DEEP READ · 54
Philosophische Untersuchungen · Ludwig Wittgenstein · 1953 (posthumous)
You carry a default picture of how language works: behind every word stands a thing, the word is that thing's name, the meaning is the thing named—and "understanding a sentence" is some inner event in which you lay hold of it. Wittgenstein's claim is not that this picture is false but that it takes one small corner of language for the whole of it—after which most of philosophy's hardest perennials (what is meaning, what is understanding, how do I know you are in pain, what makes a rule bind the next step) turn out not to be deep questions awaiting answers but illusions extorted by the picture itself. Meaning does not sit behind the word; it sits in front of it, in what people do with it and in the life they do it in. So philosophy's job is not to discover new facts or erect a new system but to cure: to prise apart the picture that has us in its grip, and show the fly the way out of the fly-bottle.
Wittgenstein (1889–1951) came from one of Vienna's richest families, trained as an aeronautical engineer, and at twenty-two turned up in Cambridge with a head full of logic and a claim on Russell's time. He published exactly one philosophy book in his lifetime—the Tractatus Logico-Philosophicus (1921)—which announced that philosophy's problems had been solved once and for all, whereupon he left philosophy and became a village schoolmaster in rural Austria. He returned to Cambridge in 1929 and spent the next twenty years dismantling himself. The Investigations had a preface by 1945, but he never let it go to press; students edited it into print two years after his death. The history of philosophy contains few episodes like this: a man builds one of the twentieth century's most influential systems, then takes it apart himself—and the demolition proves more influential than the building. It is the second half of analytic philosophy's "linguistic turn," and the common source of ordinary-language philosophy, later philosophy of mind, and a good deal of social theory besides.
Six hundred–odd scattered remarks circle three things:
Section 1 opens not with a logician but with a passage from Augustine's Confessions, recalling how he learned to speak: adults pointed at a thing, made a sound, and the child grasped that the sound was that thing's name. Wittgenstein's comment: this gives us a particular picture of the essence of human language—every word is a name, every name has an object, and a sentence is a combination of such names.
Why is that picture toxic? Because it is so natural that you cannot see it as a picture at all; it feels like language itself. And it fits only a narrow strip of language—table, apple, red. Move it anywhere else and strange things start happening. What object does "five" stand for? Or "not," "probably," "pain," "good," "tomorrow"? To save the picture, philosophers are forced to invent peculiar entities to serve as the referents: numbers become abstract objects, pain becomes a private mental object, meaning becomes a something hovering behind the word. A great many philosophical predicaments are extorted from us this way, by an unexamined picture. Hence the often-quoted remark (§115, paraphrased): a picture held us captive, and we could not get outside it, because it lay in our language, and language seemed to repeat it to us inexorably.
Casualty of the same demolition: the supposed foundational role of ostensive definition—teaching a word by pointing at what it names. Point at a pen and say "this is called tove": how does your pupil know you mean this pen rather than its colour, its shape, its material, or the number one? The gesture cannot by itself fix which dimension it picks out. Pointing only works once the learner already knows roughly what slot the word occupies—whether we are naming a colour or a shape. Ostensive definition, then, is not language's bedrock; it already has to stand on language's floor.
He offers a replacement image. Imagine a stripped-down language (§2): a builder A calls out "Slab!", "Pillar!", "Block!", "Beam!", and his assistant B brings the corresponding stone. Are these four words names? You could say so, but they behave more like four calls—their meaning simply is the whole trained routine of shouting and fetching. Wittgenstein calls such a whole, "the language and the activities into which it is woven," a language-game (Sprachspiel).
The word "game" is chosen deliberately: games have rules and moves, they require players acting together, and they mean nothing away from the occasions on which they are played. He once reeled off a long list of language-games (§23)—giving orders, describing an object, reporting an event, making up a story, acting a play, guessing riddles, cracking a joke, translating, requesting, thanking, cursing, greeting, praying—and then said, in effect: compare that variety with what logicians have told you about the structure of language.
