Linguistic Mistakes and Semantic Rules

1 Introduction

Especially after Saul Kripke’s Wittgenstein on Rules and Private Language (1982), many philosophers have found compelling the view that meaning is essentially tied to rules that govern the use of expressions. According to this view, just as games’ rules determine which moves are correct and incorrect, arguably, semantic rules determine the right and wrong ways to use language. Let us call this view “the rule-governance view.”^[1]

However, the current debate has shown that the precise content of these alleged rules is far from obvious. At the heart of this ambiguity is the notion of semantic correctness. According to the orthodox construal, which can be traced back to Paul Boghossian’s seminal article (1989), correctness is explicated in terms of truthful application. The orthodox construal of semantic rules could therefore be called “the truth rule of meaning.” It is most prominently endorsed today by Daniel Whiting (Whiting 2007, 2009, 2016, 2024).^[2]

The truth rule has garnered criticism from both the opponents and defenders of the rule-governance view. The opponents argue that while the notion of semantic correctness – understood in terms of application – is intrinsic to meaning, the truth rule only applies to speakers who want to speak the truth and, therefore, is extrinsic to meaning (e.g., Horwich 1998; Wikforss 2001; Hattiangadi 2006). The proponents of the unorthodox construal^[3] argue that correctness of application – which can be explicated in terms of truthful use – should be distinguished from correct use, which is use in accordance with meaning (Buleandra 2008; Glock 2019; Millar 2004; Reiland 2023a). The difference between the orthodox and the unorthodox interpretations of correctness conditions is commonly demonstrated using the notion of linguistic mistakes. However, critics have suggested that the notion of linguistic mistake is unclear (Whiting 2016; Wikforss 2001).

This paper has two goals. The first is to defend the notion of linguistic mistake against the claim that the notion is inherently unclear, and the second is to argue against the defenders of the unorthodox view that this notion cannot be captured by semantic rules. To achieve the first goal, I will introduce cases which are intuitively classifiable as linguistic mistakes. I argue that classifying these cases as permitted, and inversely, forbidding uses where no clear case of linguistic mistake occurs, are reasons to take the rule to be extrinsic to meaning. Since truthful use does not guarantee the absence of linguistic mistakes and vice versa, the truth rule is not intrinsic to meaning. While this might seem to count in favor of the unorthodox reading, I contend that candidates for the unorthodox rule of use also fail to align with some intuitive cases of linguistic mistakes. In particular, I will focus on the rule candidates recently proposed by Indrek Reiland (Reiland 2023a, 2023b).

Nevertheless, to show that no rule can capture the notion of linguistic mistakes, a more robust account of linguistic mistakes must be developed. I will argue that the intuitive notion of a linguistic mistake is best captured in terms of a discrepancy between what an expression is used for – its linguistic function – and what a speaker tries to use that expression for. I argue that for rules to correctly classify uses as linguistic mistakes according to this definition, they must depend on speakers’ intentions and therefore are instances of a general rule-like relationship between means and ends rather than being intrinsic to meaning. Therefore, linguistic mistakes are problematic for both the orthodox and the unorthodox construal of the rule-governance view.

First, I discuss the orthodox and unorthodox construal of semantic correctness. Next, I explore the differences between these construals by introducing the pre-theoretical notion of linguistic mistakes and showing why the orthodox account conflicts with intuitions about which uses warrant linguistic criticism. I will then discuss how Reiland’s view could clarify the notion of linguistic mistakes in terms of use-conditions and show why the rules he proposes misclassify linguistic mistakes. Finally, I propose an account of linguistic mistakes independently of the rule-governance view and argue that understood this way, linguistic mistakes cannot be captured by semantic rules.

