Belief Ascription and Context Dependence

This article considers the question whether belief ascriptions exhibit context dependence. I first distinguish two potential forms of context dependence in belief ascription. Propositional context dependence (PCD) concerns what the subject believes, whereas attitudinal context dependence (ACD) concerns what it is to believe a proposition. I then discuss three potential sources of PCD and two potential sources of ACD. Given the nature of this article, my discussion will provide only an overview of these various forms and sources of context dependence. Along the way I will identify areas where continued research is needed. 1. Belief Ascription To make a belief ascription is to say of someone that he or she believes something. One might, for instance, use (1) (1) Jones believes that Obama is an American. to ascribe to Jones the belief that Obama is an American, or alternately put to ascribe to Jones belief in the proposition that Obama is an American. Propositions are (or at any rate determine) sets of possibilities and the possibilities determined by the propositions a subject believes are the believer’s epistemic possibilities: they are those ways that things might be according to his or her conception of the world. The aim in ascribing a belief to someone is in part anyway to identify or describe that person’s epistemic possibilities. As (1) illustrates, belief ascriptions can be made using a sentence of the form ‘S believes that p’, where the verb ‘believes’ is used to refer to a propositional attitude and the sentential complement (the ‘that’-clause) is used to express (or describe; see Bach 1997) a proposition. So to say (1) is to say that Jones bears the attitude of belief to the proposition that Obama is an American. According to some theorists a belief ascription aims to say more than this, and we will look at some proposals below. What does it mean to ask whether belief ascriptions exhibit context dependence? A belief ascription is context dependent just in case contingent features of the context in which the ascription is made are relevant to determining either which proposition is identified in the ascription, call this propositional context dependence (PCD), or which propositional attitude is ascribed in the ascription, call this attitudinal context dependence (ACD). So a belief ascription made using (2) (2) believes that Obama is an American Philosophy Compass 6/12 (2011): 902–911, 10.1111/j.1747-9991.2011.00447.x a 2011 The Author Philosophy Compass a 2011 Blackwell Publishing Ltd would be context dependent if (i) which set of epistemic possibilities or (ii) which attitude is identified in the ascription could vary depending on the context in which the ascription is made. A belief ascription could exhibit both forms of context dependence. Everyone should agree that belief ascriptions can exhibit PCD of an entirely uninteresting form. A belief ascription made using (3), for example, (3) Jones believes that Obama is here now. could identify different propositions on different occasions, because the referents of the words ‘here’ and ‘now’ could vary from one context to another. So belief ascriptions made using it are context dependent. But this source of PCD is uninteresting because it results from linguistic context dependence that has nothing to do with the nature of belief or with the special sort of tasks involved in belief ascription. There is lively debate over the extent of linguistic context dependence. Some theorists hold that any ‘that’-clause, even the one occurring in (1), would exhibit context dependence, expressing different propositions in different contexts (see, for instance, Travis 1989; Récanati 2004). If these theorists are right, then virtually all belief ascriptions will be context dependent too. Other theorists, by contrast, hold that linguistic context dependence is very limited, occurring only with sentences containing words like ‘here’ and ‘now’ (Cappelen and LePore 2005) We will bypass this debate since all sides in it seem willing to agree that which attitude or proposition the speaker identifies in ascribing a belief to a subject using a given belief ascription sentence may depend on (and vary with) context even if the content of that belief ascription sentence does not so vary. That is, all sides seem to agree that belief ascription speech acts may exhibit context dependence even if the belief ascription sentence does not. So my question here is whether belief ascription speech acts exhibit either propositional or ACD that derives from the nature of belief or from the distinctive tasks involved in belief ascription itself. 