Some actions are intentional and some are not. For example, Sarah's action of stepping on Susy's toes was not intentional, but Susy's subsequent action of stepping on Sarah's toes was intentional. Whether an action was intentional or not can be important. It can make a difference in our attitude toward the action and in our evaluation of it. What makes an action intentional? Let me rephrase the question. Given that ‘S's V-ing at t» is a canonical designator of an action, what is necessary and sufficient for the truth of ‘S's V-ing at t was intentional (SV-ed at t intentionally)»?

RELATIVITY TO DESCRIPTION

The same action can be intentional under one designator of it and not intentional under another. For example, ‘S's pirouetting at t» and ‘S's pirouetting clumsily at t» designate the same action. But it may nevertheless be correct to say that ‘S's pirouetting was intentional but S's pirouetting clumsily was not intentional.

Some canonical action-designators imply that the action they designate was intentional (at least under that designation), for example, ‘S's calling for help at t», ‘S's looking around her at t», ‘S's groping for the light switch at t», ‘S's deciding to run away at t», ‘S's mentally saying “nonsense” at t», or any of the form ‘S's intentionally V-ing at t». Other designators imply that the designated action was not intentional (at least under that designation), for example, any of the form ‘S's unintentionally V-ing at t».

Our definition of an action in Chapter 1 tells us how to select from among canonical personal-event-designators those that designate actions. But to know what makes an action-designator is not yet to have a complete understanding of what makes an action. For canonical action-designators are not correlated one to one with actions. Distinct designators do not always designate distinct actions. For example, it is natural to think that although «S's willing to exert force with her hand at t» and «S's raising her hand at t» designate distinct actions, «S's raising her hand at t» and «S's slowly raising her right hand at t» designate the same action (given that each of these designators is canonical and thus uniquely picks out a single action). But our criterion for picking out action-designators does not tell us what guides such a judgment. This chapter takes up the task of developing a criterion for deciding when distinct (canonical) action-designators designate distinct actions.

We need to be concerned here only with designators in which the type of the action designated is made fully explicit. Consider a designator of the form «S's doing this morning the same thing she did yesterday that R complained about», one that uniquely picks out a particular action. It picks it out as being of a certain type (the type S did yesterday that R complained about), but it does not, just in virtue of its content, tell us what that type is.

Consider Mutt and Jeff, who agree on what sentences Odile accepts. They agree about her dispositions to behavior. They agree on just about everything that seems relevant to the question Does Odile believe that Twain is dead?

They don't agree on the answer. When Mutt was asked, it was because someone wanted to know whether Odile would list Twain under dead Americans. Mutt knew she accepted ‘Twain is dead’ and thus said yes. Jeff was asked by someone who couldn't understand why Odile, who is pointing to Twain's picture, wants to meet him. Doesn't she realize that Twain is dead? Jeff knew she rejected ‘He's dead’. He answered that, no, Odile didn't believe that Twain was dead.

What are we to make of this? Observe that Mutt and Jeff's utterances are acceptable in the quasi-technical sense of Chapter 2: A party to Mutt's conversation, who knew the relevant facts, would accept his utterance as correct.

Propositions are the sorts of things named by ‘that’-clauses (t-clauses). What would a concisely stated theory about propositions be like? It might begin by saying something about how the simplest expressions in a t-clause contribute to the proposition it names. Usually, a theory does this by explaining how to assign certain things – let's call them ‘contents’ – to such expressions. Next it gives a recipe for getting from the contents of the expressions in a t-clause, and the clause's syntax, to the proposition named.

Chapter 1 said a good deal that was relevant to the second part of such a theory, but not much that was relevant to the first. The conclusions of Chapter 1 strongly suggest that to the question

we should give an answer that goes roughly

It's time to turn to the question

Much of the debate over propositions has revolved around this question. It has generated a distressing number of proposals about content.

Assume that in an attitude ascription, that is, in sentences such as

the ‘that’-clause (t-clause) functions as a term. Call what such terms name ‘propositions’.

One would like to know what sorts of things propositions are. This chapter makes a start at an account; it takes up the question whether propositions have structure. Two views on this predominate. One is that propositions have no interesting structure, even though their canonical names, ‘that’-clauses, do. This view is held, for example, by those who take t-clauses to be names of sets of possible worlds. The other view is that propositions have a structure that more or less recapitulates the structure of a sentence. Those who identify propositions with sentences, Russellian propositions, structured intensions, or certain sets of equiformed sentence tokens are in this pigeonhole.

