Individual-level Predicates and When-Conditionals David Chabot Abstract: The present thesis, as the title suggests, is primarily concerned with the distinction between stage-level predicates (henceforth SLPs) and individual-level predicates (ILPs) and particularly with what we might call when-conditionals. The distinction between SLPs and ILPs was first proposed in Carlson's dissertation in which he divides the class of predicates into individual-level, stage-level and kind-level, roughly pertaining to how a given predicate relates to its subject. According to Carlson, the class of individual-level predicates consists in the stative verbs, all predicative NPs (`be a man', `be a mammal', etc.) and adjectives like `intelligent', `tall ', `blue' (vs. `drunk', `available', etc.). Roughly speaking, individual-level predicates are thought has expressing `permanent' properties whereas stage-level predicates are assumed to denote `transient' or `episodic' ones. For instance, to take only one example from the data, `John is tall in the car.' is odd because tall is an individual-level predicate i.e., roughly speaking, it denotes a permanent property (and so, as has been assumed, cannot be restricted to a location). On the other hand, an adjective like `drunk', which is stage-level, is perfectly fine in a similar sentence: `John is drunk in the car.' Since this first proposal, the SLP/ILP distinction has received a considerable amount of attention and different key properties have been identified in the literature as criteria for the characterization of individual-level predicates. Several authors (including Carlson, Diesing, Kratzer and Chierchia) have claimed that SLPs and ILPs have fundamentally different grammatical properties and that the distinction has repercussions in several modules of grammar. The first two sections will be dedicated to their proposals. In section 2, I will briefly present the various contrasts that have been proposed as characterizing the SLP/ILP distinction. Those contrasts have been considered as genuine grammatical differences by the aforementioned authors. In this section, these contrasts will be presented under this perspective, i.e. the contrasts are presented as being the data to be explained. In section 3, I will present the three main theories aimed at giving a uniform account to the data, namely Carlson's, Kratzer's and Chierchia's. We will begin with Carlson's theory which consists roughly in a sortal distinction. Whether a predicate is stagelevel or not boils down on what it takes as its argument, a stage-level predicate takes stages as its arguments and an individual-level predicate takes individuals; where a stage should roughly be thought as a temporal slice of an individual. In other words, Carlson proposes an ontological difference between two kind of entities in the domain: individuals and stages. The former being `four-dimensional worms' made up of stages. Kratzer and Chierchia's theories do not make an ontological distinction. Kratzer's theory is that stage-level predicates, as opposed to individual-level predicates, have an `extra' place for a Davidsonian argument that should be thought as a spatiotemporal location variables. For example, the logical form of `Mary is drunk' would be something similar to `D(m,l)' (as opposed to `D(m)'), where `l' is the Davidsonian argument. Chierchia, on the other hand, assumes that all predicates have a place for a Davidsonian argument; the difference being that the lexical entry of an individual-level predicate triggers a phonetically covert quantifier in its semantical representation that bounds it. Two main theses will be defended against the aforementioned authors. The first is that the contrasts that have been subsumed under the SLP/ILP distinction form a disparate set. Section 4 will be dedicated to this issue, where the data identified as belonging to the SLP/ILP distinction will be analyzed in detail. As I will show, the observed contrasts are not uniform but rather a collection of related but different distinctions. The second thesis is that the predicates called SLPs and ILPs do not differ in their grammatical properties. As I will argue, the main difference between the two kinds of predicates resides in world knowledge: there is no genuinely grammatical distinction between SLPs and ILPs that reflects some fundamental conceptual split. As a matter of fact, the classification of a predicate as belonging to one class or the other turns out to be highly dependent on world and contextual knowledge, while lexical properties are commonly considered to be far less flexible. The second part of the present thesis will be focused on when-conditionals, one of the main contrasts subsumed under the SLP/ILP heading. When-conditionals are sentences where the when-clause does not set a topic time, as in `When I was a kid,...', but rather serves as a protasis (i.e. the subordinate clause of a conditional sentence). A prototypical example of when-conditionals is a sentence where the situation described in the when-clause held at least twice at different moments and the consequent is said to hold at these times. For instance, (1-a) is a when-conditional whereas (1-b) is not: (1) a. When I go to my office, I usually take my umbrella with me. b. When I went to my office yesterday, I realized I had forgotten my umbrella. In section 6, we will thus briefly discuss some aspects of when-constructions and which properties they have. This should give us a minimal set of requirements for their semantics. Finally, I will propose a minimal account for when-conditionals in a dynamic framework close to Groenendijk and Stokhof's Dynamic Predicate Logic and Veltman's Update Semantics. The idea is to use DPL in a framework able to model world knowledge. The account is minimal in multiple respects. For instance, when-constructions, as will be explained in section 6, have very particular properties with respect to their behavior with tenses and adverbs of quantification like `usually', `always', `sometimes', etc. Problems with the formalization will be pointed out along the way and possible extensions of the framework will be discussed. The formalization of when-conditionals that will be presented here is not aimed to be definitive or complete but a step forward in their understanding.