1 The metalinguistic account

We all know that our way of talking is not without flaws. Sometimes a word we use would be better not used at all, while sometimes “only” the way we use an expression or the meaning that is expressed by it is problematic. How we should talk is a contentious issue and the topic of many disputes between (what I shall call) linguistic reformers and linguistic conservatives. For present purposes, we can understand reformers and conservatives regarding a particular expression e as systematically applying e to different sets of objects. While linguistic conservatives hold on to how e has often been applied in the past (by them and others), reformers apply the expression differently, e.g. to a wider group of objects.Footnote 1

Let us consider an example of a dispute between two speakers, Raphael and Connie.

Raphael: “Trans women are women”

Connie: “No, trans women are not women”.

Raphael is a reformer regarding the term “women”. He applies this term more inclusively than Connie, who is a linguistic conservative regarding the term and systematically applies “women” to a more restricted group of people. How should we interpret Raphael and Connie’s dispute?

According to Plunkett and Sundell’s metalinguistic account (see esp. Plunkett & Sundell, 2013; Plunkett, 2015), the interlocutors in this dispute mean different things by the word “women” such that the contents communicated with their statements are not in any conflict.Footnote 2 Raphael and Connie communicate the propositions that trans women are womenref and that trans women are not womencon respectively, where WOMENREF includes and WOMENCON excludes trans women. In their dispute, the speakers do not mention the term “women”. Still, they convey metalinguistic propositions about the usage of this term, according to the metalinguistic account. Not all disagreements about words have to be expressed explicitly (see esp. Grice, 1975). Regarding the above dispute, the idea, then, is this: Raphael uttering “trans women are women” communicates the proposition (p1) that trans women are womenref while also pragmatically conveying that the term “women” should be used in such a way as to include trans women. Connie replying “No, trans women are not women”, on the other hand, uses “women” to express womencon and thus communicates a proposition that is not in any conflict with p1. However, Connie also conveys the metalinguistic proposition that “women” should not be used in such a way as to apply to trans women. The speakers thus pragmatically convey conflicting propositions about how best to use “women”.Footnote 3,Footnote 4

Interpreting disputes like the above as metalinguistic negotiations goes against how many philosophers in conceptual engineering would interpret disputes between linguistic conservatives and reformers (see e.g. Cappelen, 2018; Ball, 2020; Sawyer, 2020). The metalinguistic interpretation has various merits, though, indicating that the approach deserves (at least) substantial discussion. Let me briefly mention four of these merits before moving on.

Interpreting Raphael and Connie metalinguistically means to charitably interpret them as merely using terms differently instead of communicating any a priori, conceptually false assumption (see also Hirsch, 2005, on charity to understanding). According to Plunkett and Sundell’s metalinguistic approach to the above dispute, the disputants apply the word “women” differently and both of them communicate thoughts that are true. Consequently, it is not the case that the two parties employ the same concept of a woman and Connie simply makes an a priori conceptual mistake about whether trans women fall under that concept or not. Instead, the speakers are merely employing different concepts. As a result, the metalinguistic account can stick to the idea that “typical speakers of a language have a sufficiently adequate grasp of their linguistic and conceptual resources so that they don’t generally make a priori (conceptually) false assertions” (Hirsch, 2005, p. 72).

Nonetheless, proponents of the metalinguistic picture can identify Connie as making a significant mistake of the right kind. This is a second merit of the account. Intuitively, linguistic conservatives are making a moral mistake when refusing to call trans women “women”. It is morally wrong to use “women” trans-exclusively and this kind of mistake is exactly what the metalinguistic account is able to deliver. Interpreting the speakers as pragmatically communicating propositions about how we should use “women” provides the possibility to understand them as having a morally significant disagreement about concept choice.Footnote 5

Thirdly, the metalinguistic account can easily accommodate our feeling of genuine disagreement between Raphael and Connie.Footnote 6 According to proponents of the metalinguistic account, the speakers’ disagreement is to be localised at the level of pragmatics. While Raphael pragmatically conveys the metalinguistic thought that “women” should be used in such a way as to apply to trans women, Connie disagrees. She pragmatically communicates the opposite. Hence, even though the speakers communicate compatible propositions about womenref and womencon, they also clearly pragmatically convey conflicting metalinguistic contents. Their dispute thus reflects genuine disagreement about concept choice. Although they are using “women” differently, Raphael and Connie are not merely talking past each other.

Fourthly, the metalinguistic account has a straightforward story to offer on why Raphael and Connie are talking and disagreeing about the same topic if engaged in a metalinguistic negotiation. The issue of topic continuity, going back to Strawson (1963), has been widely discussed in the literature on conceptual engineering (see Haslanger, 2012, ch. 7; Cappelen, 2018; Prinzing, 2018; Sawyer, 2020; Knoll, 2020; Nado, 2021; Flocke, 2021; et al.), so let me briefly elaborate on this advantage. The question of which topic speakers are concerned with in a dispute is a matter of pragmatics – or so proponents of the metalinguistic picture could argue (see also Jenkins, 2014, pp. 27–28).Footnote 7 To see why this assumption sits well with (at least some of) our intuitions about disputes and their topics, let me briefly illustrate this point by considering the following example:

Bib: “The smartest philosopher of all time requests a chocolate cake for her birthday.”