I think that what a t-clause names is sententially structured; in this chapter I attempt to make a case for that view. I carry on the argument within the framework of possible-worlds semantics. This is mostly for the sake of convenience, although I am sympathetic with the view that the framework is as good a one for natural language semantics as we have.

This is very much a work in the tradition of Frege and Russell. Perhaps as good a way of ending it as any is to remark on a few of the ways in which its conclusions fit in with Russell's and Frege's views of attitudes and their ascription.

Painting with a broad brush, we can contrast Frege and Russell's views on attitudes as follows. Frege thought that attitudes were not the sort of thing that could be characterized simply in terms of the objects and properties that they are intuitively about. Indeed, he thought ordinary objects and properties were simply the wrong sorts of things to use, in saying what someone believes or wants. Russell, by contrast, thought that workaday objects and properties were essential to attitude characterization. In fact, some of his comments about the multiple-relation theory of belief occasionally suggest the view that a correct characterization of an attitude refers to nothing but the objects and properties that it is about.

These differences are reflected in their views of the behavior of expressions within attitude ascriptions. For Frege, expressions governed by an attitude verb shift their reference from workaday objects and so on to the sorts of things appropriate for attitude characterization.

This book is about propositional attitudes – believing, saying, desiring, knowing, and so on – and how we talk about them. Its primary goal is to give an illuminating answer to questions like the following: When someone says

Maggie thinks that Odile is tired

or

Maggie said that Clark Kent is Superman

or

Maggie wishes that Greg would leave her alone

how do things have to stand with Maggie in order for what is said to be true? What makes sentences like these true or false?

This, then, is a book about the semantics of attitude ascription. But it would be difficult to say anything illuminating about the meaning of ‘believes’, ‘desires’, and their friends without saying something substantive about belief, desire, and the other propositional attitudes. And so this book addresses topics in the philosophy of mind, as well as ones in the philosophy of language. Discussed in the following pages are, for example, the nature of the psychological states that are beliefs and desires; the beliefs and desires (or lack thereof) of speechless nonhuman animals; the distinction between tacit and explicit belief; the idea that the sentences of a natural language play, for all its speakers, more or less the same cognitive or conceptual role.

Chapter 3 set out the rudiments of a view of the semantics of attitude ascriptions. It suppressed discussion of complications and subsidiary issues. In this chapter, I address a few of the issues ignored in Chapter 3.

I begin by discussing quantification. The account of quantification into attitude ascriptions I have given violates Leibniz's law, the principle that universal closures of

If x = y, then if … x …, then … y …

are invariably true. This principle has been said to be fundamental to objectual quantification; thus, the fact that my account violates it might be thought to be a defect. I argue that violating the principle is no defect, since Leibniz's law is no law of quantification theory: A language's quantifiers may one and all be objectual without the law being true of it.

INTRODUCTORY REMARKS

In this chapter we examine a serious difficulty for Combinatorialism.

The Combinatorialist scheme depends on all combinations of universals being compossible. It should be possible for a single individual to instantiate any such combination, provided only that the universals so combined are wholly distinct, having no common constituents. For if we do not have this promiscuous compatibility, then we get logical incompatibility of a sort not envisaged by the theory.

If we consider what passes for properties and relations in our ordinary thinking, however, then we find that failures of cornpossibility abound. Consider properties first. These characteristically appear in ranges, so that they form classes of determinates falling under the one determinable. An individual can, at one time, instantiate only one member of this given range. The colour incompatibilities are a notorious instance of this phenomenon. Historically, they seem to have furnished one of the reasons Wittgenstein had for abandoning the metaphysics of the Tractatus.

Problems also arise in the case of relations. Restricting ourselves for the sake of simplicity to dyadic relations, problems are raised by symmetrical, by asymmetrical and by transitive relations. If a has symmetrical R to b, then it is entailed that b has R to a. If a has asymmetrical R to b, then it is excluded that b has R to a. If, finally, a has transitive R to b, and b has R to c, then these two states of affairs entail a third state of affairs: a's having R to c.

PRELIMINARY REMARKS

The Naturalist theory of possibility now to be advanced will be called a Combinatorial theory. It traces the very idea of possibility to the idea of the combinations – all the combinations – of given, actual elements. Combination is to be understood widely. It includes the notions of expansion (perhaps ‘repetition’ is a less misleading term) and also contraction.

