Conventions, cognitivism and necessity.

Modal locutions are common enough to give modality an air of legitimacy. Frequently, we talk of things being necessary, possible, inevitable, or avoidable. To point out just a few cases, it is reasonably uncontroversial that Jeanna could have scored no better on a particular examination regardless of how hard she tried, having received full marks. Elana might choose hobbies different from those of her sister. Neither deserves their parents, disapproval for actions that were unavoidable. Further examples are easily produced.

It is very hard, though, to see what in the world accounts for necessity. While we readily acknowledge that Socrates was human, whether he must have been human is much less obvious. Could he have had some slight genetic difference, for example, that would have excluded him from humanity but would not have prevented him from performing as he did in Athens? Furthermore, we have defined "human," and some necessities turn on the substance of that definition. This suggests that modality is, in part, a linguistic phenomenon. We construct the categories and set their admission standards. So, modality not only appears to contain a submerged linguistic component, it also seems to contain a conventional element.

The general theory of modality examined in this paper is one in which modality is not something in its own right, but a feature of the world parasitic upon more basic linguistic features. I will call this "The Linguistic Theory" of modality.(1) This theory proposes one kind of answer to the metaphysical question of what, if anything, makes something necessary. For this discussion I take the following questions to be roughly equivalent. What in the world accounts for modality? What are the truth-conditions of modally qualified statements? What is the ontological ground of modality? What is the nature of modal facts? Is modality reducible?

These questions make clear that the sense of "reducible" that interests me is wholly metaphysical.(2) I am interested in the structure of the modal component of modal facts, i.e., modality's hold on reality. Thus, in this sense, I will examine reductive theories of modality in which the reductive base draws upon only nonmodal linguistic phenomena of various types. According to such theories, modality has a hold on reality only to the extent that some, perhaps complex, set of linguistic conditions holds. What makes something necessary or possible is something linguistic. I will argue that Such reductive accounts inevitably fail. Section I presents a general argument to show that modality, if at all linguistic in nature, must also be conventional -- not a function of abstract linguistic entities like concepts or propositions. In Section II argues that the considerations of section 1 generalise to cover even linguistic conventions. Section Ill examines latter-day successors of conventionalism and conclude that they present no refuge for the conventionalist. Finally, Section IV diagnoses two dubious assumptions which account for The Linguistic Theory's attraction and which seem to make epistemological problems for primitive modality loom larger than they ought. Once these assumptions are exposed, the attraction of linguistic theories diminishes greatly. This paper, then, is a partial defense of modal primitivism -- the view that the world has a genuine modal character and that it does not possess this character in virtue of any nonmodal character it possesses.

I. Abstracta

Suppose that modality is a genuine feature of reality inasmuch as there are significant objective modal facts. Suppose further that modal facts are not absolutely basic or primitive facts. That bachelors are unmarried or that Socrates is human are each necessary (perhaps), but not irreducibly so. One version of the Linguistic Theory holds that the objective ground of necessity in the world is found in abstract linguistic entities -- concepts or propositions, which reside in a linguistic version of Plato's heaven.

For de dicto necessities, like Necessarily, all bachelors are unmarried, it is relatively easy to see how the reductionists theory might go. If we think of propositions as structured objects with component concepts, then in this simple case the necessity is grounded in the nonmodal but intensional fact of containment. The concept of bachelor contains the concept of being unmarried. Extensionalists can mimic this explanation because the extension of being unmarried contains the extension of being a bachelor. A general account of possibility can be formulated in terms of the instantiation of abstract concepts and their combinatorial coinstantiations.

Two sources of puzzlement about this version of the Linguistic Theory bear mention, though the central difficulties lie elsewhere. First, standard epistemic difficulties arise for any theory framed in terms of abstracta. Why should we think there are any abstracta? Our utterances, in the main, are meaningful and in some way the linguistic significance of the parts contributes to the significance of the whole. If, as is common, we allow that different words can be used to say the same thing,, then, clearly, the thing said is not merely the words but something "behind" them that they express. Calling the thing said, and differently expressed, a proposition is not metaphysically illuminating, since it is really just an objectual way of saying that distinct linguistic tokens can perform the same linguistic function. Likewise for concepts. Objectifying this phenomenon and, further, placing it in a realm causally disconnected from speakers requires some account of (what must turn out to be) our quite pedestrian access to it. Postulating a faculty of intellectual intuition without some account of how that faculty works serves in this context only to name, but not to explain, the phenomena of linguistic understanding.