Bob: “I disagree. If I remember correctly, the smartest philosopher of all time actually wants to eat cherry cake on her birthday.”

Bib and Bob agree that their five-year-old daughter is not the smartest philosopher of all time (at least not yet). They agree that this is Plato, and let’s assume – for the sake of argument – that Bib and Bob are right about this. Still, as their daughter recently started to ask a lot of puzzling philosophical questions, Bib and Bob ironically mean to refer to her by using “the smartest philosopher of all time” in the context at hand. What, then, is Bib and Bob’s dispute about? Which topic are they concerned with: Plato or their daughter’s wish for cake? The answer seems clear: Bib and Bob are disputing about which kind of cake their daughter wants (although there is also an important sense in which they, or their sentences, are saying something about Plato). Coming back to our dispute between Connie and Raphael, we can now see why interlocutors in metalinguistic negotiations can easily be construed as being concerned with the same metalinguistic topic. Seeking to identify the topic of a dispute amounts to finding out what people pragmatically convey with their utterances. Topics of disputes are a matter of pragmatics, not semantics. To be sure, Connie and Raphael plausibly agree that trans women are womenref and not womencon. According to the metalinguistic approach, however, what Connie and Raphael also mean by uttering “Trans women are not women” and “Trans women are women” respectively, is how they think speakers should use the term “women”. Consequently, if we interpret the speakers as being engaged in a metalinguistic negotiation, then we can easily interpret them as genuinely disagreeing about the same topic: the topic of how best to use a particular term.

Adopting the metalinguistic account, then, combines various merits (for more on its upsides, see also Sect. 5). This is not to say that the metalinguistic account answers all relevant questions about disputes and disagreements in connection to conceptual engineering (e.g. questions on the nature of concepts/meanings in general). Moreover, some of its assumptions, especially the ones on topic continuity, have been challenged in the literature (see e.g. Sawyer, 2020; Ball, 2020) and the account might also face certain problems.Footnote 8 Given its merits, though, I think that the metalinguistic approach to disputes between linguistic conservatives and reformers is worth detailed discussion. In what follows, we need not assume that all disputes between linguistic conservatives and reformers are metalinguistic negotiations. To get the discussion going, I will only assume that a notable number of them are. I think that disputes and disagreements come in all kinds of shapes and colours and conceptual engineers should generally be open to adopting different interpretations for different disputes – including metalinguistic interpretations.Footnote 9

In what follows, I will discuss a major objection to the metalinguistic account raised by Cappelen (2018). This objection is meant to cast doubt on metalinguistic interpretations in general and appears to spell trouble even for the moderate idea that many disputes between linguistic conservatives and reformers are best interpreted as metalinguistic negotiations. As a corollary, Cappelen’s objection also threatens metalinguistic interpretations in other areas of philosophy, such as metaethics. As will become clear, Cappelen’s objection against the metalinguistic account does not succeed. Discussing the objection in full detail, however, further develops the metalinguistic approach beyond existing discussions of it in the literature.

2 The Babel objection

In his seminal Fixing Language (2018), Cappelen rejects the metalinguistic approach to conceptual engineering. He raises the following concern.

For the sake of argument, assume that the dispute between Raphael and Connie considered above is a metalinguistic negotiation. Let us say, it reflects disagreement on how best to use the specific English term “women”. According to Cappelen, it is then hard to explain why the disputants do not see their “concerns and arguments […] [as] irrelevant to someone who speaks, say, Icelandic, Chinese, or Russian” (2018, p. 174; emphasis added). Why do Connie and Raphael take “themselves to be agreeing and disagreeing with those talking about the same issue in one of those other languages” (2018, p. 174)? And why are we of the impression that there is disagreement between Connie and, say, a German reformer uttering “Trans Frauen sind Frauen”? Intuitively, “there is dis/agreement between speakers of different languages” (2018, p. 174). Yet, according to Cappelen, the metalinguistic account fails to explain why.

Cappelen has a point. We certainly want an account of disputes between linguistic conservatives and reformers that is able to construe e.g. German reformers uttering “Trans Frauen sind Frauen” as being in relevant disagreement with English conservatives (such as Connie) replying “Trans women are not women”. On Cappelen’s Austerity Framework, for instance, these speakers disagree about women and what they are. His account thus avoids the problem of accommodating disagreement across languages – and so do all other accounts which do not construe reformers and conservatives as “merely” negotiating the usage of a particular word. On the metalinguistic account, however, the German and English speakers are concerned with different words: “Frauen” and “women” respectively. While the English conservative uses “women” to express WOMENCON and pragmatically conveys the proposition that the English term “women” should not be used such as to include trans women, the German reformer uttering “Trans Frauen sind Frauen” uses “Frauen” to express WOMENREF and pragmatically conveys the proposition that the German term “Frauen” should be used such as to include trans women. On the metalinguistic account, the speakers are thus “talking about how to define different words” (2018, p. 174), which is why this account is unable to explain the intuition of disagreement across languages, according to Cappelen. I shall call this worry the Babel objection.

3 Generalising the Babel objection

The Babel objection not only matters with respect to metalinguistic approaches to disputes between linguistic conservatives and reformers. That is, Cappelen’s objection not only matters for conceptual engineering. It also generalises to metalinguistic accounts in other areas of philosophy, such as metaethics.