2. Propositional Context Dependence Belief ascriptions aim to identify a subject’s conception of things, how she takes the world to be. This means identifying or describing the epistemic possibilities determined by the proposition that is the object of the subject’s belief. The fact that the speaker’s own conception of the world may diverge in fundamental ways from that of the subject whose beliefs he is describing provides one potential source of PCD. Another involves the idea that believers have ways of thinking about the objects and properties their beliefs are about. A third concerns the possibility that certain propositions may themselves be in some way context dependent. I consider each potential source of PCD in turn. 2.1. DIVERGENT EPISTEMIC POSSIBILITIES In ascribing a belief to someone a speaker aims to describe how things are according to the subject’s beliefs. Inevitably, the speaker does this by making reference to objects, properties and relations that he believes in. The speaker aims to characterize the believer’s epistemic possibilities in terms of his own. But in the typical case the speaker knows that he and the subject do not have the same beliefs. It is not just that the speaker and subject’s epistemic possibilities need not be compatible: they need not even contain the same objects, properties and relations. This fact is a potential source of PCD as the speaker tries Belief Ascription and Context Dependence 903 a 2011 The Author Philosophy Compass 6/12 (2011): 902–911, 10.1111/j.1747-9991.2011.00447.x Philosophy Compass a 2011 Blackwell Publishing Ltd to characterize a set of epistemic possibilities that are alien to his own. I will sketch three cases in which PCD seems to arise in this way. In one sort of case, the speaker knows that he has richer conceptual capacities than the subject whose beliefs he is trying to characterize. This may happen when we ascribe beliefs to non-human animals. To say what Fido the dog believes about the bone in the yard, we use our words to try to identify a set of possibilities, but it is not the set of possibilities we would identify were we using those words to express our own beliefs (the example is from Stalnaker 1984). Since we know that Fido cannot tell the difference between a real bone and a manufactured one, we would fail to capture his conception of the world if we characterized it as one where there is a difference between real and manufactured bones, since this is a difference that we know Fido knows nothing about. So we use our words ‘that the bone is in the yard’ to identify the set of possibilities determined by Fido’s beliefs, even though that is not the set we would identify if we were using that same ‘that’-clause to express our own beliefs. Belief ascriptions may thus exhibit PCD when the speaker ascribes beliefs to a subject he takes to have conceptual capacities that are weaker than his own. In a second sort of case, the speaker knows that the believer is mistaken about what objects or properties there are or about their natures. Frege (1892 ⁄1960) told the story of the ancient astronomer who believed, as the astronomer would have put it, that Hesperus is visible in the evening but that Phosphorus is not, when in fact Hesperus and Phosphorus are one and the same planet, Venus (this story is relevantly similar to those of Ortcutt in Quine 1956 ⁄1966 and of Pierre in Kripke 1979). Stories of agents who are mistaken or uncertain about their own identities (see Castaneda 1966; Lewis 1979; Perry 1979 ⁄2000) or about the essential properties of things (Kripke 1980) are also relevantly similar. To characterize the astronomer’s beliefs about Hesperus, a speaker who knows the astronomer’s mistake needs to identify or describe a set of possibilities containing two objects, named respectively ‘Hesperus’ and ‘Phosphorus,’ where in the speaker’s own conception of things there is but one, bearing both names. The set of possibilities that the speaker aims to identify in characterizing the believer’s conception is not the set he would identify if he were expressing his own belief using the name ‘Hesperus’, since not only are the celestial facts different in the believer’s conception but so are the linguistic ones. Belief ascriptions will thus exhibit PCD when the speaker knows that his conception of the world differs in terms of its objects from that of the believer (for more on this, see Stalnaker 1984; Crimmins 1992, 1998; Jackson 1998; Chalmers 2002). Though PCD is perhaps clearest in cases of divergent conceptual capacities and disagreement over identities, the phenomenon may even occur in cases where the relevant divergence involves only ordinary factual matters. In reporting someone’s belief we aim to identify a set of possibilities that is compatible with what else we know (or think we know) about what the subject believes, and this makes a difference to what words we can use to say what the subject believes. If I know that Jones believes that Obama is not President of the US, then the belief I ascribed to him using (1) cannot be ascribed to him using (4) (1) Jones believes that Obama is American. (4) Jones believes that the President of the US is American. 904 Belief Ascription and Context Dependence a 2011 The Author Philosophy Compass 6/12 (2011): 902–911, 10.1111/j.1747-9991.2011.00447.x Philosophy Compass a 2011 Blackwell Publishing Ltd even though I could express my own belief that Obama is an American using either set of words. Given that Obama is President in all the worlds constituting my epistemic possibilities, I can use either set of words to identify the subset of those epistemic possibilities in which Obama is an American. What linguistic resources a speaker can use to say what someone believes can thus depend on what else the speaker takes that person to believe. This is just another side of the fact that ascribing a belief requires trying to say from one’s own point of view how things are from another’s. This picture of the way the shifting in perspective required for belief ascription can generate PCD is not uncontroversial and it raises many theoretical questions. Not all theorists agree that the cases do involve agents with radically divergent epistemic possibilities (Soames 2005 develops a powerful