Non canonical Case licensing and the event temporal interpretation

Dans le document Event building, selection and non-canonical Case: fi insertion in Tunisian Arabic (Page 179-195)

Chapter 4: Event decomposition and causativized stative verbs

4.4. Non canonical Case licensing and the event temporal interpretation

In the previous subsection I illustrated how the distribution of marked objects in Tunisian correlates to the relationship between the (sub)events, aspect and time. In this section, I will discuss the correlation between variable binding and the interpretative properties of the sentence.

The observed relationship is that subevents whose event run time are perceived as occurring at the same time and for the same time pattern with the presence of accusative arguments, whereas (sub)events whose respective extension is only partially overlapping, meaning that the subevents are understood as having different durations, license fi arguments.

Interestingly, Icelandic presents a distribution of Case on objects that is reminiscent of Tunisian and, in this subsection, I will illustrate Svenonius’ account of the relevant facts and his analysis, before proceeding to the comparison with Tunisian and to the discussion on the correlation between Case and the aspectual interpretation.

Svenonius’ account for the alternation between accusative and dative objects in Icelandic relies on the notion of run time. I will look closely at his proposal, summing up the relevant aspects of a small body of literature he published in the early 2000s (Svenonius 2002a; 2002b;

and 2013 first published in 2001).

Svenonius’ theory argues that the Case of the object in Icelandic is always structural;

contrary to the traditional view, in fact, he proposes that the distribution of dative and accusative objects is not lexically pre-specified on the verb; rather, such information is syntactically derived as a result of a temporal evaluation process. According to his analysis, dative arguments are predictably attributed to verbs whose subevent structure presents mismatching sub-temporal extensions; vice-versa, accusative objects are licensed when the subevent presents overlapping temporal extensions.

Svenonius suggests that the link between Case and aspect is in the uninterpretable nature of the aspectual feature that Case conveys. Such an aspectual feature is realized on the nominal phrase, but remains invisible at the level of the DP on which it occurs; consequently, syntax operates a feature-checking mechanism, as in Chomsky (1995), that pairs the argument with a predicate in order for the derivation to be legible at the interface with LF.

This theory comes with two desirable theoretical consequences. Firstly, it solves one of the critical aspects of the checking theory, namely the assumption that there can be strictly uninterpretable formal features (Chomsky 1999). Secondly, it allows us to treat nominative and accusative Case in a uniform way: just like Pesetsky and Torrego’s (2001) proposal that nominative Case is the uninterpretable realization of tense on the DP, accusative Case is the realization of uninterpretable aspect according to Svenonius. Together, the two theories shape a coherent picture in which structural Cases are consistently associated with the temporal/aspectual properties of a sentence.

The formulation of the theory is innovative, although the idea that aspect is influenced by the properties of the direct object is not new: quoting Blake (2001), Svenonius explains how accusative Case is robustly associated across languages with affectedness of the direct object;

moreover the nature of this relationship is explained indirectly by the idea that the physical properties of the direct object make it possible to temporally map the duration of the event (Tenny 1989; Krifka 1992). According to these hypotheses, the object is a measurer that allows us to determine the duration of the event and whether the event has culminated.

Based on this background, Svenonius puts forward the proposal that accusative Case in Icelandic is licensed whenever the subcomponents forming an event are mapped to equal temporal extensions; conversely, in the absence of such a requirement, dative Case is licensed.

Evidence of various kind supports this approach and makes it more appealing than the alternative lexical/semantic explanation. One of the arguments Svenonius proposes in his analysis is the existence of the phenomenon referred as “Dative Sickness” (Halldórsson 1982;

Svavarsdóttir 1982). Icelandic Dative Sickness refers to the progressively expanding use of dative objects in the language: verbal neologisms taking dative are formed and a growing number of dative objects are consequently registered in the language. According to his view, the rising number of non-canonically marked objects indicates the presence of a highly productive underlying phenomenon. The lexical approach, conversely, would predict a small number of vestigial forms that would be the “remnants of some moribund historical system” (Svenonius 2001, 4).

