« -: 1 ...»
In a memorable children’s story, Winnie-the-Pooh follows the tracks of what he thinks might
be a woozle; until he realizes that he has been ‘Foolish and Deluded’ and that the tracks are his
own.2 Similar ideas recur in many stories from classical mythology to science ﬁction.3
Characters take various attitudes towards people, whilst not realizing that they are themselves the very people concerned.
Stories like these do not just appear in ﬁction; they appear in philosophy too. Let us start with two examples (we shall see more later). John Perry recalls following a supermarket shopper who was spilling sugar from his shopping cart, only to realize that he, like Winnie-the-Pooh, was following his own trail. More fancifully, David Kaplan imagines seeing the reﬂection in a window of someone whose pants are on ﬁre, whilst failing to realize that it is his own pants that are on ﬁre.4 We can see that such stories might be engaging; but why are they philosophically interesting?
e reason is that they highlight a diﬀerence between knowledge about the world, impersonally conceived, and knowledge about our place in the world. As David Lewis graphically puts it, they highlight the diﬀerence between the information given by a standard map, and that given by a map that is erected in a public place with a ‘You are here‘ arrow. One tells you about the nature of the world; the other tells you, in addition, how you ﬁt into that world. Following Lewis, call the former de dicto knowledge, and the latter de se.
e challenge posed by such a distinction is to come up with an account of the de se. Standard theories ﬁnd it hard to do so. Standard theories treat our knowledge (and our beliefs, desires etc.) as involving attitudes towards propositions. ere is some disagreement as to quite what 1anks to audiences at , Paris and Cambridge, and to Rae Langton, Steve Yablo and Seth Yalcin.
2A.A. Milne, Winnie-the-Pooh (London: Methuen, ) Ch. ‘Pooh and Piglet Go Hunting and Nearly Catch a Woozle’.
In one version of the story of Jason and the Argonauts, Jason suggests that a would-be assassin should be sent to fetch the Golden Fleece without realizing that he is the would-be assassin. In a rather more recent story, Robert Heinlein has a time-travelling character consistently failing to recognize himself: ‘By his Bootstraps’ in e Menace from Earth (New York: Gnome Press, ).
4 ‘Demonstratives’ in Joseph Almog, John Perry & Howard Wettstein (eds.), emes From Kaplan (New York: Oxford University Press, ). p. propositions are, but on none of the main approaches is there an obvious way to understand de se knowledge.
One approach takes propositions to be structured entities. For Russell, the structured proposition corresponding to Kaplan’s belief that that man’s pants are on ﬁre is the ordered pair of the man in question, and the property of being on ﬁre. Crucially, given that the man in question is Kaplan, nothing changes when we turn to the proposition corresponding to Kaplan’s belief that his own pants are on ﬁre. e object of that belief is again the ordered pair of the man in question—i.e. Kaplan—and the property of having burning pants.
ere is more opportunity to make a distinction on Frege’s account of structured propositions, which involves not just the objects and properties concerned, but also the ways that these things are thought of: in Frege’s term, ‘modes of presentation’. Here then the proposition corresponding to Kaplan’s belief that that man’s pants are on ﬁre might be diﬀerent from that corresponding to his belief that his own pants are on ﬁre, since they might involve diﬀerent modes of presentation of the same object. But what is the mode of presentation that one has of oneself? It doesn’t look as though it could correspond to any normal description, since the same worries would apply: one could think that someone met that description, without realizing that one met it oneself. Frege concluded that ‘Every one is presented to himself in a particular and primitive way, in which he is presented to no one else’.5 But it is unclear quite what this primitive way is, and, as Frege acknowledged, it makes communication indirect: I cannot communicate my de se thought to you, since you cannot grasp it.
