• Aucun résultat trouvé

This completes our exposition of the propositional language LV. In this chapter, we have presented Gibbard’s notion of hyperplans as a way of accounting for the normative content of judgments. We explored Yalcin’s compositional and intensional implementation of Gibbard’s approach, and we found some problems with it. To overcome those problems, we proposed to think of hyperplans as functions from sets of alternatives to subsets thereof representing the preferred, or chosen alternatives within those sets. We then saw how this hyperplan framework can be offered as a semantics for a simple propositional language enriched with evaluative operators representing evaluative properties and relations over propositions. In the next chapter, we will use this semantics as a template from which to start looking at the behavior of evaluative adjectives in natural language.

Chapter 4

From L V to natural language

Summary

This chapter aims to apply the scalar hyperplan framework presented in Chapter3to evaluative adjectives across the board. The chapter is divided in three parts. The first explores how to map the propositional language explored in chapter 3 onto all-purpose evaluatives likegood,bad, andbetter, when these take propositional complements. The second part is devoted to both applying this semantics

to uses of those adjectives with other types of arguments. The last part of the chapter is dedicated to exploring further semantic properties of evaluative adjectives, such as the relative-absolute distinction

and the internal properties of their scales.

4.1 Introduction

The formulae that result from applying the operators introduced by LV, namely ⇑, ⇓, ≥, >

and≈, are intended to represent evaluative sentences of natural language. Two features ofLV introduce an initial constraint in this project: first, the evaluative operators ofLV behave like relative-standard adjectives. This is shown, for example, by the fact thatϕ>ψ does not entail

ϕnor ⇓ψ. Indeed, we have been assuming that evaluative adjectives are relative-standard, and we have used that observation to argue for our hyperplan semantics (later in this chapter we will find reasons to challenge this assumption, but for now let us set it aside). Secondly, the arguments of the evaluative operators ofLV are propositions. This means that the evaluative formulae of LV most accurately represent evaluative adjectives’ lives as sentential operators, such as in the following examples:

(4.1) It is bad that you sent that e-mail.

(4.2) It would be good if you also took the trash out one of these days.

(4.3) What he did to her is so much crueler than what she did to him!

In all these cases, the arguments of those adjectives are propositions (or sets thereof). Our claim in the first part of this chapter is that the evaluative formulae ofLV represent an essential semantic component of those sentences, so we will start by giving entries for thin evaluatives (good, bad) taking propositions as arguments, based on the operators ofLV.

However, evaluation in natural language takes many forms, and in the second part of this chap-ter we will see that considering other kinds of evaluative sentences quickly complicates things.

We will enrich our approach in three (interrelated) directions: the first is to consider how to apply the hyperplan semantics to objects of evaluation other than propositions, chiefly action types and individuals. The following are examples of evaluatives applied to action types:1

(4.4) Volunteering is good.

(4.5) Stealing is bad.

This extension of our proposal is unproblematic. It simply requires treating the grammatical subject of those sentences as containing a silentP ROpronoun, which we can fill out in various ways. A sentence like (4.4) will turn out to be equivalent to a sentence likeit is good thatP RO volunteers, whereP ROmight be interpreted in various ways.

Applying the hyperplan semantics to individuals, on the other hand, requires further tweaks. In a nutshell, we will say that a sentence where an evaluative adjective is applied to an individual does two things: (i) it predicates certain non-evaluative and potentially very underspecified properties of that individual, and (ii) it avows a preference for those properties—importantly, not for the individual nor for the proposition that the individual has those properties—over certain alternatives.

(4.6) This car is good.

(4.7) The mushrooms are even worse than the spinach.

In other words, the evaluation of an individual will be factored out into a regular, descriptive statement about the properties of that individual, together with an evaluation of those proper-ties, which essentially follows the model for the evaluation of action types. An utterance of (4.6) attributes certain underspecified descriptive properties to a car and expresses, or avows, a preference for those properties over certain alternative properties.

The second direction in which we will enrich our proposal is applying it to evaluatives of other flavors, in particular aesthetic adjectives and predicates of personal taste (PPTs), such as the following:

(4.8) This car is beautiful.

(4.9) This soup is disgusting.

