4.1 Barnes on the Connection between the Metaphysics of Gender and the Meaning of Gender Terms

Elizabeth Barnes (Reference Barnes2020) has posed an interesting challenge to the kind of meta-metaphysical framework I wish to defend. She argues that in order to make progress in the metaphysics of gender, it is not useful to focus on giving accounts of the truth-conditions of sentences involving gender terms such as ‘man’ and ‘woman’. But this seems to be in tension with Thomasson’s framework, and in particular with schema (E) earlier. As I said, on this view, in order to evaluate ontological claims about the nature and existence of a certain kind K, it is a useful first step to examine the application conditions for term ‘K’, in order to know what it would take for ‘K’ to refer, and whether those conditions are actually satisfied in the world. But Barnes (Reference Barnes2020) says:

Why does Barnes think that in order to develop a theory of gender, we should not engage in the project of giving application conditions for gender terms, contra Thomasson? The main reason, as I understand it, is that in Barnes’ view, it is possible that the social structures that underlie our gender practices and give rise to gender categories do not correspond to neat metaphysical categories such as ‘manhood’ and ‘womanhood’. That is to say, the metaphysics of social reality might not contain gender categories such as ‘man’ and ‘woman’. In her view, the best explanation of our gendered social practices might not invoke metaphysical categories that correspond to our terms ‘man’ and ‘woman’. It might be the case that the truth conditions for our sentences involving gender terms such as ‘man’ and ‘woman’ do not really match up with the underlying gender categories that better explain social reality.

As Barnes puts it:

This gives rise to the question: In what cases would the project of giving application conditions for gender terms, and the project of theorizing the nature of gender, come apart? Barnes (Reference Barnes2020) suggests three different reasons, as I understand her view. As a first approach, the three reasons are as follows: (i) gender terms might refer to different things in different contexts, (ii) the basic social structures explaining our gendered social practices might not correspond to the meaning of gender terms such as ‘man’ and ‘woman’, and (iii) we might have ameliorative reasons for using gender terms such as ‘man’ and ‘woman’ in ways that depart from what those terms actually pick out. As a first response to each reason (I will elaborate these responses next): against (i) I would say that this point is compatible with Thomasson’s framework. If the term refers to different kinds in different contexts, then the two-step process has to be applied in each context, yielding different answers. Nothing in Thomasson’s view requires that a given term refers to the same kind in all contexts. In response to (ii), I would say that even if we can usefully distinguish between the reference of our gender terms such as ‘man’ and ‘woman’, and the bedrock social categories that explain our gendered social practices, this does not mean that the application conditions for gender terms are irrelevant in metaphysics. In particular, in order to figure out those bedrock social categories, it is indispensable that we figure out the application conditions of gender terms, and the connection between the actual referents of gender terms, and those bedrock social categories. There are different metaphysical views here: gender categories (i.e., the referents of gender terms) might be constituted by, or grounded in, those bedrock social categories. Or perhaps the metaphysical connection is weaker than that: maybe there is a mere causal connection. However, if those bedrock social categories are to explain our gendered social practices, we need to figure out the metaphysical connections between those social categories and our gender categories. And for this inquiry, figuring out the application conditions of gender terms is an indispensable first step. Furthermore, in response to (iii), I would say that this is a good point, but very similar to Haslanger’s ameliorative project. Indeed, we need to distinguish between what gender terms actually refer to, and what gender terms should refer to (perhaps in different contexts), but Haslanger (Reference Haslanger2000) already taught us this. This is compatible with Thomasson’s conceptual engineering version of her deflationist view.

Let’s examine in more detail Barnes’ objections. She says:

She suggests in this passage that the metaphysics of material objects like tables might come apart from theories about the application conditions of ‘table’. This is in tension with Thomasson’s approach earlier, and in particular with schema (E). However, Barnes does acknowledge that in the case of gender, we do want the metaphysics of gender to say something about our use of gender terms. But she adds: ‘on the best interpretation of social position accounts of gender, they shouldn’t be thought of as giving us a metaphysical analysis of what it is to be a man or a woman, or of giving us straightforward application conditions for gender terms like “man” and “woman”’ (714). By social position accounts of gender, she refers to accounts along the lines of Haslanger’s ameliorative account of gender in terms of someone’s social position in a social hierarchy, as I explained earlier. Here again Barnes emphasizes the idea that the metaphysics of gender, and analyses of the application conditions of gender terms, can come apart, contra meta-metaphysical views like Thomasson’s. What are Barnes’ reasons for this claim? She says:

This passage contains the core idea behind Barnes’ argument, as I understand it. First, she claims that gender is not one thing but many things. Second, she suggests that whereas there are certain bedrocks social categories that offer the best explanation of gendered social practices (namely, Haslangerian social structures), the metaphysics of gender is not exhausted by them. There is more to gender than those bedrock social categories, although these social categories offer the best explanation of gendered social practices and hence the metaphysics of gender should point that out. But at the same time, it seems likely that our gender terms do not currently pick out exactly those bedrock social categories. And indeed, focusing on what gender terms should pick out, it might not correspond to those bedrock social categories either, but rather, say, to gender self-identification. Hence, Barnes argues, the metaphysics of gender is ultimately about those bedrock social categories, but our gender terms might not (nor should they) pick out those bedrock social categories. Hence, the metaphysics of gender comes apart from the meaning of gender terms.

