I. The ‘straight reading’ and its attractions

I herein offer a re-reading of Anscombe’s ‘The First Person’. I hope to sow serious doubts about a certain standard Straight Reading. I will also tentatively sketch an alternative interpretation.

Understanding the paper aright is valuable in itself, but my re-reading has the additional advantage of acquitting Anscombe of obvious empirical mistakes about language. Moreover, it has her not only anticipating important insights into the ‘mode of meaning’ (p. 55) of the first person singular pronoun, but also offering up sage methodological advice.

I begin by expanding upon the Straight Reading. I hope to make clear that Straight Readers are not obtuse: their interpretation is indeed initially attractive. I will then sow the doubts. I end with (what I take to be) a superior, revisionist, alternative.

Some terminology. By the Radically Non-Referring View (RNR), I shall mean a conjunction of two theses. RNR’s first plank is that in no way does ‘I’ refer. This seems outrageous on its face, given the surface grammar of ‘I’. That pronoun certainly seems to play a referring role in (1)-(2), for instance:

1. (1) I am hungry

2. (2) I shaved Bill

The second plank attempts to explain this away by analogizing ‘I’ to expletive pronouns. On RNR, the surface grammatical subject of (1)-(2) is to be treated in a way comparable to (3)-(5):

1. (3) It is raining

2. (4) It seems that Juana is happy

3. (5) There is no largest prime

Spelling this out, ‘I’ turns out to be a dummy syntactic element with no logico-semantic role. It is present in (1)-(2), despite contributing nothing to logico-semantics, merely because, as in (3)-(5), English demands a nominal in subject position.

The relationship between the two planks is slightly complicated. Plank one clearly does not entail plank two. Instead, plank one makes it urgent to reanalyze surface grammar: one cannot simply assert plank one and refuse to say more. Urgency isn’t entailment, however, not least because plank two isn’t the only explaining-away option. For instance, and simplifying greatly, Doyle (Reference Doyle2018, §10.4) suggests that ‘I’ functions like a peculiar second-order adverb: e.g., ‘I’ in (1) takes the property as argument and yields SELF-ASCRIPTION() as value. Here we get a re-analysis without treating ‘I’ as occurring only with zero-place predicates in the way that genuine expletives do. Nonetheless, the most familiar explaining-away; plausibly the best one; and the one hinted at in Anscombe’s text, is the proposed analogy with the expletives ‘it’ and ‘there’.Footnote ¹ I thus build this into RNR itself.

What I call the Straight Reading attributes the Radical Non-Referring View to Anscombe in her ‘The First Person’. It shows up repeatedly. Gareth Evans (Reference Evans1982) derides RNR as Anscombe’s ‘extraordinary thesis’. Saul Kripke says that it ‘is in and of itself incredible, difficult to understand at all’ (Reference Kripke2011, 312). Most strikingly, Peter van Inwagen (Reference van Inwagen2001, 6) writes:

Others who find something like RNR in ‘The First Person’ include: Clarke (Reference Clarke1978), Garrett (Reference Garrett1994, Reference Garrett1997), Hamilton (Reference Hamilton1991), Hinton (Reference Hinton2008), Peacocke (Reference Peacocke2008), Taschek (Reference Taschek1985), Teichmann (Reference Teichmann2008), White (Reference White1979) and Wiseman (Reference Wiseman2017).

These philosophers do not speak with one voice. Some explicitly attribute both planks; some explicitly mention only plank 1, but seemingly read in plank 2 implicitly; some, including Doyle (Reference Doyle2018), attribute only plank 1. Also, most think of RNR as a ‘bug’: they may recognize her article’s influential insights; they may acknowledge that it sparked an important literature; but, ultimately, they consider RNR to render Anscombe’s 1974 Wolfson College Lecture unconvincing. Others, including not just Doyle (Reference Doyle2018) but also Rachael Wiseman (Reference Wiseman2017), think such a radical stance on the self is a ‘feature’. I set aside these in-house minutiae and focus on RNR as defined above.

There are, I readily concede, several compelling reasons to adopt the Straight Reading. First, there are a handful of spots in which Anscombe seems explicitly to embrace various formulations of RNR. She outright says or clearly implies:

1. (6) Logically speaking, ‘I’ is not a name (pp. 53 and 56)

2. (7) ‘I’ does not involve singular reference (p. 53)

3. (8) ‘I’ does not refer to the ‘I’-user (p. 56)

4. (9) ‘I’ is not a singular term whose role is to make a reference (p. 56 and 58)

And she infamously concludes her critical discussion of ‘I’ by writing:

1. (10) ‘I’ is neither a name nor another kind of expression whose logical role is to make a reference, at all (p. 60, her emphasis).

Second, two like-minded philosophers allegedly shared both RNR-planks. According to many, her husband Peter Geach (Reference Geach1957, Reference Geach1962, Reference Geach1969) did. So, it is sometimes said, did Anscombe’s greatest philosophical influence, Ludwig Wittgenstein. He writes in The Blue and Brown Books remarks such as ‘The mouth which says “I” … does not thereby point to anything’ (Reference Wittgenstein1958, 68). (See also Wittgenstein’s lectures in 1930–33, as reported in Moore (Reference Moore1954).) If Anscombe’s nearest intellectual peers held the Radically Non-Referring view, she likely shared it.

A third attraction of the Straight Reading is that Anscombe’s central argument seemingly aims for RNR as its conclusion. Apparently, her argument consists in observations about the peculiar functioning of ‘I’ and a complex reductio: given these observations, assigning a reference to ‘I’ would lead to (something in the neighborhood of) Cartesianism about the self.