2 Correctness Conditions

In one sense, language use is indisputably governed by rules, but this is not what the defenders of the rule-governance view have in mind. Namely, it is an uncontroversial fact that there are rules that, for example, require that one should end requests with “please,” that ethnic slurs should not be employed, and that unnecessary digressions from a topic of conversation should be avoided. However, these are the rules of etiquette, morality, and pragmatics, which show how to use those expressions in a way that makes one polite or moral or a better conversationalist. While these rules govern language use, they are extrinsic to meaning. For the rule-governance view to rise above mere trivial observation, it must be shown that language use is governed by rules intrinsic to meaning. Such intrinsicality makes the rules that govern language use semantic in nature rather than moral, pragmatic, or customary. In other words, the rules that can vindicate the rule governance view must be derivable solely from the meaning of expressions without requiring further rules as premises.^[4]

While this distinction is commonly drawn in the literature (e.g., Horwich 1998, 186; Hattiangadi 2007, 181; Glüer and Wikforss 2009, 38), it is sometimes difficult to assess whether a rule is intrinsic or extrinsic. We could argue, for example, that there is nothing to the meaning of “please” beyond what is contained in the rule of etiquette (see also Section 4). However, this does not show that the distinction is not useful. While it cannot be denied that at least the rule of etiquette, as an extrinsic rule, governs the use of “please,” this is not enough to establish that this rule is also intrinsic to the meaning of “please.”

What rules, if any, could be entailed by the meaning of expressions alone? A common route for determining which actions are permitted and which are forbidden by the rules intrinsic to meaning is via the notion of semantic correctness. However, this notion itself is contested. According to one prominent reading by Paul Boghossian, semantic correctness can be captured by the following principle:

(Correctness of Application) e means F→∀x (e applies correctly to x ↔ x is f),

where e is an expression, F stands for the semantic content of the expression and f is a feature x must have for F to truthfully apply to it.^[5] To use Boghossian’s example: if “green” means green,^[6] then “green” applies correctly to x if and only if x is green (1989, 513). It has been argued that correctness conditions can be interpreted as conditions of permissibility (Peregrin 2012; Reiland 2023a; Whiting 2009, 2024). Thus, the correctness of application suggests that the use of an expression such as “green” is governed by a rule such as:

(Truth Rule) ∀S ∀x: ((S may apply “green” to x) ↔ (x is green))

However, some normativists, including Millar, Buleandra, and Reiland, distinguish between the correctness of application and a more general correctness of use, which can be understood in terms of use in accordance with meaning.

(Correctness of use) e means F→∀S (S uses e correctly ↔ use accords with F)

They argue that while correct use sometimes corresponds to truthful use, this is not always the case. Therefore, in place of the truth rule, they advocate more nuanced rules of use.

However, if we adopt truth-conditional semantics, then the most straightforward interpretation of the phrase “use in accordance with meaning” is simply applying expressions in accordance with its truth-conditions, but this would mean that the correctness of use does not differ from the correctness of application (cf. Whiting 2016, 230). Nevertheless, the defenders of the rule of use maintain that these notions of correctness are distinct. Buleandra, Glock, Millar, and Reiland rely on the distinction between linguistic mistakes and factual mistakes to illustrate the differences between the “orthodox” and the “unorthodox” reading. Therefore, the needed clarification could be achieved by shedding light on the notion of linguistic mistakes (Buleandra 2008, 180–81; Glock 2019; Millar 2004, 162–66; Reiland 2023a; for classic discussion of linguistic mistakes, see Burge 1979).^[7]

3 Linguistic Mistakes

To grasp what type of mistakes linguistic mistakes are, let us consider the following story. Suppose John, a mostly competent English speaker, sees a fox in the woods. He says to Jane, “Look! There is a vixen.” Jane, who happens to be familiar with foxes, replies, “No, that is a male.” A disagreement seems to have arisen regarding the animal’s sex. Assuming that Jane is right, we can speculate as to whether John misperceived or had false beliefs about sexing foxes. Of course, Jane could have made a mistake instead. Regardless, further observation or expert opinion could reasonably be seen as resolving this disagreement.