In response: this last inference is too quick. I agree that it is useful to think about the bedrock social categories that could offer the best explanation of our gendered social practices, and that they might come apart from either the descriptive meaning or the ameliorative meaning of gender terms. But this does not entail that our two-step process as described by Thomasson is irrelevant. Indeed, I believe it is indispensable. The reason is that in order to figure out what the bedrock social categories underlying gendered practices are, we cannot do without analysing the meaning of gender terms. As I suggested earlier, if those bedrock social categories provide the best explanation of our gendered social practices, this might be due to either a constitutive relation or a causal relation between those bedrock social categories and gender categories such as ‘man’, ‘woman’, and ‘genderqueer’. And, crucially, in order to ascertain which relation holds, we need to engage in the two-step process a la Thomasson. For we need to figure out what our current (or revisionary) gender categories are, in order to figure out what are the bedrock social practices that better explain our gender categories. That is, we need to know what our gender terms pick out (or should pick out), in order to figure out what bedrock social categories either constitute or causally explain our current (or revisionary) gender categories. It is not clear to me how we could figure out what bedrock gender categories could better explain our gendered social practices if we do not know what our current gender categories are. Or alternatively, in case the descriptive question was too indeterminate to figure out, we should at least figure out what our revisionary gender categories should be, and what bedrock social categories better explain them.

Barnes adds:

This is compatible with the idea I just sketched. Even if Barnes is right (as I am sympathetic to) that there are bedrock social categories (a la Haslanger) that explain our gendered social practices, there are still many kinds in the vicinity of gender, and hence our gender terms probably do not (and should not) pick out only those bedrock social categories. I completely agree. But as I have argued, this does not rule out the relevance of Thomasson’s two-step inquiry. On the contrary, this only calls out for a context-sensitive application of the two-step inquiry.

4.2 Dembroff on the ‘Real Talk’ Assumption about Gender

In this section I will examine another interesting challenge to the version of metaphysical deflationism I am advocating. Dembroff (Reference Dembroff2018) has argued that it is a mistake to try to account for the nature of gender by assuming what they call the ‘Real Gender’ assumption. This is, roughly, the idea that we should use gender terms in a way that closely matches gender kind membership facts, that is to say, facts about who belongs to which gender. This assumption, Dembroff argues, is common both in anti-feminist critique and in feminist arguments. As Dembroff puts it:

Dembroff claims that many feminist scholars also make this assumption, since they aim to give accounts of how we should use gender terms (and what should guide our gender classifications, which are linguistic practices) in terms of the real gender facts, that is, in terms of the real nature of underlying gender kinds. Gender classifications include gender terms such as ‘man’, ‘woman’, and ‘genderqueer’, as well as other gendered expressions such as pronouns and other gender markers in our natural languages. Dembroff claims that this assumption is common both in those anti-trans discourses that attempt to argue that gender terms should track facts about biological sex because these are the real membership conditions for the operative gender kinds, and in the work of feminist philosophers that aim to defend the opposite view. For instance, Dembroff mentions Mari Mikkola as a trans-inclusive feminist philosopher who also makes the ‘Real Gender’ assumption:

Dembroff’ central idea is that the methodological assumption to the effect that metaphysicians should offer metaphysical accounts of the nature of gender that track what we take to be just gender classifications, is not justified. The main reason is that it might be the case that the metaphysical nature of gender kinds does not actually match up with what we take to be just gender classifications, because it might be the case that the nature of the operative gender kinds is oppressive, and hence does not match up with just gender classifications. Dembroff puts the core argument as follows:

Dembroff’s main argument for premise 2 goes as follows: ‘dominant gender kinds systematically oppress persons who claim trans and non-binary identities. They do not reveal what gender classification practices should be; they reveal what these practices have been’ (Reference Dembroff2018: 35). As we can see, Dembroff’s central idea is that since dominant gender kinds systematically oppress people who do not fall under binary gender categories, then these dominant gender kinds cannot be a guide for our gender classifications, since if we base our gender classifications on the operative gender kinds in our vicinity, then our gender classifications will be oppressive since the underlying gender kind membership facts are oppressive. And the other way around: if we try to investigate the nature of operative gender kinds based on what we take to be just gender classifications, as Mikkola suggests (on Dembroff’s interpretation), then we will not get a justified account of the metaphysics of gender kinds, since we are assuming that operative gender kinds track just gender classifications, but this might not be true. And given the oppressive nature of operative gender kinds, it is likely that this assumption will be false.

Dembroff’s argument is useful and illuminating, and it is especially interesting for our purposes since it seems to be in tension with the meta-metaphysical framework I have been advocating so far. In particular, Dembroff seems to suggest that we should not investigate the nature of gender kinds on the basis of what we take to be just gender classifications. But on the two-step metaphysical inquiry that I have been defending, that is exactly what the metaphysician of gender should do. The first step is a project in conceptual engineering where we investigate the application conditions that gender terms should have, and then we are in a position to undertake the second step where we investigate empirically what individuals if any satisfy those application conditions (and therefore belong to the corresponding kind).

In other words: from Thomasson’s schema (E) earlier it follows that a term ‘K’ refers only if Ks exist. This seems to suggest that a term ‘K’ refers to a group of individuals only if that group of individuals instantiate the property of being K. Applied to gender, this in turn seems to entail something like the ‘Real Gender’ assumption, to wit: someone should be classified as falling under the term ‘K’ only if they belong to kind K, that is, if they meet the membership conditions.

