1 Introduction

This Element examines Wittgenstein’s views on sensation, perception and private language at §§243–315 of the Philosophical Investigations, drawing on The Big Typescript and Wittgenstein’s lectures and notes beginning in 1929. Wittgenstein is widely credited with producing the ‘Private Language Argument,’ which ostensibly tries to show that a private language is logically impossible. This is often taken to mean that, if the argument works, language logically presupposes a community of speakers. It is not just paradigmatically, but essentially, social.

What the argument is and whether it succeeds are contentious, but a long-standing consensus locates it at PI §258. Here Wittgenstein imagines keeping a private diary of one’s sensations, designated by terms defined by fixing one’s attention on one’s sensations and associating names with them. Supposedly, Wittgenstein argues that this procedure must fail, either because the connection between term ‘S’ and sensation S cannot be remembered later or because no connection has been established. An influential dissenting view purports to find the Private Language Argument earlier in the Investigations in the discussion of rule-following: if rule-following is essentially social, then so is language.

I argue, to the contrary, that Wittgenstein’s critique of private language neither reduces to a single argument nor attempts a definitive refutation. Rather, we find an array of arguments and examples designed to show that a certain way of thinking about sensation and perception – as objects situated in phenomenal space – is not obligatory, at best, and has been given no clear content, at worst. Wittgenstein thus advances no general thesis about the nature of language (e.g., that it is intrinsically social).

In Section 2, I connect these claims to some general considerations about Wittgenstein’s philosophical method, arguing that his remarks on philosophy at PI §§89–133 should be understood not as aiming to do away, once and for all, with all philosophical problems but as reimagining the ongoing task of philosophy as alleviating particular philosophical confusions, as they arise, by attending to overlooked aspects of the grammar of the terms in which those confusions are expressed.

Section 3 traces concerns about the object-model of sensation and perceptual appearances to Wittgenstein’s critique of sense-datum theories, from 1929 through his lectures of the mid-1930s. Here he criticises our tendency to model visual space or tactual space on physical space and to think of sensations and perceptions as objects – sense-data – occupying these phenomenal spaces. Because my phenomenal space does not overlap with yours, my sensations seem to be private objects that no one else can have or know directly. However, physical space is a dubious model for phenomenal space, and it is unclear how I could own something that no one else could. Wittgenstein’s complaints with private language derive from these concerns, together with his realisation that naming is more complex than supposed in the Tractatus: the resources available for naming in public language are stripped away by the very hypothesis that I might learn the meanings of sensation-terms from acquaintance with my own sensations. It is thus puzzling how a private sensation-language could get started.

These considerations cast doubt on the suggestion (Section 5) that Wittgenstein is targeting the language of a solitary speaker. Rather, his critical target is a language whose terms would get their meanings from sensations construed as private objects. I further distinguish (Section 4) the ‘ordinary’ privacy of our sensations from their alleged ‘superprivacy’ (NPL, 447) and their epistemic privacy from their ontological privacy, turning (Section 6) to a discussion of first-person authority – our acknowledged entitlement to express our own experiences. Were sensations phenomenal objects, this authority might derive from our privileged access to the private objects of our experience. However, we could as plausibly see it as a grammatical characteristic of paradigmatic first-person uses of verbs of experience, which are analogous to non-verbal expressions (Sections 6 and 7).

In Sections 8–11, I distinguish four ‘waves’ of interpretation that have washed over the Investigations since 1953, carving channels through which much subsequent reading flows. A recurrent theme of the first three waves is the assumption noted earlier that Wittgenstein has a central argument – the Private Language Argument, often located in the diary-example – which aims to refute the logical possibility of a private language.

The First Wave (Section 8) presents this fabled refutation as a verificationist challenge to the private speaker’s ability to remember how to apply the terms of a private language. Such readings sometimes flirt with behaviourism or collapse the distinction between private language and solitary language. They also attribute to Wittgenstein a problematic theory of meaning and reduce the refutation to scepticism about memory.

The Second Wave (Section 9) remains preoccupied with the diary-example but correctly bypasses epistemic considerations for semantic ones. Wittgenstein’s real concern is not whether the private diarist could remember the correct use of a private sensation-term, but whether the term could be defined in the first place.

These considerations, however, are swamped by the Third Wave (Section 10), which seeks ‘the real “private language argument”’ (Kripke Reference Kripke1982, 3) in Wittgenstein’s discussion of rule-following (PI §§185–242). On this reading, Wittgenstein is a sceptic about rule-following, who holds that we can be justified (but never right) in attributing understanding to others, only if our community would agree. This view collapses the distinction between private language and solitary language, and it ignores the provenance of Wittgenstein’s discussion of privacy in his transitional writings.

The Fourth Wave (Section 11) takes many forms but is generally driven by a re-evaluation of Wittgenstein’s work as a whole. For my purposes, it matters most for rescuing the insights of the Second Wave, for eroding the idea of ‘The Private Language Argument,’ and for drawing attention to the origins of Wittgenstein’s discussion of privacy and private language in his transitional writings.

At PI §270 the diary-example of PI §258 is modified by the suggestion that I might use a manometer to discover a correlation between the sensation S and an increase in my blood-pressure. I endorse Hacker’s claim (Section 12) that this example is meant to undermine the idea that the meanings of sensation-terms rest on the (re)identification of sensations, as they would if sensations were like objects. However, I argue that the example is usefully read through the prism of Wittgenstein’s treatment of the inverted spectrum (PI §272) – the old idea that there might be undetectable differences in the ways you and I see colours. Wittgenstein argues that this seems possible only if we sever attributions of colour-perception from the behavioural criteria on which they ordinarily rest. Doing so reduces perceptions to superprivate objects with no bearing on the uses of colour-terms. It cannot, then, matter whether we correctly identify those objects or not.

Finally (Section 13), I examine the much-discussed beetle in a box of PI §293, which compares the idea that I might learn the meaning of ‘pain’ from my own sensation to the idea that everyone might own a box containing something called a ‘beetle,’ which no one else can see. This is the Investigations’ most explicit expression of doubt about the ‘object-and-name’ model of sensation-language encouraged by the misleading metaphor of phenomenal space. Misunderstandings of this passage have provoked worries about behaviourism, but such worries themselves presuppose the object-and-name model.

No topic in Wittgenstein’s philosophy has generated as much commentary and as little consensus – about understanding Wittgenstein and about whether he is correct – as the discussion of private language. Over 40 years ago, Stewart Candlish joked that ‘no one with a concern for his own health’ (Reference Candlish1980, 85) would try to master the huge secondary literature on this topic. The challenge has only grown, but I have tried to acknowledge as much of the extensive commentary as I can. I have overlooked some items for pragmatic reasons, and I am sure to have neglected others through inattention or forgetfulness. I beg my readers’ indulgence.

2 Methodological and Exegetical Commitments

At PI §§89–133Footnote ¹ Wittgenstein suggests that the philosopher’s task is not to give explanations resembling scientific hypotheses (PI §§109, 126) about the fundamental nature of language or meaning or mind but to clarify such concepts (PI §§109, 118–119, 122, 125, 133) by revealing overlooked aspects of the grammar of the language in which we formulate them (PI §§122, 129–133). Philosophical puzzlement often arises when we are misled by ‘certain analogies between the forms of expression in different regions of our language’ (PI §90), and overcoming that puzzlement demands ‘an overview of the use of our words’ (PI §122). We need ‘complete clarity. But this simply means that the philosophical problems should completely disappear’ (PI §133).

The interpretation of these and related remarks is controversial and has become a central concern for readers since what I call later (Section 11) the ‘Fourth Wave.’ According to Robert Fogelin, they express a ‘neo-Pyrrhonian’ impulse ‘to eliminate philosophy altogether’ (Reference Fogelin1994, 205).Footnote ² This impulse struggles with a ‘non-Pyrrhonian’ one, which seeks ‘to replace his earlier foundationalist theory with a nonfoundationalist theory’ (205). Pyrrhonian and non-Pyrrhonian interpretations of the text, accordingly, emphasise one impulse over the other.Footnote ³

Fogelin leaves no room for the thought that Wittgenstein wants to forsake ‘traditional philosophy in order to do philosophy better’ (Stern Reference Stern2004, 35) without simply offering another theory. An analogy from 1931 suggests such a possibility. Wittgenstein compares philosophy to an infinite strip that may be divided into problems, either lengthwise or crosswise. Traditional philosophy tries ‘to grasp the unlimited strips,’ and it seems ‘that it cannot be done piecemeal’:

The task of philosophy is ongoing, but philosophy does not grapple with timeless problems (infinite strips). Rather, its problems are continually produced and reproduced by the structure of our language (PI §115):

These problems owe their longevity to the structure of European languages. This does not make them eternal or universal. A language with a sufficiently different structure, we could conjecture, might have different philosophical preoccupations. But, for those embedded in a particular tradition, the task of the philosopher must be ‘to put up signs to help in getting past the dangerous spots’ (BT 423) – those places in the language where we are tempted by misleading grammatical analogies, for example.

The finite cross-strips of Wittgenstein’s analogy are particular grammatical confusions that must be dealt with one at a time: ‘Problems are solved (difficulties eliminated) but not a single problem’ (PI §133). Further difficulties always remain. Our language is complex. The task of getting a synoptic view of our grammar is multifaceted. The ‘ancient city’ of language is ever-expanding with ‘new suburbs’ (PI §18) for us to get lost in. (Consider the philosophical puzzles produced by the advent of the digital computer.) It should be unsurprising that the philosopher’s work has no end.

But what is that work like? Peter Hacker, a careful reader of Wittgenstein’s remarks on method (Baker and Hacker Reference Baker and Hacker2005 [Reference Baker and Hacker1983], 251–334), takes them to eschew theory-building and explanation in favour of describing grammatical rules and showing that traditional philosophy violates them – ‘that certain suppositions, certain putative doctrines, make no sense’ (287). The discussion of private language, then, tries to show that ‘the idea of a private language that is implicitly invoked by so many philosophical doctrines is … incoherent’ (287).

Interpreters like Stephen Mulhall complain, however, that, even if Hacker’s reading does not commit Wittgenstein to an alternative theory, in aiming to demonstrate that talk of private language is literally ‘nonsensical or incoherent, a violation of grammar’ (Mulhall Reference Mulhall2007, 18), it remains ‘substantial’ (18) in a way that goes beyond Wittgenstein’s intentions. By contrast, Mulhall’s ‘resolute’ (18) reading sees Wittgenstein as imagining various things an interlocutor might mean by ‘private language’ but failing to find anything that will satisfy the interlocutor (18–19), leaving it open that there might yet be some meaningful construal of the interlocutor’s words.

Gordon Baker, having earlier collaborated with Hacker, similarly repudiates the substantial approach, suggesting that Wittgenstein seeks to draw our attention to neglected aspects of our grammar, not to patrol ‘the bounds of sense, sharply reprimanding philosophers who commit offences by uttering nonsense’ (Reference Baker2004, 94). Wittgenstein’s goal is therapeutic in a way analogous to the ‘treatment of a particular patient rather than [to] a kind of campaign to improve public health’ (132). It is particular, not general, in its aims, and it ‘propos[es] other ways of seeing things’ (290), not demonstrations of the incoherence of views under criticism.

