Version classiqueVersion mobile

Penser l’histoire des savoirs linguistiques

Sylvie Archaimbault
Jean-Marie Fournier
Valérie Raby

Première partie. Histoire, épistémologie, langage

Language in its own image: on epilinguistic and metalinguistic knowledge

Talbot J. Taylor

Texte intégral

1In the introduction to his magisterial 2-volume Histoire des idées linguistiques, Sylvain Auroux raises an issue of prime importance to the history of linguistic thought: the relationship between epilinguistic and metalinguistic knowledge (see also Culioli 1968, Gombert 1992).

Le savoir linguistique est multiple et il débute naturellement dans la conscience de l’homme parlant. Il est épilinguistique, non posé pour soi dans la représentation, avant d’être métalinguistique, c’est-à-dire représenté, construit et manipulé en tant que tel à l’aide d’un métalangage (éléments autonymes et noms pour les signes). La continuité entre l’épilinguistique et le métalinguistique peut être comparée avec la continuité entre la perception et la représentation physique dans les sciences de la nature. (Auroux 1989, p. 18)

2The relationship between the epilinguistic and the metalinguistic raises a central issue in our understanding of the nature of linguistic knowledge or competence. Indeed, the importance of this issue is suggested by Auroux only a few lines below the passage quoted above: “Dans le domaine proprement grammatical, encore aujourd’hui, il n’y a pas toujours véritable solution de continuité, peut-être parce que le langage est un système régulé par sa propre image” (p. 18–emphasis added).

3Reflection on the implications of this suggestion–that language is constructed and regulated by its own reflexive image–must, I believe, lead us to conclude that the relationship between epilinguistic and metalinguistic knowledge is more complicated than might at first be imagined. To see why this is so, I propose to work through a thought experiment. As a simple, surveyable template for this thought experiment, I will use the builder’s language game described in § 2 of Wittgenstein’s Philosophical Investigations (Wittgenstein 1953). In this language game, the builder’s assistant, B, brings one of four different types of building material to the builder, A, whenever A utters one of the four signs beam, slab, block, or pillar. Wittgenstein says that this language game should be thought of “as a complete primitive language” (§ 2). Note, in particular, that no metalinguistic–or, more generally, reflexive–utterances are possible, for the only signs in this language are beam, slab, block, and pillar. We will not inquire into how this language game came into being, but simply take it for granted, just as Wittgenstein describes it. Each of A’s possible utterances is apparently a straightforward sign for the building material in question: that is, the sign slab means a slab; the sign block means a block; and so on. Alternatively, one might take the sign to mean the idea of the building item in question, or possibly even B’s response of bringing the item to A. However, in the context of the present discussion, it is irrelevant which of these albeit very different “meanings” the builder’s signs are taken to have.

4Now, in Wittgenstein’s description, the practice works well. But what happens if on one occasion, when the builder A produces an utterance, B, the assistant, does not know whether she said “block” or “beam”? What can B do? There are no metalinguistic signs in the language game with which B might initiate a repair, asking A to repeat what she said, or asking whether she said “block” or “beam”, or even just saying that he did not understand what she said. Nor can he simply say “You said, ‘block’, didn’t you?” For, ex hypothesi, the only utterances possible in Wittgenstein’s language game are “block”, “slab”, “beam”, and “pillar”. Or consider another possibility: what happens if, although he heard A’s utterance clearly, B is still uncertain whether A is asking for a particular beam–just this ⇾ à beam here on the left–or whether she is asking for any old beam? There are no signs in Wittgenstein’s language game by which the speaker might be requested to specify her referent. That is, despite his uncertainty, B cannot ask something like “Which beam do you mean?” or “Do you mean this beam here or just any one?”

5In considering these communicational dilemmas, we need to keep in mind that this thought experiment assumes that there are no alternative means by which B or A can express such metalinguistic utterances–by hand gestures, by specific prosodic forms, by facial expressions, by nonverbal grunts, and the like. For if such non- or para-linguistic signs were available to them, that would in effect be to expand the language game beyond Wittgenstein’s description of it as “complete”. Again, our thought experiment involves a language game in which no metalinguistic remarks are possible; in which case it is no good sneaking them in the back door by allowing their expression to be nonverbal.