In response, I want to argue that metaphysical deflationism, and in particular schema (E) earlier, do not entail the ‘Real Gender’ assumption in a problematic way. There are several reasons for this. The first reason is that the ‘Real Gender’ assumption, as Dembroff puts it, is about what our gender classifications should be like. The ‘Real Gender’ assumption has it that our gender classifications should be guided by the membership facts of the dominant gender kinds. And this is clearly false, for the reasons Haslanger (Reference Haslanger2000) already made clear: there might be good ameliorative reasons for why the dominant gender kinds are not the best referents for our gender terms, that is, the referents that gender terms should have (even if they are the referents they actually have). When we engage in the ameliorative project, we might have good reasons to revise the meaning of our gender terms and change the referents, so that they do no longer refer to the operative gender kinds in the vicinity, but to new gender kinds so to speak. It is clear that Haslanger’s ameliorative project is not committed to the ‘Real Gender’ assumption: Haslanger’s ameliorative project is not interested in describing the meaning of gender terms on the basis of the referents that they actually have (the dominant gender kinds), but in terms of the referents that they should have (that is, what we can call ameliorative gender kinds).

However (and this is my second critique of Dembroff’s argument), once we change the meaning of gender terms given our best ameliorative reasons, we are then in a good position to investigate the metaphysics of gender, that is, the nature of these new gender kinds. The metaphysics of gender is not only about what dominant gender kinds are like (which are oppressive as Dembroff shows), but also about what less oppressive, more liberating gender kinds could be like. This is also the business of the metaphysics of gender. Dembroff’s argument assumes that the metaphysics of gender is only about the nature of dominant gender kinds. But as Talia Bettcher (Reference Bettcher and Shrage2009) already suggested, we can distinguish between different gender claims. Bettcher (Reference Bettcher and Shrage2009) distinguished between dominant gender meanings that are common in mainstream communities, and resistant gender meanings that are operative in trans-friendly communities. The former are not inclusive of trans people, but the latter are. Bettcher (Reference Bettcher and Shrage2009) focuses on the meaning of gender terms, but in my view, her argument also extends to the nature of gender kinds. The two different patterns of use of gender terms that Bettcher (Reference Bettcher and Shrage2009) identifies give rise to two different sorts of gender facts. On the other hand, there are dominant gender kinds that have membership conditions that are oppressive for trans and non-binary people. On the other hand, there are resistant gender kinds that are trans-inclusive. An exhaustive theory of the metaphysics of gender should invoke all these kinds.

Hence, there are two ways in which schema (E) earlier does not entail the ‘Real Gender’ assumption. Firstly, even if it is true that term ‘K’ refers to Ks, and in particular, given the application conditions that term ‘K’ actually has, it follows that certain individuals (and not others) are Ks, it does not follow that we should say that those and only those are Ks. The fact that something is true does not entail that we should say it. There are truths that are such that should remain unsaid. This does not entail that we should utter the negation of those truths. Rather, it might be the case that we should just stay silent on the matter. Secondly, as I have suggested, it might be the case that even if ‘K’ actually refers to Ks given the ordinary meaning of ‘K’, we ought to change the meaning of term ‘K’ for ameliorative reasons, so that it comes to have a different referent. And once the term ‘K’ acquires the new referent, ‘K’ does no longer refer to Ks, but to a new group, K*s, so there is no longer a reason to say of those Ks that they fall under ‘K’, as the real gender assumption would have it, because this would be no longer true. That is, now the term ‘K’ refers to K*s, not Ks.

What are the consequences of this discussion with respect to the two-step method that I have advocated, following Thomasson? It is clear that Dembroff’s arguments do not pose an obstacle to our meta-metaphysical framework. We have to be aware of the distinction between the gender kinds that match up with ordinary gender concepts, which might be oppressive gender kinds, since our ordinary gender concepts are probably not just, and the gender kinds that match up with the revisionary meaning of gender terms, that is, the meaning that gender terms should have, given ameliorative reasons. But this is not new: Haslanger (Reference Haslanger2000) already made this very clear. However, the reminder is always useful.

On the meta-metaphysical framework I have advocated, we should investigate the nature of gender kinds by means of examining first the application conditions of our gender terms. But as I have already emphasized (and as Thomasson has emphasized in her recent work), the interesting project is not only to figure out what our ordinary gender concepts refer to, but crucially, what our revisionary gender concepts should refer to. This is why the first step in the two-step process is a project in conceptual engineering, not descriptive conceptual analysis, as I have painstakingly argued, drawing both on Thomasson’s and Haslanger’s work.

However, for some purposes, we might be interested in examining the nature of dominant gender kinds and the ways in which they are oppressive. For these purposes, then, Dembroff is right that we cannot base our metaphysical inquiry merely in the study of just gender classifications, because dominant gender kinds might not match (and probably will not match) just gender classifications. For this purpose, we need to investigate first the application conditions of ordinary gender concepts, and then figure out the dominant gender kinds that they refer to. But as I have argued, the metaphysics of gender is not limited to this descriptive project. The metaphysics of gender (and many other projects in metaphysics, on Thomasson’s approach) is not only about the kinds that our terms actually pick out, but also, and more crucially, about the kinds that our terms should pick out.

In addition, we can also see that we are not committed to the ‘Real Gender’ assumption in the other direction of the principle that Dembroff was also worried about. In particular, we are not assuming that our gender terms should always track the operative gender kinds in the vicinity, whatever they are. As I have explained, and as Haslanger (Reference Haslanger2000) already made clear, it might be the case that the actual referents of our gender terms are oppressive, and that we have good reasons to change their meaning.