Suppose, however, that John answers, “You are right! A male vixen.” This response would indicate that the disagreement, if there were any, did not concern the animal’s sex. Furthermore, since taking John to sincerely claim that the animal is a male vixen would seem to imply contradictory beliefs, a more charitable interpretation would reassess what he thinks the word “vixen” or “male” means. For example, attributing to John the belief that “vixen” refers to all foxes, not just female foxes, would help to rationalize his behavior. Despite having “saved” John’s rationality this way, normally, we would still say that he has made a mistake. The mistake, however, is not factual; it is linguistic, i.e., a misuse of words.

While intuitively, both of John’s utterances should count as linguistic mistakes, the situation differs somewhat when viewed through the lens of the correctness of application. John’s answer “[that is a] male vixen” involves an incorrect application regardless of the animal’s sex. If the animal was male, then John applied the word “vixen” incorrectly. If the animal was female, the application of the word “male” would be incorrect. In either case, according to the truth rule, John has done something prohibited. On the other hand, the status of John’s earlier linguistic action depends on the animal’s actual sex. If the animal was male, then John would have applied the word “vixen” incorrectly. If the animal were female, then his application would have been correct. In the latter case, Jane instead applied the word “male” incorrectly to the female animal.

This discrepancy between what intuitively counts as linguistic mistakes and what the truth rules forbid raises doubts about its intrinsicality to meaning. Arguably, Jane’s use of the word “male” accords with the word’s meaning – even if she turns out to have applied the word falsely. Correspondingly, regardless of their truth values, both of John’s uses seem to fail to accord with the meaning of “vixen.” Therefore, a rule that forbids Jane from using “vixen” in accordance with its meaning and permits John to use it in discordance with its meaning should not be taken as semantic in nature. Instead, rules that require speakers to apply “vixen” only to female foxes are arguably instances of more general principles based for example on the utility or value of truth neither of which are intrinsic to the meaning of “vixen” (Horwich 1998, 190–92).

Furthermore, while Jane and John most likely do care about the truth value of their assertions, as the defenders of the rule of use have pointed, expressions are used in a variety of ways, not just making sincere assertions, many of which cannot be thought of as being governed by the truth rule, such as in lies, jokes, irony, and queries. The truth-conditional semantics may well determine the meaning of expressions used in these practices, but the use of these expressions is not guided by the rules requiring speakers to stay truthful (Buleandra 2008, 180; Millar 2004, 177; c.f. Whiting 2024, 129). Indeed, if an insincere speaker ends up speaking the truth because she is confused about the meaning of expressions, she seems as semantically criticizable as John. Finally, some expressions, such as “ouch,” “oops,” and “goodbye,” do not seem to possess truth conditions at all, but still, many argue that they may be used correctly and incorrectly (Kaplan 1968; see also Buleandra 2008; Reiland 2023a, 2023b).

These considerations could indicate that the defenders of the unorthodox construal are in a better position to defend the intrinsicality of the rules of use. After all, since the rules of use require speakers to use the expressions in accordance with meaning, it seems more likely that these rules can be derived from the meaning itself, instead of being an instance of an extrinsic rule.

However, the intuitive classification of John’s uses as failing to accord with the meaning of “vixen” can be challenged. The meaning of John’s utterance “vixen” can be understood either in terms of his speaker’s meaning as well as of the public or conventional meaning. (Grice 1989, chap. 6).^[8] While John’s use failed to accord with the conventional meaning female fox, his use seemed to accord perfectly well with his speaker’s meaning, fox. Thus, what reason there is to regard John’s usage as a linguistic mistake if he did not in fact mean female fox? (Whiting raises a similar issue 2016, 230.).

Reiland (2023a) argues that this type of response relies on the individualist view of language. The individualist, such as Paul Grice (1989), Donald Davidson (1986), and Akeel Bilgrami (1993), treat speaker’s meaning as fundamental compared to conventional meaning. Reiland contends that this perspective assumes that meaning fox by “vixen” is something John can do independently of the word’s conventional meaning.

