ARTICLES

1. Philosophy and its language

One can attack a philosophical argument by identifying a misunderstanding of the language used to state it. These misunderstandings are identified by studying the language and how it has been used by the philosopher. It is not unusual for this strategy to go by the name “ordinary language philosophy” because often it is how language is ordinarily employed that provides the basis for the accusation that the philosopher has misunderstood something.

This argumentative strategy has been much maligned. It is often depicted as an attempt to derive what is so from what is ordinarily reckoned to be so. We find Ernest Gellner (1959: 218) claiming as against the strategy that “there is nothing sacrosanct or necessarily valuable about the ordinary view”; that is, about what ordinary people ordinarily think. Roderick Chisholm (1951: 320) complains that to show that a philosopher uses words in ways that deviate from ordinary language does not “refute him, since we have not shown that what he is saying is false”. More recently we find David Papineau (2006: 20) thankful that “the banalities of ordinary language philosophy are no longer with us, done to death by a thousand miserable attempts to solve philosophical problems by careful attention to upper-middle-class English usage”. And then there’s Timothy Williamson (2004: 128) who compares ordinary language philosophy with the pursuit of clearer vision by way of staring at a pair of eyes.

Can all offensive uses of ordinary language against philosophical claims be tarred with the same brush? My aim in this paper is to clearly distinguish between two varieties of the strategy. One is subject to a straightforward criticism that is alluded to by Williamson and company. The other variety escapes unharmed by it. The former can be found in the work of Norman Malcolm and the latter in that of Ludwig Wittgenstein. Whereas Malcolm accuses the philosopher of deviation from the proper use of words, Wittgenstein accuses the philosopher of confusion about how language works; a confusion which leads her to construct what are, unbeknownst to her, methodological artefacts.

We will proceed as follows. In sections 2 to 4 I describe Malcolm’s deployment of ordinary language against select philosophical claims and I explain why it is unsuccessful. In section 5 I give two reasons to distinguish Wittgenstein’s approach from Malcolm’s. In section 6 I describe Wittgenstein’s deployment of ordinary language against select philosophical claims. In section 7 I compare Wittgenstein’s reason for attending to ordinary language use with Alfred Schütz’s reason for recommending that social scientists attend to how the agents they study interpret their surroundings. The comparison justifies placement of Wittgenstein’s approach within the category of a reputable practice to which any inquirer ought to adhere. Thus understood, Wittgenstein is not subject to the straightforward criticism against Malcolm. In order to stress the importance of keeping the different approaches apart, I close, in section 8, with a discussion of attempts by Stanley Cavell and Hans-Johann Glock to make the distinction with less than the required severity.

2. Malcolm’s offensive: philosophical claims are false

Malcolm (1942, 1951) aimed to show that certain characteristically philosophical claims are incorrect: on one occasion he aimed to show that they are false and on another later occasion that they are nonsense. He introduces his target claims by example. One such example is (1):

• (1) All that one ever sees when one looks at a thing is part of one’s brain.

Russell (1927: 383f) thought it true because of the following argument. Given the causal theory of perception, you see what causes your perceptual state. Your perceptual states are always caused by your brain. Hence (1). However, Moore (Malcolm assures us) would have thought (1) false. Malcolm’s first attempt to vindicate Moore is sought by way of two theses which are as follows.

It may appear as though (1) is a claim about those who see. According to Malcolm (1942: 350) it is really a claim about the verb “sees” to the effect that items are to be counted as the referents of the object of this verb only if they are parts of the brain of the referent of the subject of this verb. Similarly, the claim “This desk which both of us see is not a part of my brain” is to be understood not as a claim about the desk and the two of us but instead as a claim about the verb “see” to the effect that the referent of the object of the verb is what it appears to be viz. this desk. Let’s call this the metalinguistic thesis: claims like (1) are about an expression used in the claim made (see in particular (Malcolm 1942: 356-357)).

Malcolm (1942: 358-359) claims, furthermore, that there are habitual ways in which speakers count words as correctly used: “To be an ordinary expression it must have a commonly accepted use.” The sense in which an expression has a commonly accepted use is, whether it is ever actually employed or not, that it is counted by the members of some class of speakers as properly employed in a certain “sort of situation” (Malcolm 1942: 358-359). Ordinary uses exclude philosophers’ uses, so this class of speakers must exclude philosophers (at least when at work).

If we are to say that Russell was making a claim about the correct use of “see” on the basis of what Russell actually wrote then we should add that a use of an expression involves commitment to the preservation of the truth or falsity of certain claims within which the expression figures. Whatever else might be involved, the use of “see” endorsed by Russell is one on which “A sees (or saw) B” is true if B is a part of A’s brain and false otherwise. The use endorsed by Moore is one on which “A sees (or saw) B” can be true when B is not a part of A’s brain. So understood, it is not necessary for two uses of an expression to avoid overlap in employment altogether for them to qualify as distinct uses.