6Now, suppose that, on one occasion, B notices that A’s pronunciation of “beam” holds the vowel for longer than he had been used to hearing, and this leads him to wonder if she means something different by pronunciation? That is, A says [bi::m], instead of the more familiar [bim]. The assistant cannot ask “Does [bi::m] mean something different from [bim]?” or “Is a [bi::m] the same as a [bim]?” Or what if B does not know if A is talking to him or to another worker standing nearby? What can B do? Since the language game has no metalinguistic capacity, he cannot say what one would ordinarily say in such circumstances: that is, ask a question like “Are you talking to me or him?” or “Is it me you are asking?” Or what if, for whatever reason, B is not sure if A is serious or sincere? Perhaps B suspects that A is only pretending to ask for a beam? Still, there is no way to express this uncertainty in the language game and no way to ask A to clarify. Whereas in ordinary circumstances one might say “You’re just joking, right?” or “You don’t really mean that, do you?” or “Is it true that you want the beam?”, this metalinguistic move is not available in Wittgenstein’s language game. In sum, it would seem that no explicit means are available to B by which he can clear up these types of communicational confusion. All he can do, in such circumstances of uncertain knowledge, is to stand unresponsively or make an arbitrary choice of one of the possibilities that occur to him.

7So far we have only been considering how the lack a metalinguistic capacity limits the communicational flexibility of B, the hearer. But what about A, the speaker? What, for instance, can A do if she utters “block” but B brings her a slab? In the same situation, one might say “No, I said ‘block’, not ‘slab’” or “You misunderstand. This ⇾ is what a ‘block’ is”. But these metalinguistic remarks are not possible in Wittgenstein’s language game. Similarly, what if A does not know whether B understood her utterance? She cannot say “I’ll say that again” or even simply “Do you understand?”, as we might say in the same situation. She can repeat herself, of course, but what she cannot do is to verbally identify what she is doing as “repeating” herself or as “saying it again”, and thereby distinguish this repetition of her first command from the issuing of a second command. In other words, like B, A’s hands are tied. There is nothing A can do if she is unhappy with B’s response to her utterance or if she is not certain whether B understood what she said, whether B was listening, whether B knows which beam she is referring to, whether B knows that [bim] and [b::im] mean the same thing, whether B realizes that she was only joking, and so on. The lack of metalinguistic tools in the language game drastically limits the speaker’s possibilities no less than it does the hearer’s, even for the extremely limited interactional goals which Wittgenstein’s language game is intended to serve.

8Another equally important consequence of this lack of metalinguistic possibilities is that the normative or conventional character of the language game would be limited: limited, that is, to the point of disappearance. Because A cannot speak reflexively, she cannot say things like “No, this is what a beam is” or “When I say ‘slab’, you should bring me one of these” or even “No, that’s wrong” or “Yes, that’s right”. In other words, without metalinguistic signs, if B makes a mistake, A has no way of correcting him or pointing out what he ought to have done or reminding him of what he should know about the word. She cannot even identify his mistake as a mistake.

9Consequently, there is no way for A to hold B explicitly responsible for how he responds to her utterances, for there are no metalinguistic means for expressing this responsibility. At the same time, neither can B hold A responsible for what she said. Suppose that A says “beam”, which B promptly brings to her, but A looks unhappy. B cannot respond to the effect “But you told me to bring a beam and this is a beam!” In other words, even though Wittgenstein refers to the builders’ activity as a language game, still, without the possibility of metalinguistic discourse, it cannot be anything like what we call a “game”: because it cannot be normative.

