Superlative expressions, context, and focus

Annali di Ca’ Foscari. Serie occidentale

The goal of this paper is to reconcile the definite marking with the indefinite-like semantics of those superlatives that take a relative/comparative reading. Following Szabolcsi (1986) and Heim (1999), we will assume that the difference between absolute and relative readings of superlatives is represented at the syntactic level of semantic representation, LF (Logical Form). We will however depart from Heim’s hypothesis that what raises at LF is the superlative operator itself (EST). We will instead assume a quantificational-determiner analysis of EST, which involves two raising operations at LF: EST-raising to Spec,DP and raising of the whole superlative DP (Determiner Phrase) to a scope position in the sentence. We will examine the relative readings of quality superlatives. The generalizations and the proposed analysis are not assumed to extend to quantity superlatives.

2014

Various recent discussions of comparatives (Hackl 2000, Heim 2001) note an unexplained restriction on the scope of the comparative operator, which Takahashi 2006 dubs the Heim-Kennedy constraint. (HK) A quantificational DP cannot intervene between a degree operator and its trace. As an example, Every girl is less tall than Jim is ((1) on the following page) does not have the full range of expected readings. If gradable adjectives like tall are functions from individuals to degrees, we should be able to intersect the heights of all the girls, and then take the maximum of this set. This should yield a reading equivalent to "The shortest girl is less tall than Jim". This reading is not available: that is, it is not possible for less to have wider scope than every. Interestingly, the same facts hold with respect to certain types of weak islands. Amount whquestions such as (2) How tall is every player? have a family-of-questions reading in which every scopes above how, but lack the how > every reading (2b), which should mean "How tall is the shortest player?". My modest purpose here is to show that the semantic account of weak islands of Szabolcsi & Zwarts (1993, S&Z) explains the restriction in HK without further ado. This account does not answer all questions about scope restrictions on the comparative operator, but it is a promising beginning. S&Z argue that weak islands are a semantic rather than a syntactic phenomenon. In their theory expressions denoting individuals, manners, numerals, and amounts denote Boolean objects of various types, and scopal restrictions on these expressions are (partly) explained by the operations that are defined on these structures. For instance, they propose that numerals denote in a one-dimensional scale (3a), while amounts denote in the more complex Boolean structure (3b), a join semilattice. On this account, the impossibility of every > how scope in (2) is a result of this difference. In this model existential quantification is equivalent to taking joins (the Boolean counterpart of union), while universal quantification involves taking meets (counterpart of intersection). The important difference between (3a) and (3b) is that (3a) is closed under both meet and join, while (3b) is closed only under meet, and join is undefined (unless the amounts happen to be identical). S&Z claim that this explains the unavailability of the reading in (4b) since the heights of the various students cannot be intersected. Note that this is not a claim about wh-questions, but about the representation of amounts. Thus it predicts that amount-denoting expressions should show these restrictions wherever they appear in natural language, and not only in wh-expressions. The similarities between amount-denoting wh-expressions and comparatives, then, are explained in a straightforward way: certain operations are not defined with amounts because of their algebraic structure, regardless of the other details of the expressions they are embedded in. In the case at hand, the scoping less > every girl > d-tall is not available in (1b) because the intersection of all the girls' heights-max(λd. ∀x[girl(x) → tall(x)(d)])-is undefined. Unlike (HK), this theory predicts that the existential quantifier should be an acceptable intervener. This prediction cannot be tested because ∃ and-er/less are commutative (Takahashi 2006). Several issues remain to be addressed. First, degree operators and weak island-sensitive expressions show similar behavior in a wide range of contexts (Rullmann 1994 gives a detailed typology). Similar explanations may prove fruitful in these domains. For instance, Rullmann's account of negative islands with numerical and amount expressions is replicated because (3a) and (3b) are not closed under complementation (negation). It remains to be seen whether the various semantic accounts of weak islands in the literature (e.g., Abrusan 2007) can be extended to comparative scope restrictions of all stripes. Second, it remains unexplained why certain modals and intensional verbs are able to intervene between a degree operator and its trace both in amount comparatives and amount questions, as in (4) and (5). On the assumption that must and require involve universal quantification over accessible worlds, this should not be possible. S&Z suggest that these operators are acceptable interveners because they are not Boolean in nature. This is perhaps too drastic a step, but a detailed investigation is needed to account for

Scolia

This paper proposes an account for the cross-linguistic distribution of relative readings in different types of superlative constructions. Pancheva and Tomaszewicz (2012) (hereinafter P&T) observe that a relative reading with NP internal focus is available in superlative expressions in some Slavic languages but not English. We look into this reading in greater details and provide a syntactic account for its distribution based on standard locality constraints and semantic assumptions of superlatives.

