The existential and universal quantifiers in Japanese both consist of two morphemes: an indeterminate pronoun and a quantificational suffix. This paper examines the distributional characteristics of these suffixes (ka for the existential quantifier and mo for the universal quantifier). It is shown that ka can appear in a wider range of structural positions than mo can. This difference receives explanation on semantic grounds. I propose that mo is a generalized quantifier. More specifically, I assume that the phrase headed by mo is of type \({\langle\langle\alpha, t\rangle, t}\). Because of its type, mo cannot appear in certain structural positions without causing type mismatch. Ka, on the other hand, is a choice function variable of type \({\langle\alpha/t, \alpha\rangle}\), and due to its type, its distribution is not as restricted. One of the consequences of this analysis is that there are no quantifier raising or type shifting operations in Japanese that would adequately obviate type mismatch.

The paper concerns the expression of non-strict comparison, focusing in particular on constructions of the form [no(t) . . .-er than] in modified numerals. The main empirical finding is the observation that negated comparatives contrast with regular comparatives in that the former but not the latter can give rise to (scalar) implicatures. It is shown that such a contrast falls out of theories of exhaustive interpretation that claim alternatives to form dense scales. An important result is that the paper sharpens the desiderata for theories of exhaustification.

In this paper, I argue that the temporal connective prima (‘before’) is a comparative adverb. The argument is based on a number of grammatical facts from Italian, showing that there is an asymmetry between prima and dopo (‘after’). On the ground of their divergent behaviour, I suggest that dopo has a different grammatical status from prima. I propose a semantic treatment for prima that is based on an independently motivated analysis of comparatives which can be traced back to Seuren (in: Kiefer and Ruwet (eds.) Generative grammar in Europe, 1973). Dopo is analyzed instead as an atomic two-place predicate which contributes a binary relation over events to the sentence meaning. The different semantic treatments of the two connectives provide an explanation for the grammatical asymmetries considered at the outset; interestingly, they also shed some light on other asymmetries between prima and dopo, which are known to hold for the English temporal connectives before and after as well: these asymmetries are related to the veridicality properties, the distribution of NPIs, and the logical properties of these connectives first described in Anscombe (Philos Rev 73:3–24, 1964).

I argue for three basic classes of nominals, based on the (non)-relation they encode; (i) alienable nouns, which have no inherent relation, but gain an underspecified ‘R’ relation when possessed (Higginbotham, Linguistic Inquiry, 14, 305–420, 1983); (ii) relational nouns, which have an inherent relation, defined as an ‘R’ relation restricted by the lexical meaning of the head noun (Barker, Possessive descriptions. CSLI: California, USA, 1995; Burton, Six issues to consider when choosing a husband. Doctoral Dissertation, Rutgers, New Brunswick, New Jersey, 1995); and (iii) inalienable nouns, which also have an inherent relation, defined as a material part-whole relation (Link, Algebraic semantics for linguistics and philosophy. CSLI: California, USA, 1998). I then consider evidence from the Algonquian language Plains Cree, which overtly distinguishes all three subclasses of nominals.

‘Only’ associates with focus and licenses NPIs. This paper looks at the distributional pattern of NPIs under ‘only’ and presents evidence for the movement theory of focus association and against an in situ approach. NPIs are licensed in the ‘scope’ (or the second argument) of ‘only’, but not in the complement (or its first argument), which I will call the ‘syntactic restrictor’. While earlier approaches argued that ‘only’ licenses NPIs in the unfocused part of the sentence it occurs in except in its focus, evidence from DP-‘only’ shows that NPIs are also not licensed in the unfocused part of the syntactic restrictor. The distribution of NPIs provides a test for the size of the syntactic restrictor, and this test is applied to the case of VP-‘only’. The evidence shows that (i) the restrictor can be smaller than the entire VP and is not necessarily identical to the surface complement of ‘only’; (ii) in the case of association with a head the restrictor comprises an XP containing the head; and (iii) in cases of association into an island, the restrictor comprises the entire island. Generalizations (i)–(iii) can be captured straightforwardly by a movement approach but are incompatible with an in situ analysis. Contextual domain- restriction of the kind used in in situ approaches accounts for the appropriate semantics in cases where the semantic focus is properly contained in the syntactic restrictor of ‘only’.

In earlier work, we developed an composition in which predicates can be composed with arguments by operations other than Function Application, and it makes a difference which composition operation is employed. Here we take our approach further by examining two nonsaturating operations that combine property contents: Restrict, which composes a predicate with the property content of an indefinite; and Modify, which is involved in predicate modification. Nonsaturating operations that combine property contents are often formalized in terms of predicate intersection, which is commutative. Using Austronesian language Chamorro, we argue that Restrict and Modify are not ‘commutative’, but instead incorporate an asymmetry: they take one content to supply a domain that is narrowed further by combination with the other content. Syntactically, it is transparent which category’s content supplies the domain. But semantically, this information can be recovered only from the way in which the composition operation affects the contents that it composes, since—as we show—the same contents can be composed with distinct results.

The aim of this paper is to seek the optimal way to represent time in natural language. It discusses whether or not natural language employs a temporal system that explicitly quantifies over times at the level where semantic interpretation takes place. I first argue that a single-index theory is not empirically adequate for natural language. I then propose a system in which times are syntactically represented. The system works in such a way that tense morphemes saturate the time argument slots of the predicates they attach to. Consequently it predicts that only the times of the main tensed predicates of clauses are accessible. Empirical evidence is presented showing such a distinction between tenseless and tensed predicates in terms of the accessibility to the times introduced by them.

Following Zimmermann (2000), I propose that disjunctions are to be treated as conjunctions of modal propositions, and that the essential contribution of ‘or’ is merely to present a list of alternatives. Any further ingredients in the interpretation of a disjunctive sentence (such as exhaustivity) are due to extraneous factors; they are not part of the meaning of ‘or’. My analysis differs from Zimmermann’s in that it is more general and renders the logical form of disjunctive sentences less complex, but the main innovation is that the context dependence of modality is called upon to play a leading role. The theory applies not only to disjunctions of ‘may’-sentences but also covers universal modalities and conditional disjuncts. The paper concludes with a discussion of narrow-scope ‘or’.

In this paper, I develop a novel interval-based approach to some well-known semantic puzzles related to aspect shift, in particular, to the interaction of for-adverbials with accomplishment and achievement verbs that take indefinite, bare plural, and mass noun complements. My approach is based on the insight that implicit frequentative aspect plays a central role in this interaction, a fact that was largely ignored in previous analyses. Specifically, I interpret frequentative aspect as an abstract verb-level pluractional operator that brings about aspect shift and that is responsible for the distribution of subevent times and subevent participants over the event time of an atelic sentence. What Zucchi and White (2001) call "the aspectual effect of frequency adverbs" thus becomes the general rule for all frequentatively understood for-adverbial sentences. Linguistic support for silent verb-level frequentativity in English is drawn from overt frequentative aspect marking in West Greenlandic verbs.