10Thus far, I have been considering how discourse in Wittgenstein’s builder’s language is affected by its lack of a metalinguistic capacity–how the practices of playing this language game are affected. But what about the individual signs themselves? On first glance, the fact that metalinguistic discourse is not possible in the builder’s language does not seem to have any impact on the properties of the signs themselves. After all, there are still four distinct signs– beam, slab, block, and pillar. The sign slab still means a slab; the sign beam means a beam, and so on. The absence of a metalinguistic capacity appears only to impact discourse with the signs, but not the very signs themselves. This also seems to be the case if we consider the signs from an epistemic perspective, as suggested by the distinction between epilinguistic and metalinguistic knowledge. Do the builder and his assistant know what the signs mean? Do they have epilinguistic knowledge of the signs’ meanings, even though they lack the metalinguistic tools for engaging in discourse about those signs and their meanings? Again, our first inclination is probably to respond that, yes, they do: for instance, they know that the sign beam means that thing over there. Prima facie, it does not seem relevant to the question whether they know its meaning that the content of that knowledge cannot be made explicit in metalinguistic discourse– that is, that there is no way of saying “Beam means that” or anything like it.

11In what follows, I will present two different objects of comparison whose differing analogical relationships to the builder’s language should motivate doubts about the foundational distinction between epilinguistic and metalinguistic knowledge. As the first such object of comparison, imagine the following game. A child is brought up in a non-English-speaking community in which there are no commercial or monetary practices. He undergoes behavioral training in a form of exchange involving candied sweets and four kinds of paper notes–a $1 bill, a $5 bill, a $10 bill, and a $20 bill. We will imagine that the notes have been accidentally dropped from the sky by a passing jetliner. As a result of his training, when the child is handed a $1 note, he passes back a chocolate mint; if he is handed a $5 note (or five $1 notes), he passes back a lemon drop, and so on. But that is all. There is no discourse about the notes, the objects, the practice of buying, or matters of value, being worth, cost, and so on. The game is complete just as I have described it.

12In this exchange game, does a chocolate mint cost $1? Is the value of a lemon drop $5? Is a $5 note worth five times more than a $1 note? What sense would it make for us to characterize in this way the items and the relationships-between-the-items in the exchange game? It seems to make some sense, I admit. But my question is not really about the linguistic propriety of using these English commercial terms in this way. On the contrary, the question is whether the notes and sweets do in fact have the properties attributed to them: that is, the properties of value, of costing such and such, of being money, of comparative worth, and so on. Are these properties somehow immanent in the objects themselves, that is, in the notes and the sweets themselves? Or are they immanent in the actual behaviors produced by the children? Is this the case even though it is impossible in this game ever to talk about value, being worth, costing, money, and so on? Of course, if we decide that it is not the case that a $5 note used in this game has a value, or that a $5 note is not worth more than a $1 note, then we have to conclude that such a $5 note is not the same kind of semiotic object as the five-dollar bill in my pocket. For the one in my pocket does have a value, is worth more than a $1 bill, and so on.

13Let’s examine this example within the epistemic frame. Do the children playing this exchange game know that a chocolate mint costs $1? Or that $5 will buy you a lemon drop? Or that a lemon drop is worth more than a chocolate mint? We may, again, have an inclination to say “Yes, they do”. For affirming that they know these things helps us make sense of the children’s actions in the game. But this is a projection from our own reflexive perspective, is it not? For how could the children themselves be aware of these properties of the culturally-dependent practices of commercial exchange, while never having heard any talk of such practices and their component properties? Would it not be absurd to say that the children somehow automatically, “epi-commercially” know these things about the notes and objects in their game? If so, then we have to acknowledge that what they know about a $5 note is very different indeed from what I know about the one in my pocket.

14There are a few points that this constructed analogy to the builder’s language should bring into the foreground. For, at first glance, it seems to make sense to say that the signs used by the builder and her mate have meanings, refer to particular objects, are semiotically distinct from one another, can be repeated, and so on. But comparison with the children’s exchange game may lead us to doubt whether the signs really do have those properties. No doubt, when the builder says “Block”, the assistant fetches a block and brings it to her. But does this fact necessarily entail that the sign has a meaning or reference, or that it is the same sign she uttered a moment ago? Other explanations are possible for the participants’ actions and responses. We could imagine a dog being conditioned to respond as the assistant does and yet not feel that these responses justify saying of the dog that he understands “the signs’ meanings”. The sign’s possession of those properties is not necessary for the language game to proceed just as Wittgenstein describes it. As regards our initial feeling that the builder’s signs have meanings, refer to things, can be repeated, understood, and so on: at the very least, this commonsense response stands in need of justification.