Let us start by focusing on the objection’s general relevance for interpreting disputes between linguistic conservatives and reformers. For the sake of the argument, assume that Cappelen’s objection succeeds. Proponents of the metalinguistic account would then have to ask with regard to any dispute between a conservative C and a reformer R: would C and/or R relevantly disagree with somebody (S) talking back at them in a different language? If yes, then interpreting C and R’s dispute as a metalinguistic negotiation would have to be avoided. For, interpreting their dispute as a metalinguistic negotiation means to interpret C and R as negotiating the usage of a particular word of their language (e.g. English). In what way, then, would a possible speaker S relevantly disagree with C and R if S only spoke about a term of S’s language (e.g. Spanish)? Apparently, Cappelen’s objection affects disputes in conceptual engineering in general. Whenever it seems justified to think that a conservative C and/or a reformer R would disagree with a possible speaker’s S reply in a different language, a metalinguistic interpretation of C and R’s dispute could not be right, even if C and R are speaking the same language.

What is more, the Babel objection not only generalises to intra-linguistic disputes between linguistic reformers and conservatives. It also affects metalinguistic accounts defended in other areas of philosophy. As some have argued in metaethics, for instance, the metalinguistic account proves to be a useful interpretative approach to moral disagreements between speakers of different communities. Let me give two examples for this line of argument:

  1. i.

    Plunkett and Sundell (2013, pp. 1–3) discuss Hare’s (1991) case of a missionary and an Indigenous person who apply the word “good” to different objects in the course of their dispute. If we assume that these speakers attach different semantic meanings to “good”, then, according to Hare (1991, pp. 148–149), we would not be able to accommodate our impression of relevant disagreement between them. Plunkett and Sundell disagree. As they point out, even if the speakers used “good” with different meanings, they could be engaged in a metalinguistic negotiation on how best to use this term. Consequently, the intuitive datum of genuine disagreement between the Indigenous person and the missionary does not warrant the semantic conclusion that “good” is used with the same meaning during their dispute.

  2. ii.

    As Sodoma (2021) points out, moral relativism faces the challenge of accommodating epistemically significant disagreement between speakers of different communities. She argues that interpreting such “moral inter-group disagreements” as metalinguistic negotiations on how best to use e.g. “morally wrong” (i.e. according to which set of moral standards) could do the trick. On her metalinguistic picture, moral disagreements between speakers of different communities could be construed as being epistemically significant if we understand them as metalinguistic negotiations. The metalinguistic approach thus enables moral relativists to reinforce their account, according to Sodoma.

The details of these metalinguistic interpretations of moral disputes will not matter for the purposes of this paper. It suffices to note that, much like the metalinguistic approach to conceptual engineering, Plunkett and Sundell’s answer to Hare as well as Sodoma’s relativist account depend on a successful response to the Babel objection. In fact, the specific cases under discussion in their papers provide particularly strong reasons for why the Babel objection should better be unsound. After all, speakers involved in real “moral inter-group disagreements” (Sodoma, 2021) – such as the missionary and the Indigenous inhabitant of some “distant island” (Plunkett & Sundell, 2013, p. 1), for instance – are usually speaking different languages. If authors construe such speakers as genuinely disagreeing about how best to use moral vocabulary, then there seems to be a particular need to explain how exactly metalinguistic negotiations work across languages.

4 Dispelling the Babel objection

As we have seen, accommodating disagreement between speakers of different languages constitutes a general challenge to the metalinguistic account. This challenge has so far not been taken on by proponents of the metalinguistic picture. Pace Cappelen, however, the Babel objection can be successfully answered. This section will show how. The next section will then discuss a different metalinguistic response to the Babel objection.

Let us start with a general observation. It is possible that disputants genuinely disagree about one and the same topic even when they do not address that topic in their dispute. Speakers can be in a state of disagreement without explicitly expressing this disagreement in their linguistic exchange. So, German reformers and English conservatives could well be concerned with different words in their statements but still disagree over the truth of what each of them is communicating. To be sure, on the metalinguistic account, these speakers are talking about slightly different metalinguistic topics: one of them is talking about how best to use the particular English word “women” while the other is concerned with the proper usage of the German “Frauen”. But it is an important first step to note that this divergence in topics does not imply any lack of disagreement between them. To briefly illustrate this point, consider the following dispute:

Claire: “Snow is great! It makes the world look wonderful.”

Francis: “I disagree. I used to think the same but now I know that snow isn’t a great substance at all.”

Assume that in the context of their dispute, Claire means to convey that snow is great while Francis means to convey that cocaine is not great. Hence, Claire and Francis are talking about different topics on the account outlined: snow and cocaine. It is still a possibility, however, that Claire thinks that cocaine is a great drug while Francis disagrees with Claire’s assessment of snow. Francis might hate snow. Just because speakers are talking about different topics, they need not fail to disagree on these topics. They can still be in a state of disagreement on what each of them communicates (see also Chalmers, 2011, p. 526 on this).

Coming back to Cappelen’s Babel objection, the question, then, is this: are there any reasons to assume that English conservatives and German reformers do have conflicting attitudes in virtue of which they are in a state of disagreement? And if yes, is this disagreement somehow reflected in their dispute, even though the propositions which the speakers pragmatically convey in their dispute do not conflict? Answering these questions in the affirmative, I take it, are the two desiderata that the metalinguistic approach will have to meet to refute the Babel objection. To see how the metalinguistic account can be brought to satisfy these desiderata, let us start with an example.