The observations about the specialness of ‘I’ are complexly intertwined in Anscombe’s paper. I will do some separating-out for the sake of clearer exposition. And I will elide various details because, in the end, this is not what I take Anscombe’s view to be. (See Garrett Reference Garrett1994; Teichmann Reference Teichmann2008; for somewhat parallel and very helpful summaries of Anscombe’s text.)

Observation #1: Immunity to Certain Errors. Granting for the sake of argument that it refers at all, ‘I’ seems to be immune to reference failure: ‘If “I” is a name, it cannot be an empty name’ (p. 55). ‘I’ also seems to be immune to a certain kind of mistaken identification: Guaranteed reference [in this latter sense] would entail a guarantee, not just that there is such a thing as X, but also that what I take to be X is X' (p. 57). Or again: ‘[The “I”-user cannot] take the wrong object to be the object he means by “I”’ (p. 57). For instance, apparently when Ron says ‘I’ in (11), it cannot fail to refer:

1. (11)

   Ron: I am smoking

2. Archibald: #You’re right that someone is smoking, but the person you intended by ‘I’ is actually Juanita Perez, not Ron Star

This contrasts with, e.g., [_NP The man with the hat] in [_S [_NP The man with the hat] is smoking]: a speaker could use that definite description when noticing a trick of the light and would thereby not refer to anything. More intriguingly, the absurdity of Archibald’s reply highlights that a speaker cannot wish to refer to one thing with ‘I’ and yet somehow end up referring to something else.Footnote ² In contrast, Ron could use [_NP The man with the hat] to talk about a woman with a large, geometrical hairdo.

Here is a related peculiarity of ‘I’/‘myself’. Sentence (12) is perfectly sensible:

1. (12) When John Smith spoke of John Horatio Auberon Smith, he was speaking of himself, but he did not know this

Here, writes Anscombe, speaking of oneself ‘is compatible with not knowing that the object one speaks of is oneself’ (p. 47). There is a use of ‘I’/‘myself’ which is comparable. Suppose my daughter enters me in a writing contest under the pseudonym ‘Stan Robertson’. I am eagerly following the competition, and learn that ‘Stan’ is in his 50s, teaches philosophy, grew up in Toronto, etc. Given our similarities, I announce that I hope Stan Robertson wins. Later I learn about my daughter’s charming ruse, and I say:

1. (13) When I spoke of Stan Robertson, I was speaking of myself, but I did not know this

Like the name-involving (12), (13) is a perfectly sensible sentence. However, and this is the peculiarity observed by Anscombe, there is another use of ‘I’/‘myself’ – called ‘the indirect reflexive’ use – where this kind of ignorance is ruled out. Thus, learning the facts, I cannot latterly report the events correctly with (14) or with (15):

1. (14) I wanted me, myself, to win, but I didn’t know this

2. (15) I wanted to win, but I didn’t know this

If I say truly either ‘I wanted me, myself, to win’ or ‘I wanted to win’, then I must know that (as it were) it was me who was competing.

Observation #2: Immunity to Doubt. The foregoing facts about immunity-to-error, if such they be, would yield epistemological consequences. Though I (that is, Robert Stainton) can doubt whether Robert Stainton exists, thinks, etc., I cannot doubt whether I exist, think, etc. On a related note, though I can doubt whether I (that is, Robert Stainton) am Robert Stainton, I cannot doubt (in the ‘indirect reflexive’ use) whether I am myself. ‘I’-talk apparently rules out certain skeptical worries.

Observation #3: Absence of Bodily Properties. If ‘I’ refers, one can seemingly conceive of it doing so in the absence of a body altogether, and in the absence of bodily sensations. To illustrate this, Anscombe introduces a Tank Thought Experiment. In an imagined situation of utter sensory deprivation, writes Anscombe, a person can still think: ‘I won’t let this happen again’ (p. 58). Reference, if such there be, is not here secured by anything sensory. She also introduces a Body-As-Puppet Thought Experiment wherein the following sentence could be used and understood: ‘When I say “I”, that does not mean this human being who is making the noise. I am someone else who has borrowed this human being to speak through’ (p. 60). Here, if ‘I’ referred, it needn’t be to any kind of sensing body.

Observation #4: Intimate Connections to Perception and Action. Finally, the perception that, say, the man in the hat is in danger (where the man in the hat is, as a matter of fact, me) seems quite different in its action-generating effects than the observation that I myself am in danger. Closely related to this, ‘I’ seems useable to express an intention to act in a certain way – as opposed to expressing an empirically-based prediction about how a certain body (e.g., that of Robert Stainton) will behave in the future (p. 56).

Part of what makes the Straight Reading tempting, I have said, is that Anscombe seems to offer a sophisticated argument for the Radical Non-Referring View. So far, I have surveyed the observations which play a role therein. Its reductio structure can be rendered explicit via a simplified formalization. (Obviously nothing this prosaic appears in Anscombe’s paper.) Letting ‘Obs’ stand for Anscombe’s four observations are correct, ‘Name’ stand for ‘I’ is name-like, ‘Desc’ stand for ‘I’ is a kind of definite description, ‘Dem’ stand for ‘I’ is a kind of demonstrative, ‘Cart’ stand for Cartesianism is correct, ‘Ref’ stand for ‘I’ is a referring term and ‘Exp’ stand for ‘I’ is an expletive, the reductio runs:

The Straight Reader can readily and charitably reconstruct from ‘The First Person’ a defense of each premise.