Underneath sits form of life (Lebensform). A language-game is not a free-floating rule system; it grows out of a shared way of living—out of bodily reactions, training, institutions. "To imagine a language means to imagine a form of life" (§19). Which is why he wrote (Part II, paraphrased) that if a lion could talk, we could not understand him—not a translation problem, but a life problem: too little of his doing overlaps with ours for us to get a purchase on what he is up to. That sentence is the book's method in miniature: to grasp an utterance, first look at the activity it is embedded in.
This is the book's most famous line and its most misquoted. The original carries a qualifier worth keeping (§43, paraphrased): "For a large class of cases—though not for all—in which we employ the word 'meaning,' it can be explained thus: the meaning of a word is its use in the language." He is not offering a new definition of meaning; he is moving your attention from what stands behind the word to what happens in front of it.
Two analogies do the work. The toolbox (§11): hammer, pliers, saw, screwdriver, rule, glue-pot, glue, nails, screws. All are "tools," and their functions have nothing whatever in common. Words are the same: "table," "five," "not," "ouch" are all words, but forcing them into one format—all names, all standing for something—is like concluding from the fact that tools share a box that they are all for hitting things. The locomotive cab (§12): a row of handles that look much alike—one is a brake, one a valve, one a crank. Uniformity of appearance conceals total difference of function. The uniform look of language (all nouns, all able to follow "this is a…") deceives us in exactly the same way.
The practical upshot is sharp: next time you meet an argument about what X really is—freedom, art, intelligence, fairness—stop hunting for the definition and look at what the word is actually made to do in live situations. A surprising number of apparently metaphysical disagreements dissolve at that step: the two sides were playing different language-games while believing they disputed one fact.
But surely the uses of one word share something? Wittgenstein takes "game" itself as his specimen (§66): "Don't say: 'They must have something in common, or they would not be called games'—but look and see whether there is anything common to all." Board games are competitive and have winners; a child bouncing a ball against a wall has neither winning nor losing; solitaire is one person against the deck; playing house has no fixed rules at all; the Olympics mix contest and spectacle. What you find is a network of overlapping and criss-crossing similarities: this cluster shares winning and losing, that one shares luck, another only shares passing the time—and no single feature runs through all of them.
His name for this is family resemblance (Familienähnlichkeit): lay out a family's photographs and you recognise them instantly as one family, yet you cannot name a feature all of them have—the nose links father and daughter, the chin links the brothers, the expression links grandmother and grandson. And he offers a better image still (§67): "in spinning a thread we twist fibre on fibre. And the strength of the thread does not reside in the fact that some one fibre runs through its whole length, but in the overlapping of many fibres."
Schematic: each "fibre" covers only part of the set. No fibre runs the whole length, and the thread holds anyway.
Then comes a crucial question (§71, paraphrased): is a blurred photograph not a picture of a person at all? Is replacing an indistinct picture with a sharp one always an improvement? Often the indistinct one is exactly what we need. The blade here is aimed at essentialism—the assumption that every concept must rest on a set of necessary and sufficient conditions. Fuzzy boundaries are not a defect awaiting repair; they are often the reason the concept can do its job—"somewhere over that way" beats exact coordinates for most directions actually given. The idea was later borrowed wholesale: cognitive psychology's prototype theory (Eleanor Rosch's finding that people classify by typical examples rather than definitions) is family resemblance in the laboratory, and every fight about what counts as a "vehicle" in law or as "intelligence" in AI falls into the same pit.
The sharpest and most contested stretch of the book (§143–242). Teach a pupil the series "add 2": 2, 4, 6, 8… He goes fine up to 1000, then writes 1004, 1008, 1012. You say he has got it wrong; he is genuinely baffled: I did just what you taught me—add 2 up to 1000, 4 up to 2000. How do you prove him wrong? You want to say "add 2 just means add 2"—but his procedure does accord with every example you actually gave.