critique of it, and Stalnaker 2007 offers a reply and defense). The precise linguistic and conversational mechanisms that explain this context dependence are not very well understood, and there is substantial disagreement about the contributions of semantic and pragmatic facts; (Chalmers 2002 emphasizes the semantic while Stalnaker 2007 emphasizes the pragmatic) formal models developed to represent the shift speakers make as they try to characterize the believer’s world view from within their own are very abstract and provide little illumination in actual cases (for recent presentations of such models, see Ninan 2010; Swanson forthcoming). This picture of belief ascription may also seem to imply that belief ascriptions are more tenuous (can we really identify or describe the epistemic possibilities determined by Fido’s or the astronomer’s beliefs?) and that beliefs may be more indeterminate (is there really a fact as to which object in the world as the astronomer takes it to be is Venus?) than one might pre-theoretically have thought. Further work is needed to show that such concerns are either groundless or in some way acceptable. Still, if beliefs do determine epistemic possibilities and if one aim of a belief ascription is to identify them, then it seems that belief ascriptions will inevitably exhibit some PCD. 2.2. DIVERGENT WAYS OF THINKING Some theorists claim that in addition to determining a set of epistemic possibilities an agent’s beliefs also determine ways the believer thinks about the objects, properties and relations in those possibilities. And some of these theorists make the further claim that in ascribing a belief we aim to identify these ways of thinking. If this is correct, then this will likely constitute a second source of PCD. The idea that belief involves ways of thinking is usually motivated by stories of a believer who is mistaken or confused about the identities of things. One might for instance think that Frege’s astronomer had two ways of thinking of the planet Venus, one associated with the name ‘Hesperus’ and another with the name ‘Phosphorus’, and that it is because these ways of thinking were relevantly different that the astronomer failed to realize that Hesperus just is Phosphorus. If this description of the source of the astronomer’s mistake is correct, then it might seem natural to think that a belief ascription ought to identify the way the believer thinks about the objects and properties his belief is about. In the astronomer’s case it is natural to think that his beliefs also determine an alien set of epistemic possibilities (one involving two objects where there is only one in fact), but this feature of the case is not supposed to be essential to the idea of a way of thinking. It is supposed to be possible that subjects whose beliefs determine the same epistemic possibilities may yet have different ways of thinking of the objects and properties those possibilities involve. Smith and Jones may have different ways of thinking of Belief Ascription and Context Dependence 905 a 2011 The Author Philosophy Compass 6/12 (2011): 902–911, 10.1111/j.1747-9991.2011.00447.x Philosophy Compass a 2011 Blackwell Publishing Ltd Obama and of the property of being an American even if they agree that Obama is an American, that is, even if their beliefs determine the same set of epistemic possibilities. If beliefs do determine ways of thinking, and if belief ascriptions do aim to identify those ways of thinking, then it is likely that belief ascriptions will exhibit PCD as a result. At least, they will if there is a divergence between the way(s) the speaker thinks of those objects and the way he knows (or thinks) that the subject of the belief ascription thinks of them. If I know that Jones thinks of Obama in way WJ but that Smith thinks of Obama in some other way WS, then in using (2) to ascribe to Jones the belief that Obama is an American I would need to identify WJ whereas I would need to identify WS when using (2) to ascribe Smith’s belief. In this way, belief ascriptions will exhibit PCD if belief involves ways of thinking that belief ascriptions aim to identify. As before, it is controversial whether this really is a source of PCD. Not all theorists agree that belief determines ways of thinking (Stalnaker 2008). And theorists who hold that belief does determine ways of thinking are free to deny that belief ascriptions ever aim to identify them (Braun 1998), or to hold that they do in some contexts but not others (perhaps not, e.g., in the case of shared-belief ascriptions; see the final paragraph below). Moreover, even theorists who accept both claims can disagree about the nature of ways of thinking. Some follow Frege (1892 ⁄1960, 1918 ⁄1977) in taking them to be non-linguistic, conceptual entities (Evans 1982; McDowell 1984), whereas other theorists view them as linguistic or quasi-linguistic entities either contained in the proposition believed (Larson and Ludlow 1993; Higginbotham 2006) or else involved in the subject’s belief state (Perry 1979 ⁄2000, 2001; Salmon 1986; Crimmins 1992). Theorists also differ about the linguistic and conversational mechanisms that explain how belief ascriptions succeed at identifying the believer’s ways of thinking (see Saul 1998 and Ninan 2010 for discussion of this complexity). There is thus more work to be done in clarifying the idea of a way of thinking of an object or property. 