Verbs that alternate between accusative and dative objects provide additional support to his approach. See the following examples of ballistic motion verbs4:

(67) a. skjóta fuglinn ‘shoot the bird’ (acc) b. skjóta kúlunni ‘shoot the bullet’ (dat) c. skutla hvalinn ‘harpoon the whale’ (acc) d. skutla skutlinum ‘throw the harpoon’ (dat) (Icelandic, Svenonius 2013, 4: (7)a-d)

The verbs above take objects that can correspond either to the projectile, or to the target. Under the assumption that transitive events result from the combination of two subevents, the difference in the accusative and the dative uses of the verbs skjóta ‘to shoot’ and skutla ‘to harpoon/throw the harpoon’ lies in the way the event is built: accusative Case occurs when the object corresponds to the target, (67)a-c, because these constructions convey a sense that the motion initiated by the subject persist throughout the development of the shooting/launching event; dative Case occurs when the object is the projectile and, consequently, the subject’s action does not need to extend throughout the event, as in (67)b-d.

These cases can be accounted for under Svenonius’ theory: examples (67)a-c show that the launching-by-the agent subevent is presented as corresponding in length to the instant covered by the target-hitting subevent, meaning that the event is presented in such a way that the two parts, the launching and the hitting, occur at a single time; hence they properly overlap and accusative is licensed on the object. As for (67)b-d, the launching-by-the agent subevent lasts for a fraction of the time at which the launched-projectile subevent holds. In this case, the event is presented in such a way that the launching event run time holds for a shorter time that the time needed by the projectile to end its trajectory. The two parts do not properly overlap, and the object is licensed with dative Case.

As he notices, moreover, no alternation is found with transitive verbs of assisted motion.

The list in (68) presents some verbs of the relevant class:

(68) a. draga ‘pull, drag, draw’

b. flytja ‘move, transport, carry’

c. æra ‘move, bring’

(Icelandic, Svenonius 2013, 5: (10)a-c)

The object of assisted motion verbs is always in accusative, because they allow only one reading, namely the interpretation according to which a subject must continue to input kinetic power throughout the duration of the whole event in order for the patient to move.

Case alternation in spry/load constructions bring additional support to Svenonius’ approach;

in constructions of this type if the direct object is the location or target of movement, it occurs in the accusative Case, cf. vagninn ‘the wagon’ in the following example (69)a. When the direct

4 Most of the empirical evidence produced by Svenonius comes from examples originally published in

object is the element or substance that moves to such a location, it occurs in the dative Case, as heyinu ‘the hay’ in (69)b:

(69) a. Við hlóðum vagninn með heyi.

we loaded the.wagon.ACC with hay.DAT b. Við hlóðum heyinu á vagninn.

we loaded the.hay.DAT on the.wagon.ACC (Icelandic, Svenonius 2013, 9: (23)a-b)

Svenonius presents the above examples to illustrate that accusative direct objects are conceptualized as incremental themes: an accusative selecting event is mapped onto the object.

This status contrasts with the one of dative objects that are not perceived as incremental but as

“an indivisible unit undergoing movement” (ibid. p 10). The rationale that accounts for the examples in (67) also applies to the pair under discussion, so that a non-mapping object is understood as temporally unrelated with respect to the loading event.

An important point that Svenonius makes with respect to the examples at hand is that the different interpretations of dative and accusative constructions are not reflected in the actual world. The reality of things, in fact, requires a load to be progressively transferred to its destination: the transferring may apply to part or to the complete volume, but time and distance are still incrementally mapped onto one another. The Icelandic language, conversely, conceptualizes this reality in two distinct ways corresponding to the constructions in (69)a and (69)b. Two structures are associated with the loading examples: an incremental loading whose temporal mapping corresponds to the extent to which the wagon has been filled, and a punctual loading that culminates when the locatum “materializes”, so to say, inside the locus.

Svenonius explains that despite the different semantic entailments brought about by the two types of objects, the respective syntactic behavior is identical; specifically, he points out that:

“the similarity of the patterns […]do not support any attempt to locate the dative-accusative contrast in a particular licensing position, as by a null preposition.”

(Svenonius 2013, 7).

In my understanding, this point interests us because it clarifies that dative and accusative objects concur with the different semantic interpretations, without being the trigger of the interpretation itself.

