It is hard to read Wittgenstein. Burdened with ideas about what he is saying and why he is saying it, we struggle to distinguish what he actually says from what we think he must be saying. Michael Luntley's book Wittgenstein: Opening Investigations takes this difficulty very seriously. It is a sophisticated and ambitious interpretation of Wittgenstein's Philosophical Investigations (PI). Luntley aims to stand much of what we think we know about this work on its head.

Everyone knows (or thinks they do) two things about PI. First, that it should be understood as a sustained critique of the ubiquitous conception of language encapsulated in Augustine's remarks which open PI:

The second thing everyone thinks they know is that in PI Wittgenstein doesn't propose positive theories of anything. Luntley writes against this critical and quietist reading which dominates Wittgenstein scholarship, the locus classicus being Baker and Hacker's Wittgenstein: Understanding and Meaning. Instead, Luntley provides an explanatory interpretation which takes Wittgenstein to pretty much agree with everything Augustine says (x). On this reading, PI contains explanations ‘continuous with’ contemporary work in analytic philosophy of mind and language (131).

How does an infant without conceptual or linguistic resources first acquire language? Luntley's Wittgenstein thinks that an infant's entry into language begins by grasping the meaning of individual words through ostensive teaching. This teaching connects, in the mind of the infant, a word and what it refers to, such that the infant comes to have knowledge of word reference. Grasp of word reference brings with it grasp of the word's permissible combinatorial patterns (its ‘grammar’ (xix)). For instance, the name ‘Socrates’ may combine with the predicate ‘is flying’ to form a sentence (‘Socrates is flying’), but may not combine with ‘Plato’ to form a sentence (‘Socrates Plato’ is not a sentence).

Ostensive teaching, as Luntley describes it, consists of the teaching of individual words by a competent speaker of a language by means of pointing and naming, to a pupil who cannot yet frame questions about the language, such as ‘What is the name of this?’. Ostensive teaching is contrasted with ostensive defining, which is teaching by similar means but to someone who can already ask such questions, and so who possesses a rich conceptual and linguistic repertoire which ostensive defining exploits (47). Ostensive teaching works via the triangulation of a teacher (who points and names), an object (pointed at and named), and an infant. What goes on in the infant's mind which makes such teaching successful? There are two cognitive mechanisms: an object-detection device and a pattern-detection (or pattern-imposition) device. The first is attention (126), the second is sense of fit (148). The teacher directs the infant's attention to an object by pointing, and the name of the object is uttered. The pupil is thus brought into ‘direct contact and anchorage on things independently of possession of a concept for the thing in question’ (52) – this is how the reference of a word is grasped. The pupil's sense of fit discerns or imposes a pattern on experiences of words and objects given in ostensive teaching – this is how grammar is grasped or constituted. The sense of fit is a natural inclination to discern or impose structure on experience quite generally (148). Possession of sense of fit makes it possible to cotton onto the fact that a word, an act of pointing, and the thing pointed out go together (118). Normally it can be shaped through training so that the pupil groups things together similarly to the teacher (153). In this way, the world – teacher and object – and the infant's mind jointly contribute to its grasp of grammar.

Crucially, attention and sense of fit are thought of as non-conceptual and non-linguistic. Attention, for instance, is not attention of something as falling under a concept or description. This leaves open the possibility of a reductive explanation of the acquisition of language and thought. Such a project can crudely be understood as starting from the empiricist mind – absent of language and thought – and then asking how language and thought is acquired. The answer Luntley attributes to Wittgenstein is an idiosyncratic twist on the standard empiricist answer: general pattern-seeking learning mechanisms explain language acquisition.

A doubt arises, however, as to whether Luntley's twist on this is genuinely reductive. The problem concerns how it is that knowledge of reference brings with it knowledge of grammar. Can someone know that, say, ‘Socrates’ refers to Socrates without also knowing the grammar of ‘Socrates’? Luntley thinks not. What underpins this answer appears to be this: the very idea of a referring term brings with it a grasp of distinctions, such as that between a referring term and predicate expression, and that between a well- and ill-formed sentence (xix). This full-blooded grasp of reference is however a conceptual grasp for it both requires understanding a set of distinctions, and recognising the referring term as a referring term to which those distinctions apply. So Luntley faces a dilemma: either our initial grasp of reference is conceptual, in which case it may explain our grasp of grammar but not reductively, or else our grasp of reference is not conceptual, in which case it doesn't seem able to explain our grasp of grammar. The more conceptual work reference demands, the less plausible it is that we can have knowledge of word reference without a prior rich conceptual repertoire.

