Non-Actor bivalent verb: full NP First Undergoer precedes verb Capeyo ne bulngak cet capeyo ne bulang-Vk cet hat PROX be.white-APPL paint 'The hat has gone white from paint.' (55) Transitive verb: full NP Undergoer precedes verb, pronominal Actor occurs between full NP Undergoer and verb Mon da mon da man DIST 'You hit meu hwet meu h=wet 2pl 2pl=hit that man.'

If the First Undergoer is a pronoun, it occurs immediately following the verb whether or not a full NP Actor is expressed, as shown in (56) to (58).
(56) Ditransitive verb: full NP Actor precedes verb; pronominal First Undergoer follows verb Mado npunak i mado n=pun-Vk i mado 3sg=kill-APPL 3sg 'Mado killed it with a peda peda machete machete.'

(57) Ditransitive verb: no full NP Actor; pronominal First Undergoer follows verb Npunak n=pun-ak 3sg=kill-APPL 'He killed it i peda i peda 3sg machete with a machete.'

(58) Non-Actor bivalent verb: pronominal First Undergoer follows verb Bulngak i cet bulang-Vk i cet be.white-APPL 3sg paint 'It has gone white from paint.'

As we saw in (5), (7), and (8) above, repeated as (59), (60), and (61) below, pronominal Undergoers also occur postverbally in canonical transitive and Undergoer intransitive clauses.

(59) Transitive verb: pronominal Actor precedes verb, pronominal Undergoer follows I i 3sg 'He nwet yak n=wet yak 3sg=hit 1sg hit me.'

(60) Undergoer intransitive verb: pronominal Undergoer follows verb Mot i die 3sg 'It died.' (of an animal) (61) Transitive verb: full NP Actor precedes verb, pronominal Undergoer follows Oci nwet yak Oci n=wet yak Oci 3sg=hit 1sg 'Oci hit me.'

If an overt full NP Actor is present, the First Undergoer immediately follows the verb, whether represented pronominally as in (56) above, or by a full NP, as was illustrated in (21b), repeated as (62).
(62) Ditransitive verb: full NP Actor precedes verb, full NP 1st Undergoer follows Ahmad npunak kolay peda. Ahmad n=pun-Vk kolay peda Ahmad 3sg=kill-APPL snake machete 'Ahmad killed the snake with a machete'

The preferred clausal position of Second Undergoers is somewhat more complicated, and depends on the actual semantic role of the NP concerned. Pronominal Second Undergoers are virtually never encountered since pronominal reference is impossible for inanimates, and the semantic roles of Second Undergoers (theme, instrument, location) are usually incompatible with animate referents. Close Second Undergoers (those Second Undergoers which cannot be marked adpositionally, i.e. themes) can occur preverbally if there are no other full NPs found in the clause.
(63) Ditransitive verb: full NP Second Undergoer (theme) precedes verb with no other full NP arguments Pipis pipis money 'They lloci latadak i lloci l-ha-tada-Vk i much 3pl=CAUS-present-APPL 3sg gave him lots of money.'

However, remote Second Undergoers (i.e locations and instruments) always occur clause finally, no matter whether other full NP arguments are found or not. This is illustrated for a locative Second Undergoer in (64) and (65), and for an instrumental Second Undergoer in (66) and (67).
(64) Ditransitive verb: full NP Second Undergoer (location) occurs clause finally whatever other arguments expressed Kgorco bbuk k=goras-o bbuk 1sg=scrape-APPL book 'I'm scraping (them) onto the book.' cf. (65) * Bbuk kgorco bbuk k=goras-o book 1sg=scrape-APPL [Not grammatical except with meaning 'I scraped the book onto something']

(66) Ditransitive verb: full NP Second Undergoer (instrument) occurs clause finally whatever other arguments expressed Npunak n=pun-Vk 3sg=kill-APPL 'He killed it cf. (67) * Peda npunak i peda n=pun-ak i machete 3sg=kill-APPL 3sg [Not grammatical except with meaning 'the machete killed him/her with something'] i peda i peda 3sg machete with a machete.'

