2.2 The object as an internal argument
2.2.4 Objects and case marking
In this section I further establish a connection between measurers and the direct object position. The generalisation I will draw is as follows:
( 18) All measurers are (universally) marked with accusative case.
The entailment in (18) works in one direction only. It does not assert that all arguments which are marked with accusative case are measurers (an entailment which holds or fails to hold according to the morphology o f each particular language. In section 2.2.4.1 I will discuss partitive / accusative case alternations in Finnish. Section 2.2.4.2 discusses alternations between direct objects and prepositional objects. Finally, section 2.2.4.3 discusses the object case marking o f specific verbs across four languages.
In all these cases, measuring objects are associated with accusative case, while non measurers are not.
2.2.4.1 Partitive Case
Accusative/partitive case alternation in Finnish is sensitive to telicity or change o f state. Thus, objects are measurers are marked with accusative case, while objects which are not measurers (e.g., objects of stative verbs) are marked with partitive case:
The properties suggested by Dowty are a unification o f thematic and aspectual properties, and thus cover, descriptively, a wide range o f objects. Under such a theory, the internal argument o f kick and
(19) a. Matti rakensi talon. Matti-NOM built house-ACC
b. Mina rakastan sinua / *sinut.
I love-l-sg you-PAR you-ACC
(Pylkkanen, p.c.)
It may be the case that Finnish has two object positions, one in which accusative case is assigned, the other associated with partitive. Or it could be that accusative/partitive are assigned at the same position, and are simply the reflection of the relationship between the verb and its object. We know that some languages have objects marked with cases other than accusative (genitive objects in Icelandic, nominative objects in Georgian - cf. Marantz 1991), which behave like objects in all other respects (e.g., they passivize). Other theories, in particular HK (1997), assume that there exist numerous object positions in the language, depending on the type of the verb (change of state, creation, location/giving etc.).
Note that the same verb may mark its object with either accusative or partitive case: accusative/partitive alternations seem also to be sensitive to boundedness (that is, the actual delimitedness in time, cf. Depraetere 1995, see also appendix), not only to telicity (the potentiality of having a linguistically determined endpoint). Other instances o f partitive/accusative alternations include rendering an event from a single occurrence (accusative) into an habitual event (partitive), or between a quantized object (accusative) and non-quantized one (partitive). Case marking on the object in Finnish
Grammatical aspect in Finnish is expressed through the case suffixes on the object. Accusative case entails perfectivity / completeness, while imperfectivity entails imperfectivity/incompleteness (Krifka 1992,
Pylkkanen 1997);
(i) a. Anne rakensi taloa. Anne built part-house Anne was building a/the house b. Anne rakensi talon.
Anne built acc-house
may thus be triggered either by the basic relation between the verb type and its object (stative vs. non-stative) or by processes involving a higher projection (tense/aspect) or DP-intemal properties (NP-properties o f the object). Tense/Aspect/D-properties may only turn accusative case into partitive (partitive is taken by some linguists to be the default case in Finnish), never the other way around. If a verb is lexically marked with partitive case (e.g., it is stative), then its case pattern is determined by the basic relation between the verb and its object.
What I find crucial here is that non-measuring verbs (e.g. statives) inherently mark their object with partitive case. Following Manzini and Savoia's (1998) discussion of accusative/partitive clitic alternations in Italian dialects I will assume here that accusative and partitive objects are generated at the same position. Morphological case will be a reflex o f +/- measuring property o f the object.
2.2.4.2 Direct objects vs. partitive prepositional objects
Some languages allow partitive prepositions instead o f direct objects. In all these cases accusative case is associated with the measuring object while the partitive preposition is associated with the non-measuring object:
(20) a. John ate the cake, b. John ate at the cake.