Bruno and Kim, let us assume, are botanists. They know that from a botanical point of view, strawberries are not berries while bananas are. Now, imagine a situation in which Kim is eating strawberries. “Yum…”, she says to Bruno, “strawberries are definitely my favourite berries.” Subsequently, the two of them engage in the following dispute:

Bruno: “Strawberries are not berries, Kim.”

Kim: “Oh, come on, Bruno. Bananas are not berries!”

Let us assume that Bruno and Kim pragmatically convey metalinguistic propositions on how best to use the term “berries”. Bruno communicates that the term “berries” should not be used in such a way that it applies to strawberries (p3) while Kim pragmatically conveys that “berries” should not be used in such a way that the term applies to bananas (p4). Thus, the metalinguistic propositions conveyed by the speakers do not conflict. They are consistent and concern slightly different topics. Still, we typically take Kim’s “Bananas are not berries” to be a relevant reply to Bruno’s utterance, and we intuitively presume that their linguistic exchange reflects relevant disagreement, although p4 does not directly contradict p3. How can we accommodate this intuition?

I think that there is a simple explanation available. The intuition of relevant disagreement arises because there is good reason to suspect that although the metalinguistic propositions p3 and p4 are consistent, the speakers still disagree about whether p3 and p4 are true because of a more fundamental metalinguistic disagreement between them in the background of their dispute. Interpreting the dispute, we presume that Bruno has some subjective reasons to communicate p3, and most of us will quickly conjecture that a crucial reason for Bruno to convey p3 would be the belief that the usage of “berries” should follow plant taxonomy. That metalinguistic belief in the background easily explains why Bruno advocates against using “berries” for strawberries. From a botanical point of view, strawberries are not berries. Hearing Kim’s reply, on the other hand, we charitably assume that her answer is supposed to signal some relevant disagreement with Bruno, and we note that it would do so if Kim disagreed with Bruno on whether the usage of “berries” should follow plant taxonomy. A disagreement in the background on whether it is best to follow plant taxonomy would easily explain why Kim engages in the dispute with Bruno by advocating against using “berries” for bananas and why her reply also signals relevant disagreement on whether it is best to use “berries” for strawberries. The metalinguistic disagreement in the background relevantly connects the two topics that Bruno and Kim are addressing. It is because the speakers disagree on whether it is best to follow plant taxonomy that they engage in this dispute and disagree on the truth of p3 and p4.

Coming back to the Babel objection, proponents of the metalinguistic account can base their reply to Cappelen’s worry on an analogous explanation. Consider the following dispute:

G(erman) L(inguistic) R(eformer): “Trans Frauen sind Frauen.”

E(nglish) L(inguistic) C(onservative): “No! Trans women are not women.”

On a metalinguistic reading of this dispute, GLR employs the wider concept of womanhood WOMENREF and pragmatically communicates that the term “Frauen” should be applied in such a way as to include trans women (p5). Speaker ELC, in contrast, employs the narrower concept of womanhood WOMENCON while pragmatically conveying that this usage of “women” is appropriate (p6). So, in contrast to Connie and Raphael’s same-language dispute about the usage of “women”, not even the metalinguistic propositions conveyed by GLR and ELC (p5 and p6) are in any direct conflict. Strictly speaking, the disputants are concerned with different words and slightly different topics. In this respect, then, ELC and GLR’s inter-language dispute is importantly disanalogous to the same-language dispute between Connie and Raphael. Connie and Raphael convey inconsistent metalinguistic propositions about the usage of “women” while the metalinguistic propositions communicated by GLR and ELC are consistent, according to the metalinguistic account. Does that mean that the metalinguistic account cannot explain why, intuitively, GLR and ELC disagree with what their interlocutor conveys (respectively)? I do not think so. The metalinguistic account can interpret GLR and ELC as being in a state of genuine disagreement on p5 and p6 that is reflected in their linguistic exchange.

In close analogy to Kim and Bruno’s case, the intuition of disagreement arises because we assume that there is a more fundamental metalinguistic disagreement between GLR and ELC in the background of their dispute in virtue of which the speakers engage in their dispute. Quite generally, different metalinguistic disagreements in the background are possible (see also the next section). However, in light of GLR’s utterance, I take it that we will typically interpret GLR as holding the more general metalinguistic belief that people’s self-identification should guide the usage of gender vocabulary and that any language should therefore appropriately respect how trans women self-identify. This belief in the background easily explains why GLR advocates for using “Frauen” for trans women. What is more, it also explains why GLR disagrees with the conservatives’ usages of “women” in English-speaking countries, and with ELC’s trans-exclusive usage of this term in particular. That is, the metalinguistic belief in the background accounts for GLR’s disagreement with ELC on what ELC pragmatically conveys about the usage of “women” (p6).