I can make no significant contribution to these debates here. I can only try to situate my own approach. Like Hacker, I doubt that Wittgenstein is best read as seeking to end philosophy once and for all. The work of his philosophy, however, is pursued not by replacing old theories with better ones, nor by showing extant views to be nonsense. The point of acquiring a ‘surveyable representation’ (PI §122) of the grammar of the expressions that puzzle us, I contend, is to help us to see ‘connections’ (BT 417) we have overlooked – for example, between the grammar of sensation-terms and the grammar of colour-terms (Sections 3 and 6.1) – in much the way that noticing a visual aspect consists in seeing an ‘internal relation’ (PPF xi §247) between, for example, the duck-rabbit and other images of rabbits. So, like Baker (Reference Baker2004), I think that drawing attention to overlooked aspects of our grammar is central to Wittgenstein’s approach.Footnote ⁴

There may be more than one connection to emphasise: ‘We want to establish an order in our knowledge of the use of language: an order for a particular purpose, one out of many possible orders, not the order’ (PI §132). So, again, like Baker (Reference Baker2004) and Mulhall (Reference Mulhall2007, 18), I do not think that Wittgenstein’s goal in the private-language discussion is to show the idea of a private language to be nonsense because it violates the rules of grammar. This much already follows from a conception of grammar present in 1930:

It may seem that, on my reading, Wittgenstein really is trying to refute the possibility of a private language because I emphasise an array of arguments with their roots in Wittgenstein’s reflections from 1929 onward. However, these arguments are best read as trying to show that certain received views about sensation and perception are not obviously correct – that we can look at the grammar of sensation and perception in another way, and that defenders of private language have not yet given clear content to their proposals (though I hope it will also be clear that Wittgenstein leaves these defenders little cause for optimism). Moreover, the arguments are part of the text, and much light is cast on Wittgenstein’s discussion by making them explicit.

These remarks are compatible with acknowledging that the Investigations is not written in a univocal voice – that the impulses to provide a better theory or to show that talk of private language is nonsense or to do away with philosophy completely all find expression in the complex dialogical structure of the book. ‘What we are “tempted to say”,’ in these cases, ‘ … is, of course, not philosophy; but it is its raw material’ (PI §254). There are, as David Stern argues (Reference Stern2004, 21–28), many voices in the Investigations, and it is difficult and contentious to identify any of them with the voice of the author. We do not suppose, however, that Ravel’s voice, in his 1914 Trio in A minor, is really the piano, not the violin or cello, and we need not see Yuri as merely the mouthpiece of Pasternak to find a critique of the Bolshevik revolution in Doctor Zhivago. I read the Investigations in a similar spirit – as a mature expression of arguments Wittgenstein had formulated (sometimes much earlier), which attempts to exemplify the attraction of the problems with which it deals, presenting them as the raw material of philosophy, and then showing that these problems are not inevitable.

I advance several claims informed by these thoughts. First, ‘there is no such thing as “the private language argument”’ (Canfield Reference Canfield2001, 378). Rather, we find a collection of arguments, questions, thought-experiments and so on, whose goal is to break the grip that a certain way of thinking about sensation and perception has on us in reflective moments. There is neither one central argument, nor one that aims to defeat all comers.Footnote ⁵

Relatedly, Wittgenstein advances no general thesis about the nature of language in these remarks: he does not argue that language is intrinsically social. He seeks to avoid puzzlement produced, particularly, by a misleading analogy between physical space and phenomenal space – the space(s) of sensation and perception – and to remind us of the extensive resources on which we ordinarily rely when naming something in public language. The analogy invites us to objectify our sensations and perceptions and to expect the properties of the resulting ‘objects’ to mimic the properties of physical objects – to expect these ‘objects’ to be persistent, divisible, recurrent, reidentifiable and so on. The fact that my phenomenal space does not overlap with yours then makes it tempting to suppose that I have exclusive ownership and exclusive knowledge of my sense-data. However, the more we emphasise the privacy of these objects, the more we undermine the analogy that invited us to think about them, for we have stripped away what we presuppose in our everyday applications of language, and it becomes puzzling how we could ever name such private objects.

Critics may complain that I rely too much on statements of Wittgenstein’s method from the early 1930s and that the arguments I take as basic to the discussion of privacy are similarly dated. Others will protest that I understate what has been achieved by the Investigations – that the arguments clearly refute the privacy of sensation and the possibility of a private language.

So perhaps I will satisfy no one, but many passages on method in the Investigations survive from the Big Typescript and harmonise with my reading, and the early arguments against privacy and private language are also clearly present. A proper appreciation of the text must at least make these arguments explicit, and doing so brings into focus the great coherence of PI §§243–315. At the same time, Wittgenstein’s remarks on method suggest that these arguments do not aim at categorical refutations but at helping us recognise overlooked aspects of our grammar, so that what seemed like inevitable features of sensation and perception begin to look like contingent features of our ways of describing them (PI § 104).

3 Privacy and the Objectification of Sensation and Perception

Wittgenstein’s discussion of private language does not aim to demonstrate anything about the nature of language as such, but it does invite us to reconsider the grammar of our terminology for sensation and perception: it invites us not to model such terminology on names for spatio-temporal objects. This point has been frequently recognised,Footnote ⁶ but its central importance emerges from Wittgenstein’s writings of 1929 to the mid-1930s.

Wittgenstein here first challenges both the object-model of sensations and perceptual appearances and the privacy of sensations and perceptual appearances. His criticisms, beginning in 1929, have two interrelated targets: the misleading analogy between physical space and phenomenal space, and sense-datum theories of perception. The latter target may be surprising. In his 1930–32 Cambridge lectures (LWL) Wittgenstein mentions sense-data frequently and uncritically. However, though he adopts the Cambridge vernacular for his students, his concurrent manuscripts repeatedly criticise sense-datum theories, most notably for their commitment to private objects. I begin with sense-datum theories and then turn to phenomenal space.

When G. E. Moore revived the term ‘sense-data’ (Moore Reference Moore1909–10, 57), he emphasised the ‘act-object’ analysis of sensation and perception, for which he had argued in 1903. On this analysis, we must distinguish the act of perception or sensation from its object. Otherwise, like Berkeley, we may confuse sensation or perception as an act with sensation or perception as an object and infer that the objects of sensation and perception are ideas because the acts of sensation and perception are mental (Reference Moore1903, 445–446). In 1909–10 Moore identified the immediate objects of sensation and perception as sense-data, and the mediate objects of sensation and perception as spatio-temporal objects, but he left unsettled whether sense-data were representations of spatio-temporal objects (representational realism), constitutive of spatio-temporal objects (phenomenalism), or identical with parts of surfaces of spatio-temporal objects (naïve realism). The first two answers treat sense-data as private objects.

Wittgenstein’s complaints with sense-data, beginning in October 1929, were various, but two matter here: that sense-data cannot be privately owned, and that sensations and perceptual appearances are not objects. Consider privacy.

The manuscript version of this passage (29 November 1929) continues: ‘This simply shows that something is not in order with the concept of sense-data’ (MS 107, 216). A year later he returns to the argument: ‘What is essentially private, or seems that way, doesn’t have an owner’ (BT 508). If no one else can have my sensations, then there is no sense in saying that I have them. Ownership implies that another might own them, as someone else will own my house if I sell it.

It is not just the private ownership of sensations that bothers Wittgenstein:

The idea that I might own my sensations suggests that they are objects (as sense-datum theories propose). But we might as plausibly see pains as qualities, not objects, ‘and if they coincide in intensity, etc., etc., then they are the same; just as two suits are the same colour if they correspond with respect to brightness, saturation, etc.’ (BT 510; PR §61; cf. BBB 55; RSD, 292–293). Moreover, objects occupy space: they are ‘around’ the body and ‘have an effect on it’ (BT 508). But the ‘space’ of sensation is not like this, as becomes evident when we ‘examine what sort of facts we call criteria for a pain being in a certain place’ (BBB 49).

Ordinarily, if I feel pain in one hand, I can locate it with my other hand. I feel my right hand palpating my left until the tender spot is reached, and I see the movements of my right hand. Perhaps this convergence of visual, tactile and kinaesthetic indicators encourages us to think of sensations as objects – it resembles their convergence in the case of objects in physical space. When these indicators of physical objects diverge – when I seem to see something but cannot touch it – I suspect an illusion.

No such failure of convergence may seem imaginable for sensation. However, the convergence is an empirical regularity, not a metaphysical necessity. ‘The man whose foot has been amputated will describe a particular pain as pain in his foot’ (BBB 52; cf. BT 514), but this location in pain space has no correlate in tactual or kinaesthetic space. Probing with my finger will not isolate the pain in my severed foot, and this location of the pain corresponds to nothing in visual space or physical space. Stranger cases are possible:

This would amount to feeling ‘toothache in another person’s tooth’ (BBB 53). The grammar of sensation-talk is thus unlike the grammar of object-talk, despite the ‘apparent analogy’ (BBB 49) between them, and recognising the disanalogies between physical space and the spaces of sensation (tactile, kinaesthetic, pain-) dulls the temptation to think of sensations as objects.

A related examination (BT 514) can be made of the misleading analogy between physical space and visual space. Visual space in some ways resembles space proper. Things in my visual field appear to the right or left, above or below, before or behind others. However, pushed further, the analogy collapses. I do not exist in visual space (PI §399), so nothing is behind me (BT 461) or at any determinate distance from me (BT 439) in visual space. I can stand across from a table but not ‘across from my optical after-image of the table’ (BT 439). In physical space, the relation of being the same length is transitive: if marks A and B are the same length, and marks B and C are the same length, then A and C are also the same length, and I can measure two things against a common standard. In visual space, marks A and B may appear the same length, as may marks B and C, but it does not follow that A and C will appear the same length. How they appear to me is just how they are in visual space, so they are not the same length in visual space (BT 446–447, 453–454; PR §215), and there can be no measurement of them (PR §212).Footnote ⁷

Overlooking these aspects of the ‘grammar’ of visual space, I may think that perceiving material objects consists in immediately apprehending phenomenal objects – sense-data. I may reason, ‘if something seems to be red, then something must have been red’ (BT 489) – an inference Wittgenstein later describes as the ‘objectification’ of appearances, in which ‘[w]e assimilate the grammar of appearance to the grammar of physical objects’ (RSD 312). About those phenomenal objects I will want to draw inferences like those I draw about physical objects: that they can be divided (BT 449–452; PR §139), clearly distinguished from each other, re-identified later – that I can give them names. Similarly, if I objectify my sensations, it will be easy to suppose that I ‘can point to the pain, as it were unseen by the other person, and name it’ (LPE 206).

Wittgenstein emphasises the alleged ontological privacy of sensation and perception, their ‘privacy of ownership’, but he sees this as entailing their ‘epistemic privacy’ (Hacker Reference Hacker1972, 222).Footnote ⁸ ‘ … I alone can know that I am feeling pain, etc.’ (PG 129/BT 95 r) if these happenings are private. So a successful critique of the ontological privacy of sensation and perception removes the strongest reason for accepting their epistemic privacy.

These complaints with the privacy of sensation and perception predate Wittgenstein’s explicit complaints with private language, but they quickly follow his rejection of the idea of a ‘phenomenological’ or ‘primary language’ (PR §1), which he entertained in 1929. Such a language would capture immediate experience, in contrast to a secondary or ‘physical language’ (PR §68), which would talk about the world. The idea of a phenomenological language was to answer Frank Ramsey’s criticisms of the Tractatus. Wittgenstein had contended that all sentences of natural language could be analysed into logically independent, elementary propositions about simple objects. Ramsey observed (Reference Ramsey1923, 473–474) that the logical independence of such propositions was incompatible with Wittgenstein’s contention that the only necessary statements were tautologies (TLP 6.37). That it is impossible ‘[f]or two colours, e.g., to be at one place in the visual field’ (TLP 6.3751) does not look like a tautology, nor is it obvious how to transform it into one.

In response, Wittgenstein tried to redefine ‘tautology’ and imagined a language whose elementary propositions would assign Cartesian coordinates to patches of colour in the visual field (RLF, 31–32), but he soon worried that doing so amounted to treating these patches as phenomenal objects (MS 105, 11/13; 106, 153) – sense-data. This, he decided, was ‘nonsense’ (MS 105, 29). He makes no explicit suggestion that such objects would be private, but they supposedly belong to immediate experience, and Wittgenstein’s attacks on sense-data begin (MS 107, 171) days before he renounces the project of constructing a phenomenological language (MS 107, 176).

