question my feeling tired. (Note that here questions of sincerity do not even arise: the case is one where we are supposing that I am thinking that thought to myself.)
My use of the relatively neutral notion of
ascription is intended to cover both utterances and thoughts or judgments. In keeping with common practice, I shall allow myself to speak of the "semantics", as well as "truth-conditions" of ascriptions, as well as of the "I"-component of an ascription and its "reference". I shall also use quotation marks to cite the relevant ascription, without presupposing that self-ascriptions in thought must consist in one's
talking to oneself (at least in some sense). Presumably, one can think, as well as say, something properly expressed in English by the words "I am φ", where φ refers to an occurrent mental state. That is what I have in mind when I speak of mental self-ascriptions made in thought, as opposed to speech. And I am allowing that, among such self-ascriptions, there are those that can properly count as avowals. (Exactly what qualifies them as such is something that I will take up later.) However, I do not want to suggest that there are no important differences between ascribing in language and ascribing in thought.
13 Where these differences matter, I shall indicate it.
I have initially characterized avowals in a very rough-and-ready way, as present-tense self-ascriptions of occurrent mental states. Part of my task will be to refine this characterization. A first, obvious refinement pertains to "self-ascriptions". Ascribing an occurrent (as opposed to dispositional) mental state at the time of ascription to an individual who is in fact oneself is not sufficient for avowing the mental state. It seems as though one must also be thinking of the individual as oneself. Where the context makes it clear, I will allow myself simply to speak of self-ascriptions; otherwise, I shall speak of an "I"-ascription. Furthermore, avowing seems to require ascribing to oneself the relevant state in a certain way. Saying, or thinking, "I am mad at my mother" purely in consequence of being convinced by a therapist's analysis would not constitute avowing anger, since such an "I"-ascription is made on a kind of basis available to others as well. So it is not even the case that all present-tense "I"-ascriptions of occurrent mental states will count as avowals. Thus, when I use "avowal", I intend to signal that
end p.25
the relevant self-ascription is not to be understood as a theory-driven self-ascription.
It should be noted, however, that some ascriptions of occurrent mental states that fail to use "I" may also qualify as avowals. For instance, I consider ascriptions such as "My tooth hurts", or "This hurts", or "This feels cold" to be avowals, even though such ascriptions do not contain the pronoun "I" (the last two contain no overt element referring to the ascriber). I think it is plausible to treat such ascriptions as elliptical, and as containing a covert indexical reference to the utterer/thinker. This receives some support from considering languages such as Hebrew and French, which do require an overt indexical pronoun, at least in some of these cases (cf. Hebrew: "Ko'ev li" ("It hurts to me"), or French: "Il me fait froid" ("It makes me cold")). Furthermore, it would not be unreasonable to suggest that, at least in speech, avowals can use terms other than an indexical pronoun to refer to the self-ascriber. Thus consider, "Stop it, Jenny! Daddy doesn't like this," said by a parent to a child, or "Jenny wants to eat now," said by the child of herself.
The foregoing examples illustrate what will become increasingly clear as I develop my account of avowals. The class of avowals cannot be circumscribed purely in terms of the surface grammar of certain self-ascriptions. It also cannot be delineated by attending to obvious aspects of the meaning of the relevant self-ascriptions. Indeed, this is part of our difficulty: avowals contrast in their security with ascriptions that are grammatically and even semantically indistinguishable from them. "I am very happy" can sometimes be an avowal and sometimes not. (The account of avowals I shall be developing is partly motivated by this difficulty.) Thus I will not begin my investigation with a clear definition of my explanandum. Nor will I be aiming to produce at the end of my investigation a stable definition of avowals in terms of necessary and sufficient conditions. My starting point is a familiar phenomenon, which can be characterized ostensively as it were, and which is philosophically puzzling for reasons I began to explain in this chapter. If I can offer an illuminating account of this phenomenon that will make the puzzles go away, I will be satisfied. So, I hope, will my reader.
end p.26
2 Using "I" 'as Subject': Cartesian Reference or No Reference?
Abstract: The uniquely secure status of avowals apparently has something to do with the special way an avowing subject refers to herself by the pronoun 'I', as opposed to referring to herself by some description. However, the author argues in this chapter that excessive focus on the way 'I' refers has led people astray. In particular, she criticizes the view that we must see 'I' as referring to some special object, a Cartesian ego, as well as Anscombe's view that 'I' in fact, does not refer at all. Both accounts fail to respect the Semantic Continuity of avowals with other empirical reports.
Dorit Bar-On
Self-ascriptions such as "I am thirsty", "I have a terrible headache", "I feel like taking a nap right now", "I'm thinking about my poor aunt in the hospital", "I hope this noise stops soon", "I'm so annoyed that I can't get this computer to work", and so on, as normally made, have a seemingly unique status. Whether volunteered spontaneously or offered in response to questions, whether made in speech or in thought, such avowals, as I shall refer to them, appear to enjoy a special security. Unlike typical apriori judgments, avowals concern contingent matters of fact. Yet their production does not seem to involve consulting any evidence, inference, or ordinary observation. Like perceptual reports, they are seemingly 'effortless' and are strongly presumed to be true, but, unlike such reports, they are not based on ordinary perception. In fact, they appear to be made on no basis at all. And in contrast to all other pronouncements, avowals are strongly resistant to epistemic criticism, not subject to ordinary doubts and corrections, and invulnerable to brute error.
The Use of "I" 'as Object' versus 'as Subject'
It is no doubt crucial to the special status of avowals that they are
self-ascriptions. However, recall the ascriptions in the contrast class of avowals:
(a) |
"She is very tired," said (or thought) by someone else about me at that same time;
|
(b) |
"I was very tired then," said by me at a later time;
|
end p.27
(c) |
"John is very tired," said by me about someone else;
|
(d) |
"DB is very tired," said by me about myself when I fail to recognize that I am DB;
|
(e) |
"I am very tired," said by me on the basis of, say, looking in the mirror, or inference from some test results;
|
(f) |
"I am 5′5″ tall," or "There's a scar on my forehead";
|
(g) |
"I am a very patient person," or "I have a sense of humor," or "I am attentive to my friends";
|
(h) |
"There's a red car right there!" said upon seeing a car outside;
|
(i) |
"My legs are crossed," or "I'm standing in front of the house."
|
(j) |
"I am seeing a red cardinal," or "I am hearing a loud noise."
|
Many of these are self-ascriptions. Many of them do not merely ascribe a property to someone who is in fact the self-ascriber (as in (d)). Rather, they are self-ascriptions that use the first-person pronoun "I" to pick out the self-ascriber. (Consider (b), (e), (f), (g), (i), and (j).) Even so, they do not share in the secure status of avowals. Part of the task of accounting for this status, then, is to explain how avowals are different from these other, less secure "I"-ascriptions.
