I’m grateful to Anandi Hattiangadi and Elisa Paganini for reading earlier versions of this paper and providing precious comments and suggestions. I’m also indebted to the participants and organizers of the San Raffaele Summer School in Philosophy for the helpful and lively discussions of the ideas behind this paper. Lastly, I’m grateful to the audience at the Salzburg Conference for Young Analytic Philosophy for their comments.
1The argument put forward by Kripke in his 1982 book Wittgenstein on Rules and Private Language has been interpreted in many different ways since its publication. In some ways, Kripke’s Argument (KA) has proved to be nearly as cryptic as the material he drew inspiration from, namely Wittgenstein’s Philosophical Investigations. However, a tentative consensus regarding the main driving forces behind the argument has crystallized, mainly through the overlap that can be found in analyses given by different authors. It is nearly unanimously recognized that KA is an argument against the existence of meaning-determining facts; however, it is less clear how Kripke aims to demonstrate this claim. It can be safely said that Kripke highlights two separate issues: the difficulty that extant theories have in specifying, without falling into vicious circularity, which facts determine the extension of terms; and the difficulty of extant theories in non-mysteriously explaining the normative (or guiding) force of meaning. The latter difficulty has been referred to as “the normativity constraint” (Wright, 1989; Glüer & Wikforss, 2009), “the normativity argument” (Zalabardo, 1997), and “the argument from normativity” (Guardo, 2009). The extensional and the normative sections of KA intersect and complement each other in interesting ways, but their relationship will not be the focus of this paper.
2Kripke puts forward the normativity argument by highlighting how existent theories of meaning struggle to account for the supposed “guiding” force of meaning, or the way we feel as if we “should” use words in a certain way. The clearest illustration of his strategy is the way in which he criticizes dispositionalism: he says that regardless of whether facts about speakers’ dispositions can help us fix the extensions of terms, they are nevertheless the wrong kinds of facts because they are descriptive and not normative (1982, p. 37). Facts about dispositions can tell us how speakers will or do behave, but not how they should behave; because of this, they cannot be an adequate determiner of meaning. This is taken to be a crucial aspect of Kripke’s argument and has been discussed at length by Paul Boghossian (1989), Allan Gibbard (2012), Hannah Ginsborg (2011), Anandi Hattiangadi (2007), and many others.
3As a response to KA, some philosophers have given an interpretation of the normativity requirement now commonly known as “the simple argument” (Boghossian ,1989; Blackburn, 1984; Whiting, 2016). Proponents of the simple argument argue that the normativity requirement, or the assumption that meaning is normative within KA, should be understood as the recognition that correctness conditions exist and have to be accounted for in a theory of meaning; we ought to speak in a certain way because the expressions we use have correctness conditions. In other words, I should say that 5+7 equals 12 because it is correct to do so, in the sense that it is true that 5+7 equals 12. Meaning is normative precisely because there are correctness conditions.
4This interpretation of the normativity of meaning flattens the distinction between the extensional and the normative challenges of KA – providing an answer to the former automatically provides an answer to the latter. If the normativity of meaning can be understood as the existence of correctness conditions, explaining the possibility of determining the correctness conditions (i.e. the extensions) of terms will also immediately explain how we should use them. This simplifies the work of Kripkenstein’s opponent: if the simple argument is correct, providing an appropriate example of how correctness conditions are fixed would solve the skeptical paradox.
5Hattiangadi (2007) has provided a more developed version of the normativity argument and its implications, following considerations by Boghossian (1989). She proposes that a charitable way of interpreting KA is to see it as analogous to certain arguments against moral realism, and in particular with a revised version of Moore’s Open Question Argument (1903, pp. 10-21).