The remote Undergoers of non-Actor bivalent clauses behave just like the remote Undergoers of ditransitives with respect to word order, always following the verb whether an overt First Undergoer is present or not. This is illustrated in (68) and (69).
(68) Non-Actor bivalent verb: full NP Second Undergoer (instrument) occurs clause finally whatever other arguments expressed Bulngak cet bulang-Vk cet be.white-APPL paint 'It was whitened with paint.' cf. (69) * Cet cet bulngak bulang-Vk

paint be.white-APPL [Not grammatical except with meaning 'paint was whitened with something']

One final piece of evidence now needs to be presented before we provide a summary of Undergoer properties in all clause types: the ordering of remote Undergoers in semi-transitive clauses. Such Undergoers exhibit basically the same behaviour as do the remote Undergoers of other sorts of clauses. Examples (70) and (71) show that applied instrumental remote Undergoers must occur after the verb which licenses them.
(70) Derived semi-transitive verb: full NP Second Undergoer (instrument) occurs clause finally whatever other arguments expressed Lagawlik hopan l=ha-gawil-Vk hopan 3pl=CAUS-swim-APPL plank 'They swam with planks.' cf. (71) Hopan lagawlik hopan l=ha-gawil-Vk plank 3pl=CAUS-swim-APPL * 'They swam with planks.' [can mean 'planks swam with something']

Examples (72) and (73) show that the same is true of applied locative remote Undergoers.
(72) Derived semi-transitive verb: full NP Second Undergoer (location) occurs clause finally whatever other arguments expressed Nbattalono n=battalon-o 3sg=sit-APPL 'S/he sat on cf. (73) Kurusi nbattalono kurusi n=battalon-o chair 3sg=sit-APPL * 'S/he sat on a chair.' [can mean 'chairs sat on something'] kurusi kurusi chair a chair.'

And finally, examples (74) and (75) show that the remote Undergoer of a root semi-transitive must also occur after the verb, no matter what other NPs are present.

(74) Root semi-transitive verb: full NP remote Undergoer (location) occurs clause finally whatever other arguments expressed Ncung um da (li) n=sung um da (li) 3sg=enter house DIST (LOC) 'S/he entered the house.' cf. (75) Um da ncung um da n=sung house DIST 3sg=enter * 'S/he entered the house.' [can mean 'The house entered something']

4.2.3. Undergoers in Taba: an overview We have now assembled enough data to characterise the full range of Undergoers found in Taba. An initial split between close Undergoers and remote Undergoers has already been made and can be characterised according to the properties listed below: Close Undergoer:

never marked adpositionally occurs preverbally when no competing full NP argument found in the same clause includes semantic roles of theme, recipient, object of affect and stimulus of emotion

Remote Undergoer:

may be marked adpositionally always occurs postverbally, no matter what other NPs found in the same clause includes semantic roles of instrument and location

Each type of Undergoer retains all of the definitional characteristics listed above, no matter what type of clause it appears in. Remote Undergoers occur in remote ditransitive clauses, in nonActor bivalent clauses, and in semi-transitive clauses. Close Undergoers occur in canonical transitive clauses, in Undergoer intransitive clauses, in non-Actor bivalent clauses, and in both types of ditransitives. Importantly, verbal arguments with the roles of location and instrument always occur as remote Undergoers, no matter what type of clause they occur in; Undergoers with semantic roles other than instrument or location always occur as close Undergoers, again no matter what type of clause they occur in. It could be said, in fact, that the semantic role of the Undergoer(s) involved in any type of clause including Undergoers forces the selection of a clause type suitable for its appearance. Taba does not just exhibit characteristics of split intransitivity with respect to the realisation of inanimate arguments; it also exhibits characteristics of what we could call split