P2 may be supported by a dilemma. Suppose ‘I’ were comparable to an ordinary kind of name: ‘Archibald Mustafa’, ‘Juanita Perez’, ‘Ron Star’, etc. Some of Anscombe’s observations would not be met. Starting with Observation #1, there would exist no guarantee of reference for ‘I’, because names can fail of reference: ‘Homer’ and ‘Socrates’ may so fail, as may my three inventions above. There would be the possibility of error through misidentification: Archibald can say, ‘Juanita is smoking’ and be talking about Ron; so, ‘I’ would wrongly be predicted to behave this way too. And ‘I’, like names, would have only a ‘direct reflexive’ use. Turning to Observation #2, if ‘I’ were like an ordinary proper name, there would also exist the possibility of various doubts. Anscombe points out in particular that ‘René Descartes exists’ is open to doubt even by Descartes (p. 46). In a related vein, the sense of ‘I’ should be sensory/bodily: insofar as living body, animal, human, etc., are built into the conception/sense of ‘René Descartes’, they would be built into the first person singular qua synonymous name (pp. 45–46 and p. 55). Finally, recalling Observation #4, ‘I’-qua-name would exhibit a disconnect – of the sort that can arise with ‘The man with the hat’ – between third-personal perception and first-personal action. So much for the dilemma’s first horn. Suppose instead that ‘I’ is assimilated to a very special kind of name: ‘I’-qua-name has a private conception/sense, involving self-consciousness; one which would endow the ‘I’-sense with the powers necessary to deal with Anscombe’s observations (p. 52). The problem here is that the sense of ‘I’ would be private and yield infallibity in a way that Wittgenstein (Reference Wittgenstein1953) argued to be incoherent. Indeed, it would smack of something very like Cartesianism. In short, the first horn establishes that the ‘ordinary’ option leads to ~Obs; the second horn establishes that the ‘special’ option leads to something along the lines of Cart. ‘Ordinary’ and ‘special’ (in the relevant way) being the only options, P2 is true.

Still drawing charitably on Anscombe’s text, P3 may also be defended with a dilemma. Either the definite description sharing the meaning of ‘I’ is context-sensitive or it is not. Anscombe suggests that it cannot be context-sensitive – e.g., of the form ‘The sayer of this’ – because ‘this reference could only be sure-fire if the referent of “I” were both freshly defined with each use of “I”, and also remained in view so long as something was taken to be I’ (p. 57). (See also p. 58.) Leaving us without an enduring ‘self’, this horn is unacceptable. There remain two context-insensitive sub-options. ‘I’ could be a ‘special’ definite description, with an immediate-infallible-private-subjective conception/sense. But then we essentially arrive at Cart again. (Relatedly, no one else could fully grasp the meaning of ‘I am in Ottawa’ as uttered by, say, Justin Trudeau.) Horn two, sub-option two, is that ‘I’ is semantically akin to an ‘ordinary’ context-insensitive description. But then, as with names, some of Anscombe’s observations aren’t met. Thus, P3 is established: every description-based path is otherwise philosophically unacceptable or leads to (~ Obs v Cart).

Regarding P4, there is yet another dilemma. Either ‘I’ is likened to a totally bare demonstrative or to one with an implicitly-intended sortal. The former looks promising because it might sidestep some of the troubles tracing to descriptive content. Anscombe, however, taking off from Wittgenstein’s (Reference Wittgenstein1953) critical discussions of ostensive definitions, knowledge by acquaintance and such, claims that this route forward is illusory: even with seemingly bare demonstratives, the sortal must be present. ‘Where “I” is a name or a demonstrative, there is the same need of a “conception” through which it attaches to its object’ (p. 55). Or again: ‘Assimilation to a demonstrative will not – as would at one time have been thought – do away with the demand for a conception of the object indicated. For, even though someone may say just “this” or “that”, we need to know the answer to “this what?” if we are to understand him…’ (p. 53). The other demonstrative-based horn, building in an implicit sortal, faces the familiar conflicts with the observations – along with, surprisingly, the possibility of reference failure: ‘Thus I may ask “What’s that figure standing in front of the rock, a man or a post?” and there may be no such object at all…’ (p. 54).

Given her quick dismissal of it, the Straight Reader may reasonably suggest that Anscombe thinks P5 obvious. Most of her audience would have agreed.

As for P6, she apparently urges that the burden falls on the proponent of a referential view to find another model beyond names, descriptions and demonstratives. These three look exhaustive pro tem:

Finally, given that ‘I’ is not like a name, definite description or demonstrative, something must explain away its surface grammar. As we saw at the outset, a natural idea is that ‘I’ is an expletive. Hence P7.

Straight Readers may also underscore how forcefully Anscombe phrases her ultimate conclusion:

II. Sowing doubts about the ‘straight reading’

Notwithstanding all this, there are serious doubts about the Straight Reading. It faces at least three exegetical obstacles: ascribing to Anscombe an obviously false view about natural languages; contradicting passages in her text; and taking her to advance a very problematic argument form and premises. We’ll also see in the latter half of this Section that the evidence presented in Section I can be rebutted.

One way to present the idea that the mind is a mere bundle – e.g., a Humean or Buddhist cluster of mental happenings – is to draw an analogy between ‘I’ and expletive pronouns. However apt, to go beyond this philosophical metaphor to contend as an empirical linguistic hypothesis that ‘I’ simply is Radically Non-Referring, in the sense of behaving like the dummy ‘it’ in (3)-(4), should raise eyebrows.

