• Aucun résultat trouvé

3.4 Previous analyses of Finnish additivity

3.4.1 Vilkuna, 1984

Vilkuna (1984) proposes an analysis of the bound additive–kinthat is couched within Carlson’s (1983; 1984) theory of dialogue games. Vilkuna argues that the common core of all uses of–kin lies in its function as signalling thecompletion,developmentorsubstitutionof a previous public or private game move (background). In what follows, I briefly sketch out what the account looks like, and how Vilkuna’s work advanced the analysis of additivity in Finnish.

Let us begin with Carlson’s theory of dialogue games. In this framework, discourse partic-ipants can be thought of as players involved in a game. Assertions, questions, and orders – in other words, speech acts – function as the moves of the game. Vilkuna (1984) proposes that in such discourse games, the role of the additive–kinis to elaborate and complete previous moves, made either by the speaker themself or by another player (i.e. discourse participant). The formal-isation of this function of–kinis shown in (152). In the example,X–A–Y represents a sentence schematisation with a background frame (X– ... –Y) and a focalised elementAin the middle.

(152) Dialogue game rules for–kin(Vilkuna, 1984, p. 395, 400)

a. When another player has performed a background move of the sentential formX–A–

Y, any player can complete this move with a move of the formX–B kin–Y

b. A player who wishes to make the move X–A–Y may choose to develop this private move further and make the stronger, more informative moveX–B kin–Y instead The additive inference is derived using (152a). Thus, the use of–kinsignals that a previous move was incomplete, and provides a completion for it. An example of this use – a basic use, to be precise – is spelled out for (153a) in (153b).35

35Vilkuna proposes that the elaboration of a private move, derived using (152b), is the essence of scalar additivity;

(153) Analysis of standard useKINfollowing Vilkuna 1984

‘Joni fixed a bike (too)’

b. There was a previous public move of the form ‘x fixed a bike’; this move is completed The perspective adopted in the dialogue theory approach to additivity is very similar to the one adopted by Beaver and Clark (2008) (see section 3.3.2). Clearly, when Vilkuna (1984) discusses

"completing" moves, the completion happens with respect to the current QUD, to which the com-pleting move adds another answer.

The main innovation that Vilkuna proposes concerns the interpretation of–kinin double con-trast contexts (see section 3.2.3; cf. the variation question). Vilkuna’s analysis of double concon-trast additivity relies on the idea that a move can be seen as completing another move if and only if there is some common premise that both moves exemplify or defend. For example, (154) can be analysed as signalling that a previously made move that is an exemplification of a common premise was either incomplete, or not the only possible exemplification. Formally, Vilkuna sug-gests that this use of–kininvolves multiple foci (i.e. two alternative-inducing expressions). The shared premise evokes an explanation, which in turn evokes a main question. The main ques-tion can be answered by answering two subquesques-tions, which are formed by using the subjects that are implicitly present in the explanation. These subjects –motherandfatherin (154) – act as thematic foci, which provide new information in the context of the main question, but are still im-plicitly present in the context. In (154), they could be considered to be contrastive topics; indeed, the two-level question structure shown in (154) is one of the defining properties of contrastive topicality (Büring, 1997, 2003, 2014; see section 2.1.3). The rhematic foci, i.e. the discourse-new material, which in (154) correspond to VPs, answer the subquestions.

(154) a. Kukaan

‘Nobody will play with me. Father is sleeping, and mother is washing windows’

in this case, the new move provides information that is more informative than the initial private move. Informativity is inversely correlated with likelihood: the morepis likely, the less it is informative – and the less it is surprising. Thus, the inference that results from the use of (i-a) can be spelled out as in (i-b):

(i) a. Joni-kin

‘Joni fixed a bike (too)’

b. There was a possible private move of the form ‘xfixed a bike’ which is less informative than the chosen, developed move (given that Joni is a very unlikely bike-fixer compared to allx)

3.4. Previous analyses of Finnish additivity

b. [Premise:] Nobody will play with me

[Explanation:] Everybody is doing something else [Main question:] What are they doing?

c. [Subquestion 1:] What is father doing?

[Subanswer 1:] Father is sleeping (Father is doing something else) [Subquestion 2:] What is mother doing?

[Subanswer 2:] Mother is washing windows (Mother is doing something else) As will be shown in section 4.6, Finnish is particular in that it requires some extra context for dou-ble contrast additivity to be felicitous; in Turkish, for example, doudou-ble contrast additivity seems to be felicitous even in its absence. It is possible, then, that Turkish double contrast additivity is more closely related to the conjunctive or coordinating use of additives, and perhaps involves broad focus. In section 4.6, I argue that Finnish double contrast additivity does not involve broad focus. However, broad focus additivity also exists in Finnish, as the next section will show.