The interpretation for Svenonius depends on the properties of the v° which introduces the external argument and licenses accusative arguments. In the system he develops, there is not just a single type of little v°; but depending on the type of VP, different v° projections are added to the structure. Certain verbs, cf. ballistic motion verbs in (67)a-d, select dative and accusative objects, depending on the semantic relationship between V°, i.e. the “root” in Svenonius’ work, and its internal argument. In my understanding, the type of v° is determined by selection, since Svenonius proposes that different little v° types are chosen in relation to the properties of the VP, which is the combination of a verbal root and its direct argument. Measuring objects are involved in the event for its entire duration, thus they must occur under a v° that can bind a V° understood as co-temporal to one another. Dative objects, conversely, do not measure out the event, therefore

they must occur in the domain of a v° which is not a V° binder, i.e. under a v° that does not restrict the run time of the embedded subevent.

Thus, summing up Svenonius’ proposal: accusative and dative Case are structural. They are assigned in the same syntactic position. The distribution of the two competing Cases depends on the relationship between v°, which introduces the external argument, and V° responsible for the selection of the internal argument. Accusative Case is licensed only if the two subevents are understood as part of a “temporally indivisible single event” (Svenonius 2002b, 201), which means that accusative is selected whenever the appropriate mapping v° is present in the structure and that v° encodes certain aspectual properties interpreted on the verb. In Svenonius’ account, accusative Case patterns with the overlapping run time of the subevents, but it does not directly encode the co-temporal reading, since it only becomes interpretable once it enters into a relationship with the appropriate v°. If the overlapping relationship is not established, then only a dative object can be licensed. This part of the analysis is not fully developed. In my understanding, however, dative is the unmarked competitor that surfaces whenever the elsewhere condition applies.

According to Svenonius’ theory, the relationship between Case and event run time is mediated by the type of v° and Case checking is implemented as variable binding where the variable is a Davidsonian argument associated to a subevent. What is not completely clear is how binding operates. Svenonius suggest that the way v° binds the event variable in V° determines whether accusative Case can be licensed. However, he does not explain in detail how or why binding is or is not realized.

4.4.2. Icelandic vs Tunisian

The distribution of dative and accusative Case in Icelandic, according to the account of the facts proposed by Svenonius (o.c.), depends on the relationship between the run time of two subevents. The subevents’ run time is measured by evaluating whether the subject is involved for the entire duration of the subevent by introducing the object so that, if the subject is continuously involved throughout the duration of the event, the run time of all the subevents described by a predicate are understood as being equally long. The object is assigned with accusative Case in the context of these properties.

In contrast, when the initiator is not involved throughout the duration of all the subevents, the run time of the various subevents do not overlap, and dative is assigned to the object. This analysis allows us to explain why ballistic motion verbs and spry/load alternation verbs present two Case frames depending on the semantic role of the object, and why accompanied motion verbs strictly occur with accusative arguments.

The temporal overlapping of the two subevents can fail due to the properties of the Case licensing head v° or due to the properties of V°. Thus, the conditions that govern Case distribution in Icelandic pertain to the verbal domain and, more specifically, to the aspectual properties which are generally referred to as Aktionsart or Inner Aspect.

Turning now to Tunisian, the distribution of unmarked and marked objects can also be linked to the way the event-internal temporal organization is presented; the temporal overlapping of two (sub)events patterns with the presence of an accusative/unmarked direct object and, vice versa, (sub)events which are not understood as properly overlapping, i.e. non-cotemporal subevents, are found with fi-marked objects.

The similarity between Icelandic and Tunisian stops when the conditions for the evaluation of temporal overlapping are considered. While Icelandic evaluates the subevents’ temporal extension at the level where Aktionsart properties are introduced, Case alternations in Tunisian are determined marginally by Aktionsart. i.e. states, but mostly by viewpoint aspect. For instance, marked objects in causativized stative constructions require the predicate to occur in the perfective form, which is a fact that shows the existence of a correlation between viewpoint aspect and Case.

The following Tunisian examples illustrate that the distribution of Case in spry/load alternation verbs depends on aspect but not on the thematic roles, and consequently on the aspectual relations within the vP. As we can see, in fact, the insertion of aspectual fi patterns with the presence of progressive qāʕed, cf. (72)a and (73)a, and not with simple alternation in the argumental structure of the verb phrase:

(70) a. semi ʕabba l-ḥažar fī-l-berwiṭa

Semi load.perf the-stones in-the-wheelbarrow

‘Semi loaded the stones in the wheelbarrow.’

b. *semi ʕabba fī-l-ḥažar fī-l-berwiṭa

Semi load.perf fi-the-stones in-the-wheelbarrow (71) a. semi ʕabba l-berwiṭa be-l-ḥažar