6The argument against the existence of moral facts can be summarized as follows: moral facts would have to be prescriptive or inherently motivating, and natural facts we have access to do not seem to be prescriptive or inherently motivating. So moral facts cannot be reduced to natural facts. Following a line of argument defended by Mackie (1977), one can move from the conclusion that moral facts are irreducible to the conclusion that there are no moral facts at all. If moral facts cannot be reduced to natural facts, they are a sui generis class of unnatural facts that has some peculiar and unexplained property, namely inherent action-guidance (or something sufficiently similar). These facts are “queer,” create epistemic problems, render some causal relations mysterious, and lack appropriate philosophical explanation. The conclusion, again, is that we should forgo moral facts altogether.
7We can see how this line or argument can easily be applied to semantic facts if we assume that they are normative. If semantic facts are normative, that means that they are prescriptive or inherently motivating, and so they cannot be reduced to natural facts that lack these properties. But the existence of sui generis, “queer” semantic facts is unacceptable (and creates epistemic and causal problems). We are left with no possible candidates that could take on the theoretical role we wanted, and we are required to give up on semantic facts. The simple argument upholds one of the assumptions that is needed for this argument to go through, namely the premise stating that semantic facts are normative.
8There have been critics of the simple argument. Specifically, it has been pointed out that correctness conditions do not necessarily have any true normative force (Hattiangadi, 2007; 2009; Glüer & Wikforss, 2015), i.e. they do not determine what speakers ought to do in a significant sense. For example, Hattiangadi (2007, pp. 181-183) argues that semantic correctness creates no categorical obligations; it can only create obligations in connection with external desires or motivations, that is, hypothetical obligations. In contrast with cases that concern categorical normativity, hypothetical normativity seems to be reducible to non-normative facts. For example, the “ought” mentioned in “if you are going outside in the rain, you ought to bring an umbrella” is reducible to descriptive facts and does not seem to give rise to any genuine normativity. This is important because robust normativity is the kind of normativity needed for KA to go through, at least if we interpret it as has been described above – as analogous to certain arguments for moral antirealism.
9As noted previously, KA can be interpreted as an argument against the existence of semantic facts due to their normative character. Critics of the simple argument show that the existence of correctness conditions is not a sufficient reason for assuming that meaning is irreducibly normative.
10A pressing question, at this point, is whether we have any reason to believe that meaning is, in fact, strongly normative. In a sense, the simple argument can be reinterpreted as an oblique rebuttal of this assumption: if the normativity of meaning simply is nothing more than the existence of correctness conditions, and correctness conditions do not provide us with robust normativity, our intuitions regarding the existence of semantic oughts rest on shaky grounds. The normativity requirement itself, then, is dependent on a misleading intuition. Semantic correctness is not sufficient for normativity.
11Throughout the discussion surrounding KA there have been mentions of the fact that semantic correctness might not be the only type of correctness worth considering. If there were a different kind of correctness that could fill the appropriate theoretical role – that is, a correctness that could provide us with purely semantic and categorical norms – that could offer a viable alternative for the normativists. The rest of this paper will be dedicated to exploring whether linguistic correctness could be used as a basis for the normativity of meaning.
- 1 The quoted philosophers have not used the exact terminology I am using in this work but have made (...)
12Some philosophers have suggested that there are two different types of correctness: semantic and linguistic1 (Buleandra, 2008; Millar, 2002; Reiland, 2023). This is interesting because the main focus in the literature has been on semantic correctness as a possible source of normativity. In other words, one might want to defend the idea that the source of genuine linguistic norms is linguistic correctness.
- 2 This distinction has been made at least as early as in Anselm of Canterbury’s De Veritate: “Theref (...)
13Defenders of this distinction between semantic and linguistic correctness point out that there are intuitions supporting the idea that we can use language correctly even if we stray from true application. For example, whenever I lie there is a sense in which I am using language correctly, even though I am saying something false, as I am speaking in accordance with the words’ established meaning.2 I can also be said to use an expression incorrectly even if I am saying something true – typical examples include speakers who misspeak, e.g. someone who uses “arcane” instead of “ancient” (substituting the two can contingently generate true statements which are nevertheless incorrect). Another sense in which one’s use of language can be correct without being true is related to the fact that we can distinguish between appropriate and inappropriate ways of asking questions, giving orders, greeting someone, and so on.
