Synthese, Volume 166, Number 2 (January 2009), pages 231–250.

Main authors discussed: Herman Cappelen, Keith DeRose, Nikola Kompa, Ernie Lepore, Jason Stanley

On what is arguably the standard view in contemporary philosophy of language, an expression is context sensitive if its semantic contribution to what a sentence containing that expression says or expresses when asserted in a context of use depends on features of that context. Personal pronouns are a prime example. When I utter:

‘I am hungry’,

I express that Iris is hungry. When Paul utters the same words, he expresses that Paul is hungry. Our utterances have the same linguistic meaning but express different contents. Contextualism is of interest beyond the narrow confines of the philosophy of language because it has been maintained with respect to less obvious kinds of expressions, such as moral terms (“is morally permissible”), metaphysical vocabulary (“exists,” “is an object”), epistemological vocabulary (“knows”), and evaluative vocabulary (“beautiful,” “tasty,” “funny”). Thus, when I utter:

‘Pre-marital sex is morally permissible’,

I say something like pre-marital sex is morally permissible on the moral standards in place in my community. When I utter:

‘Sets exist’,

I say something like sets exist according to my conceptual schema. When I utter:

‘Paul knows he has hands’,

I say something like Paul knows he has hands on the justificatory standard currently in place.

Whether, and in what form, contextualism is correct about some philosophically significant range of vocabulary has obvious implications for those areas of philosophy that essentially draw on that range.

John MacFarlane diagnoses that the standard conception of context sensitivity conflates two concepts that are profitably kept apart:

1. An expression is context-sensitive iff its extension at a context depends on features of the context.
2. An expression is indexical iff its content at a context depends on features of the context.

He shows that both proponents and critics of contextualist accounts tend to assume that the only way the extension, i.e.  the truth value of a sentence, can depend on the context of use is by expressing different propositions in different contexts where these different propositions determine different truth values. Teasing apart indexicality and context sensitivity opens conceptual space for a new position, non-indexical contextualism, according to which a sentence expresses the same propositional content in different contexts of use where that content determines different truth values in different contexts.

The distinction is nicely illustrated by contrasting an indexical contextualist (”eternalist”) with a nonindexical contextualist (”temporalist”) account of tensed sentences. Suppose Socrates is sitting at t and not sitting at t’ and consider two utterances of

‘Socrates is sitting’,

one utterance (Ut) at t , the other utterance (Ut’) at t’. According to eternalists, Ut expresses that Socrates is sitting at t – a proposition that is eternally true – and Ut’ expresses that Socrates is sitting at t’ – a proposition that is eternally false. According to temporalists, Ut and Ut’ express the same time-neutral proposition – one which is true at t and false at t’.

MacFarlane goes on to show how the conflation of the two concepts has muddied the disputes between proponents and critics of contextualism. For example, critics charge that epistemic contextualists cannot do justice to cross-context reports of knowledge-attributions. Suppose Sam says in a context governed by one epistemic standard, “I know that my car is in the driveway.” If Barry says, in a context governed by a different standard, “Sam asserted that he knew that his car was in the driveway,” we take him to correctly report what Sam asserted. But, says the critic, the contextualist has Barry attribute to Sam the assertion of a different content than the one that Sam expressed. The nonindexical contextualist can avoid this charge. Another example: Some contextualists take so called context shifting arguments to establish that a given sentence expresses different contents in different contexts. Suppose S is true when uttered in C1 but false when uttered in C2. Doesn’t that show that S expresses different contents in C1 and in C2? “No,” says MacFarlane. All it shows is that S is context sensitive. And that is compatible with expressing the same proposition in both contexts.

MacFarlane considers a series of considerations against the idea, crucial to nonindexical contextualism, that propositional truth may be relative to more than a world. For instance, against the Fregean charge that such a proposition would be “incomplete” he argues that we cannot in general require that all the features that propositional truth depends on be factored into propositions. For instance, the truth of a sentence that expresses a contingent proposition on an occasion of use depends on what world that use of the sentence occurred in, yet worlds are typically not factored into propositions (except when the sentence expressing the proposition involves, say, an actuality operator). If they were, every proposition would be necessarily true or necessarily false. But then incompleteness as such cannot be a problem. Against Jason Stanley’s objection that we should accept features that truth is relative to only when the language we are interpreting contains operators that can shift this feature, MacFarlane argues that there is no good reason for such a requirement. On the contrary, even if we spoke a language without any modal operators, we’d still want propositional truth to be relative to worlds.

Finally, MacFarlane argues that the nonindexical, unlike the indexical, contextualist has to replace the principle:

(P) An utterance of S at C is true iff the proposition expressed by S at C is true,

with a relativized version:

(P’) An utterance of S at C is true iff the proposition expressed by S at C is true relative to C.

This is in order to avoid contradiction: If Janet’s utterance of ‘S’ in C1 is true and Barry’s utterance of ‘S’ in C2 is not true, where both utterances of ‘S’ express the same proposition p, then (P) would imply that p is both true and not true.