In
The Blue and Brown Books, Wittgenstein offers the following distinction:
There are two different cases in the use of the word 'I' (or 'my') which I might call 'the use as object' and 'the use as subject'. Examples of the first kind of use are these: "My arm is broken," "I have grown six inches," "I have a bump on my forehead," "The wind blows my hair about." Examples of the second kind are: "I see so and so," "I hear so and so," "I try to lift my arm," "I think it will rain," "I have a toothache." (1958: 66 f.)
At first glance, it is not clear what the distinction amounts to. For, in both "I think it will rain" and "I have grown six inches", "I" serves as the grammatical subject. As will become clearer below, however, the distinction does not pertain to a difference in surface grammar, but to the way the subject of predication is picked out. Wittgenstein is pointing out that certain uses of "I"—its uses 'as subject'—in some sense do not involve picking oneself out as just another item in the world, a particular object or individual, of which something is to be predicated. As Wittgenstein explains:
One can point to the difference between these two categories by saying: The cases of the first category involve the recognition of a particular
end p.28
person, and there is in these cases the possibility of an error…. It is possible that, say in an accident, I should feel a pain in my arm, see a broken arm at my side, and think it is mine, when really it is my neighbour's…. On the other hand, there is no question of recognizing a person when I say I have tooth-ache. To ask 'are you sure it is you who have pains?' would be nonsensical. (1958: 67)
Using Wittgenstein's distinction, it may be suggested that avowals are self-ascriptions that use "I" 'as subject', not 'as object'. This may provide an important insight for understanding the distinctive security of avowals. It may be suggested that, in trying to understand this security, we should focus attention on the "I"-component of avowals. Though avowals are not alone in being self-ascriptions, and though other self-ascriptions also use the grammatical pronoun "I", avowals should be marked by the fact that they involve a special use of "I": its use as subject. Thus, consider the ascriptions in the contrast classes for avowals (i.e., (a)-(j) above). Ascriptions of present mental states to others (c), or by others to oneself (a), or by oneself to oneself in the past (b) and so on, are all subject to various kinds of potential errors, corresponding to possible aberrations in the 'epistemic routes' that lead one to the judgments they express. (Perception, observation, inference, memory, etc. are all notoriously subject to various failures.) This explains why such ascriptions do not enjoy the maximal degree of epistemic security. By parity of reasoning, we might take the security of avowals to be due to the fact that they are somehow protected from these kinds of epistemic failure. Following Wittgenstein, we may also notice that avowals involve uses of "I" as subject, whereas the ascriptions in the contrast class involve uses of "I" as object. A proper understanding of what is characteristic of such uses will show why ascriptions involving them are not vulnerable to certain errors. And this may go a long way toward explaining why avowals seem to enjoy such a high degree of security.
Cartesian Egos as the Targets of Reference for "I"
What kind of protection from error or failure is afforded by using "I" as subject? Wittgenstein suggests that the protection has to do with the special way we identify the individual of whom we predicate
end p.29
something. When using "I" as subject, "there is no question of recognizing a person.... To ask 'are you sure it is you who have pains?' would be nonsensical." The errors that we seem to avoid when using "I" as subject are, then, errors of recognition. But we must ask why, and in what sense, one is protected from such errors when using "I" as subject.
As Descartes famously recognized, just thinking an "I"-thought seems to guarantee both the existence and the presence of the referent of "I". It makes no sense to suppose that one has used "I" to refer to no object at all. Moreover, at least with certain uses of "I", it seems impossible to doubt that one has got 'the right object'—namely, oneself. So the invulnerability to error seemingly involved in the relevant uses of "I" is rather remarkable. It begs for explanation.
Now, according to Wittgenstein, the historically familiar Cartesian account is an attempt to explain the seemingly guaranteed success of "I" in a particular way:
We feel then that in the cases in which "I" is used as subject we don't use it because we recognize a particular person by his bodily characteristics; and this creates the illusion that we use this word to refer to something bodiless, which, however, has its seat in our body. In fact this seems to be the real ego, the one of which it was said, "Cogito, ergo sum." (1958: 69)
On Wittgenstein's diagnosis, since uses of "I" as subject do not seem to require recognition of bodily features in picking out, or identifying the referent of "I", the Cartesian is moved to supply a special object—'something bodiless'— that each of us could refer to without relying on the use of such features, namely, her Self. Uses of "I" as subject then turn out to involve reference to this special object.
On this reading of the Cartesian view, when I say, or think, "I feel overcome by joy," for instance, I am guaranteed success in identifying the subject of my ascription, since I cannot be in error regarding which object I am trying to pick out. But we may wonder how it is that I can be guaranteed success in picking out the Ego, understood as one object (or individual substance) among others. Do I perhaps have a special access to my own Ego—a special way of recognizing it, or some kind of direct acquaintance with it? But what could that amount to? Talk of recognition of, acquaintance with, or access to implies an epistemic relationship between an epistemic subject (someone who is
end p.30
a potential believer or knower) and some object (an individual or thing, or perhaps a fact). In paradigmatic cases, there is a kind of 'epistemic gap' between the two relata, which the relationship allows the subject to bridge. But it is very unclear what could be made of these ideas within the Cartesian framework, in which I simply am my Ego and every belief, thought, impression, etc. that I have is necessarily a feature of it. (Talk of my Ego seems equally misleading.)
Suppose, on the other hand, it is suggested that when "I" is used as subject, we pick out the subject of the ascription (viz., the Ego) neither through features of the body associated with it nor through direct acquaintance with the Ego itself. Instead, we pick it out as the bearer of certain features, which present themselves to us directly. The suggestion is that we have direct access to (or awareness of) our present mental states, and we use "I" to refer to the subject of these states, namely, our Ego. But it is not clear why identifying a subject of ascription in this way should afford more security, more protection from epistemic error, than identifying a subject of ascription through its bodily features. We would still be left with the question why "I", understood as referring to the subject of certain present mental states, should be a more secure referential device than other referring expressions, or than "I" understood as referring to one particular object among others. The detour via Cartesian Egos will seem to have got us no further.
1
Be that as it may, notice that the Cartesian explanation (on the above diagnosis) takes it for granted that uses of "I" as subject, perhaps like uses of other referential expressions, must rely on some form of access to, or recognition of, the referent of "I". (This idea will be criticized in the next chapter.) This, in turn, presupposes that "I", in its use as subject, is a referential expression. It presupposes that, like uses of other subject-terms in grammatically similar ascriptions, uses of "I" as subject
purport to refer to an object of which something is to be predicated.
2 On the present diagnosis, the Cartesian begins with the correct observation that certain self-ascriptions are not subject to the same
epistemic failures as other ascriptions. Seduced by surface grammar,
end p.31
he tacitly makes the tempting semantic assumption that "I" (and its equivalents) in such self-ascriptions is an ordinary referring expression whose function is to identify some object of which something is then to be predicated. (As noted above, he further assumes that performing this semantic function requires some form of epistemic access to the referent.) But then, to reconcile this semantic function with the role of "I" in epistemically secure self-ascriptions, he is moved to make an ontological postulation, assigning as the referent of "I" something quite extraordinary—a Cartesian Self or Ego.