3.4.2 Vilppula, 1984

Within the same issue of the journalVirittäjäas Vilkuna’s dialogue-game analysis of–kin, Vilp-pula (1984) proposes an analysis of–kinthat attaches different conventional and conversational implicatures to it (Grice, 1975). The empirical focus of Vilppula’s paper is on cases where–kin cannot be paraphrased with the unboundmyös. On a closer look, all of Vilppula’s examples il-lustrate the possibility of broad focus association with Finnish bound additives. In (155), for ex-ample, Vilppula argues that the bound additive signals that the leaving of the stork is one among many signs of spring (i.e. the relevant QUD could beWhat signs of spring are there?). Crucially, (155) does not necessarily mean that somex(distinct from the stork) left.

(155) Broad focus use of–kin Kurki-kin

stork.NOM-ADD

jo already

läht-i

leave-PAST.3SG

‘The stork left’

As mentioned above, the analysis that Vilppula provides of additives involves both conventional and conversational implicatures. Let us begin with the former. Vilppula argues that–kinhas two conventional implicatures. The first encodes the comparability or parallelity between the host sentence and some other parallel sentence, where the latter could only be known to the speaker. If the possible ‘privateness’ of the parallel sentence is not considered, this conventional implicature corresponds to what we have seen in previous sections. The second conventional implicature, however, encodes the familiarity of both the speaker and the hearer with the host sentence of the additive. This implicature is related to section 2.3.3.5; as Vilppula notes, it relates

bound additives to the discourse particle–hAn, which has been argued to mainly express the familiarity of a proposition among the discourse participants (e.g. Hakulinen, 1976).36

The conversational implicatures that Vilppula (1984) discusses include neutrality, temporal-ity, and face-saving. The neutrality implicature arises because the proposition expressed by the host sentence is marked as ‘one among many’. Specifically, Vilppula argues that signalling the presence of other alternatives downgrades the importance attributed to the prejacent. The tem-porality implicature is present in examples with past tense; given that the host sentence is con-ventionally implicated to be familiar, it is taken to express events that happened longer ago than those that are described by sentences without the familiarity-marking additive. Finally, the face-saving implicature that Vilppula proposes for–kinis again related to the familiarity implicature.

If a speaker is unsure whether the hearer knowsp, it is safer to statep with a familiarity marker:

in case the hearer does knowp, they could be offended by being assumed to not knowp. While the propositional content of the host sentence of bound additives may indeed be fa-miliar to the speaker and the hearer, as proposed by Vilppula, it must bediscourse-new; in other words, it may not have been uttered earlier in the discourse. Moreover, the information contained in the host sentence may also be previously unknown to the hearer. One example of such use of broad focus–kinis given in (156) (example from Vilkuna, 1984, p. 401).

(156) Hearer-new host sentence of broad focus–kin Anni says to the Frog Queen:

Rauhoitu.

calm.down-IMP.2SG

Kruunu-si-kin

crown.NOM-PX/2SG-ADD

o-n

be-PRES.3SG

vino-ssa crooked-INE

‘Calm down. Your crown is crooked’

Both Vilkuna (1984) and Vilppula (1984) propose that in cases such as (156), the speaker has mul-tiple signs of unrest in their mind, but decides to make one of them public; the others may or may not be as important (by the neutrality implicature). However, it is not necessary for the Frog Queen to have already noticed that their crown is crooked for (156) to be felicitous.

In sum, Vilppula (1984) provides an insightful discussion of the discourse effects of bound additives, relating their use to another discourse particle,–hAn. One important observation that Vilppula makes is that bound additives may be used to express parallelism between what may be privatepropositions and the content of the host sentence; such use of–kinthen either prompts the hearer to ask what else the speaker had in mind, or to simply note that the speaker did have some parallel alternative in mind (cf. section 3.3.1).

36Vilppula notes that the speaker may sometimes take advantage of the familiarity implicature: for example, the speaker might use bound additives as a means to present news that are unfamiliar to the hearer as actually being familiar, in which case the ultimate purpose of the speaker could be to induce surprise or astonishment in the hearer.

3.4. Previous analyses of Finnish additivity

3.4.3 Holmberg, 2014

The most recent and modern investigation of the syntax of two focus-related particles in Finnish, i.e. the question particle–kOand the additive–kin, is due to Holmberg (2014), who adopts a Min-imalist perspective on the topic and proposes an analysis that relies on feature-checking (Chom-sky, 1995, 2001).