Of expressions that are ordinary in this sense Malcolm (1942: 357, 362) claims “ordinary language is correct language”. For example, if someone calls something a “wolf” when, as she acknowledges, everyone else would call it a “fox”, then, simply because of this, she speaks incorrectly. We will refer to this as the ordinary-is-correct thesis: the correct way to use an expression is to use it as it is ordinarily used.

When combined with one additional premise these two theses entail the falsity of claims like (1). If the metalinguistic thesis is correct then philosophical disputes over claims like (1) are disputes over the correct use of the word “sees”. If the ordinary-is-correct thesis is correct then the disputant whose view conforms to the conventional employment of that word is the winner of the dispute. Having identified Moore as one with ordinary usage and Russell (or anyone who believes (1) read according to the metalinguistic thesis) as a linguistic deviant, Malcolm (1942: 358, 362) concludes that Moore is the victor. He does so without addressing the truth of Russell’s premises or the validity of his inferences.

3. The trouble with Malcolm

One could attack one of Malcolm’s three assumptions. My preferred objection manages, in a way, to leave them as they are. It is voiced by Chisholm (1951) in the earlier quotation. Malcolm’s ordinary-is-correct thesis does not force a verdict on a metalinguistic dispute if there are two linguistic practices in play. When there are two conventional or ordinary ways to use a word the thesis (rightly) does not require that one rather than the other be followed. Just because a U.N. official is using “torture” in accordance with the U.N. definition and that deviates from how U.S. government officials use “torture” (viz. in accordance with the U.S. definition) does not (by itself) mean that the U.N. official is making a mistake. Indeed, disagreements about the truth of sentences that include “torture” as a constituent (e.g. “waterboarding is torture”) could legitimately continue despite mutual acknowledgement of the different uses being made of “torture”.¹ So in uttering (1), insofar as the philosopher is making a metalinguistic claim, and insofar as he is making a proposal about the proper use of “sees” rather than a proposal about how ordinary speakers ordinarily use the word, the fact that he deviates in his view of the proper use of the word from that of some group of speakers does not (by itself) count against him in any way whatsoever. Nonetheless Malcolm wrongly supposes precisely this; that such a deviation by Russell makes Moore victorious in the relevant metalinguistic dispute.²

4. Malcolm’s offensive: philosophical claims are nonsense

Responding to Chisholm, Malcolm (1951: 329) eventually came to refer to his earlier paper as “dubious remarks of some years ago”. Though he does not concede that he was wrong to make them, he does make a fresh attempt at specifying a way in which ordinary language shows philosophical claims like (1) to be incorrect. He ceases his appeal to the metalinguistic thesis and employs instead the more obvious proposal that the dispute between two philosophers over whether all one ever sees is one’s own brain is a first order dispute about what one can see. The ordinary-is-correct thesis is exchanged for an alternative. Call this the deviant-is-nonsense thesis: if a speaker operates words in a way that deviates from the ordinary usage then her words lack sense (Malcolm 1951: 337-339).

A new argument can be formulated against the correctness of claims like (1) if we make the further supposition that Moore uses “see” in accordance with ordinary usage and Russell uses “see” in a way that deviates from that. Given this and given the deviant-is-nonsense thesis, it follows that Moore’s claim has sense whereas Russell’s does not. Given that anyone who speaks nonsense in a dispute loses to someone who speaks sense, Malcolm can announce Moore the victor.

This second proposal fairs no better than the original. A philosopher can legitimately formulate alternative usages of words that deviate from a standard usage. It is not as though someone who uses “torture” in sentences in a way that conforms neither to the U.S. nor the U.N. definition is (simply because they do so) uttering nonsense. Similarly, sentences that include “see” used in such a way that (1) is true are not inherently nonsensical.

I conclude then that Malcolm’s deployment of ordinary language against select philosophical claims is a failure. In both attempts Malcolm supposes that ordinary language provides us with a conventional usage that is violated by the philosopher and that it is because of this violation that the philosopher is in error. But in each case this last assumption (that deviance entails error) is false.

Malcolm is not alone in the use he makes of “ordinary language”. Commentators on John Austin’s method often interpret it as Malcolmian (e.g. Urmson 1969; Thomasson 2007) and there are very recent renditions of the Malcolmian approach such as Bennett and Hacker’s (2003) attack on neuroscience. The prominence of this way with ordinary language goes a long way toward explaining why attacks against philosophical claims that appeal to ordinary language face immediate disapproval.