To conclude our discussion so far: when it comes to claims about what should guide our gender classifications, this is clearly an ameliorative question about the meaning that our gender terms should have. Dembroff (Reference Dembroff2018) is obviously right that the usage of gender terms should not be constrained by the dominant gender-kind membership facts, because these gender kinds might be oppressive (and probably will be). The meaning that gender terms should have probably departs from the meaning that gender terms actually have. On the other hand, it is not true that the metaphysics of gender is only about the dominant gender kinds in our vicinity. This is an important aspect, but not the only one. With respect to the project of investigating the nature of dominant gender kinds, it is true we cannot be guided by just gender classifications, since there might be a mismatch. We should investigate the application conditions of ordinary gender concepts in order to ascertain what gender kinds they actually track, and how they are oppressive. Moreover, we should also investigate the application conditions that our gender terms should have, that is, those that are more conducive to social justice, and the corresponding gender kinds that they should track. The metaphysics of gender is also concerned with investigating the nature of these ameliorative gender kinds that our gender terms do not actually track but should track. And in order to figure this out, our two-step process is indispensable.

4.3 Further Objections to the Role of Application Conditions

According to the proposed methodological approach, we need to engage in the two-step inquiry as advocated by Thomasson. But an important clarification is in order. Firstly, as I have said, in many cases in the first step we will not be able to provide full accounts in terms of necessary and sufficient conditions for someone to fall under the concept (in this sense I agree with Mikkola). In many cases, partial analyses in terms of either necessary or sufficient conditions will be all we can achieve. But secondly, and more crucially, these accounts in terms of either necessary or sufficient conditions are still not accounts of the membership conditions, that is, the conditions that determine whether someone falls under the concept or belongs to the corresponding kind. This is the business of the second (empirical) step, not the first (conceptual) step. Let me explain in more detail.

As I said earlier, the first (conceptual) step is concerned with providing necessary and/or sufficient conditions regarding what it would take for something or someone to fall under the concept. What does this mean? Let’s use an example. In the case of many natural kind terms such as ‘water’ or ‘tiger’, the first step can only deliver information about what it would take for something to fall under the concept, in a given scenario. In particular, we know that for something to fall under ‘water’, it would have to be watery stuff (that is, the odourless, colourless liquid that falls from the sky and fills rivers and lakes). But this does not mean that for a liquid sample to be water, it is necessary and sufficient that it is watery stuff. As is well known, something can be watery stuff and not be water (and vice versa). What is going on? The answer, as explained by Chalmers & Jackson (Reference Chalmers and Jackson2001), is that the first step of the inquiry can only deliver application conditions regarding what it would take for something to fall under the concept, not membership conditions. Let me elaborate: application conditions are necessary and/or sufficient conditions for something (a feature, a property, a kind) to fall under the general term, given different scenarios, but these are not yet membership conditions for the corresponding kind in the actual world. The membership conditions for the kind are discovered by means of the second, empirical step, where we discover empirically what kinds or features, if any, satisfy those application conditions in the actual world. For example, going back to the case of ‘water’, in the first (conceptual) step we can discover that for something to fall under ‘water’ (in a fully described scenario), it has to be watery stuff. That is, we would apply the term ‘water’ in a possible scenario (where do not know what the watery stuff in front of us is made of) to a certain sample of liquid only if this is watery stuff. And once we figure out which stuff in our actual world is watery stuff, we can then apply the second step, by investigating empirically the nature of the (actual) watery stuff in front of us. That is, we can discover empirically that watery stuff in our vicinity is made of H₂O. (But if we had been born and raised in Twin Earth, we would then have discovered that the watery stuff in front of us is made of XYZ instead.) Hence, the application conditions for the concept ‘water’ amount to something like being watery stuff. But the membership conditions for X to be water (in the actual world) is that X is H₂O. Hence, the application conditions themselves cannot tell us what the nature of the (actual) referent is like. We need to engage in the second (empirical) step for that result. This is so especially in the case of terms that are rigid designators, like ‘water’ and ‘tiger’, that is to say, terms that refer to the same property or kind in all possible worlds where they refer at all. For this reason, there could be possible worlds where water (that is, H₂O) is not watery stuff. Thus, only by means of investigating empirically the nature of what happens to be watery stuff in our vicinity can we learn what the nature of water is like. However, for other terms that are not rigid designators, like ‘triangle’, and others (more controversially) such as ‘friend’, ‘chair’ and ‘dancing’, the application conditions and the membership conditions can coincide. (In what follows I will put aside this complication unless it is relevant for the argument.)

Another objection to the idea that we need to provide accounts of the application conditions of gender terms in order to provide accounts of the metaphysics of gender has been presented by Mikkola (Reference Mikkola2016). She argues that feminist metaphysicians do not need to offer substantive accounts of the application conditions of gender terms. In particular, she argues that we do not need to provide substantive accounts of the meaning of gender expressions in terms of necessary and sufficient conditions. She argues that participants in the debate can rely on their extensional intuitions about gender terms, and that this suffices for the purposes of feminist metaphysics. She suggests that given the serious difficulties facing the project of giving an account in terms of necessary and sufficient conditions for the application of gender terms, we can make methodological progress by means of relying only on extensional intuitions, about which there is not so much disagreement. In response: Mikkola is right that it is extremely difficult to offer an account in terms of the necessary and sufficient conditions for someone to fall under a certain gender concept. There will be a lot of disagreement about what the necessary and sufficient conditions would be. But as we saw earlier, according to Thomasson (Reference Thomasson2008), even if it is the case that competent speakers do not usually have explicit knowledge about the necessary and sufficient conditions for someone to fall under a certain concept, competent speakers do often have some implicit knowledge, at least partial. And it is possible, in Thomasson’s view, to make this implicit knowledge explicit, by means of philosophical reasoning, analysis of thought experiments, and so on. However, Mikkola would reply that in the case of gender concepts, a consensus is not forthcoming given the recalcitrance of the disagreements. I agree that in the case of gender, disagreements about the application conditions of gender concepts are especially recalcitrant. But as Thomasson (Reference Thomasson2008) explains, we often do not need to give full accounts of a concept in terms of necessary and sufficient conditions for someone to fall under the concept, in order to obtain interesting metaphysical conclusions. In particular, as we saw earlier, sometimes we can provide partial application conditions in terms of either some sufficient conditions for someone to fall under the concept, or some necessary conditions for someone to fall under the concept, and these partial analyses are sufficient to draw interesting metaphysical conclusions. For example, from the claim that it is sufficient that X is P, for X to be G, we can conclude that given that we have empirical evidence that some individuals instantiate property P, we can infer that those individuals are G and hence Gs exist. Likewise, from the claim that it is necessary for X to be G that X is P, we can conclude that some individuals who clearly lack feature P, are clearly not Gs. Hence, partial analyses are sufficient for substantive metaphysical claims.