transitivity: instruments and locations are consistently realised in a different manner from Undergoers with different semantic roles. However, just as Taba's typology for basic transitive and intransitive argument alignment was characterised as a mixed accusative / split-S system, so too must its typology for the alignment of the full range of arguments in multivalent clauses be characterised as a mixed one. Transfer ditransitives exhibit a pattern of argument alignment with the same sorts of properties as Dryer's 'primary object' structures. Perhaps, though, given that there is little evidence for any subject or object categories in Taba, the argument types found here should be more aptly labelled 'Primary and Secondary Undergoers'. With transfer ditransitives, recipients are 'Primary Undergoers', occuring in the immediately postverbal argument slot, while themes are 'demoted' to the Secondary Undergoer position, occuring after the Primary Undergoer if one is expressed. We characterise Taba, then, as a mixed split-S / accusative language at the level of basic transitive and intransitive argument alignment, and as a mixed split Undergoer / Primary Undergoer language at a level which takes account of a fuller range of multivalent constructions. The relative semantic transparency underlying the morphosyntactic realisation of arguments in basic transitive and intransitive clauses is actually maintained across the full range of clause types found in the language. Structures we have encountered which may at first blush have looked unusual, such as the non-Actor bivalent clause, can now be seen as quite straight-forward manifestations of this basic underlying semantic transparency of argument realisation found in the language. 5. Conclusions A major difficulty that has been encountered in attempting to describe the full range of multivalent constructions found in Taba is that there is a lack of appropriate, commonly accepted terminology for talking about the full range of 'Undergoer-like' or 'O-like arguments' encountered in the language. Over recent years, descriptive linguists have often adopted either or both of Dixon's (1972, 1979, 1994) 'basic syntactic relations' of S, A, and O or Foley and Van Valin's actor and undergoer macro-roles as convenient descriptive devices allowing a simple characterisation of the properties of different sorts of arguments in different kinds of clauses. While these commonly used terms have certainly proved their usefulness for describing the alignment of basic transitive and intransitive arguments, they are clearly not fine-grained enough systems to fully describe the range of argument types found in the whole panoply of mutlivalent constructions encountered in a language like Taba. This paucity of descriptive tools for discussing multivalent constructions is reflected in discussions of applicative constructions written by the originators of the terminology, and by people associated with them. Van Valin and La Polla (1997: 338), for example, describe applicatives as forms in which 'a non-argument of the verb appears as as undergoer, displacing the default choice for undergoer if the verb is transitive'. Dixon and Aikhenvald (1997: 80), when discussing applicative derivation with transitive stems say that 'in the applicative derivation the original A is retained, an erstwhile peripheral argument is promoted to be O, and the original O is pushed out onto the periphery (typically marked by dative or locative case)'.

Clearly, the most productive use of the applicative in Taba - instrumental applicativisation of transitive stems - does not work in the manner discussed by Van Valin and La Polla, nor by Dixon and Aikhenvald. While the instruments marked by the applicative are promoted to 'remote Undergoerhood', the close Undergoer of the transitive stem is neither 'displaced' as Van Valin and La Polla suggest it should be, nor 'pushed out onto the periphery' as Dixon and Aikhenvald would predict: the underived 'close Undergoer' remains the close Undergoer of the derived form. A cross-linguistically satisfactory account of how best to understand what different kinds of undergoers or O arguments need to be accounted for by linguistic theory will need to address a much fuller range of data than has been presented from just one language here. The Taba data, though, does point to some of the complexities which might be involved in sorting out the problem. This paper, then is presented in the hope that the kinds of patterns unearthed in Taba multivalent constructions will contribute to an eventual better understanding of 'undergoerhood' cross-linguistically.