Hence the paradox (§201, paraphrased): "No course of action could be determined by a rule, because every course of action can be made out to accord with the rule." Why? Because you can only explain a rule with another rule or another sentence, and that explanation stands in need of explanation in turn—an infinite regress of interpretations. A rule is like a signpost, and a signpost cannot compel you to walk the way the arrow points; "follow the arrow" is not itself written on the arrow.
His resolution is not a deeper rule but a change of foundation, from interpretation to practice (§202, paraphrased): obeying a rule is a practice; and to think one is obeying a rule is not to obey a rule—hence it is not possible to obey a rule "privately." The reason you write 1002 after 1000 is not that you deduced something inwardly; it is that you were trained that way, that your community does it that way, and that the agreement in doing is uncanny in its stability. Push for justifications far enough and you hit hard ground (§217, paraphrased): "Once I have exhausted the justifications I have reached bedrock, and my spade is turned. Then I am inclined to say: this is simply what I do."
Why does this deserve so much ink? Because it relocates necessity itself. We assume the compulsion in mathematics, logic and meaning comes from something transcendent—Platonic numbers, an algorithm in the head, the essence of language. Wittgenstein's answer is that it comes from a trained, frighteningly stable collective practice. That is not relativism (see the criticisms below), but it does close the metaphysical strongbox we assumed was there. In daily life: a statute cannot govern its own interpretation; a company's stated values are defined by how they are actually enforced; a specification is never finished—not because it wasn't written in enough detail, but because no rule comes with its own application attached.
He then turns the same blade inward (§243 ff.). A private language is not a code or an undeciphered diary; it is a language in principle intelligible only to its speaker, its words referring directly to his own immediate sensations, to which no one else has access. Most people's first reaction is: what's the problem? Of course I can name my own sensations.
His counter is the diary experiment (§258). You resolve to call a recurring sensation "S" and to write S in a diary whenever it turns up. Now: on what basis are you sure today's sensation is the same as last time's? Your only evidence is that you remember it is. But nothing independent exists against which the memory can be checked—it is like buying a second copy of the same newspaper to confirm what the first one said (§265, paraphrased). Hence (§258, paraphrased): "Whatever is going to seem right to me is right. And that only means that here we can't talk about 'right.'" A word that cannot distinguish being used correctly from seeming to be used correctly is not yet a word.
Better still is the beetle in the box (§293). Suppose everyone has a box with something in it that we all call a "beetle." Nobody can look into anyone else's box, and each person says he knows what a beetle is only from looking at his own. Then the contents could differ from person to person; someone's box could be empty; the thing inside could be constantly changing—and as long as the conversation runs on as before, what is in the box makes no difference at all to the use of the word "beetle." The conclusion (paraphrased): if we construe the grammar of sensation-words on the model of object-and-name, the object drops out of consideration as irrelevant.
Crucially, this is not behaviourism, and he heads off the charge himself with a piece of self-mocking dialogue (§304, paraphrased): "It is not a something, but not a nothing either." He does not deny that you are in pain; he denies that the meaning of "pain" consists in referring to a private object. What holds it up instead is public criteria: a child hurts himself and cries, and adults teach him to replace the crying with the word—the verbal expression of pain does not describe the crying, it takes its place (§244, paraphrased). The pain language-game is built on natural human expression—wincing, crying out, guarding the injury—which is why it can be taught, understood and used at all. From which the flat verdict (§580, paraphrased): "An 'inner process' stands in need of outward criteria."
The same blade cuts "understanding." We take "now I've got it" to be some special inner process or state: a flash, a picture rising, a mechanism engaging. Check honestly and you find that nothing fixed need happen at all—and even if something did, that would not be your warrant: what settles whether you have understood "add 2" is whether you go on correctly, not what flickered through your head. Hence his advice (§154, paraphrased): try not to think of understanding as a mental process at all. "Now I understand" behaves more like a signal or an announcement, redeemed by what you do next.
