The proposal: intervention and fi insertion

Dans le document Event building, selection and non-canonical Case: fi insertion in Tunisian Arabic (Page 146-158)

Chapter 4: Event decomposition and causativized stative verbs

4.2. The proposal: intervention and fi insertion

4.2. The proposal: intervention and fi insertion 4.2.1. Intervention and Case licensing

This chapter addresses the link between the accusative Case and aspect. Much syntactic and semantic research has been devoted to this relationship. A significant body of literature argues that the properties of the direct object, i.e. Case, interact with the aspectual properties of the sentence, typically telicity (Tenny 1994; Wechsler and Lee 1996; Kiparsky 1998; Svenonius 2002b; Kratzer 2004; Csirmaz 2008). Other approaches argue that Case and aspect have an independent syntactic relationship (MacDonald 2006). The results presented in this chapter support the latter view.

My main claim is that the aspectual properties of a sentence and the Case assigned to the direct object are independent syntactic operations which can be affected in parallel in the presence of an intervening aspectual operator. The analysis developed in this chapter, therefore, accounts for the distribution of unmarked and fi-marked objects in Tunisian in terms of syntactic intervention.

I assume the following definition of the intervention principle (Rizzi 1990; 2004):

(1) Relativized Minimality:

In the configuration...X...Z...Y...

a local relation connecting X and Y is disrupted when there is a Z such that i. Z is of the same structural type as X, and

ii. Z intervenes between X and Y (Rizzi 2018, 341)

Intervention effects occur in contexts where two aspectual operators compete for the licensing of the event variables present in the vP.

The theory of Case licensing developed in this work assumes that accusative objects are licensed in situ by a verbal head, as in Svenonius (2002a; 2002b; 2013). Like Svenonius, I adopt a split approach to the VP structure; however, I do not assume that the lexical decomposition of the event directly determines the Case of the object, as he does. Instead, I propose that accusative Case licensing in Tunisian is subject to a syntactic condition which involves the binding of the event variable. Namely, accusative Case is attributed to a DP whenever the event variable associated to the licensing verbal head carries the same value as the one assigned by the aspectual operator which is closer to T°. This configuration is represented in (2):

(2) TP



  

OPx vP  v° VP (ex) 

 DP V’

 

[acc] 

    V°  …

A verbal head does not assign the accusative Case if a second operator occurs in the syntactic structure, i.e. OPz. In other words, if an operator of the same kind as OPx, i.e. aspectual, intervenes between OPx and the event variable (e), the v° is not able to license Case even if the associated variable is properly licensed. This configuration is illustrated by the syntactic representation in (3), whereby the intervention principle applies, since OPz binds the event variable associated to v° and OPx c-commands OPz.

(3) *TP

The event variable (ez) is not valued by the syntactically higher operator due to the presence of the intervening element OPz; consequently, v° does not fulfil the conditions that allow it to assign Case and, therefore, the argument fails to be licensed in the structure.

In order to prevent such illicit configuration, Tunisian resorts to the insertion of a Case licensing preposition, namely aspectual fi. Aspectual fi licenses oblique Case on the DP in its complement position and allows the derivation to progress, as shown by diagram (4).

(4) TP

This chapter is devoted to the development of this theory which crucially explains the alternation between accusative marked DPs and fi-marked ones as a consequence of the intervention effect just discussed. The chapter will also discuss the correlation between aspect and Case, arguing that the interpretative properties associated with the presence of accusative and fi-marked objects are derived from their syntax and from the presence or absence of multiple aspectual operators in the same clausal domain.

4.2.2. Core syntactic claims

The goal of this chapter is to provide an analysis that accounts for the presence of two unmarked DPs in object position, in sentences whose main predicate is a causativized stative predicate which occurs in the imperfective form, as in examples (5), and for the presence of an unmarked DP and a DP preceded by aspectual fi in sentences of the same type in which the causativized stative predicate displays perfective morphology, as in(6):

(5) le-ktob kif haḏaya yḥabbebū-ni l-muṭālʕa the.books as this make_love.imp-me the-literature

‘Books like this make me love literature.’

(6) l-ktob kif haḏaya ḥabbebū-ni fī-l-muṭālʕa the.books as this make_love.perf-me fi-the-literature

‘Books like this made me love literature.’