Hearing ELC’s reply, on the other hand, we charitably assume that her answer is supposed to signal some relevant disagreement with GLR, and we quickly get that it would do so if ELC disagreed with GLR on whether languages in general should reflect gender self-identification. This metalinguistic disagreement in the background would easily explain why ELC’s reply signals relevant disagreement with GLR on whether it is best to use the German “Frauen” for trans women and why she thinks that the term “women” should not be used for trans women. To put it more generally, we assume that the fact that ELC sees good reasons to pair “women” with WOMENCON is reflected in her advocating for this very word-concept pair. And we intuitively conjecture that her subjective reasons to pair “women” with WOMENCON also support “Frauen”-WOMENCON.

In a nutshell, then, the idea of a metalinguistic answer to Cappelen’s Babel objection is this. The speakers disagree on the truth of p5 and p6 because they disagree on whether languages should appropriately respect how trans women self-identify. More generally speaking, in most salient contexts, we intuitively assume that GLR and ELC hold some relevant and more fundamental conflicting metalinguistic beliefs in the background of their dispute. This disagreement in the background grounds the speakers’ metalinguistic disagreement on p5 and p6 and it explains why the speakers engage in their dispute. It is true that on the metalinguistic account, ELC and GLR pragmatically communicate consistent metalinguistic propositions (as do Bruno and Kim). As they are speaking different languages, they are speaking about slightly different metalinguistic topics. Yet, even on the metalinguistic account, our intuition of genuine disagreement can be explained. When encountering a dispute such as GLR and ELC’s, we typically interpret the speakers as disagreeing on a more fundamental metalinguistic question that grounds the speakers’ disagreement regarding p5 and p6. Encountering the dispute, most of us will naturally assume that GLR wants any language to be trans-inclusive while ELC does not. This explains why we intuitively assume that their linguistic exchange signals relevant disagreement.

To be sure, this explanation of disagreement is not quite as straightforward as the explanation of disagreement in same-language cases like Connie and Raphael’s. After all, the explanation of inter-language disputes holds on to the assumption that the metalinguistic propositions pragmatically conveyed by the speakers in such cases are consistent. Still, it is important to note that proponents of the metalinguistic account can accommodate relevant disagreement even in inter-language cases, and that they can do so by drawing on assumptions that also explain our feeling of disagreement in many same-language cases like Kim and Bruno’s. So, even though ELC and GLR negotiate the meaning of two different words, they are engaged in a dispute that is clearly not a merely verbal dispute on the metalinguistic picture.

What is more, the present response sufficiently generalises. Assume that another speaker SLR joins the discussion who speaks a third language and advocates pairing WOMENREF with a third term t that is a translation of “women”/“Frauen” into SLR’s language. (For instance, SLR might say: “Las mujeres trans son mujeres.”) On the picture outlined, then, SLR is likely to relevantly disagree with ELC but to agree with GLR even if we interpret all speakers involved as conveying consistent metalinguistic propositions. That is because it seems very likely that SLR has some subjective reasons to advocate the word-concept pair t-WOMENREF which generalise to pairing “Frauen” as well as “women” with WOMENREF, and which are not shared by linguistic conservatives such as ELC. (Probably, SLR, too, thinks that people’s self-identification should guide the usage of gender vocabulary of any language, which gives her a reason to reject p6.) Thus, intuitively, SLR appears as signalling relevant disagreement with ELC’s position on “women”-WOMENCON but as being on GLR’s side regarding “Frauen”-WOMENREF.

Coming back to possible examples in metaethics, the present answer to the Babel objection also covers disputes between speakers of different moral communities. Take two speakers engaged in a moral dispute about animal ethics. One (E1) utters “Eating animals is wrong” while the other speaker (S1) replies in Spanish: “No! Comer animales no está mal”. On Sodoma’s (2021) relativist proposal, both statements express a truth, and E1 and S1 are engaged in a metalinguistic negotiation. E1 advocates using “wrong” with reference to one set of moral standards #1 while S1 pragmatically conveys that “mal” should be used with reference to another set of moral standards #2. How, then, are the speakers disagreeing if they are talking about different words and thus different topics? Again, the answer can be developed in analogy to my reply above. What Sodoma and colleagues need to identify is a relevant metalinguistic disagreement between the disputants in the background of their dispute which explains why the speakers disagree on the specific metalinguistic proposition that their interlocutor is pragmatically communicating. Generally speaking, the two speakers have some respective reasons for using “mal” and “wrong” according to different sets of moral standards, and these reasons clash. Thus, E1’s reasons in the background for using “wrong” with reference to #1 are reasons to use “mal” with reference to #1, too. And the reasons that S1 has in the background for using “mal” with reference to the other set of moral standards #2 are also reasons to use “wrong” with reference to #2. This, then, accommodates our impression of relevant disagreement between S1 and E1.

5 Translations: a different approach?

In Fixing Language (2018), Cappelen briefly discusses a different metalinguistic solution to the Babel objection than the one presented in the last section. He suggests that “Plunkett and Sundell could liberate their view from th[e] focus on specific lexical items if it appealed to translations” (2018, p. 175). Regarding the dispute between GLR and ELC

G(erman) L(inguistic) R(eformer): “Trans Frauen sind Frauen.”

E(nglish) L(inguistic) C(onservative): “No! Trans women are not women.”

this possible variant of the metalinguistic account would claim that what “the speakers disagree over are all the lexical items that are translations” (2018, p. 175; my emphasis) of the expressions under discussion (i.e. “Frauen” and “women”).

