Against Family Resemblance
After the well-known section on games in Philosophical Investigations(section 66), Wittgenstein writes: “I can think of no better expression to characterize these similarities than ‘family resemblances’; for the various resemblances between members of a family: build, features, color of eyes, gait, temperament, etc. etc. overlap and criss-cross in the same way. And I shall say: ‘games’ form a family.”(67) He then goes on to suggest that the kinds of numbers “form a family in the same way”. He gives no further examples, though he employs the concept to characterize language: “I am saying that these phenomena [linguistic phenomena] have no one thing in common which makes us use the same word for all, but that they are relatedto one another in many different ways.” (65) There is not much to go on here, given how important the notion of family resemblance is to Wittgenstein’s later philosophy, but I think we have enough to raise some serious objections to what he says. I am going to argue that the concepts in question are not family resemblance concepts, and indeed that there are not and cannot be any such concepts. The whole idea is a mistake—and not for any particularly profound reasons. Wittgenstein’s laconic remarks are full of errors, confusions, and non-sequiturs.
The first point I want to make is that the concept of a family resemblance concept is not itself a family resemblance concept. Consider the class of family resemblance concepts—say, game, number, languageand art(not mentioned by Wittgenstein): do they have any common feature? Yes, they are all such as not to be defined by a common feature but by a series of overlapping similarities—that is the common feature they all share. Wittgenstein has given us a general definition of the term “family resemblance concept” such that anything denoted by it satisfies that definition; it is not just that the concepts are similar in certain ways, as games are said to be. So we know there is at least one non-family resemblance concept, viz. family resemblance concept. This is a bit ironic and certainly not noted by Wittgenstein: it should make us wonder about the generality of the idea. It renders contradictory the claim (not made by Wittgenstein) that all concepts are family resemblance concepts, since thatconcept isn’t a family resemblance concept. It also raises the question of how widespread family resemblance is according to Wittgenstein, and why it has the extension it does. Does he think it is limited to the examples he mentions? And what is it about these examples that requires the use of family resemblance concepts instead of common feature concepts? It would be odd if such a heterogeneous group were the only instances of the phenomenon, with all other concepts proudly possessing uniting common features. The notion doesn’t seem to have been systematically thought out and looks suspiciously ad hoc.
Now consider what is meant by “family resemblance” (elsewhere Wittgenstein speaks of “family likeness” and we could also say “family similarity”). This is not the same notion as that of belonging to the same family: that is a matter of parentage and genetic transmission, and is presumably not counted by Wittgenstein as a family resemblance concept. He is talking about various observable traits typically shared by members of the same biological family, as his list suggests. He doesn’t confine the notion to visual appearances, since he includes temperament, and he could have included aptitudes, intelligence, religious beliefs, etc. There are a great many respects of similarity between human beings. Nor does the notion coincide with membership in a biological family: some members of a family are not similar to any other members, and some people from outside the family look just like people in it. The class of people who are notably similar to members of a given family is distinct from the class of people actually making up the family. These are really completely different concepts. In fact, Wittgenstein need not have invoked familyresemblance at all; he could have just spoken about similarity in general. Take three cars, a,b, and c, where aand bhave the same color but are not the same model, while band care the same model but not the same color. Car a is similar to car bin respect of color but not similar to car cin that respect, whileband care similar in respect of model but not color.Thus we see the non-transitivity of similarity: ais similar to band bis similar to c, but ais not similar to c. The same could be true of individuals exhibiting family resemblance. So Wittgenstein is really drawing attention to the way the concept of similarity works—not a very startling insight. The question is whether similarity relations can ground a unified concept.
Notice that there is no unified (non-disjunctive) concept that corresponds to the list Wittgenstein offers. What concept do we have that expresses likeness of eyes, mouth, gait, and temperament? We might try to manufacture a concept, say Smith-ish, to characterize a certain family’s appearance, where not all members of the Smith family are Smith-ish and some non-members are Smith-ish: but this a pretty feeble concept with little internal unity. And the reason for that is that mere similarity in a respect is no basis for a concept, since everythingis similar to a given thing in somerespect. The class is simply too heterogeneous to be worth picking out. This is not so for the concept of a family, which is far more constrained; and one can’t help suspecting that Wittgenstein’s choice of familysimilarity illicitly trades on this other source of conceptual unity. For familyis a unified (common feature) concept while family resemblanceis not—it could include members of (say) my family as well as assorted people distributed across the globe who look or behave like me in some respect or other. To look like someone that looks like Winston Churchill is no basis for a usable concept (it includes certain breeds of dog or even clouds). That is why there simply is no concept that corresponds to Wittgenstein’s list. It amounts to an uninteresting disjunction: xlooks like a member of family Fif and only if xeither has the eyes of For the nose of For the mouth or the gait or the temperament or the size or the color or the religious beliefs of F, etc.