If Frege’s approach leads to obscurity, the alternative approach that takes propositions to be unstructured is in even worse trouble. e central proposal here is Lewis’s own, according to which proposition are classes of possible worlds. But if names and demonstratives are rigid, referring to the same thing in each world, then, if it is Kaplan’s pants that are on ﬁre, the class of worlds in which ‘that man’s’ pants are on ﬁre is just the same as that in which Kaplan’s pants are on ﬁre. We have nothing to distinguish the de se belief from the de dicto.
In response to these diﬃculties Lewis proposes an account that, characteristically, is at once simple and radical.6 Do not think of de se thoughts as attitudes to propositions at all. ink of them rather as self-ascriptions of properties. When Perry realizes that he is following his own trail, he self-ascribes the property of spilling the sugar; when Kaplan realizes that it is his own reﬂection that he can see, he self-ascribes the property of having ﬂaming pants. Moreover, the de dicto can then be thought of as a special case of the de se: the case in which what is selfascribed is the property of membership of a world.
Lewis’s account thus has two parts. e ﬁrst part, which is the focus of most of his discussion, and of most of the discussion that has followed, involves treating the objects of the attitudes, 5 Frege ‘e ought’ in Collected Papers (Oxford: Blackwell, ) 6 Lewis, ‘Attitudes De Dicto and De Se’ Philosophical Papers I (New York: Oxford University Press, ) pp. –. See also On the Plurality of Worlds (Oxford: Blackwell, ) and ‘Individuation by Acquaintance and by Stipulation’ in Papers in Metaphysics and Epistemology, (Cambridge: Cambridge University Press, ) pp. –.
—2— not as propositions but as properties. e second part, which is much less to the fore in Lewis’s presentation and in the subsequent discussion, involves treating our attitude to these properties as that of self-ascription. I hyphenate the term to stress that this is a primitive notion for Lewis.
To self-ascribe a property is not just to ascribe a property to oneself, as one might ascribe it to someone else. at would just give us back the question of what it is to think of oneself in the right way. After all, when Kaplan ascribes the property of having burning pants to the ﬁgure in the window, he does ascribe that property to someone who is indeed Kaplan; in one sense then, he ascribes the property to himself. What Lewis wants to insist is that he does not self-ascribe that property. e self and the ascribing do not detach.
I will argue that, as a result of underplaying its second part, the radical nature of Lewis’s account has been largely missed. In particular, much recent literature has tried to incorporate his account simply by treating the objects of the attitudes as centered worlds, where a centered world is an ordered pair of a possible world together with a spatio-temporal location. But centered propositions just correspond to properties. So whilst they give us the right objects for the Lewis account, they go no way to providing the right attitude. (Moreover, I shall argue that they provide an unnecessarily baroque way of specifying contents.) As a result of this neglect, many of the diﬃculties that Lewis’s approach faces have been overlooked. e idea of primitive self-ascription is an obscure one. Our natural grasp on it is via the general idea of ascribing a property to an object, but this is exactly what we are not allowed to do here. Instead we have to treat the relation as fundamental, something that becomes increasingly hard to do when we consider ﬁrst person plural ascriptions, and ascriptions where the ﬁrst person pronoun is not in subject position. ese diﬃculties do not render Lewis’s account totally unworkable; but they do make it much more problematic than has been generally recognized.
’ Let us start by getting clear on exactly what Lewis takes himself to be doing, and on the techniques he uses. His main contention is that propositions cannot be the objects of the attitudes, since there are situations in which an agent has a great deal of propositional knowledge, and yet lacks some further de se knowledge. e idea then is that this further knowledge cannot be knowledge of a proposition, i.e. it cannot be de dicto. (I will follow Lewis and treat talk of propositional belief and of de dicto belief as equivalent for now. e obvious problem with doing so is that many have thought that there exists a further category, de re belief, which is propositional without being de dicto. As we shall see, Lewis denies that this is really a separate kind of belief; but the issues there are tangential.) One way of arguing for this makes use of examples like those that we have seen from Perry and Kaplan. ese examples certainly establish that you can have some propositional knowledge about a situation and yet lack further de se knowledge about it. And the Kaplan cases shows, beyond that, that you can have direct perceptual knowledge that a property obtains of a person who happens to be you, without knowing that you—you yourself as Castañeda would say— have that property. What the cases don’t show, at least not without a number of further —3— assumptions, is that this de se knowledge cannot be understood as propositional knowledge.