Reiland argues that if we regard John’s utterances as tokens of English-language expressions, John cannot simply choose to change their meaning. Reiland calls this perspective the public-language view and argues that, from this perspective, linguistic mistakes seem more plausible.^[9] As long as we can confidently say that John is, indeed, speaking English, John’s use could reasonably be called a linguistic mistake. At least, we need not worry that John’s private meanings would affect the assessment of whether he has used an English expression correctly.^[10]

Even if we accept the public-language view, Whiting questions whether this shows that John’s mistake is anything but factual. That is, John used the English language expression “vixen,” but simply had a false belief about its conventional meaning. He draws the analogy to chess, where a player might mistake a bishop for a rook and in an attempt to cheat moves it diagonally. Did the player break the rules of chess because she moved a piece she believed to be a rook? Obviously not. She made a factual mistake, but because the piece was in fact a bishop, she did not break the rules of chess.

Mistaken beliefs can factor into explanations of why players break the rules of chess, but it does not affect whether a rule is broken. Similarly, John may be mistaken about what the English expression “vixen” means, but Whiting argues that this should not have a bearing on the assessment of semantic correctness, which should depend solely on whether the fox is actually male or female (Whiting 2016, 230–32.).

Although I do not share Whiting’s final conclusion, one important takeaway from his considerations is that while false beliefs about the public meaning may often be present in the intuitive cases of linguistic mistakes, and even might explain why a mistake has occurred, the explanation of an action should not be conflated with the assessment of the action. The question of whether or not John’s use of the word “vixen” counts as a linguistic mistake depends on whether his use satisfies the necessary and sufficient conditions of linguistic mistakes and not the reason why John failed to satisfy these conditions.

Of course, speakers’ false beliefs about the meaning of the words they are using could arguably be the necessary or sufficient condition for the occurrence of linguistic mistakes. Nevertheless, this does not lead to the conclusion that linguistic mistakes are a subset of factual mistakes. Moreover, the intuitive notion of linguistic mistakes does not seem to require speakers to have false beliefs about the meanings of words. It is common that speakers immediately admit their misuse right after using an expression by saying something like: “Sorry, I meant to say ‘fox’ rather than ‘vixen’.” Since the reaction is immediate, it seems unlikely they had false beliefs about the meaning of the words, but it is still reasonable to count these as misuses. Similarly, we often use expressions without fully knowing the exact meaning of them, but the conclusion that most of the uses of the expressions count as mistakes borders on absurdity. Consider, for example, newscasters who often must report on things they have limited knowledge of. When a newscaster utters, “The Higgs boson was discovered in CERN,” she is not making a linguistic mistake just because she does not know the precise meaning of “Higgs boson” as long as she knows it well enough. Although it might be tempting to treat linguistic mistakes as a graded notion, this solution would not do justice to the semantics of mistakes. One either makes a mistake or not. There is no slightly mistaken use of an expression.

So, the arguments presented against linguistic mistakes do not conclusively establish that the notion is inherently unclear. There is reason to take John’s uses as instances of linguistic mistakes. However, establishing that there are clear cases of linguistic mistakes does not help to determine what are the general conditions for the occurrence of linguistic mistakes and therefore does not reveal what are the unorthodox correctness conditions of use. Therefore, we still have not determined what the unorthodox rules of use prohibit speakers from doing.

Unfortunately, instead of offering general conditions, many advocates of the rules of use tend to rely on examples of linguistic mistakes to demonstrate what the rules permit and prohibit. Even when offering these general conditions, the conditions themselves tend to require clarification, which in turn is offered in terms of individual cases. For example Millar talks about requiring speakers to “respect the conditions of true application,” but what counts as and what does not count as respecting the conditions is clarified using individual cases (e.g., Millar 2004, 162–63). However, as a notable exception, Reiland’s account offers more concrete examples of rules of use in terms of conditions of permissibility. Therefore, I turn next to his proposal of how the rules of use could be formulated.