The gist of the proposal that Holmberg puts forth is that the two particles are very much alike in their featural make-up: both have an uninterpretable or unvalued focus feature,[u F]37, which needs to enter in an Agree-relation with an interpretable counterpart,[i F]. For this to be possible, Holmberg proposes that–kinand–kOmust both c-command an expression carrying[i F]. The main difference between–kinand–kOis that in addition to the uninterpretable focus feature, -kOalso has an interpretable interrogative feature,[i W h], which enters in an Agree-relation with a[u W h]-carrying interrogative C0. This head forces the–kO-constituent to move to its specifier due to an[E P P]feature (Chomsky, 2000).38

In short, Holmberg proposes that–kinis able to merge anywhere in the syntactic structure as long as it c-commands an expression that carries[i F](which we may assume translates into F-marking, at least roughly (Selkirk, 1996)). This constraint is thus in line with the standardly ac-cepted idea that focus-sensitive operators must take scope over their F-marked associate (Jack-endoff, 1972; Rooth, 1985; Tancredi, 1990). Holmberg mainly discusses the placement of–kin inside complex KPs, although he notes in passing that–kincan also attach to other categories, for example to the finite verb.

Holmberg’s syntax for–kinis shown in (157). The example involves a complex KP (KasePhrase) whose head is the nounisä‘father’, and a possessor KPOllinsits in its specifier. I have kept Holm-berg’s original labels in the tree shown in (157b). In this particular example,–kindoes not directly adjoin to the F-marked associate, but to the more complex KP. It could, of course, adjoin toOllin directly. In both cases, the c-command requirement would be fulfilled.

(157) Syntax of–kin (Holmberg, 2014, (47))

a. Olli-n Olli-GEN

isä-kin

father.NOM-ADD

‘Olli’s father too’

b.

37Holmberg uses the[F o c]-variant of the focus feature; for consistency, I retain the earlier[F]-notation here.

38Holmberg suggests that the two particles also differ slightly in where exactly in the structure they may be ad-joined (although the c-command requirement between[uF]and[F]must always be established). As we are here more interested in–kinthan–kO, I refer the reader to Holmberg 2014 for the technical details regarding this difference.

kin[u F]

Although Holmberg is not very explicit on the issue, the assumption seems to be that–kinis pro-nounced as attached to the category it finds itself adjacent to (i.e. no movement of F-marked NPs or KPs is assumed). One potential problem with the structure shown in (157b) – which Holmberg acknowledges – is that this structure involves right-adjunction, which is banned under the Linear Correspondence Axiom (Kayne, 1994).

Under Holmberg’s analysis,–kinis not restricted to adjoining directly to itsF-marked asso-ciate. This gives rise to the following attachment possibilities in complex KPs with narrow focus on one part of the KP (the parantheses indicate possible separate attachment sites for–kin, but only one–kincan be realised at a time):39

(158) Position of–kinin complex KPs (Holmberg, 2014, (42)–(44)) a. [[[ Olli-n-(kin)

‘in Olli’s father’s car too’

b. [[[ Olli-n-(*kin)

‘in Olli’s father’s car (too)’

39One question that these data raise is whether it is possible for F-marking to project in e.g. (158a). For some reason, the answer seems to be negative: as (i) shows, the additive presupposition of–kinseems not to be satisfied when the antecedent is "fully" different from the complex KP, unless the whole KP isF-marked.

(i) a. #Voi-mme

Int. ‘We can travel to Lapland by plane. In fact, we can also travel there by Olli’s father’s car’

b. Voi-mme

Olli-GENfather-GENcar-ADE-ADD

‘We can travel to Lapland by plane. In fact, we can also travel there by Olli’s father’s car’

3.5. Summary

c. [[[ Olli-n-(*kin) Olli-GEN-ADD

] isä-n-(*kin) father-GEN-ADD

] auto-lla-(kin) car-ALL-ADD

]

‘in Olli’s father’s car (too)’

Although the constraints on specific placement within complex KPs will not be discussed in detail in the rest of this dissertation, and the exact syntax that Holmberg proposes for bound additives is modified in chapter 5 (in particular, in section 5.1.2, I propose that the surface realisations–kin and–kAAndo not correspond directly to[u F]-carrying operators, but simply mark the presence of such an operator), Holmberg’s analysis is nevertheless a clear source of inspiration for the pro-posal put forth in this dissertation.

3.5 Summary

I began this section by introducing eight issues that are relevant for the investigation of additivity:

1. Additives arefocus-sensitive, which means that the form/content of an additive presup-position depends on the placement of F-marking within the prejacent (or host sentence).

This type of dependency can be captured formally in e.g. Alternative Semantics (Rooth, 1985, 1992).

2. Many approaches to additivity assume that they make reference to anantecedent. The de-bate is not centered around the question of what form of the antecedent has – which is a dependent onF-marking – but what kinds of antecedents are acceptable for the felicitous use of additives, if they indeed are required at all. Analyses vary in whether they simply state the additive presupposition as an existential statement, or whether some salient an-tecedent (individual or proposition) must be present in the context (which for most authors means the common ground).