14These examples indicate that there are two different senses in which we might speak (in)correctly: one that is directly linked to true and false application and one that is not. To clarify this distinction (whether it turns out to be substantial or not), we can preliminarily define the two types of correctness as follows:
SC: A statement S is semantically correct if S is true.
LC: A statement S is linguistically correct if S is used in accordance with its meaning.
15We saw that semantic correctness cannot account for the normativity of meaning because it does not provide us with anything beyond instrumental obligations. Norms related to whether we should use our terms in a semantically correct way are dependent on our desire to tell the truth. For example, a norm derived from semantic correctness – “you ought to use ‘green’ correctly” – does not seem to be in force unless it is accompanied by an external desire such as “if you want to tell the truth, you ought to use ‘green’ correctly.” Instrumental obligations are not proof of language’s normative character, as anything can be instrumentalized relative to our desires. For example, if I want to stay dry, I ought to bring an umbrella, but that does not mean that I have a general obligation to bring an umbrella; instrumental obligations are not a sign of genuine normativity.
16On the other hand, linguistic correctness could have more profound normative consequences that are constitutive of or linguistic practices. The standard definition of constitutive rules characterizes them as creating or making possible new types of behaviors (Searle, 1969, p. 35). In the case of meaning broadly understood (i.e. conventional meaning, whether it can be analyzed truth-conditionally or not), the use conditions for a linguistic expression make possible the meaningful utterance of that expression – I cannot even participate in the practice of language if I don’t speak in accordance with the expressions’ use-conditions. Thus, the use conditions of an expression are constitutive of the conventional meaning of that linguistic expression.
17To summarize what has been said until now, the distinction between linguistic and semantic correctness could have meaningful consequences for KA because the existence of linguistic correctness may vindicate the intuition that meaning is robustly normative, unlike semantic correctness.
18The debate surrounding correctness and its normative implications has seen both supporters and deniers of the idea that there might, in fact, be two different types of correctness that are relevant to meaning. Typically, those who disagree that the notion is ambiguous insist that it is impossible to distinguish linguistic from semantic correctness (Reiland, 2023). As we have mentioned, this type of correctness is thought by many not to be robustly normative. To answer Kripke’s sceptic, then, a supporter of the notion that only semantic correctness is relevant to meaning can hold an anti-normativist position and argue that semantic correctness – which is reducible – is all there is to support the intuition that meaning is normative.
19Reiland (2023) argues that the resistance to the idea that linguistic correctness is separate from semantic correctness comes from the implicit assumption that people can privately imbue words with meaning through their intentions (pp. 2201-2202). The reasoning of deniers of the distinction can be summed up as follows: there can be no linguistic error because people always mean what they intend to mean, and if they stray from publicly established norms for the usage of an expression this should always be interpreted as a type of linguistic innovation. In other words, “misuses” do not exist: if I say “good morning” to my husband right before we start eating dinner, what I am doing is not using the words “good morning” incorrectly, but trying to introduce a new word from a sort of individual language into English. The meaning of this individual language is determined by my intentions.
20However, if one is to grapple with the skepticism inherent to KA, one cannot assume that individual intentions have a role in the determination of meaning: all content-laden states (such as belief, thought, and in this case intention) can be targeted by the skeptical argument in an analogous way. This is due to the fact that if we take, e.g., intentions to be the determiners of meaning, the sceptic can reply by pressing us regarding what exactly determines the content of the relevant intentions, leading us into a vicious regress. This type of argument can be easily generalized to all contentful states.
21In any case, the idea that meaning is imbued into words via intentions is certainly not a necessary background assumption for the discussion at hand. This suggests that the distinction between LC and SC should not be flattened until someone provides further reasons to deny its existence.