Additionally, his critique of private objects and phenomenal space suggests a complaint against private language. If we think of sensations and perceptual appearances as objects, it will seem that we can name them and speak of them, as we do public objects, even if they are supposed private. However, we thereby underestimate the resources needed to name even a public object – resources unavailable in the private case. The Tractatus had taken the meaning of a name to be its bearer (TLP 3.203), but by 1931 Wittgenstein concluded that pointing at an object could not by itself determine the meaning of a name:

Pointing requires disambiguation if a name is to be defined. This can be done in language by specifying whether we are defining the name of an object, a colour, a shape, a location, a number, and so on. For someone learning the rudiments of a language (BT 31 v), however, all these concepts must themselves first be learned before ostensive definition is possible.

What holds for ordinary public language we should expect for a private language, too. Thus, if we try to imagine a child-genius who ‘invents a name for the pain himself even though he wasn’t taught one,’ we forget ‘that all sorts of things in the language have to have been prepared in advance … for the mere act of naming to make sense’ (BT 209 v).

This argument from 1934 reappears at PI §257 (see Section 11), and appreciating it is essential to understanding Wittgenstein’s critique of private language in the Investigations. Wittgenstein’s doubts about the ‘object and name’ (PI §293) model of sensation-vocabulary are of central importance to his arguments in the Investigations, and those doubts are rooted in his criticisms of the metaphor of phenomenal space, which encourages the objectification of sensation and perceptual experience. These assumptions guide my discussion.

4 Private Language

Meaning in a private language – if the idea can be made intelligible – inherits its privacy from the alleged privacy of sensations and perceptual appearances. ‘Private language’ is thus not a general category for Wittgenstein, of which sensation-language (conceived a certain way) is an instance. These points often eluded early readers.

From about 1954 to 1972 the ‘First Wave’ of interpreters of the Philosophical Investigations, who faced a daunting task, were often puzzled about private language. Was it the language of a solitary speaker isolated from birth (Ayer Reference Ayer1954, 70)?Footnote ⁹ –A code that no one ever deciphers? –A language that describes or names one’s sensations (Strawson Reference Strawson1954, 84)?Footnote ¹⁰ –A sensation-language the learning of which depends on knowing no other language (Todd Reference Todd1962, 206)?

Confusion surrounded the term ‘language’. Might Wittgenstein’s view entail the ‘absurd demand’ that a private language ‘have no regularities whatsoever, either of usages or constructions’ (Hervey Reference Hervey1957, 71)? Surely, to assume such a lack of rules of use and syntax would be to give ‘an unfair reductio ad absurdum’ (Castañeda Reference Castañeda and Rollins1962, 96)!Footnote ¹¹

However, this preoccupation with syntax and vocabulary,Footnote ¹² because it assumes that sensations and appearances are private objects to which we can refer unproblematically, misses the fundamental lesson that the very idea of a private language remains unclear. Wittgenstein wonders whether a clear story can be told about even the rudiments of a private language.

Early puzzlement about language was encouraged by puzzlement about privacy. Thus, P. F. Strawson contends that a private vocabulary might equally ‘stand … for things like colours or material objects or animals’ not just for ‘the sensations of the user of the language’ (Reference Strawson1954, 84).Footnote ¹³ A. J. Ayer correctly distinguishes a language that is ‘necessarily private’ because ‘used by some particular person to refer only to his private experiences’ (Reference Strawson1954, 64) from a contingently private language, ‘intelligible only to a single person, or to a restricted set of people’ (63). However, he then argues that a ‘Robinson Crusoe’, raised by wolves, could invent a necessarily private language because he could invent a solitary language to name and describe things on his island (70).

Wittgenstein himself distinguishes contingent privacy from necessary privacy:

Asking myself a question or keeping a coded diary involves ‘contingent’ or ‘ordinary’ privacy, whether because others do not hear my question or because they happen not to understand my code. A language that others cannot understand because its terms refer to things of which they can have no experience or knowledge involves ‘necessary’ or ‘logical’ privacy. In his ‘Notes for the “Philosophical Lecture”’ (c.1942), Wittgenstein calls it ‘superprivacy’ (NPL, 447), and this is the clear target of his criticisms.

These criticisms readily allow the occasional, ordinary, epistemic privacy of our sensations:Footnote ¹⁴ I can sometimes conceal my sensations from you. Moreover, my sensations are the ones that I feel, the ones to which I give expression. Such facts, Wittgenstein suggests, belong to the grammar of sensation-vocabulary: ‘The sentence “Sensations are private” is comparable to “One plays patience by oneself”’ (PI §248). In learning to apply sensation-terms to ourselves and to others we learn that the possibility of concealing our sensations is as central to the concept of sensation as the idea of solitary activity is to the game of patience.Footnote ¹⁵

Without the distinction drawn at PI §243, one might see no problem in speaking intelligibly of private experience (see, e.g., Maddell Reference Madell2018, 53–54). PI §248 might seem to suggest that sensations are superprivate, or that Wittgenstein thinks them so (e.g., Wellman Reference Wellman1962, 446). From this initial mistake others follow. Perhaps Wittgenstein allows that we have superprivate sensations but denies that we can name them (Strawson Reference Strawson1954, 86), or express them (Hadot Reference Hadot2010 [1959], 77–78) or say that we have them (Pitcher Reference Pitcher1964, 298–299; Mundle Reference Mundle1966, 35; Gram Reference Gram and Klemke1971, 303) because, although sensations are superprivate, language cannot be. Perhaps Wittgenstein conflates different senses of ‘privacy’ (Mundle Reference Mundle1966, 37) or holds that others might have no sensations at all (44)!

Wittgenstein’s critique of private language, however, denies neither that we have sensations nor that we can speak of them to others. His target, as in late 1929, is the idea that talk of sensations is usefully modelled on talk of spatio-temporal objects.

5 Solitary Speakers

Wittgenstein’s remarks about philosophy and philosophical problems (PI §§89–133) and his anti-essentialism about language-games (PI §§65–88) suggest that we should expect no lessons about the nature of language in general from his treatment of private language. Indeed, we should expect that treatment, grounded in the considerations adduced in Section 3, to criticise the objectification of sensation and perceptual appearances and its consequences for thinking about our vocabulary of sensation and perception.

Nonetheless, some commentators take Wittgenstein to hold ‘that a private language contradicts general principles for the meaningful use of linguistic expressions’ (Tugendhat Reference Tugendhat1986, 84) – especially the general principle that language is essentially shared. This idea was first advanced by Strawson (Reference Strawson1954, 84), but Ayer’s exchange with Rush Rhees shaped the ensuing debate.

Ayer acknowledges that a private language, for Wittgenstein, is one whose speaker ‘would have no meaning to communicate even to himself’ (Ayer Reference Ayer1954, 65), but he criticises this implied conclusion by imagining ‘a Robinson Crusoe left alone while still an infant, having not yet learned to speak’ (70). Young Robinson survives to invent a language to describe ‘the flora and fauna of his island’ (70) – empirically improbable, but not logically impossible, says Ayer. But if Crusoe can manage this, why not a language for his own sensations? Wittgenstein’s argument, thinks Ayer, requires justification for one’s application of a word, and that requires ‘some independent test for determining that the sign is being used correctly’ (Ayer Reference Ayer1954, 67). Such a test is supposedly unavailable in a private sensation-language, thinks Ayer, because if I cannot recognise my sensation, then there is no reason to think that I could recognise another inner episode – for example, a memory of the sensation – that could justify it. The problem, Ayer complains, is that ‘unless there is something that one is allowed to recognize, no test can ever be completed: there will be no justification for the use of any sign at all’ (68; cf. Ayer Reference Ayer1973, 95; Reference Ayer1986, 76). If the argument works, then public language is impossible. The argument thus makes a faulty presupposition – that no memory of a past sensation can justify my applying ‘S’ to a recurrence of this sensation. This is just scepticism about memory.

Ayer was ‘on the right lines’ (Hervey Reference Hervey1957, 71) according to some, and the impression that Wittgenstein doubted the possibility of a solitary speaker was reinforced by his early defenders, Rhees (Reference Rhees1954, 88) and Winch (Reference Winch1958, 33–34).Footnote ¹⁶ Even when Crusoe goes unmentioned, his airy presence persists in the conviction that Wittgenstein has tried (perhaps, failed) to show that language is essentially shared.Footnote ¹⁷ Many later commentators endorse some version of this ‘community view’, motivated by something like Claudine Verheggen’s complaint that if Wittgenstein is not essentially embedding meaning in communal practices, then he is not saying ‘anything illuminating about the nature of language’ (Reference Verheggen2007, 615).Footnote ¹⁸

Another tradition noted the ‘difference in logical character between’ (Hardin Reference Hardin1959, 521) an isolated speaker’s words and a private language as intended by Wittgenstein.Footnote ¹⁹ On this view, for Wittgenstein, an argument against private language requires only that language be in principle shareable, not necessarily shared.Footnote ²⁰

Wittgenstein clearly took language to be paradigmatically shared,Footnote ²¹ but finding evidence that he took language to be necessarily shared is more difficult.Footnote ²² Comparing linguistic practices to games is meant to discourage the thought that there is some set of necessary and sufficient conditions for applying the term ‘language’ (PI §65), and the metaphilosophical remarks at PI §§89–133 discourage seeking an argument with such a conclusion (Schroeder Reference Schroeder2006, 200). As John Canfield contends, for Wittgenstein, wolf-child Crusoe would be ‘a borderline case of language use’ (Reference Canfield1996, 485). If we want to describe the concept of language, we should not start with such improbable cases, but Wittgenstein’s reluctance to circumscribe the concept sharply makes it unlikely that he means to rule them out. It is not Wittgenstein’s task in the Investigations, pace Verheggen, to say ‘anything illuminating about the nature of language’ (Verheggen Reference Verheggen2007, 615).

But what of this passage?

This remark, quoted by Kripke (Reference Kripke1982, 3) and others, resonates with a line from the diary-example (see Sections 8, 9, and 11): ‘One would like to say: whatever is going to seem correct to me is correct. And that only means that here we can’t talk about “correct”’ (PI §258). Supporters of the community view think that ‘privately’ inherits its sense here from the description of rules as ‘customs’ and ‘institutions’ (PI §199) and ‘practice[s]’ (PI §202).

These terms sound plausibly communal in English, and Martin Kusch (Reference Kusch2006, 248–252) argues that this reflects the predominant usage of the original German terms. Native anglophones may feel abashed before such claims, but some prominent germanophone interpreters are unmoved.Footnote ²³ Furthermore, claims about standard use do not settle how Wittgenstein uses these terms. As Baker and Hacker argue, there are numerous passages in Wittgenstein’s writings in which such words as ‘custom’ (‘Gepflogenheit’), ‘institution’ (‘Institution’) and ‘practice’ (‘Praxis’) occur with no evident connotations of community (Reference Baker and Hacker2014 [Reference Baker and Hacker1985], 121–122, 140–143). The fact that ‘privately’ appears in scare-quotes at PI §202 emphasises that it is not obvious how to interpret it.Footnote ²⁴ The remark that rule-following is a practice that cannot be engaged in ‘privately’ originally belonged to an argument for thinking that I need not recognise my image of red before I can judge that it is red (Ms 129, 119) – a discussion that survives as PI §§377–381 (Baker and Hacker Reference Baker and Hacker2014 [Reference Baker and Hacker1985], 126–129). To interpose a moment of recognition is to treat my knowledge of my sensations and perceptual experiences as a datable achievement. But if my knowledge of my own sensations and appearances need not be such an achievement (Section 6.2) – if it could be an aspect of my linguistic competence (PI §381) – then the connection between rule-following and the critique of private language is more like this: ‘there is a way of grasping a rule’ for the use of a sensation-term ‘which is not an interpretation’ (PI §201) – not, that is, a matter of recognising my sensation and inferring that it should be categorised as pain or as the appearance of redness, as the objectification of sensations suggests.