Putting the above reasoning more explicitly, we get:
(1) |
"I" is a referring expression whose semantic function in all its uses is to pick out, or refer to, a particular object.
(This is the semantic presupposition identified above. It can be seen as an application of the Semantic Continuity claim to "I". For it may seem that, unless "I" were taken to refer to a particular individual, we could not accommodate the logico-semantic continuities between "I"-ascriptions and other ascriptions.)
|
(2) |
Certain "I"-ascriptions—where "I" is used as subject—enjoy a distinctive security: the user of "I" in such ascriptions cannot make an epistemic error in picking out the referent of "I". Uses of "I" as subject are guaranteed correct reference.
(This is the epistemic observation attributed above to Descartes. It can be seen as an application of Epistemic Asymmetry to the "I"-component. For it manifests a contrast between certain "I"-ascriptions and various other ascriptions.)
|
(3) |
If we took "I" to refer to an 'ordinary' object, say one's body (or, more fashionably, one's brain), there could be no apriori guarantee for the user of "I" that she is referring to what she thinks she is referring to.
So,
|
(4) |
Uses of "I" as subject cannot be taken to refer to any ordinary object, such as one's body, or a distinct part of it (e.g., one's brain).
|
(5) |
Given (2), the intended referent of "I" must be an object whose presence can only be ascertained in the act of making the relevant "I"-ascription.
|
(6) |
For (5) to hold, the essential characteristics of the referent of "I" must be accessible to the user of "I" in the act of using it.
|
(7) |
The only object that meets this condition is a Cartesian Ego.
|
end p.32
|
|
(8) |
The referent of certain uses of "I" must be a Cartesian Ego.
|
The foregoing is an attempt to spell out in an explicit argument form the reasoning imputed to the Cartesian by Elizabeth Anscombe, in her celebrated article "The First Person" (1975).
3 The basic idea behind the argument can be presented more simply, in terms of the following dilemma. Adhering to Semantic Continuity and accepting that "I" is a referring expression, the referent of "I" must be either an ordinary object (such as a human body) or some extraordinary object. Given the special epistemic security of avowals as captured by Epistemic Asymmetry, "I" cannot be taken to refer to an ordinary object. So, the referent of "I" must be an extraordinary object: an object securely knowable in a way no ordinary object can be. The Cartesian Self or Ego can be seen as an ontological invention tailored to fit the bill.
4
The Cartesian strategy may seem a promising one for explaining the contrasts between avowals and other ascriptions. Avowals, unlike ascriptions that are superficially similar to them, involve reference to a special kind of object, Egos. That is what sets apart uses of "I" as subject. The nature of Egos and our own special epistemic relation to them guarantee that the ascriptions we make to them when using "I" as subject are not vulnerable to any of the ordinary epistemic failings. Adhering to this picture, we can explain how your ascription, "DB is feeling very sad", can be exchanged for my self-ascription, "I am feeling very sad". We both ascribe a certain state to a particular subject of mental states, namely, me (= my Ego). However, you make the
end p.33
ascription on the indirect basis of observations and inferences regarding my bodily behavior, and your ascription is subject to all sorts of epistemic failures. I, on the other hand, make the ascription in a way that is protected from all such failures.
Accepting Cartesian dualism, with its commitment to mind-body interaction, requires embracing an ontology that includes immaterial particulars that exert causal influence on, and are causally affected by, goings-on in the physical world. The difficulties with this idea are widely appreciated. It also requires making good sense of the idea of contingently existing, objective particulars whose modifications can be known infallibly. We have seen some difficulties with this idea earlier. However, even setting aside qualms about Cartesian ontology and epistemology, one may question the adequacy of the above Cartesian account as an acceptable semantic account of the reference of "I".
First, we may observe that the account requires compromising Semantic Continuity to a certain extent. For, although the account preserves the idea that "I" refers, it requires postulating a systematic semantic ambiguity in uses of "I". To explain the security of uses of "I" as subject, the Cartesian proposes that we see them as referring to a different (kind of) object from uses of "I" as object. But then it is not clear how we can preserve the palpable grammatical and logical continuities across the two types of uses. If I say to someone (or just think to myself), "I'm just getting out of bed, but I feel so tired," I seem to be ascribing two things to a single individual (myself): getting out of bed and feeling tired. Yet, on the Cartesian picture, the first ascription has as its subject a particular body, or perhaps a particular human being, whereas the subject of the second ascription is a particular immaterial Ego. This means that my conjunctive attribution will involve a certain—unmarked and unnoticeable—shift in reference. A corresponding shift will presumably occur in the third person: in "DB just got out of bed, but she feels tired". "DB" will presumably refer to the human being DB, whereas "she" will have to refer to DB's Self (where "She feels tired" will be taken to express a fallible conjecture about a state of that Self). But if we can no longer take for granted the notion of pronominal preservation of reference, it is not clear how we are to preserve the validity of everyday inferences that seem to trade on such preservation. (e.g.: "If someone has had a lot of sleep,
end p.34
then she is not likely to be tired. DB got a lot of sleep. So she is not likely to be tired.")
5
We may assume that this problem can be circumvented; still, there is a difficulty with the Cartesian way of explaining Epistemic Asymmetry. The Cartesian explains the security of avowals as a matter of the guaranteed semantic success of "I" in referring to the 'right object'. This guaranteed success he explains by invoking the guaranteed epistemic success each of us enjoys when we identify ourselves as the targets of reference. And this success in turn is explained by the existence of certain objects that are, by their nature, susceptible to such unfailing identification. We can summarize this by saying that, for the Cartesian, the target of reference of "I" is (semantically) unmissable because (epistemically) unmistakable. But perhaps we ought to be wary of the very idea of unmissable targets of reference.