Reason upon reason can be adduced; I will content myself with a few. To be clear, some apply directly to plank one of RNR; some apply directly only to plank two, but straightforwardly raise troubles for plank one; others apply directly only to plank two. Even the latter, however, leave the proponent of plank one with no good explanation of the surface grammar of ‘I’. I leave it to the reader to decide which consideration belongs where: the crucial result is that RNR, understood as a claim about natural language syntax and semantics, is rightly rejected as ‘extraordinary’ (Evans) and ‘incredible’ (Kripke).

The morphosyntax of the first person singular pronoun, in English and related languages, is massively unlike that of true expletives. Not only can it be marked for accusative/dative (‘me’) and appear in direct and indirect object position, the first person pronoun can be governed by prepositions (‘for me’, ‘from me’, ‘towards me’), and it has both a possessive (‘my’/‘mine’) and plural form (‘we’/‘us’/‘our’). Again, whereas ‘It is raining’ with expletive ‘it’ is bizarre, the first person pronoun can readily receive stress: ‘I won the race, not Archibald’. It can also appear unembedded – e.g., someone may call attention to themselves with the accusative ‘Me! Me!’ in English, with the disjonctif case ‘Moi! Moi!’ in French, or with the nominative ‘Yo! Yo!’ in Spanish. ‘I’ licenses aphonic gaps: [I₁ want [e₁ to dance] or [e₁ to leave]] is fine. And ‘I’ coordinates with patently referential nouns to form nominal compounds as in [[_NP [_NP John] and I] love jazz. We go every Friday]. Patently, ‘it’ qua expletive has none of these properties.

Note too that some languages have a first person pronoun but, being ‘pro-drop’ (or ‘null subject’) languages, do not permit expletive pronouns along the lines of (3)-(5). Examples include several varieties of Romance (including Latin, which Anscombe not only knew well, but mentions as permitting subjectless sentences) as well as many Balto-Slavic tongues. Anscombe, if she really held the Radical Non-Referring thesis, would have to maintain, about these, that the first-person pronoun is the sole exception to their ‘no-expletives’ generalization.

Notice now a range of logico-semantic contrasts between ‘I’ and genuine expletive pronouns. As noted by Doyle (Reference Doyle2018), Harcourt (Reference Harcourt and Teichmann2000), Kripke (Reference Kripke2011) and Teichmann (Reference Teichmann2008) among others, ‘I’ licenses existential inferences: ‘I am a philosopher. I am an angler. Therefore, there is at least one philosopher who is an angler’ is a valid inference. (Compare: ‘It is raining in Toronto. It is snowing in Calgary. Therefore, there is at least one thing raining in Toronto which is snowing in Calgary’.) Relatedly, ‘I’ falsifies universal quantifications: e.g., ‘Every Canadian philosopher specializes in ethics’ is falsified by ‘I am a Canadian philosopher and I do not specialize in ethics’. In transparent contexts, ‘I’ can be substituted salva veritate with (seemingly) co-referring terms:

1. (16) Question: Did Rajni Smith attend the dinner?

   Answer: I attended dinner. I am the woman named 'Rajni Smith'. So, yes, Rajni Smith attended the dinner

Manifestly, the expletive ‘it’ in (3) and (4) has none of these logico-semantic properties.

Another problem for RNR. As hinted when introducing the very notion, if ‘I’ is an expletive in [_S I am sitting down] then [_VP sitting down] must be a zero-place predicate therein. That’s because, in the allegedly parallel ‘It is raining’, there is no argument to the verb phrase at all. But, as Edward Harcourt (Reference Harcourt and Teichmann2000, 33) underscores, [_VP sitting down] is clearly a one-place predicate in [_SJones is sitting down] the argument being Jones. This yields the empirical prediction that every predicate which can take ‘I’ as subject has variable arity. On a related note, when Ron says ‘Archibald is sitting down’ and Archibald says ‘I am sitting down’, they could not be agreeing about what property holds of a given referent.

As a final example, Taschek (Reference Taschek1985, 635ff) follows Anscombe and others in holding that (17) can be read in two ways:

1. (17) I wanted myself to win

It has both a ‘direct reflexive’ reading (the one which accurately portrays my situation with respect to the writing contest) and an ‘indirect reflexive’ reading (the one which would not capture my cognitive situation vis-à-vis my daughter’s trick). Paraphrasing, what Taschek queries is why, as uttered by him, there seems to be the entailment of (17), on both readings, that Taschek wanted Taschek to win. How can this be, if ‘I’ in no way refers, and is instead an expletive pronoun?

The problem for the Straight Reading, it should be stressed, is not that it identifies an error; nor even that it attributes one to a great philosopher. Plato, Locke and Kant, etc., all made mistakes. The problem is that Anscombe was one of the leading philosophical minds of the 20^th century, whose work in ethics and action theory is foundational for entire sub-fields. Moreover, she was a rare woman among the grand pioneers of the Analytic tradition, equaling the stature of J.L. Austin, Rudolf Carnap, Donald Davidson, H. Paul Grice, Peter Strawson, etc. And the Straight Reading has her making a blunder about natural language grammar and logico-semantics. (Imagine exegesis of, say, Michael Dummett or G.E. Moore which had them overlooking existential generalization. That is the level of uncharity at work in the Straight Reading.)