Second, there is an oddness in accounting for de re necessities via abstracta. If concepts and propositions are abstract, especially if their abstractness partly consists in their being apart from our space-time world, it is hard to fathom how they could make any difference whatsoever to what does or does not, could or could not, happen to Socrates. Socrates is a spatio-temporal being subject to the normal causal influences of the natural world. Abstract entities, by their nature, are supposed to be causally inert. So, they cannot constitutively or causally prevent Socrates from being Xanthippe, nonhuman, the number 7, or even his own singleton set. If we take seriously the causal ineffectiveness of abstracta, it is hard to make anything of the idea that such entities can, in any related sense, make a difference to the life of Socrates and make it the case that Socrates is necessarily human. Thus, it is hard to grasp how the necessity of ordinary de re claims could consist in facts about abstracta. The linguistic reductionist must minimise this puzzling feature of the program.

The central metaphysical difficulties concern the adequacy of abstracta to provide the nonmodal basis for necessity. Both the intensionalist and the extensionalist must determine whether it is happenstance that the appropriate containment relations hold between the relevant concepts or extensions. If so, the fact that is to ground necessity threatens to be amenable to alteration and thereby undermine the necessity it grounds. If not, the troublesome question is merely postponed. It would be necessary that all bachelors are unmarried because the relevant containment relations hold necessarily. Thus, containment simpliciter is insufficient to ground the necessity of necessary truths and necessary containment yields a circular reduction.

In contending that modal facts reduce to the nonmodal facts of abstract entities, the linguistic reductionist must face the following question. What makes this set of entities the right set for reducing modality? This question is wholly metaphysical and requires no account of our modal knowledge, if we have any. Our question is: In virtue of what are these entities up to the task of grounding necessity? How could abstract concepts, as opposed to any other entities, provide the reductive basis for necessity and possibility? Any reduction of modality proposes that the reducing feature of the world is more basic than that which is reduced. In this case, abstract concepts provide the (relative) groundwork of modality.(4) To do so, though, linguistic facts must limit the facts of modality while facts of modality impose no limit on linguistic reality. Per impossibile, if there were no abstract entities inhabiting a Platonic heaven, there would be no modality. Facts about these entities must be "prior" to modal facts. Thus, there can be no modal restrictions on the number of, nature of, and relations between the relevant abstract objects. These are the facts that dictate which modal facts obtain and upon which necessities and possibilities are parasitic.

The key concern for this approach, indeed for any approach framed in terms of a set of objects, is: What have these objects to do with modality? Why are they even so much as relevant? Simply admitting that there is more to the world than our spacetime continuum no more automatically accounts for necessity than does admitting that there is more to the world than the contents of my toolbox. Surely any theory which reduced modality to the existence of and the relations between objects in my entire garage (rather than just my toolbox) would be a theory which effectively stipulated a meaning for "necessity" and "possibility" which would have no bearing on our initial understanding of those terms. This is just to say that it would not be a proper theory of modality at all. So, not just any old extension of the ontology of space-time will do. What reason can the Linguistic Theorist provide for thinking that this particular expansion of the domain is any more useful than the expansion from my toolbox to my garage?

To highlight the significance of this question, suppose there are unrealizable abstract concepts.(5) A theory which reduced modality to the instantiation of the abstract concepts there happen to be (some of which are unrealizable), or the content of propositions framed in terms of those concepts (some of which are impossibly true), is a theory doomed to go wrong. The theory would sanction things as possible that are not. Thus, a theory, by including "too many" objects in its ontology, may be inadequate to the task of reducing modality.(6) The totality of abstract concepts would form an overly generous reductive base.

The reductionist cannot antecedently restrict the reductive base to realizable concepts and possibly true propositions alone. The circularity is obvious. A plausible defense of the claim that only instantiable concepts and possibly true propositions exist requires a modal theory of abstracta. Thus, to …