Naively, the notion of a target of reference suggests the idea of someone
aiming at and/or
hitting some object. Conceptually speaking, the possibility of hitting a target requires at least the possibility of missing it. Correlatively, it ought to be at least conceptually possible for one who aims at a target to
think she has hit the target without her
actually having succeeded. With ordinary targets of reference—by which I mean people, and everyday physical objects, but also things such as hurricanes, countries, wars, and numbers—these conceptual possibilities seem preserved. For such entities have identity conditions that are independent of any particular individual's success in identifying them. Not so with Cartesian Egos. To play the role of securing apriori guaranteed success of reference, Egos are required to be targets that each of us
cannot—logically or conceptually—fail to recognize, and is thus able to name unerringly using "I". This means that their identities are essentially tied to individuals' successful acts of identification. As Anscombe remarks, "I" is supposed to have 'sure-fire' reference. But this could only be so "if the referent of 'I' were both freshly defined with each use of 'I', and also remained in view so long as something was being taken to be
I" (Anscombe
1975: 57).
end p.35
With ordinary referring expressions, there has to be what I shall call "semantic distance" between applications of a referring expression and its putative target of reference. A referring expression has an extension: a range of things to which it is supposed to apply. Some things will be in the extension of the expression, others will not. But the expression is of course distinct from its extension. And it must always be in principle possible for the expression to be misapplied to a given item. That is to say, it must be possible to use the expression to refer to something outside its extension. The expression (or its user) aims to pick out an item, but may—at least as a matter of logical or conceptual possibility—fail to 'latch onto' the right item (or to any item at all). This implies that whenever a speaker uses a referring expression, there ought to be a conceptual distinction between her thinking she applied the term correctly and her actually having applied the term correctly. For example, we take it that the word "chair" is a referring expression; it is a term that purports to refer to some items—viz., the chairs of this world. To say this is to say that the word is supposed to apply to chairs only. But all this could mean is that it would be incorrect to apply it to something that is not a chair (as when someone says, pointing to a hammock, "Don't sit on this chair"). And this, in turn, implies that it is in principle possible to apply it to a non-chair, even if no one ever actually mistakenly calls a non-chair "chair".
Now, the Cartesian picture painted so far portrays "I" as an ordinary referring expression—indeed, very much like a name. However, "I" on this picture has an extraordinary item as its putative target of reference: an unmistakable Ego. Though each Ego is to be a genuine target, to be hit by particular uses of "I", it is logically guaranteed to be unmissable. Failure to hit this kind of target is logically ruled out, and so we may wonder whether we can speak of attempting to hit a target in the first place.
6 If success in picking out the referent of "I" is logically guaranteed, the semantic distance required for ordinary reference seems to be eliminated. Thus Descartes's "I" is a very peculiar referential device: in using it, we supposedly try to single out an item in the world (an Ego) the way we do with proper names or general
end p.36
terms. But, unlike in these cases, our uses of "I" are guaranteed success. I think we may legitimately wonder whether the Cartesian picture manages to preserve the idea of "I" as a referring expression on the model of ordinary referring expressions.
7
The foregoing can serve to fill out Anscombe's own rejection of Egos.
8 A natural reaction to this reasoning is to suggest that "I" does not qualify as an ordinary referring expression. Later on, I will consider various possibilities of altering the requirements on reference so as to accommodate the special behavior of "I". Some of these possibilities may well be available to the Cartesian, in which case the above argument against reference to Egos will be blocked. However, it may also turn out that such possibilities are consistent with taking all uses of "I" to refer to a more ordinary object, such as a human being. In that case, the Cartesian would no longer be able to cite the special security of "I" as a motivation for postulating Egos.
The Wittgensteinian Rejection of Private Sensations
Anscombe's argument against Cartesian Egos can be seen as running parallel to Wittgenstein's rejection of private sensations (encapsulated in his so-called Private Language Argument). On one plausible, if limited, reading of the passages surrounding Wittgenstein's
Philosophical Investigations §258, the so-called Private Language Argument is an attempt to reduce to absurdity a very specific conception of sensation talk, one which can be described as Cartesian.
9 The conception under attack is a combination of four substantive assumptions regarding the character of sensations and sensation terms.
(a) |
that sensations—the putative referents of the diarist's private signs—are objective particulars of sorts, which can be recognized,
|
end p.37
|
grouped together into kinds, and named, just like ordinary physical objects. However, their recognition is subserved by 'inner sense', rather than by ordinary perception;
|
(b) |
that sensations are logically private objects: only the 'host' of a sensation can possibly recognize it or verify its presence (from which it follows that she is incorrigible);
|
(c) |
that the 'host' of a sensation can recognize it transparently as well as infallibly: a sensation is the sort of thing which, if it is present, its presence will be immediately recognizable by its host. And if it seems to the host that she has it, then she does have it. (See Wittgenstein 1953: §§244, 246, 256.);
|
(d) |
that sensation terms, like all "individual words in language name objects", and that the meaning of a word "is the object for which it stands". (This is the core assumption of the Augustinian model of language described at the start of the Philosophical Investigations.)
|
The alleged trouble with setting up a language for which these four assumptions hold is that it will turn out that, in such a language, it would be in principle impossible to
misapply words. To borrow a metaphor from David Pears, applying words in such a language would be like throwing stones and declaring the target to be where the stones land. Where putatively objective targets are thus locked on to individual acts of shooting, the whole idea of
aiming at a target collapses. The private speaker, whose words are supposed to name (and thus have as their meanings) logically private objects, is like the idle stone-thrower: his objects of reference are necessarily tied to his individual acts of references. In this way, whatever is going to seem right to him is right. "And that only means that here we can't talk about 'right'" (1953: §258).
10
How did we get into this trouble? Let us imagine, with Wittgenstein, a diarist whose task is to set up a language with terms which name his private sensations. He begins by trying to set up some symbol, say "S", as a meaningful sign in his own language, which purports to name one of his sensations. According to the Augustinian assumption (d), the act of endowing "S" with meaning is the act of singling out some object which would be its meaning. Since "S" is supposed to be
end p.38
a term in a sensation language, the relevant object to be singled out would be some sensation, S. By assumption (a), S is an objective particular, a recognizable individual item in the world. However, by assumption (b), the putative referent of "S"—some particular sensation—will be a logically private item. So, if the diarist could succeed in setting up "S" as a name for some sensation, he would have introduced a term with a private meaning, since (by assumption (d)) "S"'s meaning is the object for which it stands. How could the diarist achieve this feat? Let us suppose the diarist has some sensation. By assumption (c), the diarist will recognize that he has the sensation, perhaps by introspection. Having introspected its presence, our diarist is in a position to make it the referent (and thus, by (d), the meaning) of "S", perhaps by some kind of inward ostensive definition (he focuses his attention on S and dubs or tags it "S"). Furthermore, by assumption (b), on any future occasion of using "S", our speaker is, logically, the only one in a position to determine whether "S" was correctly applied. Now, suppose the diarist's 'inner sense' tells him that he has S (i.e., that S is present now), and he writes down in his diary "S" (or 'tokens "S" in his language of thought'). In that case, he is using the term "S" because it seems to him that he has S. However, by the infallibility part of assumption (c), if it seems to him he has S, then he does have S. Which means that the private diarist cannot, logically, go wrong in his linguistic applications of "S".