4 Rules of Use

For a possible model for the rules of use, Reiland suggests turning to David Kaplan’s account of expressives. Kaplan argues that the semantics of expressive terms such as “ouch,” “oops,” and “damn,” as well as greetings such as “hello” and “goodbye,” can be analyzed by the rules of their use. For example, the word “ouch” is used when one is in pain, so the rule governing its use could be something like this:

(Ouch) ∀S: S may use “ouch”↔S is in pain (Kaplan n.d.; Reiland 2023b, 575.)

Whiting’s objection that rules of use are disanalogous to the rules of chess does not apply to the rule of “ouch.” Whether the speaker is in pain at the time of the utterance does not depend on what she means by “ouch” or what she believes the word means. Thus, the rule of “ouch” does justice to the chess analogy at least just as well as the truth rule, and violations of the rule are clearly distinct from factual mistakes.

Reiland’s suggestion, then, is to generalize Kaplan’s analysis of expressives to all expressions. The rule for “ouch” suggests a general form for rules of use for expression e:

∀S: S may use e ↔ C

where C is what Reiland calls use-conditions. (Reiland 2023a, 2023b.) According to Reiland, the use of an expression while failing to fulfill the use-conditions results in a linguistic mistake (Reiland 2023a, 2198).

While expressives demonstrate that some words have plausible candidates for semantic rules, this does not mean that the same holds for every meaningful expression. Common to “ouch” and other examples used by Kaplan, such as “oops,” “hello,” and “goodbye,” is that they lack truth-functional content.^[11] Since “vixen” and “Russell” have truth-functional content, expressives offer only little help in formulating rule candidates for them. To put this more concretely, the use-conditions for “ouch” refer to the mental state of being in pain, but this is no help in determining what mental states could be used in the rules for “vixen” or “Bertrand.”

Unfortunately, Reiland has yet to answer what he takes to be the use-conditions of truth-functional expressions such as kind terms, proper nouns, or declarative sentences. He has only offered a few suggestions he takes to be plausible candidates. The first suggestion is requiring a speaker to think about an expression’s referent or a property.

∀S: S may use “vixen” ↔ S is thinking about the property female fox

∀S: S may use “Bertrand” ↔ S is thinking about Russell

(Reiland 2023a)

As was the case for the “ouch” rule, these rules make it possible to distinguish between factual mistakes and semantic mistakes. However, these rules fail to do justice to the intuitive notion of linguistic mistakes. Everything the speaker is thinking about at the time of the utterance simply has no bearing on the legitimacy of the language use. The speakers’ rich inner lives can include a whole host of objects and properties they are thinking about, regardless of what they are trying to express. For example, when thinking about all foxes, John could have also been thinking about the property of female foxes, but this does not affect whether he intended to use “vixen” to refer to all foxes. Similarly, suppose Jane utters “Gottlob found a paradox in Bertrand’s logic.” If Jane knows that it was Russell who found the paradox in Frege’s logic, then she appears to be mixing up the names of the logicians and thus making a linguistic mistake. Nevertheless, since she was thinking about both Russell and Frege, she satisfied the proposed use-condition.

Perhaps Jane’s use could still count as a mistaken use of the declarative sentence. Reiland has suggested the following as the candidates for the use-conditions of declarative sentences:

∀S: S may use, “Foxy is a vixen” ↔ S believes that Foxy is a female fox,

∀S: S may use, “Foxy is a vixen” ↔ S judges that Foxy is a female fox,

∀S: S may use, “Foxy is a vixen” ↔ S entertains the proposition that Foxy is a female fox

(Reiland 2023b, 578).

However, these rules do not fare any better at capturing the intuitive notion of linguistic mistakes. Speakers can believe, judge, and entertain various propositions at the time of the utterance, but this does not guarantee that linguistic mistakes have not occurred. A speaker who mistakenly believes that “vixen” means fox could believe that Foxy is a female fox, yet this belief is not what the speaker wanted to express by the sentence.