15Or, to put this puzzle in an epistemic frame: Do the builder and his mate know that the sign beam means a particular type of building material? Do they know that the builder’s utterance of this sign refers to that object over there? Again, our first response may well be to answer “yes” to these questions, for that response helps us to make sense of what happens in the language game. But we have to ask ourselves how the builder or her mate could know these things in the absence of a capacity for metalinguistic discourse: i. e., what knowing these things epilinguistically could possibly consist in? Is it not equally questionable to assume that they automatically and implicitly attribute these semiotic properties to the vocalizations as it is to assume that the children attribute the properties of value, price, being worth, and so on to the notes and objects in the exchange game? From the perspective afforded by this comparison, then, we can see how different the signs in the builder’s language are from the signs exemplified by the words familiar to English speakers: block, beam, slab, and pillar.

16Let us now turn to a second object of comparison–one that may help us see the builder’s signs in yet another light. As an infant learns about a particular perceptual object, she learns what in the ecological psychology initiated by James Gibson is called the object’s “affordances” (1979). Gibson applies the concept of affordances to many kinds of phenomena in the child’s developmental environment. For our purposes, hand tools will serve to illustrate the concept. A child’s learning about a particular tool is seen as inseparable from learning what the tool is “good for”, what can be done with it: that is, what it affords for action. Learning what a saw is involves learning that it is good for sawing. Similarly, you do not yet know what a drill is unless you know that it is good for making cylindrical holes in objects. As the ecological psychologist Ed Reed puts it: “When one learns about the affordances of things one is learning about properties of objects, events, and places with respect to one’s own actions” (Reed 1993, p. 52). Moreover, Gibson argued that cognitive development crucially involves the child’s coming to grasp what things afford for others as well as for herself. Thus, ecological psychologists also apply the concept of affordances to sociocultural objects. Reed explains: “A bat affords hitting a ball by seeing it as a component of a complex [cultural] event in which balls are thrown and swung at” (Reed 1993, p. 53).

17A verbal sign can also be conceptualized in terms of its affordances. Knowing what block means in the builder’s language is a matter of knowing what it is “good for” in that practice. So, the builder knows that block is the word to utter in order to have a block brought to her; that this is what this word affords in the language game. And the assistant knows that block has this affordance for her. Consider the affordances of a sign used in a verbal practice which–unlike the builder’s language–possesses a metalinguistic capacity. Such a sign can be explained; its identity and its meaning can both be queried; its reference can be made more specific (“I mean that block”), as can its addressee (“It’s you I am talking to”); it can be characterized as a repetition or a correction of an earlier sign; its production can be characterized as a particular kind of speech act (“I’m asking you, not ordering you”); the speaker’s purpose in using it can be characterized as of a particular cultural type (“I’m only teasing”); the hearer’s response to it can be characterized as correct or incorrect, a mishearing, a misunderstanding, and so on. These, we might say, are some of the signs’ reflexive affordances (Taylor 2010). A sign used in a practice without metalinguistic capacities, such as the signs in Wittgenstein’s builder’s language, does not have reflexive affordances.

18Furthermore, because the builder’s language lacks the capacity for metalinguistic discourse, its signs cannot have normative properties. There is no way to say things like “That’s wrong” or “This is what you should say”, or “When X occurs, then you ought to do Y”. Metalinguistic remarks of this normative kind are not among a sign’s affordances in the builder’s language. The result is that the signs in this language game are what we might call “frictionless”, a metaphor by which I mean that the properties of a given sign can slide this way or that–freely–with nothing to hold it in place. A may use it or B respond to it arbitrarily, there being no means of applying or manifesting any normative expectations or consequences. In such a language, if Humpty Dumpty decides to use the word glory to mean “a nice knock-down argument”, then there simply is nothing Alice can say in reply, and her objection about whether “one can make words mean whatever one wants” cannot even be expressed. In other words, in ecological psychology’s sense of knowing what an object is “good for”, because the use of a sign in the builder’s language cannot be normatively regulated, it is good for anything whatsoever–which is to say, for nothing at all.