What is the alleged problem with this? In my discussion of Cartesian reference to Egos in the previous section, I suggested that, in the case of ordinary linguistic expressions, there is a conceptual distinction between correct and incorrect applications of the term. There is what I called a "semantic distance" between applications of a referring expression and the items to which it is used to refer. To recapitulate, referring expressions have an extension: a range of things to which they correctly apply. Some things will be in the extension of the terms, others will not, and it ought to be possible in principle for the expression to misapply to a given item. Thus, whenever a speaker uses a referring expression, there ought to be a conceptual distinction between him thinking he applied the term correctly and his actually having applied the term correctly. Now, on a Cartesian way of construing sensations, the subject of a sensation is infallible about the presence and character of his sensations: if it seems to the subject that he
end p.39
has a particular sensation, then he has it. Furthermore, sensations are self-intimating: if the subject has a particular sensation, then he is bound to recognize that he does. Sensations, as putative targets of reference, would thus be objects that are essentially tied to the subject's recognition of them. There would be, even in principle, no epistemic distance between a subject's sensations and his affirmations of their presence—no possibility of a mismatch between the subject's judgment about the presence and character of a sensation and the relevant facts. However, if all there is to the meaning of "S" is (as per (d)) "the object for which it stands", then affirmations of the presence of S would be all there is to meaningful uses of the term "S". But then it seems that there will be no semantic distance between S's and applications of "S" by a speaker, either.
Even supposing there could be infallibly recognizable objects, they could not serve as appropriate targets of private linguistic reference. To reiterate, suppose that our private diarist experiences some sensation. How can he undertake to use, say, the sound "S" as a name for that sensation? That is, how can he make it so that, from that point on, "S" should be correctly applicable only to that same sensation (or, perhaps, to all and only future sensations that are 'relevantly similar to' the sensation he dubbed "S")? The problem is that our diarist would only be inclined to use "S" to refer to something if he thought he was experiencing a sensation that is the same as the one he originally dubbed "S". But, by the infallibility assumption, if he thinks he is experiencing a sensation of a certain character, he would be right. Which means that our private diarist cannot go wrong in his linguistic applications of "S". Whenever he is inclined to apply "S", his application will be correct. But this means, as Wittgenstein says, that here we cannot speak of correctness or incorrectness. And the diarist's original "S" would turn out to be no more than an empty mark made on the occasion of something going on in him.
The parallel to the argument against Cartesian Egos can be brought out as follows. The epistemic security of present-tense "I"-ascriptions of sensations may encourage the idea that avowing speakers have an unfailing way of successfully picking out their sensations, ostensively perhaps. The Cartesian recognizes that, with ordinary objects, there can be no apriori guarantee of successful recognition. So, to secure such guarantee, the putative referents of sensation terms would have to be
end p.40
quite extraordinary items. But, on the above reading of the Private Language Argument, Wittgenstein argues that no language could be set up to talk about sensations understood as such items, if we think of the words of such a language as having these items as their meanings (as per the Augustinian assumption (d)). For the conditions necessary for genuine reference will not be met by the terms of such a language. The analogous argument in the case of the Ego was that no word intended to name an object whose successful identification was logically or conceptually guaranteed could be a genuine name referring to that object.
For all that this version of the Private Language Argument shows, it may be possible for someone to use an existing public language to speak of things about whose occurrence he is epistemically infallible. It may even also be possible to invent one's own private language—a language governed by completely idiosyncratic rules—to speak of publicly accessible entities. What is unique to the case of the private language Wittgenstein envisages is that it is set up to talk about logically private items that are also taken to provide the meanings for its terms. It would follow that the meanings of terms in such a language would be logically inaccessible to anyone but their inventor. In a truly logically private language, the extensions of terms like "S" would be fixed by items whose presence and character the speaker in principle cannot fail to recognize, neither at the time of defining "S" nor subsequently. To summarize, the problem with setting up (and using) such a language is that the lack of epistemic distance needed to secure genuine logical privacy undermines the semantic distance required for linguistic reference. As targets of reference, private sensations would be completely locked onto the individual applications of the terms putatively referring to them. Whatever practice the private-sensation diarist will manage to set up using terms like "S" is bound to be an idle practice, and emphatically not one involving Augustinian linguistic reference.
It is important to recognize how different the private linguist's use of "S" would be from our use of an ordinary referring expression, such as "tiger". Someone who uses "tiger" to refer to something is potentially subject to at least the following three kinds of errors. Setting aside slips of the tongue, a speaker may call a leopard "tiger" due to a semantic error, because he thinks that the word "tiger" covers (or is a word for) both leopards and tigers. Or, the speaker may make
end p.41
a conceptual error. She may call a leopard "tiger" because she has in her conceptual repertoire no classification that separates tigers from leopards; she simply lumps them together. (This of course means that she also has the meaning of the word "tiger" wrong, but not because she matches it with a conjunction of two distinct concepts she possesses.) Finally, the speaker may make an epistemic error: she may use the term "tiger" to refer to a non-tiger because she thinks mistakenly that there's a tiger in front of her, due to some perceptual defect or poor observation conditions. Alternatively, her error may be brute: though her recognitional faculties are intact, and conditions are favorable, she may be confronted with a perfect counterfeit. The differences between these kinds of errors can be appreciated, inter alia, by considering the kinds of corrections that would be appropriate, and the revisions that the subject in error would be susceptible to.
Note that, in the ordinary case, the possibilities of making these types of errors are intertwined. Thinking that someone has made an epistemic error when calling a non-tiger "tiger" depends on crediting her with the relevant concepts and on the meaning of the word "tiger" being in place. Ascribing to someone a purely semantic error presupposes that they possess the relevant concepts. And taking someone to be making a conceptual error involving "tiger" would strongly tend to undermine the attribution to her of thoughts or judgments about tigers (though of course she may still use the relevant words, even successfully).
Now, in the private case we are trying to imagine, there are no meanings in place. The private linguist's task is to set meanings up. Presumably, however, we cannot suppose that the relevant concepts are in place either. At least we should not suppose that the private diarist shares our sensation concepts. For if he did, then the invention of "S" would be like the invention of some new word to express the familiar concept table. Such an effort would hardly qualify as the invention of a private term in Wittgenstein's sense. On the above reading of Wittgenstein's Private Language Argument, the conceptual difficulty arises when trying to make sense of someone setting up "S" as a genuinely private term designed to refer to items concerning which the possibility of epistemic errors is ruled out (by the Cartesian hypothesis), but where all there is to the semantic application of "S" is the marking of the presence of such items. For, by the Augustinian assumption (d),
end p.42
"S" is to be a term whose only meaning is the object for which it stands; and that object is to be an item whose presence is infallibly recognizable by the putative user of "S".
It might be thought that the Cartesian need not be accused of trying to establish private meanings for sensation terms. Why couldn't the Cartesian preserve the semantic distance, allow for the possibility in principle of semantic and conceptual mistakes in uses of sensation terms, but insist only on closing the epistemic distance, by denying the possibility of epistemic errors? After all, even the most staunch Cartesian, it seems, would agree that a person can utter a false self-ascriptive sentence about her present state of mind through a slip of the tongue or a misunderstanding. Linguistic competence is always presupposed as a background condition of claims about infallibility and self-intimation. In addition, it is typically presupposed that the subject is sincere and is not engaged in deliberate lying, deceit, or dissimulation. (This is where direct consideration of thought rather than speech comes in handy, as it seems to eliminate these kinds of distortion.) And, clearly, possession of the relevant concepts must also enter into the background conditions. The Cartesian, it seems, need only insist on one's infallible recognition of the presence (and character) of one's own present sensations. Thus she need only rule out the possibility of epistemic errors about them.