2.3. CONTEXT DEPENDENT PROPOSITIONS In section 1 we saw that a belief ascription made using a sentence like (3) (3) Jones believes that Obama is here now. will be context dependent simply because its ‘that’-clause is itself context dependent, expressing different propositions on different occasions. Some theorists have suggested that certain propositions may themselves be essentially context dependent, either because they can only be expressed or identified using essentially context dependent linguistic resources or because they are only available to be believed or expressed in some contexts and not others. If this is so, then ascriptions of belief in those propositions will also be essentially context dependent. Some theorists have argued that certain propositions can only be expressed using words that are themselves context dependent, including some propositions about oneself, about one’s location in space or time, or about objects and properties of current perception (see Perry 1979 ⁄2000; Evans 1982; McDowell 1984). The idea is that Obama would express different propositions by the following sentences, even if those propositions determined the very same set of possible worlds. (5) Obama is an American. 906 Belief Ascription and Context Dependence a 2011 The Author Philosophy Compass 6/12 (2011): 902–911, 10.1111/j.1747-9991.2011.00447.x Philosophy Compass a 2011 Blackwell Publishing Ltd (6) I am an American. The information that Obama would convey using (6) would be somehow essentially first personal in a way that the information he would express using (5) would not be. The information conveyed by (6) would be essentially indexical information, not conveyable in any non-context dependent way. If this is right, then to ascribe that essentially indexical belief to Obama we would have to use something like (7), not (8). (7) Obama believes that he himself is American. (8) Obama believes that Obama is American. It is controversial just what it would be for information to be essentially indexical. The idea is sometimes explained in terms of essentially indexical ways of thinking: some ways of thinking of objects or properties can only be expressed using words that are themselves context dependent. But, as we have noted, the notion of a way of thinking is controversial and its application continues to be a matter of dispute. Still, if there is such essentially indexical information, and if it can be the object of belief, then it would seem that ascriptions of such beliefs would themselves be essentially context dependent. A closely related (though more controversial) proposal holds that certain propositions are context dependent, not because of the way they are expressed, but because they are only available to be believed or expressed in some contexts and not others. One version of this view starts from an idea we considered in discussing Fido, namely that what propositions are available to a believer depends on what possibilities he or she can distinguish. If an agent’s capacities to distinguish possibilities can vary depending on her situation or context, then what propositions there are for her to believe will also vary depending on her situation or context (Stalnaker 2008, forthcoming). This might be so if, for instance, an agent’s capacity to distinguish certain possibilities for some object required perceptual contact with that object. The idea that what there is to believe – what propositions are available to an agent – may depend on context is more than a little controversial. But if propositions are essentially context dependent in some such way, then ascriptions of belief in them would themselves likely be context dependent. Because this context dependence concerns the proposition believed, I have classified this as a form of PCD. But since it also concerns the limits of belief itself, it could equally well be classified as a form of ACD, and so this discussion provides a natural segue to this form of context dependence. 3. Attitudinal Context Dependence I began by noting two ways that a belief ascription made using a sentence like (1) might exhibit context dependence. (1) Jones believes that Obama is an American. One concerns the proposition identified by the ‘that’-clause and was the subject of the previous section. The other concerns the relation identified using the word ‘believes’. This sort of context dependence concerns what it is for someone to believe a certain proposition, say that Obama is American. That is, it concerns what sort of state a belief state is. This is a matter of considerable debate. But it is widely accepted that beliefs are (or essentially involve) dispositional states and in this section I consider two potential sources of context dependence that derive from this fact. Belief Ascription and Context Dependence 907 a 2011 The Author Philosophy Compass 6/12 (2011): 902–911, 10.1111/j.1747-9991.2011.00447.x Philosophy Compass a 2011 Blackwell Publishing Ltd Most theorists agree that belief states are or essentially involve dispositions (for an opposing view, see Strawson 2010). Some theorists hold that beliefs states just are dispositional states (Ryle 1949; Price 1969; Schwitzgebel 2002), whereas others hold that beliefs are the underlying internal states that explain or ground the dispositions (Armstrong 1973; Lewis 1966; Stalnaker 1984). On either view, though, it is natural to think that to ascribe a belief to someone is to say something about her dispositions and this is what I will focus on in what