If the speakers make a claim about all those words, then we would guarantee disagreement between speakers of different languages and the [Babel] objection […] would be circumvented. (Cappelen, 2018, p. 175)

I interpret this idea of how to modify the metalinguistic account in light of the Babel objection as follows. According to Cappelen’s suggestion, GLR employs the wider concept of womanhood WOMENREF and pragmatically communicates that the term “Frauen” as well as all the lexical items that are translations of “Frauen” (including “women”) should be applied in such a way as to include trans women. Speaker ELC, in contrast, employs the narrower concept of womanhood WOMENCON while pragmatically conveying that this trans-exclusive usage of “women” as well as a trans-exclusive usage of all the lexical items that are translations of “women” (including “Frauen”) is appropriate. Hence, the speakers pragmatically convey conflicting propositions about how to best use terms of different languages. Let us call this possible variant of the metalinguistic account the (metalinguistic) translation account. At least at first glance, this account circumvents the Babel objection. It seems to deliver a straightforward explanation of our intuition of disagreement between ELC and GLR.

According to Cappelen, however, the metalinguistic translation account runs into a problem, which he briefly describes as follows:

[I]t won’t work to appeal to the idea of translation. We have to choose which meaning to translate from – if we pick two different meanings, then they won’t be picking out the same set of expressions. If we pick the same one, we will bias the debate in favor of one speaker. (2018, p. 175)

This section discusses the translation account and Cappelen’s objection to it. In my opinion, the discussion adds some welcome complexity to the metalinguistic picture. It provides the opportunity to further illustrate the last section’s answer to the Babel objection (an answer that relevantly differs from the translation solution suggested by Cappelen), and it points to various important but as yet underexplored features of metalinguistic negotiations (such as the role of context).

5.1 Cappelen’s objection to the metalinguistic translation account

On the metalinguistic translation account discussed by Cappelen, GLR and ELC use “Frauen” and “women” to express different concepts, WOMENREF and WOMENCON. (That assumption is in line with what Plunkett and Sundell’s (2013) account suggests.) Moreover, GLR and ELC pragmatically convey conflicting propositions about how best to use “Frauen” and “women” as well as all those other terms that are translations of “Frauen” and “women” (respectively). Despite its initial appeal, though, we should not adopt the translation approach to dissolve the Babel objection, according to Cappelen. To further illustrate his concern, it is helpful to briefly discuss another inter-language case first (see also Plunkett & Sundell, 2013, p. 27):

Britney (pointing to French fries): “These are chips.”

Gerhard: “Nein, das sind keine Chips.”

Let us assume that Britney (who is British) uses the English term “chips” to express CHIPSFRIES. Moreover, suppose for the sake of argument that what Britney pragmatically communicates with her utterance is the metalinguistic proposition that “chips” as well as all translations of “chips” should be used in such a way that they apply to fries. Gerhard, on the other hand, is German. He uses the term “chips” to express CHIPSCRISPS. Furthermore, say, by uttering “Nein, das sind keine Chips” Gerhard pragmatically conveys a metalinguistic proposition about which usages of “chips” as well as all translations of “chips” into different languages are appropriate.

In my opinion, this example helps to illustrate Cappelen’s point on why proponents of a metalinguistic view should not bring translations into the picture to answer the Babel objection. As Cappelen points out, “[w]e have to choose which meaning to translate from” (2018, p. 175), i.e. regarding the Britney-Gerhard case, we have to decide on whether to choose CHIPSFRIES or CHIPSCRISPS as the meaning to translate from. If we picked CHIPSFRIESand CHIPSCRISPS as the meanings to translate from, then we would pick out two different sets of expressions. On the other hand, if we picked e.g. only CHIPSFRIES as the meaning to translate from, then Britney would communicate the metalinguistic thought that “chips” as well as all translations of “chips” meaning CHIPSFRIES (i.e. “Pommes”, “frites”, “French fries”, “patatas fritas”, etc.) should be used in such a way that they apply to fries. Gerhard, on the other hand, would convey that all translations of “chips” meaning CHIPSFRIES should not be applied to fries. In the scenario detailed, we would consequently bias the debate in favour of Britney, just as Cappelen cautions against, provided that terms such as “frites”, “Pommes”, “French fries” etc. should in fact be used for fries. So, if we only picked CHIPSFRIES as the meaning to translate from, then Gerhard would convey a metalinguistic proposition that seems to be false. (The same would hold, mutatis mutandis, for Britney and CHIPSCRISPS.)