However, the Cartesian conception requires more than the merely contingent reliability of subjects in recognizing their sensations (and other occurrent states of mind). It requires some conceptual guarantee of correctness or impossibility of epistemic error. And it is this idea that, according to the above reading of the Private Language Argument, undermines the conditions necessary for genuine reference to sensations understood as private objects—objects infallibly recognized by their host. For it seems as though the only way to secure the conceptual impossibility of epistemic error in the case of sensations is by essentially tying their presence and character to acts of recognition by their hosts. If we suppose in addition that the meanings of the relevant terms are to be logically private, then it would turn out that the host of a sensation not only has the final word on whether a particular sensation of hers falls within the extension of a sensation term on a particular occasion of use; she also has the final (and only) word on what sensations are supposed to fall within the term's extension. Add to
end p.43
that the Augustinian assumption (d), that the meaning of a sensation term is exhausted by its extension, and the requisite semantic distance between uses of the relevant terms and what they are supposed to apply to would seem to vanish.
The Cartesian may try to secure the semantic distance between sensation terms (as used in avowals) and sensations that is required for reference by insisting on the independent existence of sensations as states of subjects' minds. This seems to be the point of invoking minds as immaterial substances. However, it is not clear how this can be reconciled with the conceptual impossibility of a recognitional failure. For, on this option, a subject who ascribes a present mental state to herself is making a higher-order judgment about an independently existing first-order mental state. And it is not clear why the judgment could not in principle fail, however superior the mode of access deployed in achieving it is, and even supposing the subject to be sincere and conceptually and linguistically competent.
11
Finally, it is worth highlighting the fact that the problem raised by the Private Language Argument on the above construal is not restricted to
talk about sensations understood in the Cartesian way. It extends to
thought about them as well. If the argument succeeds, it shows that we cannot as much as think about private sensations, at least insofar as thinking about an item requires referring to it in thought (i.e., singling it out, or bringing it under the relevant concept). For the argument readily generalizes to the possibility of establishing and achieving genuine reference, period. The problem facing the private diarist is not merely that of managing to communicate linguistically about private sensations; it is the problem of setting up any referential link to putatively private object. It may be, however, that the argument does not yet show that private sensations could not
exist, even if we could not talk or think about them.
12
To sum up, having construed sensation avowals on the model of ordinary descriptive reports about certain goings-on (albeit 'inside' the self-ascriber), the Cartesian still wants to maintain that they enjoy
end p.44
absolute infallibility. This must amount to the claim of the impossibility of epistemic errors of recognition. There is no possibility of a sincere, conceptually and linguistically competent subject who makes a false present-tense self-ascription of a sensation because he merely thinks he has that sensation, even though in fact he has not. In addition, the Cartesian view endorses self-intimation: the impossibility of having a sensation without recognizing it and judging that one has that sensation. But the Cartesian seems to be facing a certain dilemma. To do justice to the uniquely secure status of sensation avowals, he insists that the impossibility of recognitional error in the case of certain mental states is a conceptual one. Yet it seems as though the only way to secure such a conceptual impossibility is by essentially tying the presence and nature of sensations to acts of recognition by their hosts. The first horn of the dilemma is that doing so closes the semantic distance required for reference. On the other hand, the Cartesian may try to secure the semantic distance between the self-ascription and the state of affairs that is required for reference by insisting on the independent existence of sensations as states of subjects' substantial minds. The second horn of the dilemma is that doing so inevitably reintroduces an epistemic distance between subjects and their present sensations and opens the door for brute errors of recognition.
On the above reading, the impossibility of 'a language in which I speak of my private sensations' is either that a private linguist's efforts would fail to culminate in a sensation
language or that the language he might manage to set up would fail to be
private, in the sense that its terms would have logically private meanings, accessible only to the private linguist.
13 Recall, however, that the argument to that effect depended on four assumptions concerning the character of sensations and sensation terms. These were: (a) that sensations are genuine objects; (b) that sensations are private (only the host of a sensation can conclusively attest to its presence); (c) that sensations are transparently and infallibly recognizable by their hosts; and (d) that sensation terms are names for the relevant private objects. These are strong and, on their face, not very plausible assumptions. In the chapters to come, I will be considering, explicitly and implicitly, various ways of relaxing one or the other assumption, so that we can see our way clear to
end p.45
preserving key Cartesian intuitions about 'first-person' talk and thought about mental states without falling afoul of Wittgensteinian concerns about private language.
Getting Rid of Cartesian Egos: The 'No Reference' Thesis
Let us now return to "I". The very same feature of "I"—the apparent apriori guarantee of success in reference—may disqualify both Cartesian Egos and material human beings as potential targets of reference for "I". Material entities are apparently disqualified, because when it comes to such entities, it is hard to see how one could be guaranteed to hit the right target. Cartesian Egos, on the other hand, are apparently disqualified, because it is hard to see how they can retain their status as targets of ordinary reference, given that their identity is tied to guaranteed success in hitting them. Notice that what leads to trouble here is the attempt to preserve the combination of Semantic Continuity—treating "I" as a referring expression—and Epistemic Asymmetry—treating "I"-ascriptions as enjoying peculiar security, or protection from epistemic failure. If Epistemic Asymmetry is what we seek to explain, then perhaps we should reject Semantic Continuity.
This is precisely the strategy adopted by Anscombe in "The First Person". Anscombe suggests that the source of our problems is the idea that the security of "I" is a matter of guaranteed referential
success. Perhaps "I"-users are protected from various failures of reference
not because they are guaranteed to succeed in picking out the right object, but because they make no attempt to get hold of an object in the first place:
Getting hold of the wrong object is excluded, and that makes us think that getting hold of the right object is guaranteed. But the reason is that there is no getting hold of an object at all. With names, or denoting expressions (in Russell's sense), there are two things to grasp: the kind of use, and what to apply them to from time to time. With 'I' there is only the use. (1975: 59)
Anscombe is thus led to what I shall refer to as the No Reference View: if "I" cannot fail to hit its target of reference, it is because "I" is not even in the business of referring to an object. Grammatical appearances
end p.46
notwithstanding, " 'I' is neither a name nor another kind of expression whose logical role is to make a reference,
at all" (1975: 60).
14
If the surface subject-term in "I"-ascriptions is not a referring expression, then there should no longer be any need, or room, to seek a special kind of item that could serve as the unmissable target of reference for "I" (namely, a Self). So, once the philosophical grammar of such ascriptions is properly understood, we may be freed of the Cartesian illusion.