Then what are we to say about Wittgenstein’s alleged examples—aren’t theyexisting unified concepts that are characterized merely by non-transitive similarity? The question is whether games are linked by nothing butfamily resemblance: is there really nothing they have in common? Two points may be made. First, Bernard Suits’ analysis of the concept of a game refutes this contention: a game is a rule-governed activity in which the player voluntarily chooses an inefficient means to achieve the goal of the game. I won’t go into this here; it has been amply discussed elsewhere.What I will say is that the existence of such an analysis is entirely predictable, given the unity of the concept of a game and the complete lack of unity exhibited by the pattern of similarities and dissimilarities to which Wittgenstein draws attention. Second, why can’t we say that what is in common to all games is that they are all games? That is, we treat the concept of a game as primitive and unanalyzable (we might add that all games are played). Consider the concept of a line: there are all sorts of lines—long, short, squiggly, curved, straight, open, closed, etc.—and it is hard to find one such feature that all lines share. So is the concept of a line a family resemblance concept? Why not say—what seems obvious—that the concept of a line is primitive and cannot be explained in other terms? Wittgenstein appears to be presupposing that if a concept isn’t analyzable it must be a family resemblance concept—he conveniently forgets about the possibility that it is indefinable. Compare color and shape concepts: here too we have a great variety of things that fall under these concepts and no possibility of unifying them by citing a specific color or shape. So what? Why not accept that the concepts colorand shapehave no non-circular definition, yet apply to very heterogeneous extensions? The same goes for animalor particle: these come in great variety too and it is hard to define the concepts in non-circular fashion—but why leap to the idea of family resemblance instead of accepting indefinability? And notice that Wittgenstein never cites such examples, presumably because it is obvious that the indefinability response is plausible. So we are certainly not compelled to accept the family resemblance theory—even if it were coherent—in order to handle the case of games. In fact, the concept of a game has sources of unity quite other than that (dubiously) provided by family resemblance.
The case of numbers is even more glaring, and it seems distinctly odd for Wittgenstein to choose this example to illustrate his thesis. What do cardinal numbers, rational numbers, and real numbers have in common? Well, we could just say they are all numbersand admit that the concept cannot be defined; or we could note that all numbers are subject to mathematical operations such as addition, subtraction, and division–or that they come in different sizes, belong in number series, and can be used for measuring and counting. There is actually a lotthey have in common. More adventurously, we might also add that they all admit of set-theoretic construction. You might as well say that organisms have nothing in common or chemical compounds or trees or houses. True, there is a lot of variety in those classes, but that doesn’t preclude a common feature—either expressed by the concept in question or by some analysis of it. Wittgenstein never makes a convincing case for the thesis that any of these concepts mustbe treated as family resemblance concepts. Nor can such a treatment confer the kind of unity possessed by these concepts. How, say, can Wittgenstein explain the fact that golf is a game but hitting stones with sticks to clear a patch of land is not? There is an obvious similarity between these activities, yet one is clearly a game and the other is definitely not.Similarity in some respect is a hopelessly weak relation to ground a concept, because it is so cheaply obtained. A different principle of grouping is needed.
What about language? There are certainly different dimensions of similarity among words: words can sound alike but not mean alike and vice versa, for example. But is there really nothing they have in common? Don’t they all contribute to forming sentences, and don’t all sentences have meaning and grammar? By all means let’s recognize the variety of linguistic forms, but why deny that anything can be said more generally? Aren’t all words and sentences usable in acts of communication? Isn’t human language a species-specific biologically given trait with infinite potential? There is plenty we can say generally about language. True, it is a mistake to suppose that every sentence expresses a proposition, as the Tractatusclaimed, but there is no need to go to the other extreme and declare that there is no common feature at all to language. Moreover, many things that don’t belong to language are similar to things that do, so similarity alone cannot be grounds for inclusion under the concept. The screeching of tires could sound like a cry for help, but it isn’t a part of language. The humming of a bee’s wings is not part of its language, though its dances are; and our dances are similar to theirs but not part of our language. Mere similarity to a paradigm in some respect is nothing to the point.
In sum: there are no good examples of family resemblance concepts; the notion is incoherent; and the concept itself is not an instance of itself.
It is notable that when Wittgenstein is trying to explain his notion of a family resemblance concept he falls back on metaphors in order to capture conceptual unity, specifically the metaphor of a thread made up of overlapping fibers. But this does nothing to render his theory intelligible, since it is just a metaphor.