Now, this line of reasoning might not seem too troubling when looking at Britney and Gerhard’s dispute, which might come across as being merely verbal anyway. However, an analogous worry also affects GLR and ELC’s dispute, i.e. a dispute which clearly elicits the impression of relevant disagreement between the speakers. Which meaning expressed by GLR and ELC – WOMENREF or WOMENCON – should we pick to translate from when interpreting GLR and ELC as conveying metalinguistic propositions not only about the specific terms that they are using (“Frauen” and “women”) but also about all those words that are translations of these terms, as proponents of the metalinguistic translation account would suggest? Let us assume for a moment that WOMENREF and WOMENCON do translate differently into different languages.Footnote 10,Footnote 11 Thus, suppose that there is a language L in which speakers predominantly express WOMENREF with one term (e1) and predominantly express WOMENCON with another term (e2). In this case, we would, as before, pick out two different sets of expressions (one containing e1 and one containing e2) if we picked both meanings (WOMENREF and WOMENCON) to translate from in the case of GLR and ELC’s dispute. Then, however, GLR would convey a proposition about how best to use the set of expressions containing e1 in the dispute while ELC would communicate a metalinguistic thought on how best to use the set of expressions containing e2. Hence, the metalinguistic propositions conveyed by the speakers would concern different sets of expressions. What is more, assuming that GLR would agree that e2 should not be used in such a way that e2 applies to trans women, GLR would fail to disagree with ELC on how best to use e2 (analogously for ELC and e1). On the other hand, if we picked only one of the two meanings to translate from – WOMENREFor WOMENCON – then we would bias the debate in favour of one speaker. Picking WOMENREF, for example, we would bias the debate in favour of GLR, provided that all translations of “Frauen” typically used to express WOMENREF (such as e1) should in fact be used in such a way that they apply to trans women. For in this case, ELC’s metalinguistic proposition that we should not use all translations of “women” (including e1) in such a way that they apply to trans women would be wrong.

Consequently, the metalinguistic translation account runs into a dilemma, just as Cappelen predicts. The account is not able to accommodate the impression of disagreement in all inter-language disputes. Bringing translations into the picture is therefore of no help for proponents of a metalinguistic view to provide a general answer to the Babel objection.

5.2 What can we learn from Cappelen’s objection?

Section 5.1 elaborated Cappelen’s objection to the metalinguistic translation account. In the remainder of this section, (1) I will draw attention to an important caveat regarding Cappelen’s objection. Moreover, (2) I will revisit the solution to the Babel objection presented in Sect. 4 and detail the role of context in metalinguistic negotiations.

According to Cappelen, adopting the metalinguistic translation account amounts to biasing an inter-language dispute (such as GLR and ELC’s) in favour of one speaker if we only pick one of the relevant meanings (e.g. WOMENREF) to translate from. Regarding a good many cases, I think that Cappelen is exactly right on this. His objection to the metalinguistic translation account is therefore successful. The metalinguistic translation account cannot provide a general strategy to answer the Babel objection. However, it is worth noting that even if “women”-WOMENREF translates as e1 into L (because e1 in L is predominantly used to express WOMENREF), the best way for speakers of L to use e1 might not be to express WOMENREF. After all, it is not always best to use a term in accordance with its predominant usage. In particular, it might not always be best for speakers of L to use e1 in such a way that e1 applies to trans women. But if the best way to use e1 is not to express WOMENREF, then we would also not automatically bias GLR and ELC’s dispute in favour of GLR by picking WOMENREF as the meaning to translate from. For, even if e1 is predominantly used to express WOMENREF in L, ELC would then be right in arguing that e1 should not be used in such a way that it applies to trans women by speakers of L. Moreover, GLR and ELC could actually be in agreement on how best to use e1.Footnote 12

On a more general note, it is important to bear in mind that speakers in metalinguistic negotiations can advocate a particular usage of an expression – and rightly so – even if they know that this usage deviates from how the expression is predominantly used. To be sure, speakers can also be in genuine disagreement on how best to use a particular expression e because they disagree on how e is predominantly used. Interlocutors might think that all expressions of a language should best be used in accordance with how the majority of speakers of this language uses them, for instance. However, interlocutors can certainly also disagree on how best to use e for different reasons. In fact, interlocutors in metalinguistic negotiations can advocate a certain usage of e even when they know that their preferred usage of e would vastly deviate from how e is used in a linguistic community. Speaker ELC, for instance, might know that almost all speakers of English nowadays use “women” to express WOMENREF but still advocate pairing “women” with WOMENCON. As Burgess and Plunkett (2013, p. 1094) clarify in their introduction to conceptual ethics, the “should” or “best” in “how we should use a term” or “how best to use a term” is to be understood broadly, concerning all kinds of “normative and evaluative theorizing” (see also Thomasson, 2017, p. 12). Speakers in metalinguistic negotiations might thus disagree on how best to use a particular term because they disagree on which of the concepts under discussion would be more joint-carving, morally superior, … or superior all things considered. I think that it is important to keep this manifoldness of speakers’ intentions and attitudes in mind when interpreting metalinguistic negotiations.

Does Cappelen’s worry affect the last section’s answer to the Babel objection? I do not think so. According to the answer presented in the last section, ELC and GLR (respectively) convey metalinguistic propositions about “women” and “Frauen” only. Thus, the last section’s answer to the Babel objection does not construe the speakers as conveying metalinguistic propositions about “women” and “Frauen” as well as all the lexical items that are translations of these terms. In fact, the last section’s solution does not appeal to the idea of translation at all. In this sense, my solution is not only less demanding but also unaffected by Cappelen’s worry about the metalinguistic translation account.