22Something more should be said about linguistic correctness and what it consists in. Reiland (2023) proposes a generic definition of linguistic correctness which may be adapted to different theories: using an expression in accordance with its meaning is using it while being in its “use-conditions” (p. 2193). The relationship between linguistic correctness and use conditions could take on this form: “saying S is linguistically correct when certain conditions are satisfied.” In simpler terms, this means that there will be occasions in which it is linguistically appropriate to use an expression, and occasions in which it is not linguistically appropriate to use an expression. Reiland leaves use-conditions to be further specified.
23One way use-conditions could be fleshed-out is through reliance on use-conditional semantics. Semantics has historically been understood as the domain of conventional meaning, sometimes also called “literal meaning”, which was understood to be meaning as provided by a truth-conditional analysis. Any other meaningfulness found in language was posited as belonging to the domain of pragmatics. This basic criterion for distinguishing semantics from pragmatics has sometimes been represented as “pragmatics = meaning – truth-conditions” (Gazdar, 1979, p. 2). However, it has been argued that some aspects of meaning that have traditionally been thought to be within the domain of pragmatics should belong to semantics: in particular, some philosophers noted that there are conventional aspects of meaning that have little to do with truth-conditions. For example, “goodbye” is an expression that has a well-established conventional meaning, but whatever is expressed by “goodbye” is neither true nor false. It seems intuitive, then, that conventionally established meaning encompasses something more than purely truth-conditionally understood meaning. We may want to designate truth-conditionally based meaning as meaning in the strict sense, and conventionally based meaning as meaning in the broad sense.
24Since the appearance of Kaplan’s 2004 unpublished paper based on one of his lectures at UC Berkeley, titled “On the Meaning of ‘Ouch’ and ‘Oops,’” several philosophers have tried to bring forward the project of a use-conditional semantics, something that could help us make sense of and analyze this wider sphere of conventional meaning. Kaplan’s proposal is to provide a formal semantics that encompasses the conventional aspects of meaning which are, nevertheless, unanalyzable in truth-conditional terms. The idea stems from the simple insight that taking truth-conditionality and conventionality as the criteria for semantic relevance does not yield the same results, as we have seen – conventionality casts a wider net.
25Kaplan’s framework is designed to deal with expletives, indexicals, and other components of language which are unsuited to a truth-conditional analysis. As he notices, these types of expressions seem more suited to a use-conditional analysis: the truth-conditions of “I am blonde” change depending on who utters it, while its use-conditions – namely that the sentence is correctly used if the speaker is blonde – are fixed and seem to provide us with the meaning of the sentence in a more accurate sense. Not only that, but the use-conditions for these words intuitively provide us with information about the correct and incorrect ways of using them – and clearly, this is not semantic correctness, as there are no true or false utterances of “goodbye”. It should be underlined, then, that use-conditions as Kaplan understands them are a good candidate for what determines linguistic correctness.
26Truth-conditions and use-conditions can coexist. Following Kaplan’s basic idea, Gutzmann (2015) tries to develop a “hybrid semantics” that includes both truth-conditions and use-conditions. The goal of Gutzmann’s project is to build a framework that would enable us to apply the familiar tools of formal semantics even to non-descriptive, but still conventionally determined, features of language. In his framework, while truth-conditions of propositions are based on sets of possible worlds in which the proposition is true, use-conditions are given by the sets of contexts in which an expression is “felicitously” used (Gutzmann, 2015, p. 18). It is safe to say that felicitous usage can model what we have, up until now, referred to as linguistic correctness.
27As was mentioned previously, it seems as if the obligation to speak in a semantically correct way is dependent on desires that are external to meaning – being honest, for example. The normativity in question is extrinsic to meaning. The notion of speaking in accordance with a term’s use-conditions, on the other hand, seems to be inseparable from conventional meaning. A language is inconceivable without use-conditions and given a set of use-conditions in my language, I ought to speak in accordance with them, if I want to speak at all. This suggests that the “ought” derived from linguistic correctness is inherent to language in a way that the “oughts” derived from semantic correctness are not.