As said in the previous subsection, the proposed thesis is that aspectual fi insertions in sentences like (6) is a Case-salvaging operation that prevents a syntactic failure by licensing Case on a DP which would be otherwise unlicensed. The absence of accusative Case is due to the fact that the fi-marked argument is contained within the licensing domain of an aspectual operator which is not the same operator and does not carry the same features as the one that enters in a relation with T°.

I follow Klein (1995) and assume that tense is computed as a set of relations which order the moment when the sentence is uttered, i.e. speech time, with the interval of time which the speaker is asserting something about, i.e. reference time, and finally, the time taken for the event to develop, i.e. event time.

These time coordinates have a syntactic realization in T°, Asp° and v° respectively.

According to the system proposed in this chapter, accusative Case licensing occurs when the relation between the temporal coordinates extends all the way to the v°. This means that accusative is assigned whenever reference time, i.e. the time the sentence is asserting something about, frames either the event time or part of the event time, and orders this interval with respect to the time of the utterance.

In order for this to be possible, the event variable associated with the object licensing head, i.e. v°/V° as proposed by Svenonius (2002b), must be valued by the aspectual operator which is closest to T°. This syntactic configuration is represented by the diagram in (7).

(7) TP

T° AspP 

  

Asp° vP   OPx v° VP (ex) 

 DP V’

 



  

 

 …

   


 

 (ex)

I follow Svenonius (2002b) and assume a split vP structure. Like him, I assume that every verbal head defines a subevent and that each v°/V° projects a syntactic argument linked to a θ-role. This is also referred to in the literature as the “one argument per subevent condition”, which states that

every subevent must correspond in syntax to an XP structure, and that each XP is associated with an argument position (see Rappaport-Hovav and Levin 2001 and references therein).

I take each subevent to project a Davidsonian argument too, and that this argument is linked to an event variable which needs to be bound temporally rather than existentially, as suggested by Bohnemeyer and Swift (2004). Therefore, event variable binding is on the one hand indirectly responsible for the licensing of Case, and on the other is directly responsible for the interpretative properties of the subevent denoted by the associated head. Unlike Svenonius, I adopt a Neo-Davidsonian approach and assume that state-describing verbs project such an argument too.

A consequence of the “argument per subevent condition” is the syntacticization of the notion of events. In this view, event and sub-event stand in a relation of inclusion; an event can consist of a single verbal projection, as in the case of monadic predicates, or it can present a complex structure and include several arguments and projecting heads, namely the building blocks of dyadic and triadic predicates, with each block which is equal to the structure of a monadic verb.

The term sub-event refers to these blocks, which are simple by definition, whereas events can be simple or complex.

Svenonius assumes that stative predicates display the same structure as dynamic ones, so that intransitive statives present a simple XP structure, whereas transitive ones are composed of two verbal heads and license two θ-roles. Moreover, in line with the Neo-Davidsonian approach adopted here, I assume that they also project an event argument in association with every verbal projection.

As the set of syntactic assumptions just illustrated suggests, I therefore consider the notions of sub-event and event in syntactic terms only, meaning I apply the notions to stative verbs as well as to dynamic ones. Consequently, the expression “stative-(sub)event” does not contain an inherent contradiction, but simply refers to the syntactic complexity underlain by a stative predicate.

This does not mean that I consider the opposition between stative/dynamic verbs vacuous.

On the contrary, I follow Chierchia (1995) and assume that stative and dynamic predicates differ with respect to the way their event variable is bound. Chierchia proposes that stative predicates, unlike dynamic ones, are inherently bound by a generic operator, referred here as OPgen, and that this operator is directly projected within the verbal domain. According to his proposal, verbs of this kind are inherently bound, which entails that they can only be interpreted as generic and that their event variable is always bound by this operator only.

This aspect is crucial for the theory developed in this chapter and brings us back to aspectual fi insertion. I argue that the presence of OPgen required by a stative predicate acts as an intervener in causativized stative constructions, when the verb occurs the perfective aspect, cf. (6). Verbs of this kind have a perfective operator in Asp° realized in syntax by means of the verb’s perfective inflectional morphology and present a second operator within the vP domain. As the following tree diagram illustrates, the lower operator (OPgen) binds the variables in its domain (e+gen), so that the associated verbal heads cannot license the accusative Case on the internal object, (DP2). This configuration corresponds to the ungrammatical representation in (8):

(8) *TP

In order to prevent the derivation from failing, an alternative Case licensing strategy must be adopted.

