Dissertation: Underspecific Modals
We rarely speak with complete specificity about what we want, what is required, or what is allowed. For example, I might say ‘I want ice cream,’ and see no need to specify that certain kinds of ice cream, e.g. melted, or toxic ice cream, would not satisfy my desire. In fact, being completely specific about what I want is practically impossible, for in addition to being non-melted and non-toxic, the desire I would ordinarily describe with ‘I want ice cream’ would be for ice cream that contains no dirt, no petroleum, and so on. Being completely specific, it seems, would require explicitly ruling out an infinite number of possibilities. This gives rise to a puzzle: if speakers almost always underspecify their desires, as in this example, how do hearers manage to successfully understand what they want?
This sort of puzzle arises more generally for language in the scope of modals like ‘want’, ’must’, and ‘may’. I can say truly, for example, that you must wash the dishes, or that you may have some wine, without specifying that there are specific ways of washing the dishes, or of having wine, that are not ways of doing what you must or are allowed.
In my dissertation, I aim to give a semantics for modals and related constructions that explains what underspecific modal claims mean and how we interpret them. In doing so, I also propose new solutions to some longstanding problems in semantics and philosophical logic, including Ross’s puzzle, the problem of free choice permission, and the Samaritan paradox.
Papers (contact me for drafts):
▴ Underspecifying Desires Abstract
There is a conflict between the standard view that for any true 'want’ ascription of the form 'A wants p', A has a desire with the content p, and data suggesting true 'want' ascriptions often underspecify the content of the agent’s desire. In this paper, I show that the felicitous inference from ‘Coco wants chocolate ice cream’ to ‘Coco wants ice cream’ is an instance of a generally valid pattern, and argue that the validity of this pattern gives us reason to think that ‘want’ ascriptions can be true and underspecific. This, I argue, means that we should reject the standard view. Then, I propose a semantics for 'want' that both accounts for the validity of the pattern and provides an alternative account of the relationship between true 'want' ascriptions and the contents of agents' desires.
▴ Independent Alternatives: Ross' Puzzle and Free Choice Abstract
Orthodox semantic theories of natural language modals predict that we can make true, underspecific statements about what is necessary or possible. When combined with the standard theory of disjunction, this prediction gives rise to two notorious puzzles: first, Ross’ Puzzle, i.e. the apparent invalidity of the orthodoxically valid argument from 'Must(p)' to 'Must(p or q)'; and second, the puzzle of free choice permission, i.e. the apparent validity of the orthodoxically invalid argument from ‘May(p or q)’ to ‘May(p) and May(q)’. I argue that existing solutions to both puzzles have insufficient scope for two reasons. First, neither puzzle is restricted to deontic modals. Second, modals with disjunctive complements do not merely license the inference that each disjunct is compatible with the relevant set of worlds. Rather, they license the stronger inference that each disjunct is an independent alternative among the relevant worlds. For example, `Must(p or q)' licenses `May(p without q)' and `May(q without p)'. I develop a semantics for modals and disjunction that predicts these entailments, vindicates the intuitive judgments about the two puzzles, and does not depend on any features specific to deontic modality. Furthermore, my semantics retains the duality between possibility and necessity modals, and validates the distribution of un-necessity and im-possibility modals over disjunctions.
▴ Supposition, Presupposition, and Trivalent Conditionals Abstract
The trivalent theory of indicative conditionals — on which `if A, B' is true iff A and B are true, false iff A is true and B is false, and neither true nor false when A is not true — has a lot going for it. Yet even the most sympathetic theorists have thought the theory comes at a high price, given how it appears to interact with the trivalent theory of presupposition. In this paper, I argue that this apparent high price is the consequence of a simplistic view of the relationship between truth value gaps and presuppositions, and that a natural and more sophisticated view of their relationship lets us avoid paying it altogether. I then outline a simple formal model that illustrates this by distinguishing presupposition from supposition pragmatically, even though it treats both as necessary conditions on a sentence's being either true or false.