My attempt to sow doubts is reinforced by underappreciated passages in Anscombe’s text. She writes, regarding the syntax of ‘I’, that ‘certainly “I” functions syntactically like a name’ (p. 48). Even more telling (because clearly not about mere surface form), she grants that ‘I’ has a ‘mode of meaning’ (p. 55), something expletives proper surely lack. She also makes a series of concessions about ‘I’ which are puzzling on RNR. Though it doesn’t have the same sense as ‘Elizabeth Anscombe is Elizabeth Anscombe’, she writes that ‘I am not Elizabeth Anscombe’ would nonetheless be a lie in her mouth (p. 53). In a similar vein, she begins the paper by allowing that there is a mundane, practical, everyday sense, in which it would have been correct for Descartes to say ‘I am Descartes’ (p. 46). And, paraphrasing, she notes that when things like (1)-(2) are spoken, the issue ‘Who made the assertion?’ is ‘all-important’ (p. 60). One can, of course, ask Anscombe’s ‘Whose assertion?’ question about utterances of (3)-(5); but it typically wouldn’t be the ‘all-important’ one. Finally, near the end she writes that ‘I am this thing here’ is a real proposition which means, roughly, ‘this thing here is the person (in the “offences of the person” sense) of whose action this idea of action is an idea, of whose posture this idea of posture is the idea. And also, of which these intended actions, if carried out, will be the actions’ (p. 61). The passage is notoriously opaque, but for my purposes one can highlight a crucial lesson: ‘I’ does provide a ‘thing-here’ (viz., a person in the forensic sense). Again, contrast the expletive ‘it’.

The former passages are already difficult to square with the Straight Reading. Others strike me as flatly contradicting it. As Harcourt (Reference Harcourt and Teichmann2000, 28) noticed, Anscombe accepts that each one uses ‘I’ only to speak of himself (p. 49). Statements involving ‘I’ specify objects (p. 47) and concern objects (p. 61 and p. 63) for her. They are verified/falsified by objects (the latter typically being the speaker’s body and its actions, motions, posture, etc.) (p. 61 and p. 63). Dummy syntactic elements, in contrast, are not used to ‘speak of’, ‘specify’, ‘concern’, etc., any kind of entity. Most arresting of all, Anscombe writes: ‘And a self can be thought of as what “I” stands for, or indicates, without taking “I” to be a proper name’ (p. 52). (Despite all this, I reiterate that Straight Readers are not being obtuse: it is easy enough to miss the import of the foregoing quotations because of Anscombe’s opaque sic-et-non style.)

Finally, though many suppose otherwise, Anscombe never affirms that ‘I’ is very like the dummy pronoun ‘it’. What she stresses is that there is no easy route from, say, ‘I’ serving as the grammatical subject of a verb to its being a full-blown referring term. As she put it:

This is far from radical: it is a foundational precept of Analytic philosophy that one shouldn’t blithely infer logical role from surface syntax.Footnote ³

I have cast doubt on the attribution of the Radically Non-Referring view to Anscombe. It’s equally doubtful that she is putting forward the argument for it which I semi-formalized. Anscombe’s philosophical fellow travelers were mentioned above as evidence that she was propounding RNR; turning the tables, her influences call into serious question that the form of argument she is deploying can function as the Straight Reader supposes. Anscombe wasn’t just a leading philosopher of her age, she was a leading Ordinary Language Philosopher: she trained and taught at Oxford during the heyday of linguistic analysis; attended Wittgenstein’s lectures at Cambridge after her graduation; translated his work; and ended her career by occupying his academic post. Ordinary Language themes run throughout her oeuvre, including in ‘The First Person’ itself. And yet the Straight Reader has her drawing conclusions about how ordinary language must work on the basis of arch philosophical convictions – a cardinal sin against which J.L. Austin et al. constantly inveighed. In short, Anscombe is portrayed as taking the Linguistic Turn in the wrong direction: from the a priori falsity of Cartesian metaphysics and epistemology to empirical pronouncements about linguistic form and content.

The next step is to rebut the considerations from Section I. As a preliminary, I need to introduce, briefly, a point about ‘refer’, ‘referring’, ‘reference’, etc. As we use them in Analytic philosophy of language, these just are technical terms. In ordinary parlance, merely calling attention or alluding to something would be counted as ‘referring’ to it. Here is an attested example, which I overheard recently:

1. (18) When Trinity said she’d met two famous philosophers at dinner, she was referring to Galen Strawson and David Wiggins

In colloquial usage, there is nothing whatever awry with classing Trinity’s utterance of ‘two famous philosophers’ as ‘referring’. Another example, from a recent Internet search, comes from a woman – call her Alice Thomas – lamenting in a blog post:

1. (18) He keeps referring to me as a friend

The claim could be made true by, say, the man uttering (20) when asked why he regularly dines with the Thomas family:

1. (20) Because friends are important

As quotidian usage would have it, Alice spoke correctly: the man did ‘refer to her as a friend’ using (20).

Thus, Anscombe must – like all of us Analytic philosophers – be using ‘reference’ in some technical sense or other. (Clearly the same holds of ‘logically name-like’.) As will be explained in Section III, I myself take her to be using ‘refer’/‘name’ in a way that dominated Oxbridge philosophy at the time: a usage inspired by Frege’s works, developed by Russell and the Logical Positivists, etc. Nearly half a century later, with the writings of Kaplan, Kripke et al. firmly implanted, we might think of that usage of ‘refer’, ‘referring’, ‘reference’, etc., as idiosyncratic. But it would not have sounded special or unfamiliar then and there.