The most famous stretch of Part II (retitled Philosophy of Psychology—A Fragment in current editions) is the duck-rabbit: a line drawing you can see as a duck or as a rabbit, borrowed from the psychologist Joseph Jastrow. Not one line changes, your retinal input does not change, and yet you suddenly see it as something else. He calls this the dawning of an aspect (das Aufleuchten eines Aspekts) and asks what exactly has altered. Saying "what I see has changed" is wrong—the lines did not move. Saying "only my thought about it changed" is also wrong—this is genuine seeing. "Seeing-as" lodges precisely between our categories of sensation and thought, which shows that the underlying picture—raw material received first, interpretation added afterwards by the mind—is itself broken.
Why place it last? Because it replays the whole book's method at the level of perception: meaning is not a layer stuck on the back of the material, but a matter of what you can do with it and which activity you can fit it into. He adds a thought experiment—meaning-blindness: imagine someone otherwise normal who cannot feel "the meaning of a word" as a kind of flavour (the switch you feel when you turn a pun over between its two senses). What is he missing? Most of his language use is untouched—which is the point: the inner experiences you took to constitute understanding are its accompaniment, not its substance. And in ordinary life, seeing-as explains a great deal: the same dataset read as conspiracy by one side and coincidence by the other; the same remark heard as care by one person and control by another. Many disputes are not at the level of facts but at the level of what it is seen as—and that level cannot be settled by piling on more facts.
So what is philosophy for? His answer remains radical enough to make philosophers uncomfortable. Philosophical problems arise, he says, "when language goes on holiday" (§38, paraphrased)—when words are lifted out of the jobs they normally do and left, like an engine idling (§132), roaring away while driving nothing. "What is time?" causes no trouble while you are catching a train; ask it apart from all use and it becomes a riddle for the ages. Hence (§109, paraphrased): "Philosophy is a battle against the bewitchment of our intelligence by means of language." And its results (§119, paraphrased) are the uncovering of one or another piece of plain nonsense, and the bumps the understanding has got by running its head against the limits of language.
His method is therefore description, not explanation (§126, paraphrased: philosophy simply puts everything before us, and neither explains nor deduces anything). "Philosophy may in no way interfere with the actual use of language… It leaves everything as it is" (§124, paraphrased). The difficulty is not that things lie too deep but that they lie too close (§129, paraphrased): the aspects of things most important to us are hidden by their simplicity and familiarity. Nor does he think there is a Wittgensteinian Method (§133, paraphrased): there is not one philosophical method, though there are indeed methods, like different therapies. The whole posture compresses into one exchange (§309): "What is your aim in philosophy?—To show the fly the way out of the fly-bottle." A fly-bottle is a glass trap whose exit is where the fly came in, while the fly keeps hurling itself at the lit glass. What traps it is not lack of strength or cleverness; it is that all its effort points the wrong way. The philosopher, on this view, is that fly—and the job is not to give it stronger wings but to get it to turn around.
Six hundred remarks that look like scattered notes are one argument taken all the way down: a picture → the four illusions it produces → a replacement way of looking → a reassessment of what philosophy is.
So what does the book actually establish? Not a theory of language—he explicitly refuses theories. It establishes a reversal of method: when you meet a philosophical perplexity of the form "what is the essence of X," do not dig downward first; look sideways at how the word earns its living. Most of the perplexity evaporates at that step, and what survives is the real question.
Read the Investigations alone and you miss two things: what it is demolishing, and the step it did not live to finish.