However, Mikkola could respond that in the case of gender concepts, even partial analyses in terms of either sufficient conditions or necessary conditions for someone to fall under the concepts will be difficult to agree upon. In response, I believe that extensional intuitions will probably also be very difficult to agree upon. Extensional intuitions are supposed to be intuitions about the individuals who actually fall under gender concepts. These intuitions are supposed to be neutral regarding what features are necessary and sufficient for someone to fall under the concept. Mikkola’s claim is that whereas many participants in the debate about gender disagree about the necessary and sufficient conditions for someone to fall under a gender concept, most of those participants would agree about paradigmatic cases of individuals who do fall under the concept. For instance, everyone would agree that Madonna, Beyoncé, and Hillary Clinton are all women. Mikkola argues that relying on extensional intuitions like those is sufficient to explain our usage of the term ‘woman’, since we can all use the term competently even if we are not able to spell out necessary and sufficient conditions for someone to fall under the term. In response, I believe that even if there could be agreement about some non-controversial paradigmatic cases, there will probably be a lot of disagreement about who belongs and who does not belong to the extension of gender terms. That is, our extensional intuitions are also subject to disagreement. Extensional intuitions can vary a lot between different speakers and different communities. For instance, when it comes to contested cases, for example, whether trans women are women, unfortunately there is no agreement, and many people believe that (at least some) trans women do not fall under the concept ‘woman’.Footnote ¹⁴ Even if we focus on the more modest question of who can be posited as paradigmatic individuals, there will be disagreement. As Bettcher (Reference Bettcher and Shrage2009) argues, when it comes to analysing the meaning of ‘woman’, some theorists will want to start from paradigmatic instances that do not include trans women, but this is problematic in her view, for both descriptive and normative reasons. That is, it is wrong for descriptive reasons since there are speakers who do include trans women as paradigms (e.g., members of trans-friendly communities). And, secondly, because it could be argued, following Bettcher (Reference Bettcher and Shrage2009), that it is morally problematic to include only cis women as paradigmatic instances of women (so that trans women would still fall under the term but only as peripheral members of the kind). Given all these difficulties, I do not think that relying only on extensional intuitions regarding gender concepts is a useful methodological approach.

To finish our (brief) discussion of alternative approaches to metaphysical deflationism: Katharine Jenkins (Reference Jenkins2023) also suggests that to investigate the ontology of gender, one does not need to offer accounts of the application conditions of gender terms. She says:

In response, I agree that it is useful to think about all those possible kinds that have ‘travelled under the banner’ of gender, or kinds that are in the vicinity of our gender terms, even if those kinds might not be the actual referents of gender terms (nor the referents that gender terms should have). In this sense, I agree with Jenkins that being open to this possibility is a useful attitude for the metaphysician of gender, since there might be kinds in the vicinity of gender talk that are neither their actual referents nor the target referents, but that are important to investigate for the aims of feminist metaphysics. However, in my view, the phrase ‘in the vicinity of gender talk’ is not sufficiently clear. Jenkins says that ‘if someone claimed that one of these kinds was the referent of our … gender talk, we might agree or disagree, but we would not think that the person was deeply confused or that there had been a fundamental error of communication’ (117), but unfortunately, I believe that this reaction is very likely to happen, for several candidate meanings for gender terms. For instance, one important point of disagreement in the debate about the nature of gender is whether trans women are women (see Byrne Reference Byrne2020 and Dembroff Reference Dembroff2021 for a vivid manifestation of this disagreement). People who believe that trans women are not women and do not fall under the term ‘woman’, often do say that those of us who believe that trans women fall under the term ‘woman’ are deeply confused. Moreover, we should keep open about the possibility that those who believe that trans women fall under ‘woman’, and those who believe that trans women do not fall under ‘woman’, are using different concepts of woman, and are thus talking past each other, and not really communicating. To give another example, two other important candidate meanings for gender terms have to do with gender as a social position (along the lines of Haslanger’s account of gender as a social class as stated earlier), and gender as self-identification (see Bettcher Reference Bettcher and Shrage2009 and Jenkins Reference Jenkins2016 for two compelling defences of the meaning of ‘woman’ in terms of self-identity). Again, I believe that we should be open to the possibility that different parties in this debate use gender terms with different meanings, and in this sense, they would be talking past each other.Footnote ¹⁵ Parties in this particular debate would probably not say that the other party is deeply confused, but it is an open position in the debate that the different parties are talking past each other and in this sense, not communicating with each other. Hence, it is not clear to me that the suggested heuristics, having to do with people’s intuitions about when users of gender terms are communicating or merely talking past another, is the best strategy to identify the relevant kinds in the vicinity that are worth studying.