Applying due terminological caution, the grounds given in support of the Straight Reading thus turn out to be illusory. Reconsider, first, these seemingly decisive passages:

1. (6) Logically speaking, ‘I’ is not a name (pp. 53 and 56)

2. (7) ‘I’ does not involve singular reference (p. 53)

3. (8) ‘I’ does not refer to the ‘I’-user (p. 56)

4. (9) ‘I’ is not a singular term whose role is to make a reference (p. 56 and 58)

5. (10) ‘I’ is neither a name nor another kind of expression whose logical role is to make a reference, at all (p. 60).

Crucially, it is consistent with all of these that Anscombe is merely denying ‘naming’/‘reference’ in one historically quite specific sense.

Appeal to her influencers’ views turns out not to be probative either. Not just Anscombe but her peers may well have been using ‘reference’ et al. in a now non-standard way. For example, when it comes to straightforward sentences like ‘I shaved Bill’, Geach seems to me merely to reject Frege-inspired ‘reference’ for the first person singular subject.Footnote ⁴ Granted, Geach (Reference Geach1962) contended that the reflexive pronoun ‘myself’ doesn’t refer in a sentence like (21); but this doesn’t entail that ‘I’ fails to specify/concern an object in plain-old (2)Footnote ⁵:

1. (21) I shaved myself

2. (2) I shaved Bill

As for Wittgenstein, while he toyed early on with the Lichtenbergian thesis that ‘I think’ is like ‘There is thinking hereabouts’, by the time of the Philosophical Investigations he had long since abandoned his dalliance with such radicalism. So, insofar as Anscombe was following Wittgenstein’s lead, she would have rejected RNR.

Finally, Anscombe simply wasn’t making the argument which the Straight Reader extracts from the text. Revisiting my semi-formalization on p. 9, P5 mischaracterizes Anscombe’s stance on Cartesianism: the latter is, for her, not so much incorrect as bizarre. Relatedly, to read Anscombe as performing a modus tollens misses the nuances of her complaint: because Cart is nonsensical by her lights (p. 51), the problem is that treating ‘I’ as her interlocutors do leads to language-gone-on-holiday. As she herself couches things:

Nor does Anscombe hold P6, if ‘referring term’ is used in our 2018 sense. I will defend the point at length below, but for now notice that P6, read as pertaining to our usage of ‘referring term’, would be falsified by ‘here’, ‘today’, ‘yesterday’, etc. Anscombe, however, was aware of such words – she even cites Casteñada’s (Reference Castañeda1967) article ‘On the Logic of Self-Knowledge’ (p. 46). Finally, as noted, there is no explicit reason in the text for pinning the exceedingly implausible P7 on Anscombe.

III. My revisionist re-reading

At the risk of over clarifying, this seems an apt juncture for a synopsis. I began by introducing the two planks of the Radically Non-Referring View: i) ‘I’ in no way refers; and ii) the surface grammar of the first-person singular is to be explained away by re-analyzing ‘I’ as an expletive. What I have labeled Straight Readers take Anscombe to be arguing for both planks (though some fellow-travelers, e.g., Doyle (Reference Doyle2018), only attribute the first, explaining away the grammatical appearances in some other fashion). I will be proposing an alternative to that reading. So far, however, my focus has been on sowing doubts. On the one hand, the Straight Reading requires Anscombe to hold some startling views about natural language; and there are numerous passages in ‘The First Person’ seemingly in tension with, or even flatly contradicting, RNR. On the other hand, the alleged grounds in favour of the Straight Reading lose their force upon reflection. Remarks (6)-(10) are consistent with a much weaker reading: viz. that ‘I’ doesn’t behave in the way that prototypical mid-century Oxbridge philosophers (inspired by a Russellian and Logical Positive take on Frege) understood names/definite descriptions to do. Also, it’s unclear that Anscombe’s closest peers, Geach and Wittgenstein, actually held RNR. Finally, the semi-formalized argument doesn’t capture her overarching line of argument; so, one shouldn’t appeal to that argument as grounds for finding its conclusion in Anscombe.

The essential notion for my revisionist alternative is ‘deflated reference’. Using the via negativa, I may contrast it with the Frege-inspired notion mentioned above, which I’ll label ‘inflated reference’. The latter demands a symbol: it is expressions, not people, who ‘refer’ in this sense. That symbol must have a descriptive content which fixes its Bedeutung. This enables it to be synonymous with an equally sense-bearing name or definite description (p. 48 and pp. 51–52). An ‘inflated’ referring term’s descriptive content will analytically entail features of the referent (or, relatedly, the verification conditions encoded about it). These analyticities, in turn, yield necessities and a priori knowledge (p. 48). To take the classic example, assuming ‘Hesperus’ to have as its descriptive content first heavenly body visible at night, this would simultaneously fix Venus as referent and make the name synonymous with [_NP The first heavenly body visible at night]. [_S Hesperus is a heavenly body] will thus be analytic. Hence it is necessary that Hesperus is a heavenly body, and anyone grasping the meaning of ‘Hesperus’ will know a priori that it is. It may also be built into ‘inflationary reference’ that uttering an expression as a genuine referring term requires that the speaker intend to refer to something. (Anscombe herself seems to include this as a condition for ‘referring expressions’ at pp. 56–57.)