Many key arguments of PI §§243–315 were developed earlier than, and independently of, the discussion of rule-following. The critiques of the misleading metaphor of phenomenal space, of the privacy and ownership of sense-data, of the objectification of sensations and appearances, of privileged access, all date from 1929 to 1936. The example of the child-genius at PI §257 first appears in The Big Typescript (BT, 209 v). The diary-example itself (see Sections 8, 9, and 11) originates in ‘Notes for Lectures on “Private Experience” and “Sense-Data”’, from 1934 to 1936, where the idea of keeping a diary of my private experiences appears in the context of wondering whether I could, in a private language, track recurrences of a colour-sensation, for example, of red (LPE, 234).Footnote ²⁵ The discussion of rule-following, by contrast, largely derives from work begun around 1935 (see BBB 141–143) and mostly carried out from 1937 onward.Footnote ²⁶ Had we independent grounds for thinking that Wittgenstein treats rules as essentially communal, we might think that the overall significance of the passages on private language had been transformed by the time of the Investigations, but even this would not alter their basis in the considerations reviewed in Section 3.

I assume, then, that the objectification of sensation and perceptual appearances, supported by the misleading metaphor of phenomenal space, and the consequences of this objectification for our thinking about sensation-language constitute the target of Wittgenstein’s criticism of privacy and private language. I return briefly to Kripke (Section 10), but, otherwise, I set aside concerns about the role of community in Wittgenstein’s treatment of private language.

6 Ontological Privacy, Epistemic Privacy and First-Person Authority

The distinction between ‘ordinary’ privacy and ‘superprivacy’ (Section 4) bisects the distinction (Section 3) between ‘epistemic privacy’ and the ‘privacy of ownership’ (Hacker Reference Hacker1972, 222). Each of the latter senses of ‘private’ can be interpreted in an ‘ordinary’ way or in a ‘superprivate’ way. My sensations are epistemically private in the ordinary sense when I conceal them from others, and they belong to me in the ordinary sense that ‘the person who is suffering is the person who manifests pain’ (PI §302) (Fogelin Reference Fogelin1987, 170). That is what its being my pain amounts to. But the impulse to think of sensations as superprivate is expressed epistemically by the thought that ‘only I can know whether I am really in pain; another person can only surmise it’ (PI §246), and ontologically by the thought that ‘Another person can’t have my pains’ (PI §253). Consider the latter first.

7 Avowals

If sensations are objects, then they should get their names as objects do, albeit in private, phenomenal space. PI §244 tries to distract our gaze from this picture:

Such first-person, expressive uses of psychological vocabulary are known as ‘avowals’.Footnote ⁴² An avowal of my pain is a spontaneous, articulate expression of it – for example, ‘That hurts!’ or ‘I am in pain!’ Wittgenstein describes such uses of psychological vocabulary as ‘wrung from us – like a cry’ (PI §546). They are not assertions or descriptions in the sense of being based on evidence or observation of my sensations. ‘It is not, of course, that I identify my sensation by means of criteria; it is, rather, that I use the same expression’ (PI §290).

If we assume that only assertions – or expressions of belief – are truth-apt, then it may seem that avowals can be neither true nor false. However, we need not make this assumption.Footnote ⁴³ The antecedents of conditional statements are truth-apt, but, being conditional, they assert nothing and express no belief. To say of the antecedent of a conditional that it can be true or false is to say, roughly, that another token-sentence of the same type can be used to make an assertion. So if ‘I am in pain’ can be uttered as an assertion, then its occurrence as an avowal is compatible with its being truth-apt.Footnote ⁴⁴

But does PI §244 leave room for first-person, present-tense, indicative uses of verbs of experience that are not avowals? Some have thought not,Footnote ⁴⁵ but Wittgenstein does not ‘deny that first-person sentences about sensations may … be more or less like natural expressions of sensation’ (Malcolm Reference Malcolm1954, 542). They are less like natural expressions of sensation when they are more like descriptions of sensation.Footnote ⁴⁶ When I learn to supplement spontaneous natural expressions of pain with verbal expressions like ‘I am in pain’, ‘it is not as if the language-game ends with this: it begins with it’ (PI §290). In his 1936 lectures Wittgenstein says that ‘“toothache” is not only a substitute for moaning. But it is also a substitute for moaning … ’ (RSD, 298). He later extends the point to fear:Footnote ⁴⁷

A complaint in this case is an avowal, an expression of my pain, but not every utterance of ‘I am in pain’ is an avowal. This, I suggest, is the point of Wittgenstein’s remark that his description of first-person, present-tense psychological ascriptions as expressive is ‘Not quite right’ (RPP II §63).

Other readers have suggested that Wittgenstein is committed to an expressivist theory of sensation-talk.Footnote ⁴⁹ Such a theory might look roughly like this: pain and some other sensations (nausea, itching, etc.) have characteristic natural expressions.Footnote ⁵⁰ Learning terms for these sensations, I learn to replace or supplement their natural expressions with articulate first-person, present-tense verbal expressions that make use of these terms. My expressive uses of these sensation-terms involve no intermediate step of identifying or recognising my sensations, any more than my spontaneous non-verbal behaviour does. (We could say that my sensation is identified in the act of my expressing it.Footnote ⁵¹) However, I learn to treat these natural expressions in others as grounds for ascribing the same sensations to them, and their avowals of pain also serve me as grounds for such ascriptions. What I ascribe to myself without grounds I ascribe to others on behavioural grounds, verbal or non-verbal. Because my verbal expressions of sensation are made with declarative sentences, they are truth-apt, and they can play a role in inferences. From my expressive utterances of ‘I am in pain!’ you can infer ‘He is in pain.’Footnote ⁵² Having learned how to express (and describe) my pain, I can then apply these lessons to other sensations that lack natural expressions.Footnote ⁵³

We find hints of such a view in the early 1930s (DS, 51).Footnote ⁵⁴ Wittgenstein’s lecture notes from 1934 to 1936 are more explicit (LPE 254, 261, 262, 281).Footnote ⁵⁵ Further support appears at PI §§256 and 288, and other remarks from the 1940s reinforce the impression:

Against attributing an expressivist theory to Wittgenstein are his warnings against ‘advanc[ing] any kind of theory’ (PI §109) in philosophy, where our task is to clarify the grammar of expressions that lead us into confusion (PI §§122–133). When he introduces the idea that a child learns to replace natural expressions of sensation with a verbal expression, he simply calls it ‘one possibility’ (PI §244).Footnote ⁵⁶

However, the temptation to elaborate on this suggestion, as earlier, is compatible with thinking that Wittgenstein was not advancing a general theory. Part of the task of dispelling philosophical confusion is to clarify the grammar of the expressions that puzzle us. Wittgenstein’s remarks on first-person, present-tense psychological ascriptions aim at such clarity (see Hacker Reference Hacker2019a [Reference Hacker1990], 117–121). Moreover, to claim that the suggestion of PI §244 is a possibility is still to assume a certain burden of argument. One might try to discharge that burden by saying that the proposal is at least as plausible as the view that the meanings of sensation-terms are fixed by inner ostensive definitionFootnote ⁵⁷ (see Sections 9 and 11). Or one might try to support the claim of possibility by exploring the proposal in greater detail. What matters for Wittgenstein’s purposes is that the view be sufficiently plausible to suggest that the inner-ostension account is not obligatory.

8 The First Wave: Verification and Memory

The first commentator to mention ‘the private-language argument’ in print was Héctor-Neri Castañeda (Reference Castañeda and Rollins1962),Footnote ⁵⁸ but I assume it was already common to speak of Wittgenstein’s discussion this way. To do so is typically to assume that there is a central argument – usually sought in the diary-example of PI §258, aided by some other passages – and that this argument aims to demonstrate the logical impossibility of a private language.Footnote ⁵⁹ I reject both these assumptions, but it is important to understand their pervasive influence on the reception of Wittgenstein’s work.

Castañeda attributes the argument also to Norman Malcolm, whom he regards as a reliable expositor of Wittgenstein’s views. Malcolm is certainly among the most sensitive and astute of First-Wave readers of the Investigations, and his influential interpretation fuels the idea that Wittgenstein’s critique is a sustained argument against the logical possibility of such a language – a reductio ad absurdum (Reference Malcolm1954, 537) of the idea that there might be a language that ‘cannot be understood by anyone other than the speaker’ (530–531). The significance of this reasoning, thinks Malcolm, is far-reaching, challenging ‘the philosophy of Descartes and … the theory of ideas of classical British empiricism as well as … recent and contemporary phenomenalism and sense-datum theory’ (Reference Malcolm1954, 531).

Malcolm’s focus on the diary-example of PI §258 (supplemented by PI §§259, 265) has dominated the literature:

I take the diary-example to make the semantic point that no rule for the use of ‘S’ has been established (see Section 11). However, Malcolm helps plant the seeds of two popular misunderstandings of this passage: that it aims to show the impossibility of (1) verifying (2) the private speaker’s memory of how to apply a term in a private language (Reference Malcolm1954, 532). On Malcolm’s epistemic reading, I try to define the word ‘pain’ privately. However, my private definition succeeds only if it ensures that my future uses of ‘pain’ are consistent with my attempted definition (532). Wittgenstein’s objection, thinks Malcolm, is that I cannot ‘prove’ or ‘confirm’ (532) such consistency in the circumstances imagined. There would be no discernible ‘difference between my having used [the word] consistently and its seeming to me that I have’ (532), so the concept of correctness would not apply. I would have only ‘impressions of rules’ (PI §259) governing my use of the word.

Why can nothing prove the consistency of my later use of ‘pain’ with my definition? I can rely only on my memory, and here my memory cannot be tested for correctness by comparing it to something independent (Malcolm Reference Malcolm1954, 533–534).Footnote ⁶⁰ Malcolm calls on PI §265 for support:

I cannot confirm the correctness of my memory-impression, because the sensation is long-gone, and I cannot appeal to anyone else’s memory to confirm mine. Without this possibility of confirmation, says Malcolm, ‘there would not be, in the private language, any conception of what would establish a memory as correct’ (Reference Malcolm1954, 534). Talk of correctness would be meaningless.

Malcolm did not intend to saddle Wittgenstein with a verifiability theory of meaning, but, as Thomson (Reference Thomson1964, 29–31) complained, it is difficult to distinguish talk of ‘checking one memory against another’ (Malcolm Reference Malcolm1954, 533) from saying that my memories must be verifiable for there to be any fact about how I should employ my sensation-terms. As we saw (Section 5), that demand, which Ayer thinks is the conclusion of the argument, reduces to a scepticism about memory that would apply equally to public language.

These epistemic themes – verificationism and memory-scepticism – set the agenda for commentators for two decades. Many critics followed Ayer, arguing that Wittgenstein’s objections to private language ‘arise from the rejection of memory rather than from the privacy of experience’ (Wellman Reference Wellman1959, 225),Footnote ⁶² or complaining that ‘Wittgenstein dogmatically presupposes the verifiability theory of meaning even though one of his main theses is its inadequacy’ (230). Others, following Malcolm’s exposition, criticised Wittgenstein for confusing a speaker’s ‘inability to verify his recollection of the meaning of [sensation-term] “E” with inability to understand what it would be for his recollection to be right’ (Donagan Reference Donagan and Pitcher1966, 339),Footnote ⁶³ and some reproduced Ayer’s conclusion that ‘Wittgenstein’s private language argument [is] open to a reductio’ (Stocker Reference Stocker1966, 47) because memories of public events are no more ‘checkable’ (50) than memories of private ones.Footnote ⁶⁴

Sympathetic commentators inadvertently bolstered the verificationist interpretation. According to Ambrose, in a private sensation-language ‘there is no check on one’s apparent memory that this is the same as what one had before’ (Reference Ambrose1954, 115).Footnote ⁶⁵ Geach, denying that Wittgenstein is a verificationist, contends that he rejected the possibility of giving sensation-terms ‘a private sense … by just attending to one’s own pain-experiences, a performance that would be private and uncheckable’ (Reference Geach1957, 3–4; my emphasis).Footnote ⁶⁶ Other sensitive readers emphasise the importance of ‘criteria for distinguishing the correct from the incorrect use of language’ (Linsky Reference Linsky1957, 287; my emphasis), without which ‘the undertaking [to use a sensation-term in the same way] would be empty because you could never know whether you had fulfilled it or not’ (Garver Reference Garver1960, 394; my emphasis).Footnote ⁶⁷

9 The Second Wave: Ostensive Definition

Hints of the verificationist interpretation remain long after the First Wave subsides,Footnote ⁶⁸ and with them two philosophical problems: (1) the verifiability theory of meaning seems mistaken; (2) the considerations about memory overflow the banks of private language and flood the vast plains of public language.