A traditional issue in the analysis of comparatives is whether or not degrees are essential. In the first part of this paper I discuss the traditional analyses that account for comparatives with (Seuren, von Stechow) and without (Klein) degrees, and remind the reader that these are very similar to each other. A more recent issue is how to account for quantifiers in the than-clause. The traditional analyses account well for Negative Polarity Items in comparative clauses, but have problems with conjunctive quantifiers. The strength of the proposals of Larson (1988) and Schwarzchild & Wilkinson (2002), on the other hand, goes exactly in the opposite direction. I will discuss two types of strategies so as to account for both types of quantifiers: (i) one based on the traditional analysis, but by making use of more coarse-grained models or of intervals, (ii) one where comparatives are taken to be ambiguous between the traditional reading and the Larson-reading, and where the actual reading is selected with the help of the strongest meaning hypothesis. 2 The traditional analyses of comparatives There exist two major types of approaches to the analysis of gradable adjectives: comparison class approaches and degree-based approaches. In this section I sketch the traditional approaches along these lines, and show how close they are to each other. Intuitively, John can be counted as tall when we compare him with other men, but not tall when we compare him with (other) basketball players. Thus, whether someone of 1.80 meters is tall or not is context dependent. Wheeler (1972) and Klein (1980) propose that every adjective should be interpreted with respect to a comparison class. A comparison class is just a set of objects/individuals and is contextually given. In particular if the adjective stands alone, we might assume that the contextually given comparison class helps to determine what counts as being tall. Klein (1980) assumed that with respect to a given comparison class, some elements of this set are considered to be definitely tall, some definitely not tall, and the others are borderline cases. The truth of the positive sentence (1) (1) John is tall. * I have presented parts of this paper at various occasions. At a DIP colloquium in Amsterdam, at the CSSP 2007 workshop, and later at talks given in Utrecht and in Tokyo. I would like to thank the audiences for their remarks. I also would like to thank Chris Kennedy and Roger Schwarzchild for discussion.. Robert van Rooij depends on the contextually given comparison class: (1) is true in context (or comparison class) c iff John is counted as tall in this class. The proposition expressed by a comparative like (2) is context independent. (2) John is taller than Mary. and the sentence is true iff there is a comparison class according to which John counts as tall, while Mary does not: ∃c[T (j , c) ∧ ¬T (m, c)]. 1 According to the degree-based approaches (e.g.

There is a point of similarity between English suffixal superlatives and English past tense. There is the question whether or not comparative -er should be decomposed in parallel fashion to superlative -est, and how that fits in with FOFC effects. And then there is the importance of all the contrasts between 'most' and 'least'.

Journal of Semantics, 1998

The hypothesis that two logical schemes are, more or less directly, involved in the so-called quantificational' readings of superlatives is defended in the present paper. It is argued, in particular, that sentences like, e.g. John can solve the most difficult problem and John cannot solve the easiest problem can be associated with their corresponding quantificational' interpretations, i.e. John can solve any problem and John cannot solve any problem, only in the contexts in which they are uttered, and understood, as the key premises, A and not B, of modus ponendo ponens and modus tollendo tollens, respectively. This hypothesis, it is also argued, gains some generalizations missed in Fauconnier's (I975a, b, I979, 1980) well-known analysis of the relevant phenomena in terms of pragmatic scalarity'. In particular, (i) it can clearly distinguish between contexts in which these quantificational' readings are welcome and contexts in which they are not, (ii) it can naturally account for the alleged similarity' between (existential/universal) any and quantifying' superlatives like the most difficult, the easiest, etc., and, what is important, (iii) it can uniformly characterize the mosaic of the environments in which scalarity' phenomena occur, allowing us to explain what it is about exactly these environments (and not others) that makes them licensers of the quantificational' readings

Linguistics and Philosophy, 1995

In this paper, we examined the acquisition of adjectival superlatives such as 'the biggest painting by Picasso.' Sentences that contain such expressions, e.g., "Sally bought the biggest painting by Picasso," are claimed to give rise to up to three possible readings, cross-linguistically: (i) the absolute (ABS) reading: of the paintings produced by Picasso, Sally bought the biggest one; (ii) the relative reading with NP-external focus (REX): of the people who bought paintings by Picasso, it was Sally who bought the biggest one (and not some other buyer); and (iii) the relative reading with NP-internal focus (RIN): of the paintings purchased by Sally, the biggest one was produced by Picasso (and not by some other painter). While the ABS and REX readings are universally available, the RIN is available only in article-less languages (Pancheva & Tomaszewicz 2012; Shen 2014, to appear). In this paper, we look to corpus and experimental data to determine the kinds of readings that English-speaking children are capable of assigning to such superlative expressions. While the spontaneous production data reveal instances of the ABS reading, they tell us little about children's knowledge of the two relative readings. We thus designed an experiment to investigate the availability of the three readings, the results of which reveal evidence for all three readings in 4-year-olds.

Proceedings of SALT X, 2000