For it could be that there is some way of deducing the de se knowledge from additional de dicto knowledge, if only one had enough of it. And if the de se knowledge is deducible from the de dicto knowledge, then presumably it in turn is de dicto.
is is where Lewis’s approach is innovative. He wants to argue that you can have all of the de dicto knowledge that there is, and still lack de se knowledge.7 For this he needs some new examples, since the Kaplan and Perry cases lack this feature. If you ﬁlled in the propositional knowledge of the subjects involved—gave them information about the angle of the reﬂection, say, or about the complete extent of the sugar trail—they would be able to deduce that it was their pants on ﬁre, or their trail of sugar. at is not so say that they would have deduced such conclusions from entirely de dicto beliefs; that would still need to be investigated. But it does mean that, without further elaboration, the examples will not serve to show that one can have complete de dicto knowledge and lack de se knowledge.
e main example that Lewis gives to make this point is that of the two gods:
Consider the case of the two gods. ey inhabit a certain possible world and they know exactly which world it is. erefore they know every proposition that is true in their world. Insofar as knowledge is a propositional attitude, they are omniscient. Still I can imagine them to suﬀer ignorance: neither one knows which of the two he is. ey are not exactly alike. One lives on top of the tallest mountain and throws down manna; the other lives on the top of the coldest mountain and throws down thunderbolts. Neither one knows whether he lives on the tallest mountain or on the coldest mountain; nor whether he throws down manna or thunderbolts.8 is example is somewhat problematic, since it is rather hard to ﬁll in the details. If the gods are able to chose their actions and then implement them, can they not infer who they are from seeing the results?9 So let me give another of Lewis’s examples in which the diﬃculties, if still present, are rather less to the fore. is example, originating from Perry, is that of Lingens, lost in the Stanford Library, who has such complete amnesia that he does not know who, or where, he is. As Lewis develops the story, Lingens has an alter ego—Lauben let us suppose—who is lost, in a similar condition, in the Widener Library at Harvard. Let us further suppose that the Stanford and Widener Libraries are even more comprehensive than they are now—their works include all de dicto knowledge—and that Lingens and Lauben have read and remembered every word. Let us also suppose that the libraries are qualitatively identical inside; not only are there no ‘You are here’ maps on the walls, there are no stamps in the books saying ‘Property of Widener’ or ‘Property of Stanford’, no Windsor chairs bearing crests emblazoned ‘Veritas’ or ‘Die Luft der Freiheit weht’ and so on. en despite their complete de dicto knowledge, 7In discussing an example of Castañeda’s, Lewis writes: ‘To support this claim [that attribution of knowledge de se is not equivalent to any attribution of knowledge de dicto] we need only ﬁnd a case in which the editor knows well enough which of the worlds is his without knowing whether he is among the millionaires of his world’. ‘Attitudes’ pp. –.
8 ‘Attitudes’ p. 9 For a good attempt to make sense of the possibility, see Robert Stalnaker Our Knowledge of the Internal World (Oxford: Oxford University Press, ) p. , n..
—4— including their knowledge of there being two amnesiacs lost in diﬀerent libraries, neither Lingens nor Lauben will know which they are.
Do they lack all de se knowledge? Surely not. If Lingens sneezes he knows that he sneezes;
likewise for Lauben. (If the libraries really contain all de dicto knowledge, including reports of all the sneezing that goes on, then if one sneezes the other had better sneeze too; or else Lingens and Lauben would be able to work out who was who.10 ) And they will each know that they are either Lauben or Lingens. What each lacks is any de se knowledge that distinguishes him from the other, and so allows him to integrate his de se knowledge with the de dicto knowledge that is given to him by the books.