3. A standard assumption in the additivity literature is that the antecedent must bedistinct from the prejacent. The distinctness condition is often hardcoded into the semantics of additivity. However, the distinctness condition may also be a pragmatic implicature based on the expected informativity of utterances.

4. In some languages, the lexicalisation of additives involvesform alternation. This means that the analysis of additivity in such languages must determine the constraints and reasons behind the distribution of the different lexicalisations. The two main ways to deal with this issue are the scope approach, where e.g. tooandeitherhave the same semantics but eitherscopes over negation, and the polarity approach, whereeitheris a polarity item with a semantics distinct from that oftoo, andeitherscopes below negation.

5. More generally, we might ask whether the additives that we see in surface syntax are also the additives that are at work in the semantic representation. While some authors assume

that this is indeed the case, others have proposed that the visible additive is only a marker of a silent operator. This issue was dubbed thederivationquestion in this chapter.

6. Although almost all work on additivity focuses on the basic use of additivity, there is con-siderablevariationin the functions that additives can take on, and moreover, that range is subject to cross-linguistic variation. There have only been some formal attempts to con-nect the less-studied uses of additivity to the basic use.

7. The broad consensus in the literature on additivity is that additives arepresuppositional.

While the standard way to analyse the presupposition is as an definedness or admittance condition, analyses couched within the binding or speaker presupposition approach also exist. In recent years, proposals that take additive meaning be at-issue have emerged, some of them from the exhaustification framework.

8. Finally, although not all approaches discuss this question, the distribution of additives has been argued to be constrained not only by the form and source of the antecedent, but also by their discourse function. As similarity markers, additives seem to be obligatory in some contexts, and optional in others.

After reviewing these issues, I proceeded to present and discuss previous approaches to additivity based on a classification into existential and anaphoric approaches.

I begun in section 3.2.1 by presenting the classic existential approach of Karttunen and Kart-tunen (1976) and KartKart-tunen and Peters (1979), and the more modern existential approach of Rull-mann (2003). While the two approaches are similar in their answer to the antecedency question, they differ in their answer to the form alternation question: the former represents the scope ap-proach, and the latter represents the polarity approach. I then presented two recent defences of the existential approach against the anaphoric approach (Kapitonov, 2012; Ruys, 2015) in sec-tion 3.2.2. Of these two approaches, the former is particularly interesting in that it views additive presuppositions not as admittance or definedness conditions, but as speaker presuppositions (Stalnaker, 1973, and subsequent work). Finally, in section 3.2.3, I presented some other new in-stantiations of the existential approach (Zimmermann, 2015; Szabolcsi, 2015, 2017). I then moved on to the anaphoric approaches, beginning with two classic anaphoric approaches to additivity (Heim, 1990, 1992; Kripke, 1990/2009). This was followed by a presentation of a number of other anaphoric approaches, all couched within a different formalism (Geurts and van der Sandt, 2004;

Abrusán, 2014; Ahn, 2015; Beaver and Clark, 2008).

Finally, in section 3.4, I presented previous work on additivity in Finnish (with the exception of the proposal of Karttunen and Karttunen (1976), which was discussed in section 3.1.1). Vilkuna’s (1984) dialogue game analysis of the basic use of–kinis complemented by an insightful analysis of the double contrast use, while Vilppula (1984) focuses on discussing cases of broad focus and –kin. From these semantico-pragmatic proposals, I moved on to the syntactic analysis of–kin proposed by Holmberg (2014).

3.5. Summary

The take-away message of this chapter is the following. While additivity is generally presented as a phenomenon that has a standard analysis – either in existential or anaphoric terms – there is in fact considerable variation in how it is analysed. If one takes additive meaning to be presup-positional, for example, there are analyses relying on admittance conditions (Heim, 1990, 1992), presuppositional binding (van der Sandt and Geurts, 2001; Geurts and van der Sandt, 2004), and speaker presupposition (Kapitonov, 2012). In fact, Karttunen and Karttunen (1976) and Kart-tunen and Peters (1979) formulate their analysis in terms of conventional implicature, and not presupposition; indeed, the authors identify specific existential statements that they take sen-tences with additives to convey.40Moreover, as this chapter has shown, it is not even universally accepted that additive meaning is presuppositional or non-at-issue: Ahn (2015), for example, explicitly argues for a truth-conditional analysis where the truth of some antecedent for the pre-jacent is asserted, and not presupposed.

As this chapter has made clear, the literature also disagrees on the right solutions to the issues

As this chapter has made clear, the literature also disagrees on the right solutions to the issues