This is an an interesting paper that offers attractive semantic options not just to temporalists and relativists about matters of personal taste but also to moral and ontological relativists.

Reviewed by Iris Einheuser
Duke University

Main authors discussed: Gareth Evans

Gareth Evans’ argument against the coherence of vague identity presents a well-known obstacle for theories of vague objects.  The main purpose of Williams’ paper is to sketch a new model of ontic vagueness on which Evans’ argument can be blocked. Evans’ argument runs as follows:

1. Assume: It is indeterminate whether a is identical with b.
2. Then:  a has the property of being indeterminately identical with b.
3. Clearly: It is not indeterminate whether b is identical with b.
4. Then: b does not have the property of being indeterminately identical with b.
5. Thus: a is not identical with b.

Williams starts by laying down and defending a series of constraints on a satisfactory defense of ontically vague identity against Evans’ argument, constraints that rule out many of the extant solutions. One of the constraints–namely, that a defense should recover ontically vague identity and not locate the source of the vagueness in either semantic indecision or our ignorance–appears to rule out any solution that tries to block Evans’ argument by appeal to referential indeterminacy, because referential indeterminacy is typically viewed as a matter of semantic indecision.

Williams sketches a model on which the referential indeterminacy is instead due to ontic indeterminacy. Start by distinguishing, in actualist fashion, between the concrete universe we inhabit–Williams calls it reality–and the possible worlds that constitute the maximal properties that reality might have instantiated. Assuming that reality is vague while the possible worlds are precise, it is indeterminate which possible world is instantiated by reality. Say that a world w corresponds to reality just in case w is not determinately uninstantiated and call a world that does correspond to reality in this sense an actuality. If reality is vague, then there is more than one actuality.  We say that a sentence is true (of reality) just in case it is true with respect to all actualities. We say that a sentence is indeterminate if it is true with respect to some actualities, false with respect to others. Williams does not specify a detailed semantics, nor does he need to given the purpose of his paper, but the reader is referred to forthcoming work by both Williams and Elizabeth Barnes.

Next, the model of ontic indeterminacy is applied to what looks like a paradigm case of vague identity. An amoeba, Sue, splits into two amoebas. One of which, Sally, goes off west, while the other, Sandy, goes off east. Williams’ model allows him to diagnose the situation as one in which it is indeterminate whether Sue is identical with Sandy, and the indeterminacy arises from referential indeterminacy that is induced by ontic vagueness. Here is how. There are two relevant actualities. In the first, Sue survives as Sally, and in the second Sue survives as Sandy. Since Sue survives in all actualities, it is true that Sue survives. Yet, it is indeterminate whether Sue is identical to Sandy, because ‘Sandy’ (as well as ‘Sally’) is referentially indeterminate. In the second (first) actuality, it picks out the surviving amoeba, in the first (second), the new amoeba that splits off. ‘Sue’, on the other hand, is referentially determinate, picking out the surviving amoeba in each actuality and thus picking out the surviving amoeba in reality. The important point is that the referential indeterminacy of ‘Sandy’ does not flow from any deficiency in the semantic convention that fixes the name’s reference: ‘Sandy’ refers to the amoeba that floats off eastward. What is indeterminate is whether that amoeba is identical with Sue–the unique amoeba that exists before and survives the split.

Williams goes on to argue that this case of vague identity is not ruled out by Evans’ argument because, putting ‘Sue’ for ‘a’ and ‘Sandy’ for ‘b’, the inference from 3 to 4 fails. In every actuality, “Sandy is identical with Sandy” is true and so it is not indeterminate whether Sandy is identical with Sandy. Yet ‘Sandy’ is referentially indeterminate in reality, and so there is no object that ‘Sandy’ picks out and to which 4 can deny the property of being indeterminately identical with Sandy. Thus, Evans’ argument is invalid.

Besides this novel defense of the coherence of ontically vague identity, the paper also includes some interesting discussion of the importance of ontic vagueness as well as of the relationship between vague identity and other instances of vagueness, such as vague existence and vague instantiation. Building on work by Elizabeth Barnes and Kathrine Hawley, Williams argues that, contrary to what many authors believe, vague existence and vague property instantiation give rise to vague identity, so that Evans’ argument, if successful, rules out more than meets the eye.

One small point of criticism: Williams takes possible worlds to be maximal precise world properties. It would seem that, on the assumption that reality is genuinely vague, it does not instantiate any maximal precise world property, either determinately or indeterminately. The world might have been precise and so could have instanced one of these properties but as things actually stand, it doesn’t. While many precise images may depict vague reality equally well, none of them depicts it correctly. Elizabeth Barnes, whose closely related work Williams refers to, takes it that the world is precise but that it is indeterminate which precise world property it instantiates. This move would get around the objection, but at the cost of locating worldly vagueness not in the world but in the world’s relationship to how it might be.

Reviewed by Iris Einheuser
Duke University