The general problem common with the propositions discussed thus far is that there is no necessary link between the mental states and the use of the expression that determines which mental state is relevant for the assessment of the use. A straightforward way to achieve this link is to ask what the speaker believes herself to be doing when she uses the expression. However, this would lead to the problematic conclusion that speakers’ beliefs about the public language meaning are a necessary condition for the occurrence of linguistic mistakes.

Of course, the rules discussed here are only a few candidates for rules of use. Reiland’s goal is to argue for the general plausibility of the rule governance view. Therefore, he is not committed to any specific rules as long as some rules are plausible. However, while there is a long list of potential use-conditions, hardly any reason is given for thinking that any one of these conditions necessarily fails to be fulfilled if and only if a linguistic mistake occurs. At least, based on the examples considered thus far, the prospects do not look promising.

5 What are all These Words for Anyway?

I will now turn to what I take to be the most promising suggestion for use-conditions that Reiland has offered, which is based on what expressions are for. While I will ultimately reject his suggestion for rules of use, I will argue that this promising idea could illuminate the notion of linguistic mistakes independently of the rule-governance view. However, as I will argue, linguistic mistakes understood this way cannot be captured by any rules that are intrinsic to meaning.

In introducing the notion of linguistic meaning, Reiland suggests that it can be explicated in terms of what expressions are “semantically for doing in the language,” which can further be clarified as “what its meaning enables us to use it to do.” For example, “Vixen” is used to talk about female foxes, “Bertrand” is used to talk about Russell, and “I” is used to talk about oneself (Reiland 2023a, 2197).

Asking what the expression is semantically for doing is akin to asking what the artefact function of a tool or the biological function of an organ is. Just as a hammer is for nailing and a heart is for pumping blood, the linguistic function^[12] of “vixen” could be to talk about female foxes and “hello” to be for greeting.

The idea that expressions have functions is nothing new to philosophy. It is commonly asserted that language has a communicative function, and the claim that public language expressions also have linguistic functions can be traced back to at least J. L. Austin (1962). To put this idea more generally, the linguistic function of an expression is to perform locutionary or illocutionary acts (for a more in-depth discussion of this idea, see Millikan 1984, 2005).

Reiland suggests that these semantic functions could be taken as the use-conditions in the rules of use. That is, for a linguistic action to be legitimate, meaningful expressions must be used for their intended purposes. Indeed, Reiland considers just such rules as candidates for the rules of use:

∀S: S may use “vixen” ↔ S is talking about female foxes

∀S: S may use “vixen” ↔ S is expressing the property: x is a female fox

∀S: S may use “Bertrand” ↔ S is referring to Russell (Reiland 2023a, 2197).

The trouble is that if public language determines the meaning of “vixen” and what “Bertrand” refers to, how could one fail to talk about female foxes, express the property of being a female fox, or fail to refer to Russell? In order to fail to satisfy the use-conditions, the speaker would have to use these expressions with different meanings. Note that merely using an expression that sounds or looks the same as “vixen” but with a different meaning does not count as a violation. The expression must be taken as the English language expression for the speaker to violate these rules, and since the meanings of these expressions are determined by the public language, the speaker simply cannot use the expression and fail to satisfy these use-conditions.^[13]

Even though these rules fail to capture the notion of linguistic mistakes, linguistic functions could still help to clarify what is common in the cases of linguistic mistakes discussed thus far. John tried to use “vixen” to talk about foxes, but “vixen” is for talking about female foxes only. Therefore, John made a mistake. Similarly, if Jane confused the philosophers’ names and tried to use “Bertrand” to refer to Frege, she made a mistake because “Bertrand” is for talking about Russell, not Frege. Thus, linguistic mistakes can be defined as:

S misused^[14] expression e ↔ (S tried to use e to φ ∧ the function of e is not to φ)

This definition of linguistic mistakes escapes the problems identified thus far. Since the only thing required from the speaker is that she attempts to use an expression for something, not her success in doing so, advocates of the public language perspective can recognize speakers’ intentions while denying that these intentions determine the conventional meaning of the expressions. Secondly, the discrepancy between the linguistic function and what the speaker is trying to use it for remains, even if the speaker is bound to use the expression for its intended purpose. Thirdly, the intention to use expressions for something does not presuppose that the speaker has false beliefs about the conventional meaning of those expressions. Finally, there is a necessary link between the expression and what the speaker intends to use that expression for. Thus, although the speaker may have various intentions, only the ones that are related to the particular expression at that particular time are relevant when assessing whether the speech act commits a linguistic mistake or not.