5. Separating Wittgenstein and Malcolm

There are striking similarities between Wittgenstein’s and Malcolm’s prose. These can make it seem as though there is nothing more to find in Wittgenstein’s work than in Malcolm’s. Wittgenstein (BBB: 54-55) agreed with Malcolm’s earlier claim that sometimes what looks like a first order claim is in fact a metalinguistic claim and both Malcolm (1942: 363-365) and Wittgenstein (BBB: 16) proposed that philosophers find claims like (1) tempting because they emphasise certain uses of expressions used therein while downplaying others. Nonetheless there are two good reasons to believe that there is a sizeable distance between the two.

The first can be put with relative ease. On several occasions, Wittgenstein acknowledges being unable to make sense of an expression and then gives the expression sense. For instance, he proposed that the claim that a thought has a location is nonsensical. That the sun has set is not something, it seems, that one can find next to the piano. However, “this phrase has sense, if we give it sense”. (BBB: 7) There are ways of understanding locality and thought so that we can utter a sentence like, “The thought is here”, using “here” to identify a physical location and say something sensical. For instance, if one found a correlation between brain activity in a given location and someone’s thinking of bananas then one could say that their thought about bananas is at the part of the brain that is active. Wittgenstein believed that new uses of words are possible which have sense where, on another occasion, they did not. The same can be said of Wittgenstein’s view in the Investigations when he allowed invented language games (e.g. PI § 23, 122, 143). If so then Wittgenstein allowed that one can operate words in ways that deviate from some group’s conventional usage without becoming nonsensical. Thus Wittgenstein upheld a claim that undermines the force of Malcolm’s second attack. That is one reason to keep the two apart.

The second reason requires a rather extensive introduction. Wittgenstein identified a particular model of language as a source of trouble. He describes it in the following remarks:

We are up against one of the great sources of philosophical bewilderment: we try to find a substance for a substantive. (BBB: 1)

One of the reasons for this mistake is again that we are looking for a “thing corresponding to a substantive.” (BBB: 5)

There is....the tendency to look for something in common to all the entities which we commonly subsume under a general term. (BBB: 17)

There is a tendency rooted in our usual forms of expression, to think that the man who has learnt to understand a general term, say, the term “leaf”, has thereby come to possess a kind of general picture of a leaf, as opposed to pictures of particular leaves. (BBB: 17-18)

[On this model] in order to get clear about the meaning of a general term one had to find the common element in all its applications... (BBB: 19)

[One] sees a law in the way a word is used. (BBB: 27)

The model is as follows. Kinds of thing correspond to words. They are identified by one criterion (“the common element”) for all uses of a given word. This fixed criterion determines what counts as that which the word is about. Let’s call this model the interpreted calculus model. Why call it that? A logical calculus consists of a collection of expressions; syntactic rules which define which combinations of expressions are members of the language; and semantic rules which assign semantic values to the expressions of the language. The semantic values either are or determine what counts as the referent of a name, the extension of a predicate, the truth-condition of a sentence etc. The model described in the quotations is one on which the expressions of a natural language (e.g. Portuguese) are members of an interpreted calculus in this sense; thus the label.

Wittgenstein denied that the interpreted calculus model is an accurate model of natural language expressions:

We are unable clearly to circumscribe the concepts we use; not because we don’t know their real definition, but because there is no real “definition” to them. (BBB: 25)

There are words with several clearly defined meanings. It is easy to tabulate these meanings. And there are words of which one might say: They are used in a thousand different ways which gradually merge into one another. No wonder that we can’t tabulate strict rules for their use. (BBB: 28)

There is no one exact usage of the word “knowledge”; but we can make up several such usages, which will more or less agree with the ways the word is actually used. (BBB: 27)

Words per se may have indeterminate extensions but they can be employed, as parts of larger activities, so that in those activities and the circumstances in which they are carried out, what counts as falling in their extension is determinate (enough for that activity) (BBB: 17).

Similar remarks are nested in the Investigations. A target is set up in the opening sections (PI § 1, 2). They describe the interpreted calculus model. The meaning of a word can best be understood by attending to the wider activities (“language-games”) in which the words can be employed (PI § 7). That the diversity witnessed in words’ behaviour does not collapse into a common ingredient is flagged (PI § 10-15) and the point is run home in the discussion of naming (PI § 26-64). Despite this diversity of the word’s possible employments, determinacy in the playing of particular language-games on particular occasions is not forfeited (PI § 1, 2, 37).

I want to interpret a rejection of the model as a rejection even of a view on which there are clusters of criteria associated with given linguistic expressions, where what criterion is relevant on which occasion of use is governed by invariant rules.³ This interpretation, however, cannot be confidently adopted for the later Wittgenstein simpliciter because he was undeniably equivocal on this matter in the Blue Book but not in the Investigations.