Now, ‘reference’ has a familiar relation/relatum ambiguity: e.g., reference-qua-relation obtains between the name ‘Elizabeth Anscombe’ and the philosopher; that same woman is the reference-qua-relatum (i.e., ‘referent’, ‘nominatum’) of the name ‘Elizabeth Anscombe’. The features in the previous paragraph all have to do with the relational sense of the word. A final aspect of my ‘inflated reference’ pertains to the relatum: there must be a robust entity, ‘a distinctly identifiable subject-term’ (p. 53), corresponding to the symbol – as opposed, say, to mere Strawson (Reference Strawson1953) feature-placing. Returning to (3)-(5) as examples, though a lay speaker might colloquially say that there is ‘reference’ therein to rain, seeming happiness, and the absence of a largest prime respectively, this would not count as ‘inflated reference’.

My main contention is that Anscombe merely denies that ‘I’ refers in such an ‘inflated’ way. This leads to an obvious question: if ‘I’ is not an expletive for her, yet it also lacks ‘inflated reference’, what is the word’s ‘mode of meaning’? The short answer, as noted, is ‘deflated reference’. But, turning to the via positiva, what is that? Unsurprisingly, given the relation/relatum ambiguity, it has two aspects. Instead of a descriptive content for ‘I’, the first person singular pronoun has only a rule-of-use. Speaking through her ‘imaginary logician’ (to whom I will return at the end), Anscombe herself provides it:

This rule-of-use does not fix whether the object specified/spoken of will be a soul, a Cartesian ego, a body, etc. Per se, it is silent on that: the pronoun not expressing any conception/sense, the ‘referent’ (in my ‘deflated’ sense) will be whatever turns out to use ‘I’ assertorically. Continuing to spell out my re-reading, this proves to be ‘deflated’ too. The sorts of things which assert hereabouts, for Anscombe, are persons in the ‘offences against the person’ sense, as she aptly phrases it. (These are to be sharply contrasted with either ‘selves’ or ‘persons’ construed mentalistically, e.g., in terms of self-consciousness (p. 52); about these, Anscombe genuinely is scornful.) What’s more, persons in this forensic sense turn out to be living human bodies – men and women, animals of a certain species (p. 55 and p. 62):

Crucially, on pain of violating some observations above, none of this is analytically settled by any alleged descriptive sense of ‘I’, there being no such thing.

Finally, ‘deflated reference’ does not require that the person-asserting with ‘I’ intend a description of an entity: given an utterance by X, the rule-of-use is sufficient to fix that ‘I’ in that utterance specifies/concerns X.

Why adopt this re-reading? It is far more charitable to Anscombe, as she is acquitted of naïveté about natural language syntax and semantics. It fits with the passages in (6)-(10), once we construe them as pertaining only to logical names and referring terms understood as ‘inflated’. And it readily accounts for Anscombe’s otherwise puzzling remarks, including:

1. (22) ‘I’ is the word each one uses to speak of himself (p. 47, p. 48 and p. 49)

2. (23) ‘I’ can be used to specify an object (p. 47)

3. (24) ‘I’ has a ‘mode of meaning’ (p. 55)

4. (25) ‘I’ functions syntactically like a name, occupying the same grammatical places (p. 48 and p. 56)

5. (26) ‘I’ can be replaced salva veritate with a name (p. 56)

6. (27) Sentences containing ‘I’ concern an object, and are verified/falsified by the events/states involving that object (p. 61 and p. 63)

Beyond these now-familiar considerations, note Anscombe’s recurring insistence that if there is to be reference, there must be a conception/sense at play:

The demand is perfectly reasonable if she has ‘inflated reference’ in mind. (See also Anscombe Reference Anscombe1976.)Footnote ⁷

More pointedly, Anscombe may be adumbrating my re-reading in so many words when she insists: ‘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’ (p. 59).

The final advantage of my revisionist reading is that it coheres better with what seems the overall dialectic of the paper. Its first paragraph reads:

This is not a mere historical preamble. It lays out the very topic of the paper: Anscombe’s central aim is to defang this neo-Cartesian semantic argument for dualism.

Her paragraph is confusing. Here is my attempt to unpack it:

P1*, P3* and P4* merit little comment in the present context: are all standard Cartesian-style premises. What requires stressing, given its novelty, is the argument’s semantic nature: it represents an important departure from the parallel argument in the Meditations. Consonant with an Ordinary Language Philosophy orientation, Assumption is phrased using semantic ascent: it doesn’t directly propose, for reductio, that I am a body, but rather that ‘I’ names a body. P2* is an attempt to paraphrase formally the final sentence of the quotation. Though my phrasing is convoluted, the main idea seems to be captured, viz., a linguistic spin on Augustine’s idea that I am exactly that which I am perfectly certain of being (p. 45). More broadly, the argument is innovatively semantic in that one arrives at C1* and C3* essentially via considerations about the conception/sense of ‘I’.

Assuming my reconstruction is roughly correct, it underwrites my point about Anscombe’s ultimate quarry: she wishes to rebut this line of thought. She does not attack the ‘usual suspects’ like P1*, P3* and P4*. Instead, she refutes what she terms ‘these writers’ assumption’, namely P2*. More specifically, she rejects P2*’s first conjunct. Her argument against it is two-fold. The first half echoes, to a degree, the argument which the Straight Reader claimed to find: the various observations about the special functioning of ‘I’, if one takes its ‘mode of meaning’ to involve the kind of ‘inflated’ reference that P2* demands, lead one to talk philosophical nonsense. Thus, some thesis really is reduced to absurdity in ‘The First Person’ – not that ‘I’-use specifies/concerns an object, however, but P2*. Secondly, no good reason has been offered for taking the ‘mode of meaning’ of ‘I’ to be fully akin to names (or descriptions, or demonstratives): the only evidence presented concerns surface grammar and logical role, and neither are sufficient to establish conjunct (a). Now, if the main thrust of Anscombe’s paper is to answer this neo-Cartesian ‘corollary argument’, all she needs is that the term ‘I’ lack ‘inflated’ reference: calling P2* into question does not require commitment to RNR. I thus urge that we read Anscombe as arguing for the weakest premise sufficient for her task.