This notion of linguistic mistake is also acceptable for the individualist. Despite treating speaker meaning as basic, they typically recognize that expressions can have conventional meanings and that utterances can be seen as tokens of expression types (Davidson 1986, chapt 6; Grice 1989). Thus, individualists can recognize not just communicative intentions which determine the speaker’s meaning, but also the intention to use conventional expressions for their conventional purpose (Bilgrami 1993, 134).

Where does this definition of linguistic mistakes leave the rule-governance view? Although the notion of linguistic mistakes analyzed this way is open to both the individualist and the defender of the public language perspective, only the latter can accept rules based on linguistic mistakes. Since individualists seek to derive the conventional meaning from the speaker’s meaning, they are likely to dismiss the philosophical significance of any rules requiring speakers to conform to the conventional meanings (see Davidson 1986, 252) and instead see these rules stemming from the purely pragmatic utility of using languages as other do (Bilgrami 1993, 134–35; Wikforss 2001, 211).

Assuming then the public language perspective, what rules forbid the use of, say, “vixen” or “Bertrand” if and only if a linguistic mistake has occurred? The desired rule can be acquired from the last discussed candidates by changing the use-conditions from what the speaker does with the expression to what she intends to do with them:

∀S: S may use “vixen” ↔ S intends to talk about female foxes

∀S: S may use “vixen” ↔ S intends to express the property: x is a female fox

∀S: S may use “Bertrand” ↔ S intends to refer to Russell.

Although Reiland himself does not discuss rules such as these, they have important merits, which could make them suitable candidates for his view. First, they avoid the problem with the rules which require the speakers to use expressions with their linguistic function. While the public language picture makes it impossible to use “vixen” and not talk about female foxes, one can try to talk about both male and female foxes using “vixen” but fail to do so. Second, these use-conditions are closer to the use-conditions of “ouch,” since both intending and being in pain are part of speakers’ mental states rather than linguistic acts, such as the act of referring.

However, the problem with this option is that rules seem to be instances of a more general rule-like relationship between means and ends than just the relationship between communicative means and ends. That is, these rules do not appear to be intrinsic to meaning. Rather, they are general instrumental norms that are extrinsic to meaning.

Note that the point is not to claim that since rules are contingent on what the speaker intends to do, these rules are not normative in a strong sense (see Hattiangadi 2006, 228). Indeed, Reiland argues that the rule-governance view could be plausible even if the rules are normative in a weaker formal sense, as long as it is still intrinsic to meaning (Reiland 2023a, 2204; see also Foot 1972 and fn. 10). However, for this argument to work, the intention must occur outside of the rules and not determine whether or not use accords with the rule (e.g., If you intend to speak English, you must follow the rules of “vixen.”) In contrast, in the proposed rules, the intention is a part of the use-conditions of the rules, and thus they determine whether the speaker conforms to the rule. Since the intention plays an essential role in these rules, they should be taken as instances of instrumental rules. Regardless of their normative status, these rules are not intrinsic to meaning.

A further reason to be skeptical about the intrinsicality of rules based on speaker intentions is that they are disanalogous to the intrinsic rules of games. The intrinsic rules of games are not contingent on players’ intentions. Although one might intend and fail to castle, whether one also breaks a rule of chess does not depend on this failure to accord with one’s intentions. One could have a plan to castle but fail to do so by making a legal move with the king, after which castling is no longer permitted. Similarly, while John’s utterance, “There is a vixen,” was not appropriate for his intention to talk about both male and female foxes, it nonetheless corresponds to a meaningful sentence in English – that is, John failed to satisfy his intentions but nonetheless performed a permitted linguistic action.