Instead of appealing to ontological dualism, the No Reference View attempts to save Epistemic Asymmetry by abandoning Semantic Continuity. But this is too heavy a price to pay. As Anscombe herself points out,
all "I"-ascriptions must be seen as governed by what she calls "a logician's rule" which gives their truth-condition:
If X makes assertions with "I" as subject, then those assertions will be true iff the predicates used thus assertively are true of X. (1975: 60)
15 Thus, my "I"-ascription, "I feel really tired", will be true if, and only if, it is true of this person, DB, that she feels really tired. But how can that be if no component of my self-ascription has the semantic function of referring to me (DB)? Anscombe remarks that "[t]he truth-condition of the whole sentence does not determine the meaning of the items within the sentence" (1975: 60). Presumably, she means that the fact that the self-ascription as a whole is made true by the actions, movements, or states of some particular individual does not entail that any component of the ascription has the semantic function of referring to some distinct entity. But while this is true, it is not clear what model we are to follow here.
In his discussion of Descartes's
cogito, Lichtenberg (
1971: 412) proposes that "Cogito", which is typically translated as "I am thinking" (or "I think"), should be translated by a less committal locution that avoids a personal pronoun. He proposed
Es denkt, on the model of
Es blitzt ("There's lightning").
16 "There's lightning", "It's raining",
end p.47
"It's thundering", are examples of statements whose truth-condition does not require the grammatical subject-term to refer to any entity. Such statements, to be sure, are made true by events and happenings in a certain region of space and time. But they do not involve picking out any individual of whom something is predicated. Similarly, Anscombe may be taken to suggest that "I"-ascriptions do not contain a distinctive component whose function is to pick out a particular individual. This, despite the fact that "I"-ascriptions are made true by the actions, states, etc. of a particular individual.
However, the difficulty is that "I" in, e.g., "I am really tired" functions logically, as well as syntactically, just like a singular referring expression. For example, it can replace (in context) singular terms salva veritate. On the basis of my contextually interpreted avowal, you could infer "She (or DB) is really tired", where your ascriptions presumably do involve reference to me. If I say, "You are really tired, and so am I," you would be entitled to conclude that DB is (also) really tired, and thus that at least two people are tired. If you say to someone, "DB will be giving a talk here tomorrow," I can truly report: "She thinks that I will be giving a talk here tomorrow." And so on. Denying that "I" refers flies in the face of such continuities between "I"-ascriptions and other ascriptions. Unlike "It" in "It's raining", or "There" in "There's a storm", "I" does not behave like an idle surface element. It appears to have a robust logical function. It is hard to see how it could serve this function without being semantically on a par with referential devices.
It may be more appropriate to invoke in this connection Russell's famous analysis of definite descriptions (in his (1905)). Consider the familiar example: "The present King of France is bald." On the Russellian analysis, the surface singular term—the definite description "the present King of France"—dissolves. The proposition expressed by the sentence turns out to have no proper subject. Rather, the proposition as a whole is construed as a quantified claim, asserting the unique existence of an individual satisfying a certain description ("is a present King of France") and ascribing to him a property (being bald). Thus, on Russell's 'translation', the statement becomes, roughly: "There is one and only one individual who is a present King of France, and whoever is a King of France is bald." This translation contains no constituent that works like a singular term
end p.48
whose function is to pick out a particular individual. Surface-grammatical appearances to the contrary, Russell maintained, definite descriptions are not expressions whose function is to refer to particular individuals.
The case of "I" seems different, however. The Russellian analysis of definite descriptions still leaves some room for interchanging the surface singular term with other (genuine) singular terms that would pick out the individual uniquely satisfying the relevant description. For, provided the relevant individual exists, the unique existential claim into which the surface singular term dissolves in effect provides a recipe for getting to a particular individual. And that individual can then be referred to using genuine singular terms. (Russell himself mentions the possibility of a "verbal substitution" of "Scott" for "the author of Waverly" in his analysis of "George the IV believed that Scott was the author of Waverly" (
1905: 489).) By contrast, simply denying that "I" refers does not tell us how we could preserve the idea of the interchangeability (in context) of "I" with genuinely referring expressions, which successfully single out the user of "I".
The problem runs deeper. Accepting the No Reference thesis would have the undesirable consequence that an "I"-ascription does not really have the form "
a is F", where "
a" is taken to refer to the individual who must satisfy the predicate "F" in order for the ascription to be true. But, as emphasized by Evans (
1982: ch. 7), "I"-ascriptions are logically linked with other ascriptions which clearly do have this form. To make the self-ascription "I am happy", I must be able to conceive what it is for anyone, myself included, to be happy (I must grasp:
x is happy for arbitrary
x's). Also, at least for some form of identification, |δ|, I must grasp what it is for [δ = I] to be true.
17 In other words, my self-conscious thought requires that I be able to understand what it would take to latch onto some person in the world as myself; it is
about some particular individual. My "I"-ascriptions, like all ascriptions of the form "
a is F", conform to the Generality Constraint, which "requires us to see the thought that
a is F as lying at the intersection of two series of thoughts: the thoughts that
a is F, that
a is G, that
a is H...on the one hand, and the thoughts that
a is F, that
b is F, that
c is F...on the other hand" (for any F, G, H, and
a,
b,
c, of
end p.49
which the subject has a conception) (Evans
1982: 209). Yet the No Reference thesis seems to imply that "I"-ascriptions are not subject to this constraint, since they are not on a logical par with ascriptions of the form "
a is F".
In order to retain Epistemic Asymmetry while avoiding appeal to Cartesian Egos, one might be prepared to pay a rather high price. One might be willing to deny that "I"-ascriptions have the logical form "a is F", and to reject Semantic Continuity. But the point of doing so will be lost if the promised pay-off is not forthcoming. The Cartesian offers us reference to infallibly known immaterial Egos as the distinctive mark of avowals, which can serve to separate them from other self-ascriptions (the idea being that only when making first-person mental self-ascriptions do we refer to our Ego). By contrast, the No Reference thesis, understood as a flat-out denial that "I" refers, would fail to distinguish avowals from many "I"-ascriptions. Even if the thesis were restricted to present-tense "I"-ascriptions, it would still apply to a variety of non-mental present-tense "I"-ascriptions. Anscombe's own prime examples are all of that kind (see below). This means that accepting No Reference would not achieve for us what the Cartesian appeal to Egos is designed to do: namely, to explain why avowals should seem epistemically more secure than all other "I"-ascriptions.