The insertion of aspectual fi should thus be seen as the Case licensing strategy which allows the derivation to proceed to spell out. This mechanism is represented in (9).

(9) TP

The insertion of aspectual fi in (5) salvages the structure granting Case to the argument projected by a verbal head whose event variable is inherently bound by OPgen. From the syntactic view point, OPgen acts as an intervener because it values the variables in its domain and is c-commanded by OPperf. However, from the interpretative perspective, it affects the interpretation so that the subevent projected by the verbal heads in its domain are understood as extending beyond the sentence’s reference time.

4.2.3. The link between Case licensing and aspectual interpretation

I will adopt a tripartite system of temporal evaluation in the spirit of Klein’s (1995) development of Reichenbach’s seminal proposal (Reichenbach 2005, first published in 1947). According to Klein (o.c.) the situation-time is “the time at which the situation described by an utterance obtains” (ibid. p 687). I will refer to it as the “event run time” or, simply, as the “event time” (Ev-T).

The event run time must be differentiated from both the time of speech (Sp-T) and the “topic time”, which I will refer to with the more familiar expression “reference time” (Ref-T). The time of the utterance is the time of speech; whereas reference time for Klein is “the time for which the assertion is made (or to which the assertion is confined)” (Klein 1995, 687). Bohnemeyer and Swift (2004) adopt the same system in their work and present reference time as the time at which a proposition is evaluated. This means that, in consideration of its illocutionary force, the reference time of a declarative sentence is the time for which the proposition is asserted to be true.

Consequently, reference time is the time when the truth conditions of a given event are checked, an aspect which will become crucial later, when we look at the temporal properties of complex events.

The major advantage of Klein’s proposal is the way it characterizes the notion of reference time. His system allows us to tease apart the contributions of aspect and time. For Klein, tense linearly orders reference time and the moment of speech, whereas aspect defines whether reference time frames the entire event time, or just a sub-portion of its run time, thus it defines a relation of inclusion.

Past progressive contexts suitably illustrate how the temporal coordinates just discussed interact in practice. Consider a sentence like (10):

(10) A dog was barking when my alarm went off.

The duration of the barking-event corresponds to the event run time of the matrix sentence. The reference time corresponds to the moment at which the alarm goes off, since the modification of an activity, i.e. barking, by a punctual adverbial clause such as ‘when the alarm went off’ allows us to identify the reference time, i.e. the moment the sentence makes a statement about something, with the interval denoted by the punctual adverb (Demirdache and Uribe-Etxebarria 2007). The use of the past makes it possible to differentiate the moment of the assertion from the moment of the utterance, since past time places the reference time before the time of the utterance.

Thus, reference time in the above sentence corresponds to the moment when the alarm goes off and this time interval is also a subinterval of the run time of a barking event. The truth conditions of sentence (10) are met precisely at a temporal window when a barking event is in progress and the alarm goes off. Finally, past reference situates this specific interval, i.e. reference time, before the time of the utterance.

According to Klein’s relational approach, the run time of an event is not an absolute value but depends on aspect. We will consider it as a binary option, perfective/imperfective, whereby perfective aspect encodes proper inclusion of the event time within the boundaries of reference time, so that the latter contains the former, including the event endpoint, as in (11):

(11) Ref-T




Otherwise, imperfective aspect encodes partial inclusion, which logically corresponds to a configuration whereby reference time is “shorter” and is included in the duration of the event run time, as in (12):

(12) Ref-T




Aspects encode this relation, defining whether reference time has access to the entire event, including the event’s conclusion, or whether it can only frame a subinterval of the event’s duration.

With these premises in mind, we can now go back to the examples discussed in the previous section, copied below for ease of reference. As we can observe, examples (13) and (14) differ in overt syntax with respect to their verbal morphology and with respect to the presence of fi before the theme object (fī)-l-muṭālʕa ‘(fi) the literature’:

(13) l-ktob kif haḏaya yḥabbebū-ni l-muṭālʕa the.books as this make_love.imp-me the-literature

‘Books like this make me love literature.’