follows. It will be helpful to distinguish two sorts of dispositions associated with belief. First, having a belief involves being disposed to manifest it in different ways in different situations. If Jones believes that Obama is American, then he is likely to act and react as if it were the case that Obama is an American. Second, an agent’s conception of things is disposed to change in light of changing evidence or theory, and so her beliefs are associated with dispositions to revision. Jones is likely to change his mind about Obama’s citizenship if presented with relevant evidence or if he changes his mind about what it is to be an American. In ascribing a belief to someone we are saying something about both sorts of dispositions – manifesting and revising dispositions – and ascriptions of both sorts may involve context dependence. 3.1. MANIFESTING DISPOSITIONS The fact that beliefs involve what I am calling manifesting dispositions is not something unique to belief states. All dispositional states involve manifesting dispositions. Fragile things and generous people are apt to display their fragility or generosity in certain ways in certain situations. So ascribing a dispositional state to something always involves saying something about the sorts of situations in which the thing would manifest that state. Many theorists hold that all ascriptions of manifesting dispositions exhibit context dependence (for discussion, see Mumford 1998). The idea is that in saying that a thing will manifest some property one always presupposes standard or normal conditions for that manifestation, and that what counts as standard or normal can vary from context to context. When we say that a certain vase is fragile, we mean (let us suppose) that it will shatter when dropped from a sufficient height onto a sufficiently hard surface. Such ascriptions presuppose normal conditions, and these presuppositions may vary from one situation to another (changes in temperature might matter to the sufficient height or hardness) and may not be the same as what is presupposed in ascribing fragility to a marble statue or a dried butterfly wing. If this correct, then belief ascriptions will also exhibit context dependence of this sort. Were we to say that Jones believes that Obama is American we would usually mean that he is likely to say so if asked, but not if he is drugged, asleep, hypnotized, insincere or if the situation is abnormal in any one of countless possible ways. Since belief ascriptions also involve presuppositions about normal conditions for the state’s manifestation, if such presuppositions are context dependent then belief ascriptions will be context dependent too. It is relevant that belief involves what Ryle (1949) called multi-track manifesting dispositions. Ryle meant that a person’s beliefs may be manifested in many different kinds of ways: in the person’s intentional actions (including their speech acts), in their emotional reactions (such as surprise at counter-evidence, or indignation at others who don’t share the belief), and even in their imaginings. Presumably, each of these manifesting dispositions involves its own presupposed normal conditions, and so its own potential context dependence. Normal conditions for manifesting a belief in an assertion need not be the same as normal conditions for manifesting it in an emotional reaction. All of this suggests that what dispositions we ascribe to an agent in making a belief ascription will likely depend on context. 908 Belief Ascription and Context Dependence a 2011 The Author Philosophy Compass 6/12 (2011): 902–911, 10.1111/j.1747-9991.2011.00447.x Philosophy Compass a 2011 Blackwell Publishing Ltd Some accounts of belief suggest that ascribing manifesting dispositions will exhibit context dependence. These accounts hold that a belief is associated with a cluster of manifesting dispositions where having that belief requires only a (weighted) majority of those manifesting dispositions (Ryle 1949; Schwitzgebel 2001, 2002). The idea of such a cluster of dispositions is natural in the case of moral or social virtues, were it is natural to think that there are different ways for someone to be generous, courageous or sympathetic. Cluster theorists hold that the same is true in the case of belief, that there are different ways for someone to believe that Obama is an American, different subsets of dispositions each of which would be sufficient to count as having that belief. If this is right, then it is likely that which manifesting dispositions we ascribe to someone in saying what he or she believes will depend on context. For one thing, what manifesting dispositions we ascribe will depend on what other capacities we know the believer to have. We do not expect little children and non-human animals to have the very same manifesting dispositions as mature adults even when we think that they share the same beliefs. We might know that little Bobby agrees with Jones that Obama is an American, but we need not expect Bobby to manifest that belief in the very same ways that we expect Jones to manifest it. Jones’ failure to manifest it on some occasion might be evidence that he lacks the belief, whereas Bobby’s failure to manifest it on that same occasion might not. In Bobby’s case it might instead be evidence that he lacks some independent capacity, one that we presupposed in ascribing the belief to Jones. In this way, what we are saying about Jones and Bobby in ascribing