Semi load.perf the.wheelbarrow with-the-stones ‘Semi loaded the wheelbarrow with stones.’

b. *semi ʕabba fī-l-berwiṭa be-l-ḥažar Semi load.perf fi-the.wheelbarrow with-the-stones

(72) a. semi qāʕed yeʕabbi fī-l-ḥažar fī-l-berwiṭa

Semi prog load.imp fi-the-stones in-the-wheelbarrow ‘Semi is loading the stones in the wheelbarrow.’

b. *semi qāʕed yeʕabbi l-ḥažar fī-l-berwiṭa

Semi prog load.imp the-stones in-the-wheelbarrow

(73) a. semi qāʕed yeʕabbi fī-l-berwiṭa be-l-ḥažar Semi prog load.imp fi-the-wheelbarrow with-the-stones ‘Semi is loading the wheelbarrow with stones.’

b. *semi qāʕed yeʕabbi l-berwiṭa be-l-ḥažar Semi prog load.imp the-wheelbarrow with-the-stones

Examples (70) and (71) contrast with respect to the order of their complements: the former pair presents a locatum-direct object, i.e. ḥažar ‘the stones’ while the latter one displays the locus l-berwiṭa ‘the wheelbarrow’ in the same position. According to the previous discussion (cf.

Svenonius 2013) spry/load alternations correspond to different inner aspectual configurations but, as we can see, Case in Tunisian appears not to be sensitive to this distinction since both examples require an unmarked object despite the different thematic roles assigned to the objects.

Examples (72) and (73) are the progressive counterparts of perfective (70)a and (71)b. As they illustrate, viewpoint aspect interacts with Case; since ‘load’ requires a fi-marked object when it occurs in the progressive periphrasis and, vice versa, requires an unmarked direct object when it occurs in the perfective form. The starred examples in (70)-(73) show that whether the direct argument corresponds to the locatum or the locus is not a relevant variable in the computation of Case.

Accompanied motion verbs return the same results. Verbs of this type in Icelandic only select accusative direct objects (cf. the list in (68)), while in Tunisian they can occur with both unmarked and fi-marked arguments. As Svenonious suggests, verbs of this type imply the participation of the agent for as long as the theme is affected by the event and, therefore, the two subevents are seen as completely overlapping, which in Icelandic results in the presence of accusative Case.

Tunisian, conversely, presents accusative objects with a perfective accompanied motion predicate and fi-marked objects when the predicate occurs in the periphrastic progressive construct, see example (74).

(74) a. semi dēz marwa Semi push.perf Marwa

‘Semi pushed Marwa.’

b. semi qāʕed ydīz fī-marwa Semi prog push.imp fi-Marwa

‘Semi is pushing Marwa.’

The evidence presented in this subsection clearly shows that the distribution of Case on the direct object in Tunisian is not governed by the same principle that operates in Icelandic. The two languages, in fact, link Case to aspectual properties of different kinds, although, as I will show in the next subsection, they are similar with respect to the interpretative effect obtained in the presence of non-canonically marked objects. With this I mean that we can capture Case alternations on the direct object in the two languages by saying that accusative Case patterns with events which are understood as presenting a uniform temporal interpretation, whereas non-canonical Case is assigned in complex events whose internal temporal development is not uniform.

4.4.3. Temporal overlapping in complex events

The proposal I put forward in section 4.3 is that accusative Case licensing in Tunisian depends on the way in which the various event variables projected by a complex event structure are valued.

In this section, I discuss what this entails for their temporal interpretation and show that the interpretation reflects the way syntax operates, since the temporal relations between subintervals are evaluated with respect to a common reference point. Specifically, I suggest that reference time, intended in the sense discussed in 4.2.3, is used as tertium comparationis to evaluate the relation between subevents.

I will address this theory in two steps: I will first explain what complex events with not uniform and uniform internal temporal articulation look like and I will then explain how this evaluation process is mapped from syntax to LF in order to account for the link between Case and event structure observed in causativized stative predicates.

Let us go back to the relevant data and consider sentence (75), which was originally presented in (21) of this chapter:

(75) le-kteb haḏaya ḥabbeb-ni fī-l-muṭālʕa this make_love-me.perf.acc fi-the-literature

‘This book made me love literature.’

‘This book made me love literature.’

Dans le document Event building, selection and non-canonical Case: fi insertion in Tunisian Arabic (Page 179-195)