It is to be emphasized that the central idea is not original, although naturally I hope I will be making some contribution to the details. The central idea is in the Tractatus, and it is one of the central ideas of the Tractatus. Perhaps its charter is 3.4:

I myself encountered the Combinatorial idea, and was converted, in Brian Skyrms's article ‘Tractarian Nominalism’ (1981), which is reprinted at the end of this book as an appendix. But a Combinatorial conception of possibility was put forward earlier by Max Cress well (1972) and before that at least toyed with, quite a vigorous toying, by Quine (1969, pp. 147–52). The Combinatorial theory of Cresswell and Quine does not involve the fictionalist element that mine will have.

I shall develop the theory in a particular way, a way determined by my own metaphysical views, in particular by my acceptance of (in re) universals (see Armstrong 1978a, b). It seems a peculiarly apt way to develop a Combinatorial theory.

THEORIES OF POSSIBILITY

Every systematic philosophy must give some account of the natureof possibility. The main constraint I wish to place on such anaccount is that it be compatible with Naturalism. The term ‘Naturalism’ is often used rather vaguely, but I shall understand by it thedoctrine that nothing at all exists except the single world of spaceand time. So my objective is to give an account of possibility whichis in no way other-worldly.

A non-Naturalist theory of possibility tries to account for thenotion by postulating entities additional to the world of space andtime, for instance, possible worlds or ‘abstract’ propositions. It willbe useful to begin by criticizing some non-Naturalist theories, concentratingespecially on the account given by David Lewis.

Among the non-Naturalist theories of possibility there is one,not Lewis's, which has a central place. This is because other theoriescan be seen as reactions to it. The theory is popularly attributedto Leibniz, although the textual warrant for this is dubious. But it isconvenient to call it the Leibnizian view.

According to this view, over and above the actual world thereare an indefinite multiplicity of merely possible worlds. They constituteall the ways that the world could have been. Included in theactual world are its past, its present and its future. The actual world contains minds, perhaps matter, perhaps God, perhaps still strangerthings. The actual world is a possible world. The other possibleworlds, the merely possible worlds, are ways that the actual worldmight have been.

HIGHER-ORDER ENTITIES

In his article ‘Tractarian Nominalism’ (1981) Skyrms raises the question, ‘In what sense is Tractarian Nominalism nominalism at all?’ He then remarks, ‘It is certainly not nominalism in the sense of Goodman [“A world of individuals”] or Quine [“On what there is”], since it finds quantification over properties and relations … just as acceptable as quantification over individuals, and cashes both in terms of facts’ (p. 202).

Skyrms's facts, of course, are our states of affairs. Only the terminology is different. And since for Skyrms, as for us, different individuals may have the same property or be related by the same relation, and that in no mere Pickwickian sense, his properties and relations can fairly be described as universals. So why Tractarian Nominalism? Nominalists reject universals.

Skyrms goes on to answer his question thus:

By way of clarification, Skyrms makes an important qualification of this stand. He will admit ‘higher-order’ facts that are logically determined by first-order facts, but:

AGAINST THE LEIBNIZIAN VIEW

The view now to be criticized is that, while ours is the only actual world, there do exist, in some sense of the word ‘exist’, merely possible worlds, set in one-to-one correspondence to, indeed constituting, the ways that the actual world might have been. It may or may not have been the view of Leibniz, but it is convenient to call it the Leibnizian view. It involves two levels of existence: the actual and the merely possible.

There exists a strong argument against this view, first stated, as far as I know, by D. C. Williams. It has since been endorsed by David Lewis, who heard the argument from Williams. I was told about it by Lewis.

Williams says:

MORE UNIVERSALS?

Suppose one is a Naturalist, holding that the space-time world is all there is. Suppose further that one holds that this space-time world has an ultimate structure: It is a conjunction of states of affairs whose constituents are individuals having (universal) properties and relations, the identification of these universals being an a posteriori matter. Suppose, finally, that one holds a Combinatorial theory of possibility, holding, in particular, that all mere possibilities are (non-existent) recombinations of actual elements.

Two difficulties present themselves: First, is it not possible that there should be universals which neither are identical with actual (that is, instantiated) universals nor have as constituents actual universals? Following Lewis (1986a, pp. 91–2), call such universals alien universals. Second, is it not possible that there should be individuals which are neither identical with actual individuals nor put together out of actual individuals? Call such individuals alien individuals. Alien universals and alien individuals seem to be ruled out by our three premisses.

The present section will consider the question of alien universals. I will deny their possibility. The position concerning alien individuals is more complex. I am reluctant to deny their possibility. As a result, a modification of the Combinatorial scheme is forced on me. That will be a matter for a succeeding section; in the meanwhile we address the problem of alien universals.