that belief to them is different and this shows that belief ascriptions are a matter of context. A second source of context dependence is the fact that which of the dispositions in the cluster are the ones being ascribed to a subject in saying what they believe might depend on our reasons for ascribing the belief. This sort of context dependence is familiar in the case of the moral or social virtues. Which dispositions we have in mind in calling someone generous can depend on context (generosity in calculating a tip in a fancy restaurant is not the same as generosity in providing financial help with a basement project). Jones might count as generous in one situation but not in another, not because he has changed, but because what dispositions we have in mind in classifying people as generous depends on our interests. The same sort of context dependence can be exhibited in the ascriptions of belief. Jones might on one occasion count as matching the dispositional stereotype sufficiently for believing that Obama is an American, but on some other occasion count only as in-between believing this (see Quine 1960; Schwitzgebel 2002, 2001; Price 1969; for related work on belief, see Gendler 2008). In this way, what we are saying about a person – what manifesting dispositions we are ascribing to them – in saying that they believe some proposition can depend on context. Since this is not a matter of the proposition believed, but of what it is for them to believe it, this would be ACD. 3.2. REVISING DISPOSITIONS A subject’s beliefs are meant to keep track of the world, and we expect them to change in the face of changes in evidence or theory. We expect someone who believes that Obama is an American to be ready to change his or her mind if presented with compelling evidence from a trustworthy source. This fact about belief states – that they involve revising dispositions – is not a feature shared by all dispositional states. It is not part of being fragile that a thing will lose its fragility in certain situations, or part of being generous that a person will lose that virtue under certain conditions. Things do of course stop being fragile and generous, but in the case of belief it is part of the very idea of a belief Belief Ascription and Context Dependence 909 a 2011 The Author Philosophy Compass 6/12 (2011): 902–911, 10.1111/j.1747-9991.2011.00447.x Philosophy Compass a 2011 Blackwell Publishing Ltd that there are conditions under which a subject will change their mind. And this can give rise to context dependence in the ascription of beliefs. Unlike manifesting dispositions, revising dispositions are not multi-track. They are only manifested in one way: by the person’s losing the belief. But aside from this difference, the revising dispositions we ascribe to an agent in saying that they believe some proposition will depend on context for the same sorts of reasons we saw in discussing manifesting dispositions (for more on this, see Stalnaker 1988 ⁄1999). First, as with all dispositions, ascribing a revising disposition will presuppose something about ordinary or normal conditions and what the speaker presupposes about this may be a context dependent matter. Second, what revising dispositions we ascribe to someone may depend on what other capacities we know him or her to have. We would not expect little Bobby’s conception of the world to be as sensitive to changes in evidence and theory as Jones’. Jones’ reaction to counter-evidence might show that he did not believe after all that Obama is American, whereas little Bobby’s similar reaction might display only the absence of certain independent capacities that we knew Jones had when we ascribed the belief to him. Third, what revising dispositions we expect someone who believes some proposition to have may depend on our reasons for classifying people as having that belief. In these ways, what revising dispositions we ascribe to a person in saying them to believe some proposition – what it is we are saying about them – may depend on context. Since this is a matter not of what they believe but of what it is for them to believe it, this is ACD. In this paper I have considered three potential sources of what I have called PCD (Divergent Epistemic Possibilities; Divergent Ways of Thinking; and Context Dependent Propositions) and two potential sources of what I have called ACD (Belief-Manifesting Dispositions and Belief-Revising Dispositions). These potential sources of context dependence may be exhibited together when we ascribe a shared belief to several subjects at once. For the aim of a shared-belief ascription is to identify or describe similarities across the subjects’ conceptions of their shared world, abstracting away from known differences in their epistemic possibilities, in their idiosyncratic ways of thinking, and in their individual practical and rational capacities (this point is made in Quine 1960). We might in one context say, for instance, that Fido, little Bobby and the biology Professor all believe that there is a bone in the yard, even though we know that this ascription leaves out differences in what they believe or in what it is for them to believe it, and that these are differences that may well be relevant in other contexts. Shared-belief ascription thus constitutes, if not exactly an additional source of context dependence, at least an important and familiar occasion when those sources I have identified here can be exhibited together.