In the three decades since the publication of Kripke’s Wittgenstein on Rules and Private Language the claim that the meaning of linguistic expressions should be explained in normative terms has been one of the most debated issues in the analytic philosophy of language. Some of the main controversies in this area have been focused on questions such as whether a normative dimension is essential for meaning and determines it, or whether it is only a consequence of its more properly semantic features; whether the normativity of meaning can be accommodated within a naturalistic account of semantics, and if not, whether it makes for a case against naturalism or against normativism; what is the role of intentional concepts such as “understanding” in an account of meaning, and what is the relation between normative aspects of meaning and dispositions and regularities that can be specified in a descriptive vocabulary. However, in the recent years, a variety of arguments against the claim of normativity has gained prominence that address a more fundamental issue of what in fact the normativity of meaning means, i.e. what kind of norms—or what is the nature of the norms that are involved, in one way or another, in linguistic meaning¹. Although I will not discuss any specific arguments in detail, I want to focus on the basic question, which in a simplified form can be put as follows: are linguistic norms (the norms of meaning) constitutive or prescriptive?

The two terms of this alternative are related to two historically dominant models of thinking about normativity: rules of games and moral norms. (One should note that the idea of a specific kind of semantic or linguistic normativity is a fairly recent development in modern philosophy). Constitutive norms are understood on the model of rules of games such as chess: the norm constitutes its object, defining what is and what is not to be counted as realizing it. Prescriptive norms differ from constitutive ones most clearly in that it is possible to lapse from the norm while still being subject to it. The paradigm here is provided by moral norms: even if I do actually lie, I am still subject to the norm “one ought not lie”, although I violate it. By contrast, if I move a piece on the chessboard in a way that is not in agreement with the constitutive rules of the game, I am not violating the relevant constitutive norm—I am not making a move in a game of chess at all (I might be violating a different norm, though, such as the norm obligating me to try and play chess if I agreed to do it).

There is some intuitive appeal to the idea that linguistic norms are constitutive in nature and somehow resemble rules of games, especially (but not exclusively) if one thinks of meaning in terms of proprieties of use, in a Wittgensteinian or Sellarsian manner. However, as opponents of this line of thinking point out, a construal of linguistic norms as constitutive makes it difficult to explain the possibility of error—and to distinguish between a properly semantic error and an empirical error. A wrong use of a word is, in many cases, still a use of this very word. If I say “this is green” pointing at a red apple in front of me, I am making a mistake, but the word “green” has its proper meaning.

On the other hand, if linguistic norms are understood as prescriptive norms, serious problems arise as well. For it is difficult to establish the relevant content of a norm governing, for instance, the use of the word “green”. What exactly ought I do to act in accordance with the norm? Ought I use the word “green” only in reference to things that are green? This is obviously not a genuine norm, as I might just as well be joking or lying. Ought I use the word “green” only when I mean green by it? This, of course, explains nothing.

In the present paper, I want to suggest an analysis of the normativity of meaning that steers free of these problems. I will argue that linguistic norms should be understood as constitutive, but only in a global, rather than local sense. The consequence of this claim is that linguistic normativity is in the first place a matter of the pragmatic and practical dimension of language, rather than its strictly semantic aspect.

We should begin by observing that the proper focus of an investigation into linguistic normativity is on use of words and sentences insofar as they are expressions or applications of concepts, and not merely physical events such as uttering a sequence of sounds or making some signs on a paper. The distinction is important, because it underlies the possibility of uttering sounds that correspond to a given word or sentence without actually applying the adequate concepts: either because the sounds do not express any meaning, or, more simply, because one uses the word with an incorrect meaning. I will come back to this shortly.

As I have mentioned, the main difficulty with construing linguistic norms as constitutive is what might be called the problem of error. Let me start with a basic and simple example of an individual concept, such as the concept expressed by the word “red”. There is clearly a correct way of using this word (applying this concept)—to refer to things that are red²—and there are also incorrect ways of using it. The same concerns a simple judgment or sentence “This is red”.

At a first glance, the norm governing the use of the concept “red” is not constitutive of its object: one can use the concept incorrectly and still be taken as using it. The norm is merely regulative in this sense and it leaves room for mistake. On the other hand, however, if we consider the use of a concept in a more general way, and not just a single application of it, we can see that incorrect applications are essentially dependent on correct ones. If someone never used the concept “red” in a proper way, there would be no sense in claiming that they are using this concept at all. It would be more reasonable to assume that they express a different concept with the same word, if there was some discernible pattern of its use, or if there was none, that they are not really saying anything when uttering the sound “red”. A speaker can count as using a concept, whether correctly or incorrectly, only if he or she is able to use it correctly. In this sense, the norm governing the use of a concept is constitutive of its object.

To resolve this ambiguity, I propose to distinguish two senses of constitutiveness: local and global. Norms involved in games like chess are locally constitutive: nothing is subject to them, unless it is in accordance with them. Norms of concept-use are only globally constitutive: some performances can be both subject to a norm and incorrect according to it, but it is impossible for all performances subject to a norm to be incorrect. (Moral norms, on some views at least, are not constitutive even in this sense: something might be a moral obligation even if no one has ever done it and no one ever will).

The claim that linguistic norms are globally constitutive is based on the assumption that one could not be taken to be able to use a particular concept, if one never used it correctly. However, this is plausible only insofar as we consider examples involving merely one isolated concept. Things change when we consider the use of a concept in a broader context.

Imagine that we are dealing with someone who appears to be perfectly able of speaking a language, such as English, and correctly applying many of the concepts that this language enables her to express. She has no trouble with distinguishing apples from oranges and all sorts of fruit and correctly employing a rich color vocabulary. There is nonetheless one concept, “mauve”, that she never applies correctly (or even if she sometimes does—if she says of something that it is mauve and she is right—it is rarely enough for it to count as mere luck). Does the globally constitutive character of linguistic norms entail that she is not in fact using this concept at all?