As we saw, Jenkins is right that the actual referents of gender terms are not the only focus of interest for the ontology of gender (nor should we limit ourselves to the kinds that gender terms should refer to). However, in my view, if we want to figure out what kinds have travelled under the banner of gender terms, the most useful approach is to engage in a two-step inquiry a la Thomasson. Among other virtues, this approach has the promise of telling us which kinds are the actual referents of gender terms, which kinds are the referents the terms should have, and which other kinds may have sometimes been associated with gender terms but are not their current referents. If we dispense with this method, then we are going to be pretty unsure about what kinds have really travelled under the banner of gender terms, since the starting data we have is full of recalcitrant disagreements, and in particular, some speakers will be happy to say that other speakers using gender terms in different ways are deeply confused. Given this situation, I submit, the two-step inquiry offers a systematic approach, a useful method, and a way to make progress in an otherwise intractable debate.

6.1 Contextualism

How can we approach the question of what gender terms actually mean? A first caveat has to do with the distinction between the manifest and the operative concept that we have already explained. The ultimate aim when we inquire into the ordinary meaning of gender terms is to reveal the operative concept, that is, the objective type that is picked out by the corresponding term (if any). However, as we saw in Section 5, there is disagreement among scholars about what the best approach to questions about the operative concept is. On the one hand, internalist theorists tend to believe that we need to figure out at least some of the information associated with the term by competent speakers, since this plays a role in fixing the referent of the term. As we saw earlier, scholars that favour this approach include Jackson (Reference Jackson1998), Chalmers (Reference Chalmers2002), Thomasson (Reference Thomasson2008), Glasgow (Reference Glasgow2009), and Plunkett (Reference Plunkett2015), among others. On the other hand, this approach is in contrast with an alternative approach, namely, externalism. On this view, referents are fixed not by what is in our heads, but rather by external factors, including environmental factors and causal-historical factors that most competent speakers do not have access to.

As I suggested earlier, in this Element I aim to be as conciliatory as possible with regard to this debate. I agree with advocates of externalism that, given the problem of ignorance and error, it is very difficult to discover any identifying information in our heads that could be part of the meaning, since competent speakers are usually ignorant or mistaken about identifying information about the referent. On the other hand, I agree with the proponents of internalism that it is very difficult to explain reference-fixing without any appeal to information about the referent that is somehow associated with our terms. Given these competing reasons, I will conclude that there is a very modest, but indispensable, role played by information associated with the term regarding reference-fixing. At a minimum, many terms are associated with sortal information about the sort of thing that the referent belongs to.Footnote ²⁴ Sometimes we can identify at least some necessary conditions or some sufficient conditions for something to fall under the concept. This minimal stance is still compatible with different semantic approaches.

Often, scholars attempting to reveal the ordinary meaning of gender terms start with some basic intuitions or some basic facts about patterns of usage of gender terms, and they use these starting points to build their theories. In this section I will explore some of these approaches. For instance, Jennifer Saul (Reference Saul, Crasnow and Superson2012) starts her investigation by stating that people seem to use gender terms such as ‘man’ and ‘woman’ in different ways. Some utterances seem to be about someone’s biological sex, whereas other utterances are clearly not about someone’s biological sex. She also mentions that according to many feminist scholars, what determines being a woman has to do with the subject’s social position or social role, not their biological sex. Taking all these data into consideration, she claims that we should appeal to a ‘collection of ordinary usage data’ (Reference Saul, Crasnow and Superson2012: 200) as a starting point for our theorizing about the ordinary meaning of ‘woman’, which we can later argue should be revised given ameliorative considerations. Similarly, N. Laskowski (Reference Laskowski2020) argues that theories about the actual meaning of ‘woman’ should satisfy what he calls the ‘usage constraint’, that is, the idea that the theories should do justice to available data about patterns of usage. He also claims that given the best review of the available data, it looks like people in different contexts can use the term ‘woman’ with different meanings.Footnote ²⁵ From a different starting point, Bettcher (Reference Bettcher, Soble, Power and Halwani2013) also argues that there seems to be a diversity of usages of the term ‘woman’. In particular, Bettcher claims that we can identify a dominant usage of the term ‘woman’, according to which the term ‘woman’ aims to refer to someone’s genital status (which is a linguistic practice that perpetuates transphobic ideology), and a resistant usage of ‘woman’, which is common in trans-friendly communities, where the term ‘woman’ aims to refer to people’s gender identities, and where both cis women and trans women can be paradigmatic instances of women.

In reaction to this appearance of diversity of usages, Saul (Reference Saul, Crasnow and Superson2012) introduced a version of contextualism about the meaning of ‘woman’, as a possible view that could aim to accommodate the apparent diversity of usages. However, Saul (Reference Saul, Crasnow and Superson2012) found the view ultimately problematic and did not endorse it. But the reason for not endorsing it was normative, not descriptive. That is, Saul (Reference Saul, Crasnow and Superson2012) argued that contextualism cannot do justice to the claims of trans women, and for this reason should be rejected. Saul herself acknowledged that this consideration pertains to ameliorative projects rather than descriptive projects. As we saw earlier (drawing on Haslanger, Dembroff and others), what gender terms actually mean and what they should mean could come apart. But in order to settle this debate, it will be useful to review in more detail some of the literature on what gender terms actually mean, to see if we can make some progress. In particular, we will review two central semantic theories: contextualism and externalism, as applied to gender terms. In this section we will explore the prospects of contextualist theories of the meaning of ‘woman’, since, as many authors have suggested, it seems, prima facie, that people use the term ‘woman’ with different meanings in different contexts.Footnote ²⁶

Saul’s contextualist proposal goes as follows:

Recall that Saul’s aim is to put forward for our consideration a putative account of the ordinary meaning of ‘woman’ that can capture the apparent diversity of usages. This version of contextualism is at least prima facie well suited to do that. The account claims that utterances of the form ‘X is a woman’ have the following truth conditions: the utterance is true if and only if X is human and relevantly similar to most of those that, it is assumed, are paradigmatic cases of being a woman, namely, those who possess all of the biological markers of female sex. This view does not presuppose any specific account of what biological sex is, but it does assume that those who possess all of the biological markers of female sex will probably be paradigmatic instances of women.Footnote ²⁷ It is important to bear in mind that according to standard accounts of biological sex, it is not necessary to have all the markers of female sex in order to count as being biologically female, only many or most of those markers.Footnote ²⁸ Moreover, this contextualist account allows that someone can be a woman even if they do not have many of the biological markers of female sex. How? Because what determines that someone falls under ‘woman’ is not whether they possess the biological markers of sex, but rather whether they are relevantly similar to those who possess all of the biological markers. Saul’s intention, then, was to sketch a view of the meaning of ‘woman’ that would characterize its meaning in terms of similarity with paradigmatic cases of the referent. This is a common semantic strategy. What makes this view a version of contextualism is that on this view, what counts as relevantly similar to those paradigms can vary from context to context. That is, the criteria of similarity can vary from context to context.

Let’s consider two different examples to see how the theory would work. In a context of a trans-inclusive feminist conference where someone is taking attendance and keeping a register of the number of women attending the conference, utterances of the form ‘X is a woman’ express the proposition that X is similar to most of those possessing all the markers of female sex (from now on, those who are biologically female, for short) with respect to the criterion of self-identifying as a woman. A caveat: not all those who are biologically female identify as a woman. But arguably, many or most of those who are biologically female identify as a woman. Hence, if X is similar to that group in the sense that she also self-identifies as a woman (where X herself can be biologically female or not), then ‘X is a woman’ will be true. On the other hand, Saul considers the case of a forensic anthropologist who utters the sentence ‘this bone belonged to a woman’. In this context, arguably, the relevant criteria of similarity have to do with having certain DNA or being biologically female more generally. Thus, the referent of ‘woman’ in this other utterance will be different.

Therefore, the contextualist account seems to be able to accommodate our intuitions about different patterns of usage, at least in some preliminary cases. However, Saul (Reference Saul, Crasnow and Superson2012) argues that, after reflection, we can realize that contextualism cannot accommodate all the intuitions that a feminist scholar would want to accommodate. In particular, it cannot do justice to the claims of trans women, who would like to advocate for an account of the meaning of ‘woman’ that includes all trans women. As Saul (Reference Saul, Crasnow and Superson2012) explains, contextualism has the consequence that the advocates of trans women would express a trans-inclusive concept of woman in their utterances involving the term ‘woman’, but unfortunately, the contextualist view also has the problematic consequence that the opponents of trans women would express a trans-exclusionary concept of woman in their utterances involving the term ‘woman’. Why? Let me explain. If we apply Saul’s contextualist account to a sentence of the form ‘trans women are women’ uttered by an advocate of trans women, then the sentence will be true because ‘women’ will refer to the class of humans who are relevantly similar to those who are biologically female, and in this context, given the relevant criteria of similarity, trans women are clearly similar to most of those who are biologically female in the sense of self-identifying as women (or occupying a similar social position). However, if an opponent of trans women utters ‘trans women are not women’, contextualism would imply that this utterance is also true, since in this context, ‘women’ would refer to the class of humans who are similar to those who are biologically female with respect to the following criterion of similarity: possessing all the biological markers of female sex. For this reason, Saul argued that we should reject contextualism about ‘woman’. But as she explains, this is an ameliorative consideration pertaining to what the term should mean, not what the term actually means.Footnote ²⁹

In response to Saul (Reference Saul, Crasnow and Superson2012), in Díaz-León (Reference Díaz-León2016) I argued that there is an alternative version of contextualism about ‘woman’ that can satisfy two desiderata: it can do justice to the collection of data that is the starting point for a semantic theory of ‘woman’, and it can be inclusive of trans women at the same time. The key idea is to use a distinction that other advocates of contextualism (about other terms) have used, namely, attributor contextualism versus subject contextualism. According to attributor contextualism, a term varies its content from context to context, in virtue of who utters the term, that is, depending on features of the speaker. Attributor contextualism is a plausible view about the meaning of demonstratives and indexicals such as ‘I’, ‘you’, ‘here’, and ‘now’. The referent of different utterances of these terms depends on who is the speaker and other features of the speaker such as their location in time and space. On the other hand, subject contextualism about a term says that the meaning of that term changes from context to context, depending on features of the subject matter of the utterance that can change from context to context. For example, a plausible version of subject contextualism about ‘knowledge’ would claim that utterances of the form ‘S knows that p’ express the proposition that the subject S is able to rule out alternatives to p being true, where which alternatives S needs to rule out depends on features of the situation of S herself (not the speaker who utters the sentence). For instance, if S is at the zoo looking at zebras, then in order for the utterance ‘S knows that there is a zebra in front of her’ to be true, S would arguably need to rule out the alternative that there is a mule painted as a zebra in front of her. That is, her evidence would need to rule out that there is a mule painted as a zebra (her evidence would have to be sufficient to show that what she is looking at is a zebra, not a mule painted as a zebra). In particular, subject contextualism about ‘knowledge’ claims that if S is in an environment where mules painted as zebras are common, then she would need to rule that out, but if she is in an environment where mules painted as zebras are a very remote possibility, then she does not need to rule that out.