IV. Concluding remarks

Many thoughtful readers, including great philosophical admirers of Anscombe, have embraced the Straight Reading. I think this a mistake, but it’s an understandable one. The dialectic of ‘The First Person’ renders it difficult to discern which claims Anscombe embraces and which she presents for criticism. I end with an especially important illustration.

The paper introduces an ‘imaginary logician’. On the one hand, he holds not only that ‘I’ occupies the same syntactic place as names et al., but that it shares their general logical character (p. 55 and p. 56). Thus, e.g., ‘I’ can be replaced salva veritate by an ordinary name of X when ‘I’ occurs transparently in subject position in assertions made by X (p. 56). He also proffers the rule-of-use discussed above:

Given it, a person encountering a statement the form ‘I am F’, knowing who made that statement, knows who must be F for the statement to hold (p. 56).

On the other hand, Anscombe’s ‘imaginary logician’ maintains that the foregoing syntactic and logical features are already sufficient for ‘I’ to be a genuine referring term. As a result, ‘the way of being given’ associated with ‘I’, if there even is one, isn’t his concern. What’s more, though he grants that the ‘indirect reflexive’ exists, and that there is immunity to mistaken identification with ‘I’, the imaginary logician is indifferent to explaining these (p. 55 and p. 57). More generally, he’s uninterested in the curious characteristics which differentiate ‘I’ from, say, ‘The man with the hat’ – or from the peculiar name ‘A’ used in Anscombe’s imagined community (p. 49ff), which has every person apply ‘A’ to themselves via observation.

Most readers would agree that this is how Anscombe portrays her foil. What many miss is that Anscombe accepts the ‘first hand’ whole cloth: regarding the syntax of ‘I’; regarding its logical role being similar to names et al.; and regarding the context-sensitive rule-of-use capturing the truth-conditional contribution of ‘I’.Footnote ⁸ About the latter, she suggests that all hands – including the Cartesian – should agree that ‘[t]he rule was sound enough’ (p. 56).

What Anscombe resists is his ‘other hand’. The syntax and logical role of ‘I’ are not sufficient to establish that ‘I’ is genuinely a referring term (in her sense of that expression). That is a matter of ‘mode of meaning’ – in particular, of whether ‘I’ invokes a conception/sense. Thus, after laying out the ‘imaginary logician’s’ view, she writes: ‘But the question is, why “I” was said to refer to the “I”-user?’ (p. 56). She also insists that the ‘way of being given’ is essential: it is unacceptable that her imagined interlocutor provides ‘no views on the sense of “I” or on what the “I”-user means by “I”’ (p. 56). Equally essential is the specialness of ‘I’. As philosophers, we need to say something about the (seeming) observations she has made: either explaining them, or explaining them away as illusory. Anscombe urges, finally, that the rule-of-use for ‘I’, though true as far as it goes, is impotent when it comes to such explaining:

This make-believe interchange, properly understood, not only reinforces my main conclusions, but encapsulates them. In accepting the ‘first hand’, Anscombe rejects not just plank two but plank one of the Radical Non-Referring View. For her, ‘I’ does ‘refer’ (as many would now use that term), but only in the ‘deflated’ sense captured by the rule-of-use. I.e., despite ‘I’ lacking a sense, speakers nonetheless use it, in context, to identify an object (about which something may then be predicated). In rejecting the ‘other hand’, Anscombe denies that we have been given solid reasons to posit ‘inflated reference’ for ‘I’. Relatedly, she makes an excellent point about methodology in philosophical semantics: be cautious in drawing inferences about (things along the lines of) Sinn/Bedeutung – not just from surface syntax, but from logical role. Finally, in chastising the ‘imaginary logician’ for his lack of curiosity, Anscombe confirms why she cares about the semantics of ‘I’. The existence and nature of a Frege-inspired conception/sense is central to the neo-Cartesian semantic argument she is trying to rebut; the ‘imaginary logician’ is faulted for refusing to address this most seminal issue.

I end with some remarks on what I have and haven’t achieved. I have sown serious doubts about the Straight Reading: it is highly uncharitable; it fits ill with many passages in her paper; and the arguments in favour of attributing the Radical Non-Referring View to Anscombe, albeit initially plausible, don’t stand up to scrutiny. I have not have provided sufficient evidence for adopting my alternative as the uniquely correct reading. As a general point, the text resists that, due to its richness and style. Beyond this, there are specific passages which remain mysterious given only what I’ve said above.Footnote ¹⁰ Of particular salience, why does she say that ‘I am Elizabeth Anscombe’ is not an identity proposition? (p. 60 and p. 63). I believe I have a very promising reply, but I haven’t presented it here. Finally, no argument to the effect that Anscombe doesn’t hold such-and-such can be wholly convincing without a solid account of what she does hold. Thus a complete defense of my re-reading requires much more on the presumed ‘Anscombean reference for “I”’. I assure the reader, however, that this is addressed in depth and in detail in Botterell and Stainton Reference Botterell and StaintonForthcoming.