For a long time Wittgenstein took there to be unnoticed behaviours (logical forms or uses) of expressions that need to be studied in order to avoid philosophical confusions; e.g. (RFL: 29-30). But up to and in the Blue Book he understood the interpreted calculus model to apply to the object of a mature speaker’s competence. This is reflected, for example, in his then treatment of language games:

...we recognize in these simple processes forms of language not separated by a break from our more complicated ones. We see that we can build up the complicated forms from the primitive ones by gradually adding new forms. (BBB: 17)

The interpreted calculus model is retained but modified: the criterion for using a word is relativised to the language games one has to learn to know how to use the word in question.⁴ Things are different in the Investigations (BBB: vii). Wittgenstein recognises that language-games themselves do not sum up to linguistic competence because there is no absolute answer to the question of what counts as being the playing of the same game (PI § 62). So learning a language-game will fall short of knowing how to use words that figure within them because there will remain the question of what would count as playing a given game again. Linguistic maturity requires more than this. Indeed, language-games are introduced here as the means by which children may learn a language and not as fragments of the mature speaker’s competence (PI § 7). Thus the Wittgenstein of the Investigations denies that there is any such thing as the use(s) or logical form(s) of a given linguistic expression. In other words, though he dithered in the Blue Book, in the Investigations he robustly rejects even attenuated versions of the interpreted calculus model.

This fact provides us with the second reason to avoid assimilating Wittgenstein to Malcolm on the relevance of ordinary language to philosophy. Both Malcolm’s ordinary-is-correct thesis and his deviant-is-nonsense thesis require there to be a conventional association of a word and a proper use (in the relevant sense). If there is no such thing as a word’s use (in this sense) then one cannot criticise philosophers for deviating from words’ proper use (in this sense). It seems then that, given his recasting of language-games in the Investigations, Wittgenstein rejected a prerequisite of Malcolm’s line of attack; there is no proper use from which philosophers could be deviating.

6. Wittgenstein’s offensive

How then does Wittgenstein deploy ordinary language against philosophical claims like (1)? He certainly believed that a study of ordinary language will provide resources with which to attack. For example, he made the following recommendation for one who finds herself making claims like (1):

The thing to do in such cases is always to look how the words in question are actually used in our language. We are in all such cases thinking of a use different from that which our ordinary language makes of the words. (BBB: 56)

This passage can be read as expressing agreement with Malcolm: the philosopher errs in deviating from an ordinary usage. But it should not be so read. For Wittgenstein the role played by the examination of words in use is not as means to identifying a violated convention, as it was for Malcolm. This is made plain by the way he continues this passage. The aim of studying “our ordinary language” is instead to uncover hitherto unnoticed vacillations between different uses of a word:

When something seems queer about the grammar of our words, it is because we are alternately tempted to use a word in several different ways. (BBB: 56)

If this is the justification for studying “our ordinary language” then the recommendation in the previous quotation is not to catalogue the use of words (insofar as there is such a thing), but instead to make noticeable something we are doing with words which would otherwise go unnoticed.⁵ This is very much the point of emphasis in the methodological discussion of the Investigations (PI § 103-133). That discussion is introduced by acknowledging that some misunderstandings of the use of words caused “among other things, by certain analogies between the forms of expression” can “be removed by substituting one form of expression for another” which may be called “‘analysis’ of our forms of expression” (PI § 90). But, contra Malcolm, he immediately cautions against a misimpression this way of putting things may cast: that ordinary language is to be studied in pursuit of a “final analysis of our forms of language” (PI § 91). As we have seen, he thinks there is no such thing. The alternative point of the study comes out in the discussion on method proper.⁶ The aim of studying ways of using words as laid out in stipulated language-games is not to identify “the order” but instead “an order in our knowledge of the use of language: an order with a particular end in view; one out of many possible orders”, where the “end in view” is to understand cases of “entanglement in our rules” (PI § 125) where these rules are not the rules of (say) the English language but the rules “we lay down...for a game” (PI § 125). So the Wittgensteinian point of studying ordinary language is to notice unnoticed vacillations in how we are using our words which result in “entanglements”. An example will help focus this attribution.