Tractatus Logico-Philosophicus (1921). This is the first half of his life. Its core is the picture theory: a proposition can mean anything at all because it shares a "logical form" with the fact it depicts, the way a picture corresponds to a scene ("the world is the totality of facts, not of things"). It follows that every meaningful sentence is an empirical proposition capable of truth or falsity; logical propositions are empty tautologies; and ethics, aesthetics, the meaning of life—and philosophy itself—overstep the limits of language and can only be shown, never said. It closes with the famous line: "Whereof one cannot speak, thereof one must be silent." He also concedes that his own propositions are ultimately nonsensical, and that the reader must throw away the ladder after climbing it. Every punch in the Investigations lands on this system: language has no single essence, no logical skeleton hidden beneath the everyday, and meaning rests on use rather than correspondence. Without the earlier book, the later one looks like a man fighting the air.
The Blue and Brown Books (dictated 1933–35, published 1958). Lecture dictations from the transitional years, where "language-game" and "family resemblance" first take shape—far more plainly written than the Investigations, which is why many people use them as the ladder in.
On Certainty (1969). Notes from his last eighteen months, prompted by G. E. Moore's famous refutation of scepticism—holding up a hand and saying "here is one hand." Wittgenstein accepts neither the sceptic nor Moore's defence, and proposes hinge propositions: sentences like "the world has existed for a long time," "I have two hands," "the earth was not created the year I was born" are neither things we know nor things we doubt, but the pivots on which all doubting and testing turn—a door turns only because the hinges stay put. "Doubt itself rests on what is beyond doubt" (paraphrased). This carries the Investigations' method into epistemology: the ground of certainty is not harder evidence but the whole inherited world-picture we act from—learned, handed down, and not questioned in practice. To see the full shape of the man, this one is required.
Remarks on the Foundations of Mathematics (1956). The rule-following argument pushed into mathematics: a proof does not discover a truth already standing in a Platonic realm; it institutes a new rule of use. This is the part professional philosophers of mathematics resisted hardest—and the part he regarded as among his most important.
1. You carry a default picture—word as name, meaning as the thing named, understanding as an inner grasp of it. The picture is not false; it just takes one corner of language for all of it, and half of philosophy's classic problems are extorted from us by it.
2. "For a large class of cases—though not for all—the meaning of a word is its use in the language" (§43, paraphrased). Stop asking what stands behind the word; look at what happens in front of it.
3. Language is not a picture of the world but a mass of language-games: ordering, complaining, guessing riddles, thanking, praying. Like hammer and glue-pot in one box, they look like one kind of thing and share no function. "To imagine a language means to imagine a form of life."
4. "Don't say they must have something in common—look and see." Concepts hold together by family resemblance: as in spinning a thread, the strength lies not in one fibre running the whole length but in many overlapping. Blurred edges are not a defect to be repaired; often they are exactly what we need.
5. The rule-following paradox: any course of action can be made to accord with the rule, because every explanation needs a further explanation. A rule is a signpost, and "follow the arrow" is not written on the arrow.
6. The way out is not a deeper rule but practice: "Obeying a rule is a practice; and to think one is obeying a rule is not to obey a rule—hence one cannot obey a rule privately." Where justifications run out, "my spade is turned… this is simply what I do."
7. A private language is impossible: write "S" for a sensation only you can know, and nothing distinguishes remembering rightly from seeming to—"whatever seems right to me is right, and that only means we can't talk about 'right' here."
8. The beetle in the box: everyone has a box no one else can see into; the contents could differ, or be nothing at all, and the word's use would be unaffected. He is not denying your pain; he is denying that "pain" means by pointing at a private object—"an inner process stands in need of outward criteria."
9. The duck-rabbit: not a line has moved, yet you see it as something else—meaning is not a coating behind the material but the activity you fit it into. Many disputes are not about the facts but about what the facts are seen as.
10. Philosophical problems arise when language goes on holiday and the engine idles; philosophy is "a battle against the bewitchment of our intelligence by means of language," it explains nothing and leaves everything as it is. "What is your aim in philosophy?—To show the fly the way out of the fly-bottle." What traps the fly was never a shortage of strength, but a lifetime of pushing the wrong way.