Let me emphasise one further detail about metalinguistic negotiations that is often disregarded in the literature (e.g. by Cappelen, 2018, p. 174). So far, we have often talked as if ELC and GLR’s metalinguistic negotiation would simply concern how best to use two different concatenations of letters: W-O-M-E-N and F-R-A-U-E-N. But as Plunkett and Sundell emphasise, metalinguistic negotiations concern the question of “how best to use a word relative to a context” (2013, p. 3; italics added). Plunkett and Sundell see functional roles as part of that context. According to them, in a particular context (“setting”), “certain words (largely independent of which specific concept they express) fill specific and important functional roles in our practices” (2013, p. 20). The idea of functions or functional roles of concepts and/or terms is highly debated in the literature on conceptual engineering. (For instance, the accounts of Prinzing, 2018, Thomasson, 2020 and Nado, 2021 appeal to functions while Cappelen, 2018, ch. 16 and Riggs, 2021 take a more critical stand on the functional approach to conceptual engineering.) Luckily, we do not need to stick our necks out on this issue here. Proponents of the metalinguistic account need not commit to functions as being part of the contexts relative to which speakers in metalinguistic negotiations advocate the usage of certain terms. Independently of whether proponents of the metalinguistic account commit to functions, they can interpret ELC and GLR’s dispute as follows: GLR advocates how best to use “Frauen” within the German-speaking linguistic community – i.e. a community in which “Frauen” is typically used in a particular way and in which it has certain connotations, colourings, figurative usages, lexical effects, (functional roles?) etc. All of these semantic and non-semantic features are features of the context relative to which GLR advocates how best to use “Frauen” in his dispute with ELC. ELC, on the other hand, advocates against using “women” for trans women within the English-speaking linguistic community in which, again, “women” is typically used in a particular way and in which it has all kinds of further features, which are part of the context relative to which ELC advocates how best to use “women” in his dispute with GLR.

Hence, ELC and GLR negotiate how one should use two different words relative to two different contexts. However, as we have detailed ELC and GLR’s case so far, these two contexts, while being different, are also notably similar regarding “women” and “Frauen”. That is to say that “women” and “Frauen” have very similar – if not the same – semantic and non-semantic features within the English and the German linguistic community respectively, features which should be seen as part of the contexts relative to which ELC and GLR negotiate how best to use those terms. This similarity assumption underlies our strong feeling of relevant disagreement between ELC and GLR. If two or more speakers discuss how to use two or more expressions e1, e2, … relative to two or more different contexts c1, c2, …, then it is likely that the speakers’ dispute really reflects disagreement on how best to use e1, e2, … relative to c1, c2, … only if e1, e2, … have relevantly similar features in c1, c2, ….

Just hypothetically assume that, unlike “Frauen” in German, the term “women” in English had severely negative lexical effects and connotations. That is, assume that the contexts relative to which ELC and GLR discuss how to best use “women” and “Frauen” respectively are not relevantly similar regarding these two terms. In this case, there might actually be no disagreement between ELC and GLR on which concept of womanhood best to pair with the English “women”. Just assume that in general, say, GLR thinks that gender vocabulary should be used in accordance with how persons self-identify. However, GLR is uncertain whether calling trans women “women” would really be the best choice in a scenario in which “women” is extremely negatively coloured. In the situation as described, we might expect GLR to refuse a trans-inclusive usage of “women” within the English-speaking linguistic community.Footnote 13 Note, however, that whether ELC and GLR would really disagree on how best to use “women” eventually depends on how exactly we detail the case and on GLR’s doxastic attitudes in particular. Sometimes, for instance, speakers also want to see negatively coloured terms reappropriated or decide to deliberately use terms disruptively (see Sterken, 2020). In general, adding sufficient details to a case under discussion may dissolve the impression of disagreement. Varying details about cross-linguistic disputes call for varying interpretations. The metalinguistic account is flexible enough to accommodate that. It can react to specific details of each case and evaluate individual cases individually.

Let me briefly recap what we have learned in this section. This section discussed (what I called) the metalinguistic translation account, which Cappelen (2018) briefly presents as a possible variant of the metalinguistic account. At least at first glance, the metalinguistic translation account might seem to circumvent the Babel objection. Yet, the translation account does fall prey to Cappelen’s objection (Sect. 5.1). As a result, proponents of the metalinguistic picture should not adopt the metalinguistic translation account to answer the Babel objection.

In Sect. 5.2, I raised a caveat about Cappelen’s objection to the translation account. As I have argued, we do not automatically bias a dispute in favour of one speaker if we only pick one meaning to translate from. As we would do so in a good number of cases, however, adopting the metalinguistic translation account still does not provide a general solution to the Babel objection. In addition, I underlined the difference between the metalinguistic translation account and the less demanding metalinguistic response to the Babel objection that I presented in Sect. 4. (My answer does not rely on construing ELC and GLR as conveying metalinguistic propositions about how best to use “women”, “Frauen” as well as all translations of these terms.) Moreover, the section elaborated on the role of context in metalinguistic negotiations. I argued that varying features of the context(s) relative to which two disputants discuss how to best use (a) particular word(s) can call for varying interpretations of their dispute. As I illustrated, the metalinguistic account leaves room for evaluating disputes across languages on a case-to-case basis. The account is flexible enough to accommodate varying details of varying disputes.

6 Conclusions

The present paper defended the metalinguistic approach to conceptual engineering against one of its major critics. According to Cappelen (2018), the approach cannot accommodate disagreement between speakers of different languages (the Babel objection). As shown in the paper, the Babel objection also affects metalinguistic interpretations in other areas of philosophy, such as metaethics. However, the objection misses its target. As a result, the metalinguistic account still stands as a noteworthy contender for explaining disagreements in conceptual engineering and elsewhere.