We know that the problem Wittgenstein believed philosophers face derives from their (attenuated or full) adoption of the interpreted calculus model when it is a bad model for natural languages. Let us consider the likely predicament of someone who is thus benighted. Suppose there is a philosopher who believes that the model applies to the language she uses when really it does not. Suppose two claims are under consideration which share vocabulary and, in part because of this, are considered to be inconsistent. One such claim is philosophical i.e. analogous to our exemplar (1). The other is a sentence which is thought to express a truth. For example, compare:

• (2) I cannot know that someone else is in pain.
• (3) I know that he is pain. [said of a particular man]

Sentences (2) and (3) appear to be inconsistent.⁷ A philosopher who supposes that her words behave in accordance with the interpreted calculus model will think that words do not change their criteria across different uses. This includes the criteria for employing “know” and “pain”. So once she has identified a way one can use these expressions she will fixate upon it as if it were the use of these expressions, even though (given the falsity of the interpreted calculus model) the words per se do not have criteria which do this work. The philosopher in effect holds fast her usage of the words in ways she ordinarily would not. So if, with criteria thus held fast, it is not possible to say truly, “I know that he is in pain”, where with the exception of the speaker, this is so no matter who he is, it will then also be possible to say truly, “I cannot know that someone else is in pain.”

To see how a philosopher could (accidentally) hold fast the way she uses her expressions, compare the sentences: “He has a gold tooth” and “He has a pain.” On their surface, claims made using these sentences are similar. (Imagine the sentences translated into predicate calculus.) In both cases we are saying of someone that he has something. What is had is in one case a gold tooth and in the other a pain. This formal similarity can encourage the supposition that there are other similarities as well. For example, someone can have a gold tooth and conceal it behind her mouth. So unless she opens her mouth one cannot tell that she has a gold tooth. Because the two claims are formally similar one might suppose that this situation can act as a model for the case of someone’s having a pain. We might suppose that someone has a pain like someone has a gold tooth, a pacemaker, or a cancer. There is a container with an opaque boundary. An examination of the interior of that boundary from outside it is not possible. So one cannot tell without crossing that boundary how things are behind it. Carry this over to the case of pain: someone’s pain resides behind a barrier which blocks one’s view on the pain. The model makes it appear as though there is a certain sense in which someone has a pain and that to detect whether someone has a pain in that sense one has to be within that boundary. Since one can’t be in such a position with respect to another’s pain so understood, the model, articulated in accordance with the analogy with the gold tooth, tempts one to say, “I can only know that I have pain; that anyone else has a pain is a conjecture” (BBB: 56). This conclusion may appear to be inconsistent with all claims of the form “I know he is in pain” when in fact (many) such claims are made on different uses of the words “know” and “pain”. Thus one appears to have made a discovery that overturns what one thought one knew when in reality, whatever inconsistency we have here, it is a consequence of how the words are used in an episode of philosophising. Thus a philosopher who uses the form of words as a guide to homogenous use may accidentally perform a restriction on her usage of an expression when non-philosophical employments of that expression are not so restricted in the criteria by which they are wielded:

When words in our language have prima facie analogous grammars we are inclined to try to interpret them analogously; i.e. we try to make the analogy hold throughout. (BBB: 7)

The man who is philosophically puzzled sees a law in the way a word is used, and, trying to apply this law consistently, comes up against cases where it leads to paradoxical results. (BBB: 27)

To summarise, the Wittgensteinian point of studying ordinary language is not to identify the proper use of linguistic expressions. It is instead to shake off the idea that the interpreted calculus model accurately models natural language. Recognition of this can be a potent weapon against a philosophical claim. In supposing that expressions behave in conformity with the interpreted calculus model, the philosopher has come under the impression that her philosophical claim reveals various ordinary claims to be false when really it does no such thing. This revelatory impression is created by artificially and unintentionally holding fast ordinary vacillations in the use of expressions across different occasions for employing them and confusing this holding-fast for a self-standing feature of the expressions employed. The mistake is inevitable insofar as a version of the interpreted calculus model is false and the philosopher operates words on the assumption that it is true (if only while she philosophises).

That this was Wittgenstein’s way with ordinary language is obscured by a common reading of one of the most widely discussed tracts of his Nachlass: the rule-following discussion. The topic of that discussion is commonly supposed to be the use of a word in the sense presupposed by Malcolm; the proper use that a word has throughout one’s career as a speaker of (say) English. (See McDowell (1984) and Wright (1980) for two prominent examples.) If so, then one would suppose that Wittgenstein believed there to be such a thing. That in turn can make it tempting to read any talk of “normality” and the “ordinary” as talk of words’ use in this sense, and then in turn again, to read his calls to study normal or ordinary use as calls to the Malcolmian project (e.g. PI § 39, 81, 82, 87, 90, 120, 132, 141, 142, 156). It is important then to emphasise that the topic “rules for a word’s use” is introduced with the claim that insofar as a rule is involved in the use of a word, different rules are offered in explanation of the use of a word for different purposes: there is no one for-all-purposes rule that governs a word’s use (PI § 79, 81, 82, 87). Given this, the rule-following discussion that follows should not be read as concerning a word’s use in the Malcolmian sense, but instead, as concerning a word’s use in some particular activity.