19English, like every other human language, is a verbal practice whose signs do offer all of these reflexive affordances and many, many more. Therefore if, in keeping with this analogy, we picture knowing a sign as involving grasping its reflexive affordances, then what the builder and her assistant know about the sign “block” has to be seen as very different from what English speakers know about the word “block”–or, indeed, from what the speakers of any language know about its words, regardless of their training in literacy or schooling of any kind. To be sure, there is a superficial identity between “block” in English and “block” in the builder’s language game–they both sound the same and there is some association with a kind of building material–but because their reflexive affordances are so different, knowing the English word “block” is a wholly different matter from what the builder and her assistant know–if anything at all.

20This cursory review of two objects of comparison to Wittgenstein’s builders’ language may help us to make sense of the following claim regarding the relationship between metalinguistic discourse and linguistic knowledge. It is only because the signs of human language have reflexive affordances that they are as polished, articulated, useful, and precise as we know them to be: knowledge which we count on in our everyday verbal activities. Signs like those in the builders’ language–signs which, because of the absence of a metalinguistic capacity in the language game, have no reflexive affordances–stand proportionately to the signs of human languages as a sharp stick stands to a laser drill. Language is constructed and regulated by its own reflexive image (Taylor 2003). And yet, if we accept that the identity of a human sign stands in an internal relation to its reflexive affordances and that metalinguistic discourse has an essential and ineliminable role in fashioning the parts, powers, and uses that we know human language to have, then we must at the same time acknowledge that different cultures do not all talk the same way as we do about their verbal acts and the linguistic tools with which they perform them. As the attention given by linguistic anthropologists to this topic has shown, the items used in metalinguistic discourse–as well as the relationships between those items–are not structurally isomorphic from language to language, even though the powerful influence of the West’s historically institutionalized forms of metalinguistic discourse might incline us to think that they are. And we must go further: for we must also acknowledge that even fellow members of the soi-disant “same” culture do not all share equal access or identical positionings relative to their own culture’s patterns of metalinguistic discourse (Agha 2007). Nor can we blithely assume that those patterns are matters of implicit consensus and homogeneity. On the contrary, as sociolinguists and historians of linguistic inquiry have been showing us for decades–with Sylvain Auroux very much at the forefront of this effort–these are matters of social differentiation, contestation, mystification, and power.



Agha Asif, 2007, Language and Social Relations, Cambridge, Cambridge University Press.

Auroux Sylvain, 1989, Histoire des idées linguistiques, t. I, La naissance des métalangages, Liège, Pierre Mardaga, p. 13-37.

Culioli Antoine, 1968, “La formalisation en linguistique”, Cahiers pour l’analyse, 9/7.

Gibson James, 1979, The Ecological Approach to Visual Perception, Boston, Houghton Mifflin.

Gombert Jean-Emile, 1992, Metalinguistic Development, Chicago, University of Chicago Press.

Reed Edward, 1993, “The intention to use a specific affordance: A conceptual framework for psychology”, Development in Context: Activity and Thinking in Specific Environments, R. Wozniak, K. Fischer ed., Hillsdale, NJ, LEA.

Taylor Talbot, 2003, “Language constructing language: the implications of reflexivity for linguistic theory”, Rethinking Linguistics, H. Davis, T.J. Taylor ed., London, Routledge.

Taylor Talbot, 2010, “Where does language come from? The role of reflexive enculturation in language development”, Language Sciences, 32/1, p. 14-27.

Wittgenstein Ludwig, 1953, Philosophical Investigations, Oxford, Basil Blackwell.


College of William and Mary, Williamsburg

Le texte et les autres éléments (illustrations, fichiers annexes importés) sont sous Licence OpenEdition Books, sauf mention contraire.

Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search