The interpretative problem is that the diary-example makes a semantic argument, not an epistemic one. A clear articulation of this point was needed to exonerate Wittgenstein of the philosophical errors. Some criticisms of the verificationist interpretation appeared in the 1970s,Footnote ⁶⁹ but the philosophical world had moved on to other controversies, and philosophers at large took little notice of the ‘Second Wave’ of interpreters. Generations of philosophersFootnote ⁷⁰ remain familiar only with verificationist, often behaviourist, interpretations of Wittgenstein and with variations on Kripke’s reading of the Investigations (the crest of the Third Wave).

The charge of memory-scepticism, however, generated important early responses, which suggest a semantic alternative to the verificationist reading. Rhees replied to Ayer that the discussion of private language concerned ‘not a question of whether I can trust my memory,’ but ‘a question of when it makes sense to speak of remembering; either of a good memory or a faulty one’ (Reference Rhees1954, 83). Wittgenstein was not arguing that, without a reliable memory, there would be no criterion for correctly using a private sensation-term but that, without a criterion for correct use, there would be no such thing as remembering.Footnote ⁷¹ This insight was best formulated by Les Holborow, who explained that to say that the private speaker has ‘no criterion of correctness’, is ‘not to cast doubt on the accuracy or reliability of the diarist’s memory. Wittgenstein is claiming not that he might misremember the criterion that he has, just for this instant, fixed; but that he has not by this ceremony fixed any criterion at all’ (Reference Holborow1967, 347).

The point is pursued by Peter Hacker (Reference Hacker1972) and by Anthony Kenny (Reference Kenny1973).Footnote ⁷² Like First-Wave interpreters, both think there is a central argument aimed at demonstrating that a private language is impossible (Hacker Reference Hacker1972, 215; Kenny Reference Kenny1973, 178), and, like Malcolm, both think that this aim matters because ‘several traditional and influential theories’ (Kenny Reference Kenny1973, 179) in epistemology and the philosophy of mind entail the possibility of such a language (see Kenny Reference Kenny and Pitcher1966; Hacker Reference Hacker1972, 216–217).

However, Kenny and Hacker see clearly that a private language is one whose ‘words have acquired their meaning for each of us by an essentially private process: an internal ostensive definition … ’ (Kenny Reference Kenny1973, 179). Its privacy is thus unrelated to the possibility of a solitary speaker (Hacker Reference Hacker1972, 222). Moreover, Kenny insists that the key passages do not present a verificationist argument (Reference Kenny1973, 195) and do not concern the reliability of memory (191–192; see Hacker Reference Hacker1972, 236–237).

Belief in the possibility of a private language, says Kenny, rests on the mistaken beliefs that experience is (super)private and that meaning can be given solely by ostensive definition (Reference Kenny1973, 180; cf. Hacker Reference Hacker1972, 215–216). The second mistake is criticised by Wittgenstein’s discussion of naming, ostensive definition, and ostensive teaching (PI §§27–36), and the upshot, says Kenny, is that ‘bare ostension without training in the use of words could [not] constitute the teaching of a language’ (Reference Kenny1973, 180; see Hacker Reference Hacker1972, 235).

If this result is established early in the Investigations, it may seem that the impossibility of a private language is included as a special case. However, Kenny suggests, an aspiring private speaker might argue that, even if such considerations undermined the possibility of a language whose ‘words were learnt from private sensations by bare ostension’ (Reference Kenny1973, 180), they would not thereby rule out the possibility of a language whose ‘words referred to private sensations’ (180), because such a language might be learned ‘by some private analogue of training in the use of words’ (180–181).Footnote ⁷³

I find no such analogue mentioned in Wittgenstein’s text.Footnote ⁷⁴ However, Kenny thinks that his reading ensures that the discussion of private ostension is not superfluous (Reference Kenny1973, 181). Like Malcolm, Kenny regards PI §258 as the ‘kernel’ (190) of the argument, but he thinks that Malcolm and others misunderstand the role of memory:

Saying that ‘I have no criterion of correctness’ (PI §258) may suggest that some test of the reliability of my memory is unavailable to me. But, really, as Holborow (Reference Holborow1967, 147) argued, to lack a criterion of correctness is for there to be nothing to remember because there is no fact about what I mean by ‘S’ (cf. Hacker Reference Hacker1972, 234–235).Footnote ⁷⁵

Kenny appeals to the railway-timetable example of PI §265, but, again, his reading differs from Malcolm’s. In the timetable example I check my memory of my train’s departure-time by recalling a memory-image of a page from the timetable, and the question is why the private speaker could not do something comparable – recall a memory of a correlation between ‘S’ and S to confirm that ‘S’ is correctly applied. The table that we are to imagine matches terms in a would-be private language with memories of sensations.

The private speaker responds that, if I wonder whether I have remembered the application of ‘S’ correctly, then I can check by calling the table to mind. That will confirm whether I am using ‘S’ as I did earlier. But when I try to test my memory of the sensation by comparing it with my memory-image of the translation-table, I must know with which line in the table to compare my memory. ‘[T]his procedure must … actually call forth the correct memory’ (PI §265).Footnote ⁷⁶ In the case of the railway timetable, which line in the table I should take to confirm or disconfirm my memory is determined by the scheduled departure-time of my train. If I travel from Halifax to Quebec City on a Wednesday, I call a memory-image of the table to mind, see where Halifax is listed and see what departure-time is given. This is the right memory-image to consult if it accurately reproduces the published timetable.Footnote ⁷⁷ But which line of my translation-table of sensation-terms should I use? My finding that line presupposes that I know which sensation to pair with which term, but if I knew that, I would not need the table. ‘[T]here can be no real looking up to see which sample goes with “S”. All there can be is remembering which sample goes with “S”, i.e. remembering what “S” means. But this is precisely what the table was supposed to confirm’ (Kenny Reference Kenny1973, 192–193; cf. Hacker Reference Hacker1972, 236).

Kenny and Hacker correctly shift the problem of private language from the epistemic dimension (could a private diarist correctly remember recurring sensations?) into the semantic (has the private diarist given a sensation a name and established rules for its use?). However, their continued emphasis of PI §§258 and 265 muffles the importance of Wittgenstein’s discussion of ostensive definition. Nearly a decade would pass before the importance of this shift was appreciated, and the influence of that appreciation was diluted by the storm surge of the Third Wave.

10 The Third Wave: Rules

Saul Kripke’s Wittgenstein on Rules and Private Language (Reference Kripke1982) captured the imaginations of philosophers who had lost interest in the debates of the First Wave, if they had ever followed them, and who had ignored the Second Wave. I have already criticised the ‘community view’ (Section 5) of which Kripke’s view is a species, and his view focuses on rule-following, not sensation-language, but it requires further consideration because Kripke maintains that ‘the real “private language argument”’ (Reference Kripke1982, 3) occurs not at PI §§243–315 but in the discussion of rules.

Kripke interprets Wittgenstein as a sceptic about the possibility of following a rule, surveying one proposal after another concerning what ‘fact about me’ (Reference Kripke1982, 21) would determine that I was following, for example, the rule for addition, rather than some eccentric rule whose observance would supply the same results over a range of cases (e.g., involving small numbers) but produce radically different results if applied in cases that, contingently, I never consider (e.g., sums involving extremely large numbers). Facts about me that might seem to settle the question, but ultimately fail, include my actual behaviour, my behavioural dispositions, my having the feeling of understanding when trying to apply the rule, my associating some mental image with the rule, my intending to continue the series in the same way as its earlier terms have been presented to me, and my having an interpretation of the rule. When no fact about me can be found to determine which rule I follow, Kripke’s Wittgenstein concludes that there is no such fact. Insofar as meaning something by my words requires me to follow rules, ‘It seems that the entire idea of meaning vanishes into thin air’ (22). This meaning-scepticism is entirely general, so ‘The “private language argument” as applied to sensations is only a special case of much more general considerations about language previously argued … ’ (3).

How can we accept this sceptical result about meaning? Kripke’s Wittgenstein adopts a sceptical solution: although it is neither true nor false that I follow any rule, it may be assertible that I follow a rule, provided my teachers and peers are satisfied with my results and provided assertions about that rule have a role to play in our practices (Kripke Reference Kripke1982, 77–78). That description of assertibility-conditions rules out private language: without a community of ostensible rule-followers, I cannot be asserted to follow a rule because it is assertible of me that I do, only if my results are taken to agree with those of others. ‘The impossibility of a private language emerges as a corollary of [Wittgenstein’s] sceptical solution to his own paradox … ’ (68). The remarks ordinarily taken to provide the critique of private language, says Kripke, ‘deal with the application of the general conclusions about language drawn in §§138–242 to the problem of sensations’ (79). That explains, he says, why Wittgenstein–in the thick of his discussion of rules–remarks that ‘it’s not possible to follow a rule “privately”’ (PI §202).

Not just private sensation-terms are ruled out, but any private rule-following. This, in one sense, includes Ayer’s Crusoe and, in another, does not: ‘The falsity of the private model need not mean that a physically isolated individual cannot be said to follow rules; rather that an individual, considered in isolation (whether or not he is physically isolated), cannot be said to do so’ (Kripke Reference Kripke1982, 110).

Kripke neither espouses the scepticism or its sceptical solution, nor univocally attributes them to Wittgenstein (Kripke Reference Kripke1982, 5), but many have since done the latter. I cannot make a proper case against reading Wittgenstein this way here, so I limit myself to a few observations.Footnote ⁷⁸ First, Wittgenstein’s discussion of rule-following is better read as casting doubt on reductive accounts of rule-following or meaning. On this reading, it is theories of meaning that lead to Kripke’s sceptical results, and since we evidently do follow rules and make meaningful utterances, meaning and rule-following are no more suitable objects of deep explanatory theories than are cups and saucers. ‘The meaning of a word is what an explanation of its meaning explains’ (PI §560) – a clarification in context, not a general theory.

Kripke thinks that the discussion of rule-following culminates in scepticism about the possibility of meaning and rule-following:

But, notoriously, he ignores the next paragraph:

Only then do we learn that ‘“following a rule” is a practice’ (PI §202) and, therefore, not private. The sceptical conclusion is taken to rest on a misunderstanding, and, although something must be said about this anticipation of the theme of privacy (see Section 5), this does not entail that the real private language argument has already been made.

Kripke formulated his view as early as 1962–63,Footnote ⁷⁹ but some of his key points are made by other commentators.Footnote ⁸⁰ Most importantly, Robert Fogelin presents Wittgenstein as offering a ‘sceptical solution’ to ‘sceptical doubts’ (Reference Fogelin1976, 143; Reference Fogelin1987, 161) about the possibility of following a rule, and this sceptical solution undergirds what Fogelin later calls the ‘training argument’, which he thinks ‘establishes the contingent impossibility of a private language’ (Reference Fogelin1987, 175). Like Kripke, Fogelin began formulating his views in the early 1960s (Reference Fogelin1987, 241n10), and, also like Kripke, Fogelin suggests that ‘… Wittgenstein’s reasons for saying that obeying a rule is a practice provide the framework for examining the possibility of a private language’ (Reference Fogelin1976, 154; Reference Fogelin1987, 167).Footnote ⁸¹ However, Fogelin finds several arguments against private language.

First, Fogelin sees Wittgenstein’s discussion of privacy and certainty at PI §§244–248 as attempting ‘to diagnose the influences that make it seem natural to hold’ (Reference Fogelin1976, 159; Reference Fogelin1987, 172) that our everyday sensation-talk might be logically private. This, however, involves no attempt to demonstrate the logical impossibility of a private language.