7. A comparison with Schütz: methodological artefacts

Alfred Schütz was a sociologist preoccupied with the ordinary. Wittgenstein’s motivation for studying ordinary language parallels Schütz’s motivation for recommending that social scientists study the methods of the agent’s they study. I am going to draw out the similarity in an attempt to clarify the kind of mistake Wittgenstein attributed to the philosopher. The aim of this section is to find a description of the philosopher’s error which can be put as simply as the Malcolmian line that the philosopher errs in deviating from ordinary use.

Both Schütz and Wittgenstein were concerned with the following difficulty. It can appear as though that which one studies exhibits a certain property when in fact this is an appearance generated by one’s method and which is not exhibited by that which one studies independently of the investigation’s pursuit. For instance, one may fail to notice that one’s pipette is contaminated with bacterium X while attempting to show that bacterium X grows without exposure to oxygen. Every time one checks, one finds the bacterium but not for the reason one thinks one does. One thus creates a methodological artefact: an observation which appears to be of a phenomenon that exists independently of the investigator’s investigating but which in fact has no such independent reality.

Schütz was concerned with this problem as it afflicts the social scientist. How an agent acts will depend upon her understanding of the environment within which she acts. Schütz explains the danger that arises for the social scientist in terms of a “construct”; an object of thought such that two thinkers can examine the same scene but there be different objects of thought available to each. It may appear that, on the contrary, if there is the same scene before each then there are the same things to think of for each. We can make sense of this by treating the objects of thought (“constructs”) as things understood a certain way. For example, suppose it is getting dark. A wants her privacy. She believes that the curtains are thick enough to block out her silhouette. So she draws the curtains. Thus her action is guided by the environment as she understands it.

The social scientist’s study of such a thing (A’s action) is itself an activity and in performing that activity the social scientist will employ her own “constructs” of her subject’s environment; “constructs” which are quite alien to her subject and hence her subject’s actions. For instance, suppose the social scientist notices that outside A’s window, across the street, there lives B who is from the nation in which the curtains were made. Suppose that the curtains were made by the enemies of B. The social scientist, knowing this connection, and recognising the symbolism involved, may interpret A’s drawing of the curtains as an attack or insult to B.

Schütz (1953: 4) was concerned that the social scientist “takes for granted that the very adoption of the methods of the natural sciences for establishing constructs will lead to reliable knowledge of social reality” when in fact it “would lead far away from the constructs in terms of which men in the reality of daily life experience their own and their fellowmen’s behavior”. Returning to our example, A’s actions were not guided by the relation between the curtains and B. She was driven only by her desire for privacy and what she understood to be a means to that in the circumstances. If the social scientist misidentifies the constructs of the agent she studies (by confusing her own with those of her subject) then she will, without realising it, misunderstand what guides the agents she studies. There is a danger of methodological artefact. It is as part of an effort to avoid this danger that Schütz (1953: 34) recommends that social scientists study the methods of interpretation exercised by their subjects. Regardless of whether the social scientist believes that these methods of interpretation are ones that the subject ought to employ, when the subject matter is an agent’s actions it is nevertheless important to know what they are and to allow that they differ from those one employs in conducting one’s study.

Whereas Schütz addressed himself to the social scientist’s study of action, Wittgenstein was concerned with the philosopher’s study of her subject matters: meaning, time, virtue etc. A typical mode of philosophical investigation opens with a question of the form: “What is X?” One answers such questions with an (often complex) criterion for X-hood and one is prohibited from providing examples of X as answers to the question. (See Wittgenstein’s (BBB: 20) complaint against Socrates on this score; see also PI § 92) If all one needs in order to identify a homogeneous class of items is a word, namely “X”, because that word has its own criterion which divides the world into items that are X and items that are not, then one can raise a sensible question with a determinate answer just by uttering the words, “What is X?” Wittgenstein’s suspicion was that linguistic expressions which share a form can mislead one into thinking there is homogeneity in the extension of an expression where in fact it is absent. Instead of indicating homogeneity of use, homogeneity of form coaxes a philosopher into homogenising her use of words when she conducts her questioning and arguing (see for example: (BBB: 31) and (PI § 104, 107, 114, 175)). This presumptuous philosophical method is liable to generate illusions of discovery. Wittgenstein’s recommendation that the philosopher attend to ordinary language use is warranted in the same way Schütz warranted his recommendation that social scientists study the interpretative methods of the agents that form their subject matter. In both cases the point is not to ape the ordinary but instead to reduce the risk of methodological artefact by sensitising the investigators to unnoticed effects that their investigations may be having on their respective subject matters; a sensitivity that any good inquirer ought to foster.