Aesthetic adjectives and PPTs are for the most part predicated of individuals or action types (rather than full propositions), so our proposal will essentially follow the model of evaluatives applied to individuals or action types. The fact that these predicates introduce particular flavors of evaluation will be cashed out by introducing restrictions on the type of non-evaluative prop-erties that these adjectives attribute to their objects: whereas a sentence like (4.6) predicates certain non-evaluative and underspecified properties of a car, (4.8) does basically the same thing, but in addition it introduces certain restrictions on the type of non-evaluative properties that are predicated of the car—in the case of aesthetic adjectives and PPTs for example, those properties have to be perceivable. In other words, evaluatives of particular flavors reduce or constraint the underspecificity in the descriptive component of these adjectives.

1By action types, we mean things likestealing,loitering or volunteering. Evaluation of action-tokens, by contrast, fall under the propositional model (i.e., (4.3)).

Finally, the third way in which we are going to expand our account are thick evaluative ad-jectives. Thick adjectives, as we saw in the introductory chapter, are those that carry in their meaning a descriptive and a evaluative component that are often taken to be separable from each other. These are some examples:

(4.10) Marielle Franco was very brave.

(4.11) Nathalie is extremely generous with her time.

(4.12) Hiding your office mate’s keys is cruel.

Beinggenerous, for example, has the descriptive meaning of giving beyond what is due. But in addition to this, calling a persongenerousinvolves a positive evaluation, which forgenerous can be paraphrased asbeing good in virtue of giving beyond what is due.

To extend our account to thick terms, all we need to do is plug our story about thin evaluation into the evaluative component of these terms. So when we call someonegenerous, we are both saying that she gives beyond what is due, and in addition to this, we would be expressing a pref-erence for the type of action that makes her generous (namely, giving beyond what is due). In our view, the difference between thin and thick evaluatives (including thin adjectives of specific flavors of evaluation) is not that the former have one less component of meaning than the latter.

At least when evaluative adjectives are applied to individuals (as they are most of the time) thin and thick adjectives both involve a descriptive and an evaluative component. The difference between them, then, lies in the fact that thin evaluatives admit a much higher underspecificity in their descriptive component. Whereas agoodperson can have any combination of properties whatsoever, agenerousperson is invariably someone who gives beyond what is due. And in addition, in both cases the speaker is also expressing a preference for those sets of properties over certain alternatives.

Crucially, what we say here makes only one assumption about the debate on thick adjectives, namely that the descriptive and evaluative component of these terms can be distinguished. But it remains neutral with respect to the status of the evaluative component—in our case the hyper-plan component, that is, whether it is an entailment, a presupposition or an implicature (or yet another thing altogether; see Cepollaro and Stojanovic2016; Väyrynen2013for discussion).

Finally, in the third part of this chapter we will explore further scalar properties of evaluatives, some of which appear to be shared among them and some of which not so clearly. In particular, we will consider their behavior with respect to different types of adjectival modifiers, such as measure phrases (# $200 generous), ratio (twice as beautiful) or so-called endpoint modifiers (slightly / perfectly cruel), which offer hints about the nature of their scales and the relation to their thresholds for the positive form (whether they are absolute- or relative-standard ad-jectives). Furthermore, we will consider the behavior of these adjectives with respect to PPs denoting comparison classes, and we will tackle the question of their scale structure. The most common types of scale used in measurement theory areORDINAL,RATIOandINTERVALscales, and we will consider arguments and data points in favor of assimilating evaluative adjectives to each scale type. We will conclude that there are reasons both in favor of adopting an interval and a ratio scale for evaluative adjectives, and we will remain undecided.

This chapter is structured as follows: In §4.2, we will consider how to extend LV to natural language. We will first discuss how to move fromLV, where the alternatives relative to which the unary evaluative operators are defined were held fixed, to natural language, where the rel-evant alternatives change with context (§4.2.1, §4.2.2). Then, we will give truth conditions to

sentences containing thin evaluative adjectives taking clausal arguments (§4.2.3). In section

§4.3we will look into various possible extensions of our proposal. In §4.3.1we will see how to account for thin evaluative adjectives when they take objects of evaluation other than proposi-tions, in particular action-types and individuals. We will then see how to extend our proposal to thin evaluative adjectives of specific flavors of evaluation (§4.3.2); and then to thick evaluatives (§4.3.3). We will then discuss our proposed solution to the Frege-Geach problem in both its

“supra-” and “sub-sentential” version, which we introduced in Chapter2(§4.4). After that, in the last two sections of the chapter we will look at further scalar properties of evaluative adjec-tives, in particular the relative-absolute distinction §4.5and the question of their scale structure (§4.6). §4.7recaps and concludes.