The second and third lines of argument appear in the diary-example and the case of the human manometer at PI §270 (see Section 12). One of these, the ‘public-check argument’ (Reference Fogelin1987, 175), is a variation on Malcolm’s reading of PI §§258 and 265, which, we have seen (Section 8), does not rule out a private language any more than a public language. However, like Kenny and Hacker, Fogelin also notes Wittgenstein’s discussion of ostensive definition at PI §§27–36. The diary-example, he thinks, shows that the supposition that I might name a private sensation, merely by focusing my attention on it, illicitly presupposes a complex background of public language, without which it would be mysterious how the term ‘S’ could designate anything.

This additional reasoning does not refute the logical possibility of a private language, but, Fogelin thinks, it establishes two important points: ‘(i) the construction of a private language may seem unproblematic only because we illicitly help ourselves to the logical features of expressions that occur in everyday language’ (Reference Fogelin1976, 161; Reference Fogelin1987, 174); ‘(ii) … if we do give a symbol a public employment sufficient to fix its sense, then it is already up to the mark as far as significance goes, and there is no point in saying that it also has a private reference’ (Reference Fogelin1976, 161; Reference Fogelin1987, 175). The former point is central to PI §258 (see Section 11). I return to the latter point in Section 12.

A fourth argument, thinks Fogelin, shows ‘the contingent impossibility of a private language’ (Reference Fogelin1987, 175). The ‘training argument’ (Reference Fogelin1987, 175) contends that following a rule amounts to behaving as one has been trained to. Fogelin invokes the remark that Kripke assiduously avoids: ‘there is a way of grasping a rule which is not an interpretation, but which, from case to case of application, is exhibited in what we call “following the rule” and “going against it”’ (PI §201). Training constitutes a sceptical solution to the paradox about rule-following because it allows that any number of rules can fit any finite series of behaviour, while observing that we do not accept certain responses to our instructions to follow a particular rule as conforming to that rule. Someone who continues a series in response to the command to add 2, by writing, ‘… 1000, 1004, 1008, 1012 …’ will be held not to have followed the rule, even though there is no justification for saying this. Whoever learns arithmetic is trained to respond in certain ways, and ‘that is the end of the matter’ (Fogelin Reference Fogelin1987, 176).

However, it is a contingent fact about human beings that there are no ‘untrained trainer[s]’ (Reference Fogelin1976, 164; Reference Fogelin1987, 176), ‘no linguistic self-starters’ (Reference Fogelin1976, 165; Reference Fogelin1987, 178). So I could not develop a private sensation-language because I could not be trained in the use of its terms by someone already familiar with their application, and I could not train myself without prior expertise. This does not make a private sensation-language logically impossible, but ‘very general facts of nature’ (PPF xii §365) exclude such possibilities.

Fogelin’s reading differs from Kripke’s, but it, too, attributes to Wittgenstein a scepticism about the determinacy of rule-following that admits of only a sceptical solution, wherein ‘an unjustified (indeed, unjustifiable) belief is grounded in nothing more than a brute fact of human nature’ (Reference Fogelin1987, 175). This reading, like Kripke’s, is flawed, but it retains two significant virtues: (i) it suggests that there is no such thing as the private language argument but, rather, a network of intertwined considerations about problems raised by the idea of a private language; (ii) it emphasises the great importance for the diary-example of Wittgenstein’s remarks about ‘stage-setting’ at PI §257. Both ideas have since furthered our understanding of Wittgenstein’s investigations into private language.

11 The Fourth Wave: Stage-Setting

Counting waves eventually becomes a mug’s game, but we can usefully identify one more. While many philosophers rode the Kripkean surf, other currents stirred in the depths. Many serious readers of Wittgenstein were dissatisfied with existing treatments of his later work, and this dissatisfaction yielded various re-evaluations of the passages on private language that characterise what I risk calling the Fourth Wave.Footnote ⁸² I start with Stewart Candlish – whose views are closer to Kenny and Hacker’s than he allows.

Candlish (Reference Candlish1980) criticises the ‘Old Orthodoxy’ (Fogelin’s ‘Public Check’ argument) and the ‘New Guardians’ (Kenny and Hacker) for focusing too narrowly on PI §§258 and 265, and for neglecting the roles of other passages on private language. He agrees with Kenny and Hacker that the diary-example presents a semantic challenge about how a term in a private language could be given any meaning. However, he thinks that Kenny’s reading faces as much difficulty as the Old Orthodoxy.

On Kenny’s reading, Wittgenstein’s reasoning is ‘wildly fallacious’, says Candlish (Reference Candlish1980, 94). Kenny imagines confronting the private speaker after the attempted definition of ‘S’ by private ostension and asking what ‘S’ means. The diarist could say, ‘By “S” I mean the sensation I named “S” in the past’ (Kenny Reference Kenny1973, 194), but this would require justification by a memory of the correlation of a past sensation with a past application of ‘S’. Kenny reiterates the complaint with this procedure:

Candlish reads this as a dilemma: if the wrong memory cannot be called up, then ‘S’ is meaningless because it cannot be misapplied–no rule has been given. If the wrong memory can be called up, then the private speaker ‘does not really know what he means’ (194). The latter horn, insists Candlish, unreasonably requires infallibility in subsequent applications of ‘S’, but I can understand how to use a term even if I sometimes misapply it (Candlish Reference Candlish1980, 91).Footnote ⁸³

However, when Kenny writes, ‘If so, then he does not really know what he means’, he is not completing a dilemma.Footnote ⁸⁴ He is saying, ‘If it is so–that whatever seems right is right–then he does not really know what he means.’ That is just what we should expect, if ‘attaching meaning to a name does not mean acquiring infallibility in its use’ (Kenny Reference Kenny1973, 193). So Kenny has not assumed that ‘there is actually an application of a sign to a sensation by a private-language user, and that the problem is one of later remembering this earlier application’ (Candlish Reference Candlish1980, 91). Kenny’s point is Candlish’s: there has been no successful definition of ‘S’, and no rule has been established for the application of ‘S’.

Charles Dunlop, following Candlish, complains further that Kenny and Hacker have overlooked the crucial importance of Wittgenstein’s discussion at PI §§27–36 of ostensive definition (Reference Dunlop1984, 364n4–365n4).Footnote ⁸⁵ In fact, both Kenny (Reference Kenny1973, 182–183) and Hacker (Reference Hacker1972, 235) indicate that these remarks are significant for the critique of private language. However, beginning with the Fourth Wave, they come clearly to the surface.

The main reason for emphasising the remarks on ostensive definition here lies in the passages immediately preceding the diary-example. One of the tasks of PI §§244–256 (see Section 6) is to undermine the supposition that our ordinary sensation-terms refer to superprivate objects. However, a question remains about whether such a private sensation-language might yet be possible:

How could one acquire a grasp of sensation-terms if this were the case? Here we encounter the child-genius:

This ‘stage-setting’ argument was noted by Candlish (Reference Candlish1980, 86), Hallett (Reference Hallett1977, 337–339), Fogelin (Reference Fogelin1976, 159–162; Reference Fogelin1987, 172–175), Cook (Reference Cook, Ambrose and Lazerowitz1972, 62, 65), Goldberg (Reference Goldberg1971, 88) and even earlier by Holborow (Reference Holborow1967, 350),Footnote ⁸⁷ but it gets its due only with Stroud (Reference Stroud2002 [Reference Stroud1983]) and Williams (Reference Williams1983).Footnote ⁸⁸

As we saw (Section 3), the argument appears in the Big Typescript, immediately after a discussion of ‘the great variety of language-games’ (BT 209). There the stage-setting – what has ‘been prepared in advance’ (BT 209 v) – is found in Wittgenstein’s discussion of ostensive definition at BT 31–33.Footnote ⁸⁹ In PI, immediately following an examination of ‘the variety of language-games’ (PI §23), we find a parallel discussion of ostensive definition that suggests, first, that we distinguish ostensive teaching and learning from ostensive definition.Footnote ⁹⁰ The former is important in the acquisition of language, but the latter presupposes a grasp of language. Second, ostensive teaching and learning rely extensively on features of context, including, eventually, linguistic context, to disambiguate ostension. When I point towards a book before us, the gesture alone does not determine whether I point to the book, its shape, its size, its colour, the image on its cover, the number of books present, and so on. If I share a language with you, then I can specify: ‘Is this your book?’ ‘Like most books, this is roughly a parallelepiped.’ ‘I’m colour-blind: is this green?’ ‘Who’s that on the cover?’ If I cannot rely on such disambiguations, then I might resort to comparisons and contrasts with other things to make it clear what I am pointing at – for example, by grouping the book together with other books or other green things or other parallelepipeds, and so on, but this is gradually to fall back on ostensive teaching.

The lesson for §257 is that it is unclear how the child, however ingenious, can name private sensations, which admit of no such disambiguation – disambiguation from concurrent sensations, or from their location, duration, intensity, and so on. Concentrate on your unnamed sensation. –Now on the intensity of that sensation. –Now on its location. –Now on its duration. What did you do differently each time? How could you do it without the concepts needed to distinguish a sensation from its intensity, duration or location? The challenge is to show how disambiguation can be achieved without this stage-setting.

These themes recur in the diary-example. Barry Stroud gives a clear and direct statement:

Consequently, says Stroud, ‘there is so far nothing that even an infallible memory could remember’ (75).Footnote ⁹²

Meredith Williams emphasises similar themes. Commentators have focused erroneously, she says, on the ‘Consistency Assumption’: the terms of a private language must refer to objects ‘of the same kind as the object originally baptised’ (Reference Williams1983, 59) if they are to be meaningful. The proper focus belongs on the ‘Naming Assumption’ (58): sensation-terms get their meaning from the ‘ostensive baptism of a sensory experience’ (58–59). This assumption is undermined by Wittgenstein’s discussion of ostensive definition at PI §§27–36, to which he adverts at PI §257. The role of the diary-example, says Williams, is to show that the Consistency Assumption cannot be satisfied by a private language because ‘no standard for subsequent namings has been set … ’ (65).Footnote ⁹³

There is good reason to emphasise stage-setting in the case of the child-genius, and we might plausibly agree with Stroud and Stern (Reference Stern2004, 184–185) that the diary-example makes the same point in a vivid way, but Canfield thinks this reading involves a circularity. The last four lines of PI §258, he suggests, are ‘supposed to demonstrate the failure of the would-be private ostensive definition’ (Reference Canfield2001, 383), and a premise in the argument for that conclusion is that ‘I have no criterion of correctness’ (PI §258). However, on the no-stage-setting reading, the reason that I have no criterion of correctness for my use of ‘S’ is that no ostensive definition has been given – which was to be shown. Canfield concludes that PI §258 must add some ‘new point’ to the child-genius example, and he argues that ‘§257 says that setting up a rule requires a practice, whereas §258 (like §202) says that following a rule requires a practice’ (393).

Defenders of the no-stage-setting reading of the diary-example can deny Canfield’s assumption that the last four lines of PI §258 are ‘supposed to demonstrate the failure of the would-be private ostensive definition’ (Reference Canfield2001, 383). ‘[T]he narrator’s closing words are best read as providing a forceful restatement of the case for thinking that the interlocutor has done nothing that amounts to giving a word a meaning’ (Stern Reference Stern2004, 185).Footnote ⁹⁴ However, these lines do seem to be critical of the proposal that ‘I commit to memory the connection between the sign and the sensation’ by concentrating my attention. If that critique says only that there is no reason to think that a definition of ‘S’ has been given because the usual stage-setting for naming is absent, then it begs the question by ignoring this proposal.