Our summary description of the philosopher’s error as it figures in Wittgenstein’s offensive can now be put like so: she confuses (or at least risks confusing) artefacts of her method with objective features of the words she employs and the world described therewith. If the interpreted calculus model is false but one operates words on the assumption that it is true of them, one’s results will lack ecological validity.

This deployment of ordinary language in an attack on philosophical claims is not subject to the objection facing Malcolm. Malcolm’s approach requires us to suppose that a different usage is a broken usage. That supposition is false. The charge that (some) philosophers are unknowingly generating methodological artefacts does not require this supposition. For the philosopher to avoid Wittgenstein’s criticism it is not even necessary that her use of words be the same as that of non-philosophers. It is for this reason that the perennial charge against ordinary language philosophy does not apply to Wittgenstein’s version of that philosophical strategy.

8. A comparison with Cavell and Glock

Both Stanley Cavell (1979) and Hans-Johann Glock (1991) have offered interpretations of Wittgenstein’s use of ordinary language that are akin to my own. However, their depictions of the Wittgensteinian assault inadequately distil an interest in ordinary language owed to linguistic conservatism from one owed to a worry about methodological artefacts. In this section I will stress the importance of a sharp distinction of complaints by explaining what is suspect in the attempts of Cavell and Glock.

Cavell posed a dilemma to a philosopher who aims to show that we know nothing. Cavell’s (1979: 144) philosopher pursues her aim through exchanges of the following sort. A makes a claim:

• (4) I know that a table is there.

The philosopher asks A how she knows this. A gives a basis of the knowledge: “Because I see it.” The philosopher then provides ground to doubt the basis, e.g. “But you don’t see all of it. The most you see is its surface.” The philosopher then draws the conclusion that (4) is false because A sees no more than the surface and that is compatible with the table’s absence. The philosopher then generalises: one never knows that there is something there because one’s senses always land one in this predicament.

Cavell’s dilemma rests upon the assumption that natural language expressions do not conform to the interpreted calculus model (1979: 168-190). Given the falsity of the model, that one actually makes a claim by uttering (4) is not guaranteed by the words employed. Nor is it guaranteed what claim one would make if one did make one. Cavell’s (1979: 202-203; 220) dilemma is then this. Either the claim made in A’s uttering of (4) is one that someone might make outside of the philosophy classroom or it is an exceedingly strange claim that is only ever made inside a philosophy classroom. If it is the former then the philosopher’s ground for doubt can at best be a ground for doubt only in special cases. That is, only when A and the philosopher manage to use their words with a sense such that someone qualifies as failing to see all of the table and this qualifies as a reason to doubt (4). That is so only sometimes. There are plenty of ways in which one can qualify as seeing the table even though one cannot see the side facing away from one (for instance). So insofar as a ground for doubting (4) (thus understood) has been produced, the doubt applies only in some cases. On the other hand, if what is claimed in A’s utterance of (4) is of the strange sort found in the philosophy classroom then the philosopher’s ground for doubt may well show that a certain set of claims makeable with (4) are false. But this has no bearing on claims makeable with (4) that fall outside that set. Since the claims that the philosopher intends to target fall outside the relevant set, no discovery that overturns what was previously taken to be true would have been made. So either the philosopher achieves a general conclusion about a set of claims that does not include what the philosopher thinks it does, or else, the philosopher does overturn the claims she thinks she does, but only some of them. Either way the philosopher doesn’t get a conclusion with both the generality and interest she thought it did. The appearance to the contrary is an artefact of the philosopher’s faith in the interpreted calculus model; the philosopher is “inventing something where he supposes himself to be discovering it” (Cavell 1979: 223).⁸

Has the deviance-is-nonsense thesis been supposed in posing this dilemma? Cavell (1979: e.g. 155ff) is explicit in rejecting that thesis. The philosopher’s problem isn’t deviation but instead a loss of control (Cavell 1979: e.g. 193). I think it reasonable to equate (perhaps with some finessing) this “loss of control” to the predicament I labelled the generation of “methodological artefacts” in section 7. However, Cavell is often equivocal on the role that a standard usage plays in his discussion. True, he did explicitly distinguish the view he attributes to Wittgenstein from a view found in Moore’s work:

In them [Wittgenstein and Austin] the emphasis is less on the ordinariness of an expression (which seems mostly to mean, from Moore to Austin, an expression not used solely by philosophers) than on the fact that they are said (or, of course, written) by human beings to human beings, in definite contexts, in a language they share: hence the obsession with the use of expressions. (Cavell 1979: 206)