(21) a. axalti et ha uga. I ate OM the cake T ate the cake' (all of it) b. axalti me ha uga
I ate from the cake
I had some of the cake' (Hebrew)
Some languages which are not discussed here seem to distinguish measuring and non-measuring objects structurally (see Ramchand 1997 for Scottish Gaelic). It may be that in such cases two object positions are motivated.
With accusative case the object (cake) serves as a measurer o f the event: it terminates when the cake is completely eaten. No such measuring relation exists with the partitive objects: eating at a cake has no inherent endpoint. Consider, next, (22):
(22) a. Janet shot the bird, b. Janet shot at the bird.
The measuring properties of (22a) are somewhat less clear than in (20). The speakers' intuitions seem to be that shoot NP implies that the shooter "reached" her aim, while shoot at NP implies that she was aiming at her target, but not necessarily reaching it. Yet (22a) is a measured event: it terminates when the bird is shot. (22b) is not a measured event: shooting at a bird can go on for some time, with no (linguistically determined) endpoint.
There are other instances of this direct object/prepositional object alternation in English, which involve non-measuring verbs, like hit. This is known as The conative alternation:
(23) a. Janet hit the wall, b. Janet hit at the wall.
Here the difference between the direct and partitive object seems to be associated with a thematic rather than aspectual property o f the object. Hit the wall implies that the wall was directly affected by the hitting, perhaps moved, broke etc., while hit at the wall
implies that the hitter was hitting, but the object was not necessarily affected in any way.
' ^ Note that prepositional objects in English may count as object for grammatical processes (e.g., they passivize: The bird was shot at). This alternation may, therefore, reflect once more the difference "thematic" /aspectual relation between the verb and its object.
I will take the direct/prepositional object alternation to be another instance of the relationship between measuring objects and case marking. An object cannot be a measurer if it is not marked with accusative case.
2.2.4.3 Cross-linguistic variation in Case m arking
The most convincing evidence about the correlation between measured events and accusative case comes from a cross-linguistic examination o f object case marking. Recall Jackendoffs objection to the syntactic relevance o f measuring out, namely that many objects are not measurers, yet they fall under the thematic characterization of themes, being affected arguments. Such examples are verbs like kick, touch, drive - all o f them having an object which is affected by another arguments, while failing to measure out the event. English has the objects of these verbs marked with accusative case, but this is just a peculiarity of English. In Hebrew, these three verbs mark their objects with a locative preposition, be (at): kick at the ball, use at the knife, drive at a car, etc. This kind of variation does not exist with measuring verbs. Kill, for example, takes accusative case in both languages. Consider now the following groups of verbs: those having non-measuring objects (states and activities) in (24), vs. those with measuring objects in (25):
English Latin Classical G reek Hebrew help +acc auxilior+dat boetheo +dat azar +le (to) use +acc utor +abl xraomai +dat hiStameS +be (at) trust +acc fido +dat pisteuo +dat batax +be (at) fight +acc pugno+dat polemeo +dat nilxam +be (at) rule +acc dominer +abl arxo +gen maSal +al (upon) obey +acc pareo +dat peithomai +dat ziyet +le (to)
English Latin Classical Greek Hebrew
build +acc construe +acc oikodomeo +acc bana +acc write +acc scribo +acc grapho +acc katav +acc murder +acc occido +acc apokteino +acc racax +acc eat +acc edo +acc esthio +acc axal 4-acc wash +acc lavo +acc luo 4-acc raxac 4-acc -
Two-place predicates with measuring objects universally mark their object with accusative Case. Two-place predicates with non-measuring objects may mark their object with either accusative, dative, ablative or genitive Case, or by a preposition, depending on the particular morphological properties o f that language. Some o f the predicates in (24) have affected objects which are not measurers: help, fight, rule. The case marking of these objects may vary across languages. With measuring objects (25) no variation occurs: these objects are universally marked with accusative case. I will take this to indicate that the core properties of prototypical object/patients are aspectual properties: undergoing change, or being an incremental theme. These are the properties which seem to remain stable with respect to case marking patterns across languages.