One plausible response is that concentrating my attention depends as much on stage-setting as naming does (see PI §§33–34; Stroud Reference Stroud2002 [Reference Stroud1983], 75). Like the object of one’s pointing, the object of one’s attention is determined by the circumstances: ‘Just as making a move in chess doesn’t consist only in pushing a piece from here to there on the board – nor yet in the thoughts and feelings that accompany the move: but in the circumstances that we call “playing a game of chess”, “solving a chess problem”, and the like’ (PI §33). Attending requires as much disambiguation as does pointing, so it cannot circumvent the challenges to private ostensive definition.

However, perhaps we can reconcile Canfield’s claim (that the last four lines of the diary-example make an additional point)Footnote ⁹⁵ with the stage-setting interpretation in a way that also helps clarify the connection between PI §258 and PI §202– the passage celebrated by defenders of the community view. I suggested earlier (Section 5) that the advice of PI §202 might be to keep PI §201 in mind when we consider private language:

This remark reminds us that, when we judge whether someone has learned a rule, ‘– The application is still a criterion of understanding’ (PI §146). We have no higher court of appeal regarding someone’s understanding than our ordinary judgments about how well she applies a rule, if at all. The teacher’s judgment that a pupil can do addition is based entirely on the pupil’s performance with actual sums.

In the diary-case there is no performance, no manifestation, even to myself (see Stern Reference Stern1994, 559), of my understanding the rule for applying ‘S’. I have a sensation, I ‘focus’ my attention, and I write ‘S’ in my diary. Later, I have a sensation, and I am to remember the connection – the rule – that I was to have established earlier. The problem is that the criterion for whether I have applied the ostensible rule for ‘S’ correctly or not is to be given by the very application of ‘S’ that is to be assessed with reference to that criterion, and this renders my judgment that I have understood my rule for ‘S’ trivial. It is as though the question of my understanding were settled by my avowal of understanding (contra PI §§151–152). So, given what generally counts as a criterion for someone’s having understood how to apply the rule, there is no criterion: nothing logically distinguishes my being right from my thinking I am right.Footnote ⁹⁶ Canfield is right, then, that the diary-example expands on the example of the child genius, but the new point is that further elements of the stage-setting for naming are also missing.

The subsequent passages reinforce the doubt that any naming of a sensation or fixing of a rule for the use of ‘S’ has taken place. At PI §260 we consider whether I might at least ‘believe that this is sensation S again.’ ‘– Perhaps you believe that you believe it!’ is the sarcastic response. If I have not defined ‘S’, then I cannot believe rightly or wrongly that S is recurring (Kenny Reference Kenny1973, 194), and ‘“S” so far has’ no function. Nor, then, is it clear how I can ‘inwardly resolve to use the word in such-and-such a way’ (PI §262) for it is unclear how concentrating my attention on my feeling (PI §263) could establish a ‘technique of applying the word’ (PI §262). Without the appropriate stage-setting, I can no more give myself ‘a private explanation of a word’ than ‘my right hand’ can ‘give my left hand money’ (PI §268). Worse yet, the private diarist’s aspiration to apply ‘S’ to a sensation, assumes too much, ‘For “sensation” is a word of our common language, which is not a language intelligible only to me’ (PI §261). It is implausible to think that what the private diarist wants to do can be described without employing terms that undermine the ostensible privacy of the diary.

PI §259 suggests that the mere impression of a rule could no more function as a rule than the impression of a balance could function as a balance, a theme pursued further in PI §267, where a contrast is drawn between imagining ‘what is called justifying the choice of dimensions for a bridge’ by ‘imagining making loading tests’ and actually ‘justifying an imagined choice of dimensions’ (PI §267). This distinction, in turn, is like that between ‘look[ing] at a clock to see what time it is’ and ‘mov[ing] the hands of a clock till their position strikes me as right’ (PI §266).Footnote ⁹⁷ The timetable example of PI §265, central to Malcolm’s and Kenny’s readings, sounds a similar note: ‘Looking up a table in the imagination is no more looking up a table than the image of the result of an imagined experiment is the result of an experiment’ (PI §265).

The Fourth Wave’s interpretation of the diary-example in light of the case of the child-genius and the earlier passages on ostensive definition is a significant advance. It helps dislodge the diary-example from its traditionally central place in the critique of private language and lets us see that there are many different arguments employed in that critique. If any theme unifies these arguments, it is doubt that sensations are analogous to objects in physical space.

None of this aims to show that a private language is impossible. Rather it leaves defenders of private language with a question: given what ordinarily counts as pointing at something, attending to something, naming something, following a rule – why think that concentrating on my sensation is sufficient to give it a name?

12 The Human Manometer

At PI §270 we return to the diary of PI §258, with a twist:

I argue that we should read this passage in light of Wittgenstein’s treatment of the problem of the inverted spectrum (Section 12.1). It is an argument for thinking that the uses of sensation-terms do not depend on any prior identification of sensations, even if we think of sensations as superprivate objects. Reaching this conclusion, however, requires a detour through the existing literature.

The manometer example has inspired significant misunderstandings. Ayer (Reference Ayer1954, 67), for example, likened it to the way in which one’s sensation-terms are ‘tied up with [one’s] natural expressions of sensation’ (PI §256), as though the manometer readings were natural expressions of S,Footnote ⁹⁸ but the two cases differ. My natural expression of a sensation is what Wittgenstein elsewhere (e.g., BBB, 24–25; see Hertzberg Reference Hertzberg2023) calls a criterion that justifies others’ judgments that I am in pain. On one influential reading (Lycan Reference Lycan1971; Baker Reference Baker1974), a criterion for my pain gives others necessary evidence that I am in pain. Ordinary evidence is evidence only against background assumptions: a rash on my palms is evidence of my allergy to penicillin only given modern immunology. Without the concepts of immunoglobin, antibodies and histamines, I simply have red, itchy palms. By contrast, my cries and grimaces, on this understanding of ‘criterion’, remain evidence of my pain in any context – although other evidence (e.g., my being an actor on stage) can override their evidentiary value. According to another view (Malcolm Reference Malcolm1954, 543–547; Canfield Reference Canfield1981; McDowell Reference McDowell1982), criteria define concepts in the sense that if a criterion for my pain is satisfied (as when I cry out after an anvil falls on my foot), then I am in pain. (The actor’s behaviour on stage only seems to satisfy a criterion for pain.) The rash on my palms does not similarly guarantee my allergy to penicillin.

My hypertension is like the rash on my hands, not like my spontaneous cries of pain, whichever way we understand ‘criterion’. It does not express my sensation S. The connection between S and my increased blood-pressure is ‘just a contingent fact that happened to be discovered’ (Holborow Reference Holborow1967, 351). The manometer readings may thus be evidence for others of my having S, but not a criterion for my having S.

As importantly, evidence and criteria are evidence and criteria for others, not for me, as the asymmetry between first- and other-person uses of sensation-vocabulary suggests (see Section 6.3). I need neither in order sincerely to express my sensations, verbally or nonverbally. I do not discover that I have S by consulting the manometer. On the other hand, once I note the correlation between my sensation and the manometer-readings, my having S becomes evidence for me of my hypertension, and my writing ‘S’ in my diary becomes evidence for others of this same condition and a criterion for them of my having S again.Footnote ⁹⁹

Another early misunderstanding read Wittgenstein as arguing that sensations are irrelevant to the meanings of sensation-terms.Footnote ¹⁰⁰ This interpretation assumes that sensations are private objects and concludes that if the use of ‘S’ to report hypertension does not require recognising my private sensation, then sensations are ‘no longer important’ (Todd Reference Todd1962, 217) to the meanings of sensation-terms. Strawson (Reference Strawson1954, 84) draws a similar conclusion from the beetle-in-a-box of PI §293. However, what is in question is whether I must identify my sensations – like faces in an old photograph – if I am to express them linguistically. ‘If my use of a sensation-word satisfies the normal outward criteria and if I truthfully declare that I have that sensation, then I have it – there is not a further problem of my applying the word right or wrong within myself’ (Malcolm Reference Malcolm1954, 556).

The need to identify my sensations arose from construing first-person authority as privileged access (Section 6.3). And something like Malcolm’s point – that the manometer example concerns the irrelevance for the meanings of sensation-terms of my identifying my sensations – is widely affirmed.Footnote ¹⁰¹ This consensus, however, masks disagreement about why the supposition that I make a mistake is ‘merely sham’ (PI §270).Footnote ¹⁰²

On one popular reading, the manometer example correlates the alleged private sensation, S, with a publicly observable phenomenon, an increase in my blood-pressure. This might provide the criterion of correctness missing from the diary-example of PI §258: circumstances in which ‘S’ is correctly applied (when my blood-pressure rises) and others in which it is incorrectly applied (when my blood-pressure does not rise). If ‘S’ then has a public use but includes a private connection with S, then a private component of meaning has been preserved for sensation-terms. However, the effect of this public correlate is to render the meaning of ‘S’ public. So no private language has been established after all.

Kenny (Reference Kenny1973, 194–195) is the best-known proponent of the argument so sketched.Footnote ¹⁰³ On his reading, PI §270 argues that ‘“S” is not the name of a private object’ (Kenny Reference Kenny1973, 195) because misidentifying S would not matter to the application of ‘S’, and it would not matter because I would have no reason ‘to say that I have misidentified the sensation rather than misremembered which kind of sensation goes with the rise’ (195) in blood-pressure (see Fogelin Reference Fogelin1987, 174).

However, this argument begs the question. Consider a parallel example:

It is mysterious how I could be right about the movement of traffic with better than an even chance if my getting the colour of the light correct is irrelevant.Footnote ¹⁰⁴ Likewise, it would not matter whether I had identified the sensation correctly or not, only if it had already been shown that no such act of identification was necessary to make ‘S’ the name of a sensation.

On Hacker’s reading, by contrast, the publicity of the sign ‘S’ is not the conclusion of the passage but a reason for thinking that the meanings of sensation-terms do not require (re)identifying sensations.Footnote ¹⁰⁵ Rather than argue that the use of ‘S’ is public because no need for identification arises, PI §270 argues that no need for identification arises because the use of ‘S’ is public.Footnote ¹⁰⁶ This view is supported by earlier drafts of the example, in which the sensation correlated with the manometer-readings is pain (MS 165, 146–148) and not the recurrent private object of PI §258 (see Hacker Reference Hacker2019b [Reference Hacker1990], 74–75).

However, if S is not a private object, then the contention that it does not matter whether I misidentify it is puzzling. Why allow that an ordinary sensation might be misidentified, only to insist that it would not matter if it were, if ordinary sensations cannot be identified or misidentified in advance of their expression? If the fact that ordinary first-person uses of sensation-terms do not presuppose the application of criteria is insufficient to show that my use of ‘S’ does not require first identifying S, adding that the misidentification of S would be irrelevant to using ‘S’ to indicate hypertension will not help. As Stephen Mulhall observes, ‘Anyone who thinks that misrecognition of sensations is possible will … take it that the threat of misrecognizing the sensation will in fact undermine the usefulness of the correlation’ (Reference Mulhall2007, 125) – the same objection faced by Kenny’s reading.

Perhaps there is a purposeful ambiguity in how to understand S – as a private object or as an ordinary sensation.Footnote ¹⁰⁷ We might take Wittgenstein’s view to be that in either case identification plays no role in giving ‘S’ meaning. If S is an ordinary sensation, then the meaning of ‘S’ rests on no identification because the paradigmatic function of ‘S’ is to express my sensation. If S is a private object, then, again, the meaning of ‘S’ rests on no identification, for different reasons.

But what reasons? How is Wittgenstein not begging the question against the prospective private speaker? I do not see how to relieve this worry without going beyond PI §270, but we need not go too far afield for an answer.

13 The Beetle

The cases of the child-genius and the private diarist discourage thinking that anything has been accomplished when we try to imagine a (super)private ostensive definition of a sensation-term. If it seems that such a procedure can work, then that comes from accepting a dubious analogy between sensations and physical objects, an analogy reinforced by the misleading analogy between phenomenal and physical space. If we think of sensations as private inner objects, then we will further interpose a moment of recognition or identification between the occurrence of a sensation and its bodily or linguistic expression. However, the case of the human manometer discourages thinking that any such moment of recognition or identification occurs. This does not mean that we cannot say what sensations we have. If sensations are not superprivate objects, then we may plausibly say that they are expressed without any intermediate step, and even if sensations were superprivate objects, then by the private speaker’s own lights their recognition would play no role in their expression because it would matter nothing for the meanings of sensation-terms what those superprivate objects were, even from one occasion to the next.