28It might be helpful to rely on an example in order to clarify what this type of obligation may consist in and why it is different from an obligation to speak truthfully (i.e. in a semantically correct way). If we take a non-referring term such as “goodbye,” it’s clear that it has no truth-conditions we can adhere to. However, it is also clear that there are definite use-conditions which regulate its (linguistically) correct use: it is felicitous to say “goodbye” to people with whom we are parting, it is infelicitous to say “goodbye” when we’re sitting down to eat, and so on. The use-conditions for “goodbye” seem to provide us with something that is intimately tied with the meaning of the word and with being able to use it in the English language. If I do not adhere to the rules set by linguistic correctness, it can be doubtful that I am speaking at all and not merely making word-like noises.
29Obligations, even ones stemming from constitutive rules, can be violated. Violating our obligation to use language according to its use-conditions has some interesting parallels to violating the constitutive rules of games. While straying from use-conditions of a language may be a sign that we’re not speaking it anymore, straying from central constitutive rules of a game is equally a sign that we’re not playing it anymore; however, in both cases violations may be used to innovate or constitute a new practice. This suggests that the possibility of linguistic innovation does not interfere with the idea that use-conditions are constitutive of meaning, just like the possibility of innovation within the rules of a game does not interfere with the idea that games are constituted by their rules.
30One might object to the idea that LC is constitutive of meaning by arguing that the obligation to speak in a linguistically correct way is dependent on a desire to communicate or being understood. However, communication is widely accepted as being (at least) one of the primary functions of language. Because of this, it is difficult to conceive of participating in language in any way that precludes the desire to communicate something. When our aims are not to communicate or be understood, it is arguable that we are speaking at all. Desires related to such an integral function of language are not external to the practice. If you want your actions to count as playing chess, you should move the bishops diagonally across the board. If you want your actions to count as speaking, you should speak in a linguistically correct manner.
31The idea that LC is constitutive of meaning broadly construed can be plausibly defended. However, this is not enough for our purposes: the position we set out to explore is that there are categorical meaning-norms, and that these are provided by LC. Constitutive rules do not automatically give rise to categorical oughts; the rules of chess are constitutive of chess, but I am not in any way categorically obligated to follow them. Just because I need to move bishops diagonally in order to play chess does not mean I ought, generally, to move bishops diagonally. I may not want to play chess at all.
- 3 The account of which exact features are the agency-constituting ones will vary – this is simply a (...)
32However, there have been attempts at deriving categorical oughts from constitutive rules. Returning to the parallel between moral and semantic norms, some philosophers have suggested that in the case of categorical moral norms, they can be derived from facts about what is constitutive of being an agent. Enoch (2006) presents a survey of these attempts, where he individuates several goals of the theories he covers, one of which is responding to Moore’s Open Question Argument and reinstating the possibility of a naturalistic explanation of moral norms. For our purposes, we may lay out the structure of these arguments as follows: there are some essential normative features of agency,3 in the loose sense that without these features, agency would not be possible. Since agency is constituted by these features, one cannot be an agent and avoid the normative force posed by them.
33“The Problem” with these proposals, as Enoch calls it, is the following: what if someone isn’t interested in being an agent? What if I am a “schmagent,” i.e. something similar to an agent but which differs in term of the norms constituting it? If we can’t rule out my being a schmagent and not an agent, there is simply no way of rendering the norms constitutive of agency categorical, which was the original goal. The key notion here is that of interest. It seems that for the “rules of agency” to have any hold over me, I need to have an interest in being an agent; and that is just a different way of saying that I need to have an external reason to be an agent (and not a schmagent). Categoricity does not seem to be derivable from constitutive rules.