Suppose (contra Anscombe) that we were to restrict the No Reference View to certain uses of "I"-ascriptions only—specifically, to their uses as subject. The idea would be to maintain that the distinction between the use of "I" as object and its use as subject coincides with the distinction between referential and non-referential uses of "I". Only uses of "I" as subject would then be claimed to be non-referential. After all, as noted earlier, the Cartesian cannot offer a uniform treatment of the reference of "I" either. He must allow that at least sometimes—e.g., in "I have a scar on my arm"—"I" is used to refer to something other than one's immaterial Self. Similarly, it may be suggested that the No Reference claim should be regarded as a claim only about certain uses of "I", rather than a claim about "I" as an expression type.
In support of this move, one might invoke Peter Strawson's claim (
1950) that referring is not something that components of ascriptions do, in the abstract, but rather something speakers or thinkers do in
end p.50
context. We can see an application of this claim that is relevant to our present discussion in Keith Donnellan's familiar distinction between "attributive" and "referential" uses of definite descriptions.
18 A speaker can use a definite description such as "Smith's murderer" on an occasion attributively: to pick out
whoever satisfies the description "the man who murdered Smith". Or the speaker can use the same description referentially: to pick out a particular, contextually salient individual, who may or may not in fact be the man who murdered Smith. Donnellan suggests that Russell's aforementioned 'no reference' analysis of definite descriptions is appropriate only for the "attributive" use of definite descriptions. If so, then, in agreement with Strawson, we could maintain that whether or not a definite description has a referential function depends on its use. Definite descriptions purport to refer when used referentially; they do not purport to refer when used attributively.
Somewhat similarly, it might be suggested that there are two uses of "I", one referential, the other not. When using "I" as object, speakers do intend to single out a particular object or individual in the world, of whom they want to predicate something (being 6′ tall, or having a scar on the arm, and so on). This would constitute a use of "I" as a more or less ordinary referring expression. However, when using "I" as subject, the expression does not serve a referential function. It is not intended to pick out a particular individual; it is rather more like an idle surface element that contributes no distinct logico-semantic component to the proposition expressed.
Notice, first, that this modified No Reference View would still not allow us to preserve semantic and logical continuities between ascriptions that involve uses of "I" as subject and other ascriptions. The view would still face the question: what licenses the inference from "I have a headache" (in which "I" is presumably used as subject) and "He has a headache" to "Two people here have a headache". And so on. The modified No Reference View would still have to buy Epistemic Asymmetry at the cost of Semantic Continuity, and would require denying that uses of "I" as subject involve ascriptions of
end p.51
the general form "a is F". But, more importantly, portraying avowals as "I"-ascriptions that do not involve reference to an object still seems insufficient for explaining their distinctively secure status. The reason is as follows. Presumably, the security that would result from using "I" non-referentially would be a kind of negative security: where there is no attempt to target oneself as an object of reference, there is no room for failure in picking out the right referent. However, the security of avowals goes beyond this (a point I shall take up further in Chapters
3 and
4). It is manifested also in a relatively high degree of security in applying the relevant predicates. To understand this further security, it looks as though we must attend, in addition, to the kinds of states avowals ascribe (viz., occurrent mental states). But then the worry is that, even after we rid ourselves of the need for 'sure-fire' reference for "I", the temptation to postulate Cartesian Egos may remain. This is because Cartesian immaterial substances may be required as appropriate 'substrata' for the securely ascribed states. If that is so, then we will not have escaped the Cartesian illusion. And we might as well use Egos as referents for certain uses of "I".
We should pause to reflect on what an Anscombe-style explanation of the seeming predicative success of avowals would look like. Assuming that predicates such as "tired", "hungry", "thinking about Descartes", or "hoping for rain", etc. purport to refer to instances of the relevant mental states or events, and given Epistemic Asymmetry as applied to the predicative component of avowals, the Cartesian view would urge us to suppose that mental predicates refer to extraordinary types of property (at least as applied to oneself). Otherwise, how could the striking predicative success of avowals be explained? Rejecting this Cartesian view, the Anscombian strategy would dictate denying that this security is a matter of guaranteed success. Instead, it would portray it as a matter of impossibility of failure due to the fact that the relevant predicates do not purport to refer to the relevant states, at least not in their first-person use. This would be the No Reference View applied to the predicative component of avowals. Note that here too we have an analogue for the assumption that uses of "I" require some way of identifying the referent of "I". The analogue would be the assumption that the application of mental predicates to relevant instances always requires using some epistemic
end p.52
method for determining the presence of the relevant state, be it in the case of others or oneself. This is an assumption I shall later dispute.
19
It should be evident that putting together the two No Reference views would yield absurd results.
20 It would turn out that no part of an avowal has the function of referring to anything—either to an individual or to a state of that individual. Avowals, then, would not be
about anything, they would not
say anything. As thoroughly inarticulate noises, avowals could not obey the "logician's rule" of which Anscombe speaks, for they could not possess truth-conditions. Accepting this would amount to no mere compromise of Semantic Continuity—it would constitute a complete abandonment of it.
Toward the end of her paper, Anscombe notes that the "I"-thoughts she has been considering "have been only those relating to actions, postures, movements, and intentions"—e.g., "I jumped", or "I am standing"—"[n]ot, for example, such thoughts as 'I have a headache,' 'I am thinking about thinking,' 'I see a variety of colours,' 'I hope, fear, love, envy, desire,' and so on" (1975: 63). Her reason for focusing on the former kind of thoughts and avoiding the latter, 'Cartesianly preferred thoughts' (as she calls them) is that "only thoughts of actions, postures, movements, and intended actions … both are unmediated, non-observational, and also are descriptions (e.g., 'standing') which are directly verifiable or falsifiable about the person of E. A. Anyone, including myself, can look and see whether that person is standing" (1975: 63). If we regard these ascriptions as grammatically 'subjectless', there will be no need to invoke anything extraordinary to understand how they can be made true or false. For such ascriptions are made true or false by the readily observable "actions, postures, movements" of a particular individual—the individual who happens to make the ascriptions using the semantically idle "I".
Contrasting the "I"-ascriptions that concern Descartes with the ones that can be handled simply by treating them as 'subjectless',
end p.53
Anscombe says: "The Cartesianly preferred thoughts all have this same character, of being far removed in their descriptions from the descriptions of the proceedings, etc., of a person in which they might be verified. And also, there might not be any. And also, even when there are any, the thoughts are not thoughts of such proceedings, as the thought of standing is the thought of a posture" (1975: 63). Anscombe acknowledges that this raises a difficulty for her anti-Cartesian position, and proposes to address it on another occasion. But she does not seem to recognize that the difficulty undermines the only attraction that the No Reference proposal has in the first place. The point of denying that "I" refers (at least in some of its uses) was to allow us to avoid the temptation to postulate peculiar objects—immaterial Egos—to which uses of "I" can securely refer. But if this temptation remains even after we accept the No Reference View of "I" (on however limited a basis), we seem to have made no progress. And if, to avoid the temptation, we would also have to resort to a No Reference View of mental predicates, we ought to question the cogency of the No Reference approach and re-evaluate some of the assumptions that gave it its initial appeal.
end p.54