The manometer-example thus has affinities with the example of the beetle in a box (PI §293).Footnote ¹¹⁰ However, one task of PI §293 is to undermine the idea that I know what pain is ‘only from my own case’ – a goal not explicit in PI §270. Moreover, the beetle-example challenges the idea that a private object might play no role in the use of a sensation-term and, yet, somehow, still contribute to its meaning.

The passage begins with the thought (see PI §283) that I learn the meanings of sensations terms ‘from my own case’ (PI §293) and then combines this idea with the assumption that these sensation-terms have a use in a shared language. The conclusion is that if the word ‘pain’ belongs to a shared language, it does not get its meaning from private ostensive definition, and the temptation to think otherwise comes from objectifying sensations.

Because sensations are not irrelevant to the meanings of sensation-terms,Footnote ¹¹¹ we should not liken sensations to objects.

However, this passage also feeds the behaviourist worries of commentators who misinterpret the last lines. Like the manometer example, the case of the beetle inspired the misconception that Wittgenstein, while denying a role for superprivate sensations in our sensation-talk, nonetheless allows that we have such superprivate sensations. Thus, Pitcher remarked that Wittgenstein ‘only wants to reject the idea that … when you use the word “pain,” you are referring to your sensation and telling other people that you have it’ (Reference Pitcher1964, 299).Footnote ¹¹² The idea is foreshadowed by Strawson’s remark that Wittgenstein falsely believed that ‘one cannot recognise or identify sensations’ (Reference Strawson1954, 86) and that, therefore, ‘“pain” is not the name of a sensation’ (88).Footnote ¹¹³

One variation says that Wittgenstein views sensations on the model of the inverted spectrum. You and I might have radically different experiences when we are in pain; however, ‘What is irrelevant is not the existence of the object, but what it happens to be’ (Donagan Reference Donagan and Pitcher1966, 347).Footnote ¹¹⁴ Another says that Wittgenstein takes sensation-terms to have a dual meaning: a public meaning shared by co-linguists and a private meaning or reference, knowable only to individual speakers.Footnote ¹¹⁵

None of these variations is likely. The inverted spectrum, we saw, severs attributions of colour-experiences to others from the behavioural criteria for their usual attribution, and the idea that sensation-terms might have two components – a public one and a private one – is raised and criticised at PI §§273–279. More fundamentally, these readings take Wittgenstein to accept the analogy between beetles in boxes and sensations in bodiesFootnote ¹¹⁶ and so misunderstand the final sentence of PI §293, which invites us not to model the grammar of sensation-terms on the grammar of object-talk. So the passage brings to the foreground what lies in the background of Wittgenstein’s discussion of privacy and private language. – ‘What I do deny is that we can construe the grammar of “having pain” by hypostatising a private object’ (NPL, 451).Footnote ¹¹⁷

That we should not apply the grammar of ‘object and name’ (PI §293) to sensation-talk was noted by many First-Wave interpreters.Footnote ¹¹⁸ The leading voices of the Second Wave agree,Footnote ¹¹⁹ and many later commentators echo this point:Footnote ¹²⁰ ‘A sensation is not an object’ (1989, 51), says Malcolm Budd.Footnote ¹²¹ But there is disagreement on how to explicate the reasoning. On one account, their epistemic superprivacy makes private sensations irrelevant to public useFootnote ¹²² – we cannot know what others refer to when they speak of pain, but we can talk of pain, so the objects play no role in the meaning of this term, as they would if the object-and-name model were correct. On a related reading, the fact that the public use of ‘pain’ can be taught and learned without reference to superprivate pain renders the object-and-name model irrelevant.Footnote ¹²³ The reason that the private object can play no role in the meaning of ‘pain’ derives from the diary-example, says Hacker: ‘[“pain”] would not be used as the name of a thing; it would not be used as we use names for ordinary objects’ (Hacker Reference Hacker2019b [Reference Hacker1990], 104) because ‘[t]here is no method of comparing a sample with a private object’ (Reference Hacker2019b [Reference Hacker1990], 104).Footnote ¹²⁴ Like Hacker, Stern (Reference Stern, Kahane, Kanterian and Kuusela2007, 266) thinks that PI §§257–258 provide the reason that the private object plays no role in the use of ‘pain’. However, he thinks the objection is simply that ‘the initial ceremony’ of introducing the word ‘beetle’ for whatever is in one’s box ‘is entirely unconnected with the rest of our language …’ (Reference Stern, Kahane, Kanterian and Kuusela2007, 266), not that such inner ostensive definition has been shown impossible.

Like Stern, I think that we should not expect an attempted refutation of the possibility of private language in the Investigations. Like both Stern and Hacker, I think that the reason that the private object is irrelevant to the meaning of ‘pain’ derives from the failure of private ostensive definition. Our merely being unable to know what others refer to when they speak of (superprivate) pain, would be compatible with claiming a convergence in the shared, public use of this word, and the contention that this use could be taught and learned without the existence of superprivate pain might be thought to beg the question. By contrast, if the child-genius and the aspiring private diarist fail to define sensation-terms by inner ostensive definition, then superprivate pain does not matter for the meaning of ‘pain’.

Some readers, nonetheless, interpret the beetle-example as evidence of Wittgenstein’s ‘Linguistic Behaviourism’ (Mundle Reference Mundle1966, 35).Footnote ¹²⁵ The temptation is reinforced – it is ironic – by the fact that in nearby passages the worry about behaviourism is, at first, lurking …

… and then explicitly raised by one or more interlocutory voice:

The responses to these worries, however, reaffirm the grammatical conclusion of the beetle-example: ‘– If I speak of a fiction, then it is of a grammatical fiction’ (PI §307): namely, that sensations are to be modelled on physical objects or processes that involve them. Their reality is never in doubt.

PI §244 suggested that we see sensation-terms like ‘pain’ as extensions of natural expressions of those sensations, rather than as dispassionate descriptions, based on observation, of our inner states or processes. Some such alternative is needed to produce the ‘radical break’ required to eliminate the ‘paradox’ (PI §304) of saying that a sensation is neither a something nor a nothing.

John McDowell recognises Wittgenstein’s caution about the analogy between objects and sensations (Reference McDowell1989, 292–293), but he worries that ‘letting the sensation as particular drop out of consideration as irrelevant’ (290n10) is ‘overkill’.Footnote ¹²⁶ McDowell thinks that the relevant disanalogy between sensations and ordinary objects is not that sensations are not particulars (as objects are) but that external objects ‘are there for one’s thinking anyway, independently of what one thinks about them’ (Reference McDowell1989, 293), whereas sensations are not. This claim is supported, he thinks, by Wittgenstein’s contrast between ‘describ[ing] my state of mind’ and ‘describ[ing] my room’ (PI §290). When I describe my sensations (rather than merely express them), it is still not as though I were dispassionately observing independent objects in a private room (PI §398) or in ‘a kind of peep-show box that everyone carries around in front of himself’ (BT 463).

When dealing with descriptions, thinks McDowell, it is legitimate to speak of objects, provided we remember that describing my sensations is very different from describing my office furniture: ‘the correct point is not that sensations are not objects of reference, but that they are not objects of reference in the way external objects are’ (McDowell Reference McDowell1989, 293).

McDowell’s expression ‘objects of reference’ muddies the waters. Wittgenstein does not deny that we can name our sensations. This makes them ‘objects of reference’, but, by the same token, numbers and colours are objects of reference, and it clarifies nothing to say that naming, or referring to, or describing a colour is ‘a limiting case of bringing an object under a concept’ (McDowell Reference McDowell1989, 292) as McDowell says of sensation-talk. ‘If we say, “Every word in the language signifies something”, we have so far said nothing whatever …’ (PI §13). As we have seen (Sections 3 and 6.1), Wittgenstein encourages us to see the grammar of sensation-terms as more like the grammar of colour-words than like the grammar of names for particulars. So, although McDowell has identified one way in which sensations differ from spatio-temporal objects, he ignores the more general critique of the misleading analogy between physical and phenomenal space that underlies Wittgenstein’s discussion of private language.Footnote ¹²⁷

Other passages on behaviourism explore the implications of treating sensations as objects. If physical objects undergo physical processes, it seems that private mental objects undergo private mental processes. Since a process is something that happens, it seems preposterous to deny such processes: surely, when I feel pain or see red, something happens!

Remembering is not a sensation. Wittgenstein is drawing connections between his treatment of sensation and perception and his treatment of other ‘verbs of experience’ (RPP I §836). However, the ‘picture of the “inner process”’ derives from the view that sensations are objects in phenomenal space – objects involved in processes and events like physical processes and events, involving familiar physical objects.

We don’t want to deny them because we don’t want to deny that something happens when we think or feel pain or see red – we certainly don’t want to deny that we have these experiences. But our conception of what it is to know a process better derives from our investigation of physical processes, and this is a useful model only if there are private objects in phenomenal space. That point is difficult to keep sight of, and we easily return to a picture of experience that is framed by a ‘grammatical fiction’.

Worries about Wittgenstein’s alleged behaviourism are further reinforced by the thought that ‘someone who had never felt pain’ could not ‘understand the word “pain”’ or ‘could not imagine pain without having sometime felt it’ (PI §315). But here we have only questions: ‘[H]ow do we know? How can it be decided whether it’s true?’ (PI §315). There is no large population of people who have never experienced pain, whom we can poll or test on the matter. But many people have never experienced colours, and it seems mistaken to say that they have no concept of colour, even if they can make no empirical discriminations. Someone completely blind can still learn, for example, that red light vibrates at a lower frequency than blue light, much as we all can learn that microwaves vibrate at a lower frequency than radio waves. Such a person is like someone who, according to Wittgenstein, ‘may understand [the] explanation [of a game] but not be able to learn the game’ (LW II 75). So the inclination to insist that the eternally unsuffering cannot know the meaning of ‘pain’ is a requirement that we impose, not a discovery that we make.

Many of the remaining passages in Wittgenstein’s discussion of privacy approach this grammatical fiction from different angles. ‘I know what “pain” means only from my own case’ is neither an empirical proposition nor a grammatical one but ‘a picture [ein Bild]’, analogous to an ‘allegorical painting’ (PI §295) – like the image of justice wearing a blindfold. It holds us ‘captive’ (PI §115) but ‘is not informative’ (PI §298). It expresses only what we find ourselves ‘tempted to say’, which is ‘not philosophy; but … its raw material’ (PI §254):

The sources of our temptation to say such things may be many, but especially important is our lacking ‘an overview of the use of our words’ (PI §122).

So the object-and-name model is no discovery of philosophical inquiry but a prejudice that structures it, a ‘requirement’ (PI §107), given ‘certain analogies between the forms of expression in different regions of our language’ (PI §90). The conclusion is not that words like ‘pain’ are not names of sensations but that sensations do not get their names as spatio-temporal objects do. To reject ‘the grammar which tends to force itself on us here’ (PI §304) is not to reject the plain fact that we have sensations. The sensation is ‘not a something’ – an object that, perforce, would be superprivate, ‘but not a nothing either!’ – not a mere fiction.

14 Epilogue

I lack the space to account for every line of PI §§243–315, but the reading I have presented preserves the coherence and philosophical interest of these passages while respecting Wittgenstein’s admonitions about philosophical problems and the practice of philosophy. Our attraction to the object-and-name model of sensation and perception is driven by the misleading analogy between phenomenal space and physical space, together with an oversimplified picture of naming. Wittgenstein’s critique of these notions, beginning in 1929, is thus of central importance for understanding his critique of privacy and private language in the Philosophical Investigations.Footnote ¹²⁸