34We can see how this applies in an analogous way to the constitutive rules of meaning. If I want to speak, I should follow certain norms; but this normativity is dependent on external reasons to speak. Without these external reasons, I have no obligation to speak in any kind of way. The constitutive norms of meaning are not categorical.
35Enoch considers several possible answers to The Problem that plagues attempts at deriving categorical norms from what is constitutive of agency, the most promising of which is the idea that there is no escaping being an agent. What if it makes no sense to ask whether I’m interested in being an agent because there is simply no way I cannot be an agent? Enoch argues that even if we are forced into a practice, that does not translate into having a categorical reason to follow the constitutive norms of that practice. There is still a need for an independent reason to act. He imagines a scenario in which we’d be forced to play chess – would we be categorically obligated to follow the rules of chess in that case? It’s clear that unless we have a reason to want to play chess, we are not categorically obligated to play a certain way, even if playing is unavoidable.
36The unavoidability line of argument does not only run into this kind of difficulty but is also less applicable to the case of meaning. While we could defend the idea of choosing and acting being unavoidable, not being a speaker is clearly at least conceivable. While opting out of language altogether is rare, it does not seem to be impossible, and so we could not argue along the unavoidability line even if it were promising.
37I will now briefly return to the assumption that is the reason for our exploration of LC: that meaning is categorically normative. KA is certainly stronger if we accept this premise. However, one could argue that we do not need categorical oughts to explain the intuitions this premise is based on – namely the intuition that there is a sense in which I ought to speak a certain way. This is just another way to say that there are two possible types of strategies for contrasting the skeptical argument from its normative side: the first one is to accept its premise – that there are categorical semantic norms – and try to show that this fact is compatible with a coherent theory of meaning and does not leave us in an uncomfortable position. This work has focused on surveying a particular argument that could be made following this first strategy; namely, trying to show that there is a certain type of categorical meaning-normativity that is constitutive of language. As we have seen, this cannot work because there are great difficulties in deriving any type of categorical norms from constitutive facts.
38The second type of strategy is to fight the assumption that meaning is robustly normative at all. One can then either take the intuition that meaning is normative seriously and try to explain it without relying on robust semantic norms; alternatively, it can be argued that our intuitions about the normativity of meaning are confused, misplaced, or unjustified.
39I will not discuss these replies to KA here. The important thing to note is that distinguishing SC from LC can present an opportunity to craft an argument of this second type, too. When the sceptic asks why it feels as if I ought to use expressions a certain way, one might account for this intuition by explaining that LC exists, that use conditions are constitutive of meaning broadly understood, and that if one is to speak, then they are compelled to speak in a linguistically correct manner. We are not compelled to speak – meaning that the “ought” in question cannot be categorical – but whenever we do, we feel the force of these constitutive norms; and this is an interesting fact about meaning and the practice of language. This can be the beginning of a compelling explanation of our intuitions that meaning is normative, without granting that the normativity in question is categorical.
40Even if we take it that linguistic correctness is constitutive of meaning (broadly understood), this does not seem to generate categorical semantic norms. The existence of linguistic correctness, then, does not provide us with a reason to believe that meaning is irreducibly normative. However, recognizing that use conditions are constitutive of meaning can explain the intuitions that support the idea that meaning is normative. This may seem unsatisfactory to some, due to the fact that this framework places the source of our obligations not within standard truth-conditional meaning but within a broader conventional meaning. In a sense, this objection is warranted: the basis for Kripke’s normativity argument is that whatever facts determine the extensions of our terms cannot also account for the normative character of meaning.
41I see no harm in trying to respond to the normative requirement of KA in this less orthodox way, that is, by understanding it as highlighting the difficulty in reconciling two different aspects of meaning: a) the fact that I ought to use language in a certain way, in a sense that is purely internal to meaning, even if the “ought” in question does not represent a categorical obligation; and b) the fact that a theory of meaning should ideally rely on naturalistic, non-opaque, publicly available facts. LC can help us explain a) while keeping open a possibility for resolving b).
RSS feed