References Baker, Mark C. 1988. Incorporation: a theory of grammatical function changing. Chicago: Chicago University Press. Bowden, John, 1997a. Taba (Makian Dalam): description of an Austronesian language from eastern Indonesia. University of Melbourne: PhD thesis. Comrie, Bernard, 1978. 'Ergativity'. In Winfred P. Lehmann, ed., Syntactic typology, 329-394. Austin: University of Texas Press. Dixon, R.M.W. 1972. The Dyirbal language of North Queensland. Cambridge: Cambridge University Press. Dixon, R.M.W. 1979. 'Ergativity.' Language, 55: 59-138. Dixon, R.M.W. 1994. Ergativity. (Cambridge Studies in Linguistics 69) Cambridge: Cambridge University Press. Dixon, R.M.W and Alexandra Y. Aikhenvald, 1997. 'A typology of argument determined constructions.' In Joan Bybee, John Haiman and Sandra A. Thompson, eds. Essays on Language Function and Language Type dedicated to T. Givn, 71-113. Amsterdam: John Benjamins. DuBois, John W. 1985. 'Competing motivations', in John Haiman, ed. Iconicity in Syntax. Amsterdam and Philadelphia: Benjamins. Foley, William A. and Robert D. Van Valin, Jr. 1984. Functional syntax and universal grammar. Cambridge: Cambridge University Press. Gruber, J.S. 1965. Studies in Lexical Relations. PhD Dissertaion: Massachusetts Institute of Technology.

Margetts, Anna, 1998. 'Give' verbs in Saliba (Western Oceanic, Papua New Guinea). Paper presented to the Workshop on Cross-Linguistic Perspectives on Argument Structure: Implications for Learnability. Nijmegen: June 1998. Mithun, Marianne, 1991. Active / agentive case marking and its motivations. Language , 67: 510-546. Van Valin, Robert D. Jr, and Randy J. LaPolla, 1997. Syntax: Structure, meaning and function. Cambridge: Cambridge University Press. Van Valin, Robert, and David Wilkins, 1996. 'The case for 'effector': case roles, agents, and agency revisited', in Masayoshi Shibatani and Sandra A. Thompson, eds. Grammatical Constructions. Oxford: Oxford University Press. Zavala, Roberto, 1998. Inverse clauses with non-agentive bivalent verbs in Olutec (Mixe-Zoquean). Paper presented to the workshop on 'Cross-Linguistic Perspectives on Argument Structure: Implications for Learnability' held at the Max Planck Institute for Psycholinguistics in Nijmegen, June 26-29, 1998.

Endnotes 1 The terms S, A and O ar taken from Dixon (1972, 1979, 1994). Some authors, following Comrie (1978), adopt the label 'P' in pplace of 'O'. The terms actor and undergoer have been adopted from Foley and Van Valin (1984). 2 Although I have a preference for using naturally occurring examples of spontaneous speech whenever possible, many of the examples given in this paper are elicited. Because Taba has such productive ellipsis of arguments in natural conversation, it is rare to encounter constructions with mutiple arguments, especially ditransitives, where all of the arguments that are licensed by the verb actually appear overtly. For the sake of clarity in the exposition, then, a number of elicited examples in which all of the licensed arguments are overtly mentioned have been given. 3 Another reason for not seeing the locative expression in (4b) as an oblique is the fact that Taba has two kinds of locatives: what I have labelled elsewhere (Bowden, 1997: 115) as 'independent' and 'dependent' locatives. Independent locatives include a class of directionals as well as the set of 'place demonstratives', while dependent locatives consist of locative postpositional phrases and place names. In Taba discourse there is a strong preference for locative adjunct phrases to consist of an independent locative as head and followed by an optional dependent locative phrase. In the example above, there is no independent locative phrase to act as the head of the construction. 4 There are a few marginal exceptions to this generalisation, but in all cases where a human is treated as Undergoer, there is an implication that the human referent has less than human qualities in some respect. The example below shows one such instance, with the verb amseh 'be drunk'. Drunkenness is a state that is seen as somehow less than human in ethnically muslim Taba society.

Amseh i be.drunk 3sg

Вам также может понравиться