Reductive ambitions can also be seen to cause trouble in accounting for how sense of fit contributes towards language acquisition. Any conventional account of language acquisition must say something about what it is in the infant's mind that plays a role in it. The worry is that because so little can be said about how sense of fit operates, there is little that can be said as to how it plays this role, and so there is little reason to think it does. The Chomskyan approach to language acquisition is very different, and the contrast is instructive. The Chomskyan takes us to possess language innately, in the form of rules which explain how we learn the natural language in our locale. If like the Chomskyan you explain natural language acquisition by linguistic rules you arguably fail to provide a reductive explanation of thought and language, but you do provide a detailed mechanism which can be tested. For a reductive approach to succeed, we need a similarly detailed account of the general learning mechanism by which language is learned. In Luntley's account, the general learning mechanism is the general pattern-seeking device, viz., sense of fit. Since sense of fit is non-conceptual it is not rule-governed (13, 119), and since it is primitive it is not subject to any explanation (it is characterised as a primitive subjective-normative aesthetic sense – 148, 152). No details are offered on how sense of fit discerns or imposes word patterns – why, that is, these words and things are grouped together and why these are not – and so any explanation invoking it appears ad hoc. But this means we cannot say how the sense of fit mechanism operates. As such, we cannot assess its explanatory power and go on to compare it with rival theories of language acquisition. The upshot is that the reductive constraint on sense of fit prevents it making a genuine contribution to explaining language acquisition.

Luntley takes sense of fit to play a starring role in the explanation of language acquisition, and so a role in what distinguishes human infants from rocks and rabbits, which cannot learn language (120, 153). He doesn't always seem clear on the distinction between what enables and what accounts for language acquisition. Even if sense of fit fails to explain language acquisition, it may nonetheless come into a modest story of what makes acquisition possible for us – and so arguably into what distinguishes us from rocks and rabbits. Chomskyans can appeal to their language learning device to both account for what enables and what explains language acquisition – the innate organ doesn't exist in rabbits or rocks, and the story of how the organ grows is the bulk of the story of natural language acquisition. By contrast, while the capacities of attention and sense of fit do not suffice for an explanation of language acquisition, they may make it in some sense possible. (Though whether this Kantian question is something Wittgenstein would countenance is put into doubt by PI, §25.)

A striking limitation of Luntley's account of language acquisition is that it is entirely restricted to referring terms. One thing Wittgenstein appears to be doing in the PI is critiquing his ‘old way of thinking’ in the Tractatus (PI, Preface). Within the range of this critique is arguably the thought that language consists entirely of combinations of referring terms or that the core of language is constituted by them (PI, §39 and thereabouts). While Luntley doesn't explicitly endorse this targeted view, his account of language learning by ostensive teaching seems to force him to. For it is hard to understand how non-referring terms can be learnt by the triangulation which occurs in ostensive teaching. This requires an object which can both be pointed at and be the object of joint attention. As such, words which are not referring terms seem not to be learned this way. This includes words like ‘not’, or ‘like’ or ‘or’. Indeed, not even all apparent referring terms seem learnable through triangulation. For instance, ‘2’ (and other terms which appear to refer to ‘abstracta’) or ‘the sense of foreboding’ (and other terms which appear to refer to ‘the inner’), or ‘Sherlock Holmes’ (and other terms which appear to refer to ‘fictional entities’). For in such cases the purported object doesn't look capable of being pointed at and sustaining joint attention.

This raises the question: how close are Luntley's explanatory theses to Wittgenstein's own views? One way we can begin to approach this is to ask whether they are consistent with Wittgenstein's overall conception of philosophy – leaving aside the possibility that Wittgenstein's philosophy might itself deviate from his conception of philosophy. Luntley focuses on Wittgenstein's rejection of philosophical explanation (PI, §109). This rejection is read very narrowly as a rejection of reductive scientistic explanations in philosophy (PI, §134). And he argues that the explanations he attributes to Wittgenstein are consistent with this rejection. But even if Luntley's reading of ‘explanation’ is plausible, his attempt to locate Wittgenstein as offering positions continuous with those in contemporary analytic philosophy doesn't take seriously enough how radical Wittgenstein's conception of philosophy is. Wittgenstein asserts that ‘Philosophy simply puts everything before us, and neither explains nor deduces anything’ (PI, §126). Though this says more about what philosophy does with us than what we do with philosophy, the thought seems to be that the work of philosophy is not the work of reasoning from premises to results. It is about getting clear about what ‘is possible before all new discoveries and inventions’ (PI, §126). Whatever this may mean, it appears discontinuous with the search for results in philosophy, and to the extent to which Wittgenstein is forced into this mould, his thought – for good or ill – is distorted.

If Luntley is right that PI contains conjectures akin to those in empirical linguistics, then these conjectures would be hostage to fortune. And indeed he recognises evidence that ostensive teaching is not in fact how infants acquire natural language (81 fn11). Deeper engagement with current research on empirical linguistics would have been welcome in order to assess whether Luntley's Wittgenstein has fundamentally gone astray. Luntley's book is an ingenious attempt to read Wittgenstein against the grain. He proceeds with great care – in particular, with the opening sections of PI – and finds Wittgenstein to be endorsing the very theses he is standardly taken to reject. Though it doesn't ultimately succeed, the book will be of interest to Wittgenstein scholars, for it forces us to read PI with renewed attention and thoughtfulness. Furthermore, it forces us to ask again what relevance Wittgenstein may have for contemporary philosophy, and what the significance of Wittgenstein's uncooperativeness, in this regard, is. Luntley's engagement with questions about language acquisition, reference, realism, rules, cognitive psychology and philosophical methodology also ensures his book will be of interest to anyone thinking about them.