(14) l-ktob kif haḏaya ḥabbebū-ni fī-l-muṭālʕa the.books as this make_love.perf-me fi-the-literature

‘Books like this one made me love literature.’

Causativization creates complex events because it increases the valency of a predicate by introducing an additional head where the external causer is realized (Hallman 2006; Svenonius 2001; Ramchand 2008; Belletti 2017 a.o.). The position occupied by the external causer is represented in syntax by an additional head projecting a vP layer. Thus, under the assumption that there is a one-to-one correspondence between subevents, arguments and verbal heads, causativization always generates complex events by changing a monadic predicate into a dyadic one, or a dyadic predicate into a triadic structure, as in the examples under discussion.

Assuming that every subevent introduces its own event variable, the availability of three subevents allows the variables projected by the event to be valued by different operators. As we discussed in the previous subsection, this happens in causativized stative constructions because the predicate displays perfective morphology, but the underlying verb is stative, cf. example (14).

Let us go over the decomposition of the event described by the predicate ḥabbebū ‘made love’. Causativized stative predicates correspond to a single verbal token which projects three subevents:

(15) Ev-T1cause Ev-T2love


While the causer subevent is arguably dynamic, the two remaining subevents are stative and, therefore, according to Chierchia’s (1995) proposal, present event variables which require inherent binding by means of a generic operator.

As the author explains, inherent generic binding translates at the interface with semantics in the idea that the state of affairs described by the underlying verb tends to last over time. States, in effect, describe permanent situations which do not come to a natural conclusion. This is represented by the chart in (16).

(16) Ev-T2love




Considering the Klenian system adopted here, perfective reference time includes the entire run time associated with a stative verbal projection, because states are generically bound and, therefore, tend to last beyond the temporal window at which the sentence is evaluated.

Crucially, however, causativized stative verbs are not composed entirely by stative verbal projections because they also include the causative sub-event, which is dynamic and whose event variable does not require inherent generic binding. Thus, the causative sub-event, unlike the lower components of the event structure, relies on aspect to value its variable and to define its relationship with reference time.

(17) Ev-T1cause ?





This means that, in the presence of a perfective operator, i.e. perfective morphology, reference time will be seen as framing the run time of the causative component via event variable binding, but not the run time associated with ‘to love’, whose event variables are in the scope of the generic operator.

The following representation, cf. (18), illustrates this point and depicts the temporal development of the subevents projected by the perfective predicate ḥabbebū-ni ‘made me love’

found in example (14).

(18) Ref-Tperf








The representation stresses the ununiform way in which the run time of the cause-to-love components are understood in the sentence under discussion: this means a situation in which a given reading episode of a given book, i.e. its content, the beauty of the prose etc., caused the experiencer, identified as the first-person pronoun, to love literature.

While the causing event, i.e. the implicit reading event, is understood as being over at reference time, the love subcomponents of the complex event denote a situation which lasts in time, extending presumably at least to the moment of speech.

The claim I will defend in section 4.4 of this chapter is essentially the same as the one put forward in the previous chapter, namely, that theme DP ‘the literature’ occurs in its marked form, i.e. fī-l-muṭālʕa not because marked arguments independently encode the unbounded temporal extension of the love event, but because the event variable associated with the verbal projection that composes ‘to love’ are licensed by an operator, which attributes them a value that contrasts with the one used in the computation of the sentence’s truth conditions. From this, it is possible to state that their relation is “mediated” by syntax.

Sentence (14), in fact, makes a claim about the time when the causing-to-love event occurred; it states that the causing event ended, since perfective binding allows to see the endpoint of the event, but it claims that the love-state held at that moment, without entailing anything about the duration of the states outside of the reference time.

Sentence (14), in fact, makes a claim about the time when the causing-to-love event occurred; it states that the causing event ended, since perfective binding allows to see the endpoint of the event, but it claims that the love-state held at that moment, without entailing anything about the duration of the states outside of the reference time.

Dans le document Event building, selection and non-canonical Case: fi insertion in Tunisian Arabic (Page 146-158)