Could there be thus far undiscussed use-conditions that would manage to capture the notion of linguistic mistakes defined in terms of linguistic functions? Unlikely, since acting in accordance with teleological functions relates to the success and failure of achieving a goal, whereas acting in accordance with rules is a matter of compliance and violation. Using a hammer for something besides nailing is not a violation of the function of the tool. If failing to achieve a goal is to be treated as a violation, the goal itself must be treated as something to be complied with. However, language itself does not tell speakers what expressive goals should be pursued, and therefore the requirement for compliance must come from an extrinsic source, such as speakers’ intentions. This means that there is bound to be an extrinsic component in any rules of use which manages to capture the use in accordance with the linguistic function.

One way to avoid the conclusion that no rules can capture the notion of linguistic mistakes would be to offer a different definition for linguistic mistakes, but before such a definition is available, there is no reason to think that there could be rules of use that can capture the notion of linguistic mistakes. An alternative would be to reject the requirement that “use in accordance with meaning” requires the absence of linguistic mistakes, but this would mean that supporters of the unorthodox rule are in no better position to defend their version of rule-governance than the defenders of the orthodox construal. Recall that the truth rule of meaning was criticized because it forbids actions that seem to warrant no semantic criticism, such as reporting false beliefs. The supporters of the rule of use now must answer to the very same criticism they directed against the truth rule.

6 Conclusions

This paper sought to answer the following questions: 1. What are linguistic mistakes? 2. Can the notion of linguistic mistake be captured by semantic rules? I argued that there are intuitive cases of semantic mistakes despite the arguments to the contrary. While the intuitive cases do alleviate the fear that the notion is inherently unclear, the question remains what the necessary and sufficient conditions for linguistic mistakes are.

I have argued that the proposed semantic rules misclassify intuitive cases of linguistic mistakes. The orthodox truth rule requires speakers to always speak the truth, but some false uses, such as reporting false beliefs, are not linguistic mistakes and some uses are mistakes even if they are true. While the defenders of the unorthodox rule of use argue that their rules do better justice to the notion of linguistic mistakes, the concrete examples of the rules also misclassify linguistic mistakes. Conditions such as what the speaker is thinking about, or what she believes, entertains, or judges, all fail to establish a necessary link between the condition and the particular use. Speakers have various mental states at the time of the expression use, but only some of them are relevant for judging whether they have misused the expression.

I proposed a definition for linguistic mistakes based on the Reiland-inspired idea of linguistic function or, in Reiland’s terms, what expressions are for doing in the language. Instead, I argued that a linguistic mistake occurs when a speaker tries to use rather than what she actually uses an expression for something other than its linguistic function. This discrepancy can be used to account for linguistic mistakes. In short, the mistake was the use of a word that did not suit her intentions. However, this implies that the accordance with the rules of use depends on the speaker’s intentions, which indicates that they should be seen as instances of general instrumental rules, which are extrinsic to meaning.

Notably, the problem of extrinsicality arises not from the analysis of linguistic mistakes in terms of linguistic functions or from Reiland’s public-language perspective, but from the rule-governance view itself. While the expression’s linguistic function is intrinsic to its meaning, the functions themselves do not govern its use. Meaningful expressions are like tools with which one can achieve communicative goals, but the question of whether a use is permitted or forbidden depends on whether the goal is worth achieving. Meaning itself does not tell speakers what to express, only what it can be used expressed. If we are interested only in linguistic mistakes independent of rule governance, then the public-language perspective and what expressions are semantically for doing provide a good foundation for the analysis.

Although the pre-theoretical intuitions relating to the word “mistake” might suggest otherwise, even accepting the existence of linguistic mistakes does not force one to accept that there are intrinsic semantic rules. On the contrary, semantically forbidden uses’ alignment with linguistic mistakes becomes the criterion of adequacy for the rule-governance view.