If meaning is use, then a theory of meaning is a description of use not an analysis of sense. This is the message of Wittgenstein’s Philosophical Investigations, as commonly understood. Instead of analyzing what a sentence means, breaking it into parts, providing necessary and sufficient conditions, we should aim for a perspicuous description of use, a survey of the practical capacities that constitute mastery of a language. Not ideal composite unities (the “proposition” of the Tractatus) but the concrete multifarious phenomena of actual speech, spread out in time, located in a context: performances not propositions, deeds not definitions, actions not analyses. Thus, we abandon the quest for analysis and replace it with something completely different; we change the form of a theory of meaning, and hence the form of a philosophical production. The picture of a unified but composite sense, capable of analysis, is dropped in favor of the idea of a multi-faceted linguistic practice, susceptible only to open-ended description not scientific dissection. No more talk of combinations of objects, hidden forms, final analyses, logical pictures, parts and wholes, simples and complexes. Meaning is not a structured whole, internally articulated, but a pattern of use distributed in time and place, a multiplicity of practical moments. Wittgenstein doesn’t say much of a systematic nature about this pattern of use, but it is possible to reconstruct the outlines of what he had in mind (or what readers have taken away from his text). First, we have the antecedents to a particular occasion of use: what led up to the occasion, particularly the training the speaker received, but also what was going on that caused the speaker to speak as he did (what we might think of as the historical context of the occasion of use). Second, we have the consequences of the use in question—its effect on hearers, the way it changed the state of things (to put it as vaguely as necessary). The use has a future as well as a past. For example, the act of speech might lead to a belief formed or an action carried out or a reply made. Third, we have the criteria of assertion for the sentence involved (if it is an assertoric sentence): what made the speaker think his utterance was warranted, the evidence he had at his disposal. Fourth, we have the purpose of the act—what the speaker was aiming to achieve, the function of his words. He might be trying to induce a belief or getting his interlocuter to act in a certain way (“Please shut the door!”). So, the pattern of use includes historical context, future developments, criteria, and purpose—a motley collection of fragments quite unlike a classical proposition or thought (in Frege’s sense). As an example, I will cite a certain kind of speech act occurring during a tennis match, which puts the use theory in the best light. At a certain point the umpire says “Fifteen all”: what was the pattern of use he exemplified? The past context includes the fact that a tennis match is underway and this match led him to make his utterance—his past and present surroundings form the context of utterance. The future consequences include the players continuing to play until the game is won, the server is reversed, and a new game commences. The criteria involve where the ball landed during the point, and hence who won the point. The purpose of the utterance was to indicate the score, with the further purpose of determining who wins the match. This is all rule-governed behavior, an embedding of speech in a non-linguistic context, a customary practice, a form of life, a sequence of choreographed actions. Describing it is nothing like analyzing the momentary meaning of the sentence “Fifteen all”—its sense, propositional content, what passes before the speaker’s mind (if anything). Use is not an inner something like grasping a sense, a kind of intellectual perception; it is a practical ability, a mode of action. Thus, analysis is banished and perspicuous description takes its place. But is this really true? Has analysis been banished? Haven’t we replaced a complex of objects (Tractatus) with a complex of actions (Investigations)? For the use is itself made up of a series of connected elements: it has a kind of composite unity. Specifically, it is composed of the four elements we have identified: history, results, criteria, and purpose. We have analyzed the total use into four parts, all connected. We haven’t changed the method of philosophy, only its objects. We have introduced a new type of complex whole—a pattern of use—and analyzed it into its constituents. It has parts, components, just as a classical proposition has parts, components. Use therefore has a componential analysis. Indeed, these components can be converted into necessary and sufficient conditions: each aspect of total use is necessary and together they are sufficient. The meaning of “Fifteen all” is given via a description of these components of use, suitably linked together: the sentence would not mean what it does (according to the use theory) without the contribution of each component, and the conjunction of them is sufficient for that meaning. What we have here is a classical analysis of a concept—the concept of meaning. We have a whole with parts, capable of analytic breakdown, a composite kind. And how else could meaning be explicated? Certainly not by some sort of unstructured simple event, an unanalyzable deed: meaning is inherently complex, so any account of it must respect this complexity. It is a sophisticated organized human achievement not an unanalyzable atom of semantic goop. Nor could it be just a chaotic collection of unrelated activities: the components of use must be intelligibly related to each other. A language game is a unified entity made up of separable parts; it can therefore be analyzed (what can’t be analyzed?). It is neither random nor indivisible. Thus, linguistic use is a complex assemblage capable of analysis. If we think of use as a card-carrying behaviorist might, we could say that meaning is a matter of a structured sequence of behaviors (stimulus and response) that can be divided into sub-behaviors, which can also be further divided—that is, the relevant behavior has the usual type of analysis. There is really no alternative to analysis, unless you want to go magical and mystical (use as an emanation of spirit with neither parts nor aspects). Pragmatism is as analytical as logical atomism; it just shifts the locus of analysis (from facts to actions). This means that its methodological burdens are much the same as those of other approaches: it has to give an intelligible account of the constitution of whatever it chooses as the intellectual foundation of its theories. It cannot shirk necessary and sufficient conditions, or take refuge in airy slogans (“In the beginning was the deed”). In the case of Wittgenstein, we need to be told what the use is exactly and how it determines a unique meaning for any arbitrary sentence of the language. We need an analysis of use not merely an inarticulate pointing or positing. The whole idea of a non-analytic style of philosophy is really an abnegation of intellectual responsibility (this is no doubt why it appeals to a certain kind of mind).
 For the dedicated conceptual analyst, an adequate use theory of meaning would assume the following form: A sentence S has a (meaning-conferring) use if and only if (a) S has an appropriate history, (b) S has certain kinds of result, (c) there are criteria for the assertion of S, and (d) there is a purpose to uttering S. It would then be necessary to spell out these conditions in more detail to avoid counterexamples and circularities (as well as intolerable vagueness). None of this would be easy, but it is what theory demands. This would be good old-fashioned conceptual analysis. Then we would need to extend the theory from sentences to words—and good luck with that! A use theory will face all the challenges of any philosophical theory.