Unfortunately, Cavell goes on to obscure the contrast by expressing the alternative to a focus on “ordinariness” by drawing attention to a deviation from none other than [...] the ordinary. He claimed that for Wittgenstein:

...investigating ourselves, we are led to speak “outside language games”, consider expressions apart from, and in opposition to, the natural forms of life which give those expressions the force they have...What is left out of an expression if it is used “outside ordinary language games” is not necessarily what the words mean...but what we mean in using them when and where we do. (Cavell 1979: 207)

Similarly, when he considered Wittgenstein’s proposal that when philosophers use language it goes on holiday or is like an engine idling, Cavell (1979: 226) glossed the phenomenon by saying that “what happens to the philosopher’s concepts is that they are deprived of their ordinary criteria of employment”, thus once again falling back on the idea that the philosopher’s error lies in deviating from the ordinary.

This packaging of the philosopher’s error is recurrent (Cavell 1979: e.g. 189; 196-197; 203) and problematic. Despite the availability of a charitable reading, Cavell leaves open another (far from obscure) reading on which if only the philosopher did not deviate from what is ordinarily done with linguistic expressions she would not be impaled on the dilemma. Thus Conant (2005: 64) presents Cavell’s dilemma in precisely these alloyed terms.⁹ While this is not quite Malcolm’s deviance-is-nonsense thesis, it is a form of linguistic conservatism. Cavell, so read, would be saying: “Sure, the philosopher can construct her own uses of expressions and that in itself is unproblematic. But she is sure to tie herself in knots if she does so. So really she should not be so creative.” That variant on the deviance-is-nonsense thesis is just as distasteful as Malcolm’s own. So while Cavell certainly took the philosopher’s mistake to lie in her losing control, Cavell’s presentation of that point is blemished by the way he describes the knots the philosopher entangles herself within: namely, as due to her departure from the ordinary.

A very similar complaint can be made against Glock (1991). Glock (1991: 72) aims to explain how Wittgenstein could both: claim that he never puts forward contentious theses and, at the same time, attack philosophical claims. Surely, if your thesis is uncontentious then it cannot conflict with your opponent’s thesis. Glock’s (1991: 80) solution is to find in Wittgenstein’s method an “undogmatic procedure”: a style of argument which does not require the pushing of opposing theses. The alternative method is to uncover unnoticed mixtures of uses of expressions from philosophical claims and arguments which show the philosopher’s argument to fall short of the intended conclusion (Glock 1991: 82, 84). To do this no opposing thesis is required. One need only give an accurate description of what the philosopher is doing. Like Cavell, and unlike Malcolm, Glock (1991: 81, 84) explicitly denies that mere deviation from “ordinary” usage is problematic. Unfortunately, and again just like Cavell, Glock persists in helping himself to the idea that there is such thing as an ordinary usage in the sense defined earlier, and that it is because the philosopher deviates from this that she ties herself in knots. He (1991: 78) claims, for instance, that “...the special function of grammatical reminders is to draw attention to the violation of linguistic rules by philosophers, a violation which results in nonsense.” Which rules are these?—Those governing ordinary usage of expressions. Thus, he falls back into the Malcolmian way with ordinary language despite it being a moral of the paper that that is not the basis of Wittgenstein’s attacks on philosophical claims. The situation is not helped by his un-prefaced talk of “our language games” (Glock 1991: 76) and “our ordinary [concepts]” (Glock 1991: 81). Un-prefaced, these locutions allow readings which presuppose a Malcolmian use. This is not a sensible descriptive policy. Glock also makes use of PI § 90 but omits to mention Wittgenstein’s rejection of the aforementioned “misimpression” that it can generate (see section 5 above). The misimpression is precisely that of leaving in view a notion of ordinary use that makes talk of “the ordinary”, and its role in Wittgenstein’s philosophy, with a Malcolmian appearance.

To conclude, there is a straightforward objection to any argument against a philosophical claim that has as a premise that deviation from an ordinary usage is in itself problematic. Because of this, ordinary language philosophers must take greater care to avoid this bad assumption and boldly advertise its avoidance with unequivocal exposition. There is much for philosophers to learn by studying ordinary language but no one’s going to listen if it all sounds like linguistic conservatism.¹⁰

Biographical Note

Alex Davies recently received his PhD in philosophy of language from King’s College London. In his dissertation he argued for a hitherto unacknowledged utility of linguistic context-sensitivity by drawing an analogy between it and the operationalisation of theoretical terms in the application of scientific theories to particular experimental arrangements. He is currently working on a response to a main objection to radical-contextualism: that it renders communication impossible. This work applies developments in recent action theory and conversation analysis to Austin’s philosophy of speech action. He also has interests in philosophy of mind and feminist philosophy.