In previous work I used this distinction to defend subject contextualism about ‘woman’ (Díaz-León Reference Díaz-León2016). It is clear that a version of attributor contextualism about ‘woman’ is not going to solve the inclusion problem. The reason is that if the reference of ‘woman’ in different contexts of utterance depends on the beliefs and values of the speaker, then when transphobic speakers utter the term ‘woman’, it will not express a trans-inclusive concept. However, according to subject contextualism, it does not matter who the speaker is. Going back to the case of ‘knowledge’, if we are talking about a zoo-goer at a zoo where there are no mules painted as zebras, it does not matter what the speaker has in mind. If they utter ‘S knows that there is a zebra in front of her’, it can be true even if S cannot rule out that she is looking at a mule painted as a zebra, regardless of what the speaker believes about what knowledge requires. Given what ‘knowledge’ means, and what context the subject of utterance S is at (namely a context without mules painted as zebras), the utterance ‘S knows that there is a zebra in front of her’ will be true.

Likewise, I argued, if different people utter a sentence of the form ‘trans women are women’ it does not matter what the speakers have in mind. What matters for the truth conditions of the utterance is who are the subjects of utterance (the same for all the speakers) and what are the relevant criteria of similarity in those contexts. I suggested that the relevant criteria of similarity are determined by normative considerations salient in the context, including moral and political considerations regarding how the people that the utterance is about should be treated for the purposes that are salient in that context. Let’s examine a few examples in light of these considerations.

Imagine that we are at a clinic where the nurses received a notice asking them to send notifications to women over forty years old asking them to make an appointment for a smear test.Footnote ³⁰ Given these instructions, the nurses go through the files, putting aside the files of the people they need to send letters to. In this context, if they utter a sentence of the form ‘X is a woman’, it can be argued that the relevant criteria of similarity here have to do with having female genitalia. To consider a different example: imagine that we are in North Carolina discussing the infamous ‘bathroom bill’ that said that only people who were assigned female at birth could use women’s toilets. In this context, it is clear that it is morally wrong to exclude trans women from women’s toilets. Hence, the relevant criteria of similarity have to do with self-identifying as a woman, not with being biologically female (nor having certain genitalia). Hence, utterances of the form ‘X is a woman’ in that context will be true to the extent that X is a human and similar to most of those who are biologically female with respect to the criterion of self-identifying as a woman, which is the relevant criterion of similarity in that context.

In Díaz-León (Reference Díaz-León, Plou, Castro and Torices2022) I argued that subject contextualism can help us understand what is at stake in many contemporary debates between trans allies and so-called ‘gender-critical feminism’. According to gender-critical feminism, only those who are biologically female occupy the social position of subordination associated with being a woman.Footnote ³¹ In their view, the term ‘woman’ does and should only refer to those who are biologically female (or at least, those who are perceived to be biologically female and discriminated because of this). They also believe that trans women should not have access to women-only spaces. One of the points of disagreement has to do with the putative truth conditions of the following sentences:

1. (1) Women are more likely to be subject to strict norms of beauty than men.

2. (2) Women are likely to be paid less than men for the same jobs.

3. (3) Women are more likely to suffer domestic violence than men.

Advocates of gender-critical feminism argue that what makes those claims true is a conception of womanhood in terms of biological sex. However, subject contextualism can help us see why this is wrong. According to subject contextualism, the meaning of ‘women’ in these utterances is determined by the moral and political considerations that are the most relevant in each context. It is clear that in those contexts, the class of people that are likely to be discriminated in those ways include trans women.

Bettcher (Reference Bettcher2017), Laskowski (Reference Laskowski2020), and Dan Zeman (Reference Zeman2020) have all independently argued that my version of subject contextualism would collapse into a version of invariantism about the meaning of ‘woman’. The main reason is that given the harms of misgendering, in most or even in all contexts (at least in the actual world), the duty not to misgender others would take priority, and hence, if the truth conditions of utterances of the form ‘X is a woman’ are determined in virtue of the moral and political considerations that are relevant in the context (as I argued), then in all those contexts, the truth conditions will be the same, namely, the term ‘woman’ will refer to people who self-identify as a woman. This is a very important challenge to subject-contextualism. In response, in Díaz-Leon (Reference Díaz-León, Plou, Castro and Torices2022) I considered a variety of examples such as the following:

1. (4) ‘We have to call all women over 40 for a smear test’ (uttered in the context of the clinic described earlier)

2. (5) ‘This bone belonged to a woman’ (uttered by a forensic anthropologist as explained earlier)

3. (6) Women are more likely to have their symptoms of a heart attack ignored.

My claim is that in these cases (4–6), it is not obvious that moral and political considerations would always select a referent of ‘woman’ corresponding to the group of people who self-identify as a woman. To clarify: as I said earlier, it is clear that the view that gender-critical feminists defend, namely, a version of invariantism about ‘woman’ where the term always refers to those who are biologically female, is very implausible, and subject contextualism can explain precisely where that view goes wrong. To wit: in most contexts, the relevant criteria of similarity clearly do not amount to being biologically female (e.g., utterances 1–3). The remaining question is whether a version of invariantism where the term ‘woman’ always refers to those who self-identify as a woman would be the most plausible consequence of subject-contexualism about ‘woman’. And this is not entirely clear. Of course, Bettcher, Laskowski, and Zeman are right that the duty not to misgender is in effect in all contexts. There are moral and political considerations in favour of respecting people’s gender identities that are applicable to all contexts. The question is whether, in addition to this, there are additional moral and political considerations that would have to be weighed up against those, so that the overriding considerations could select a different referent for ‘woman’. And I believe a case can be made for this (context-shifting) view.

For example, in utterance (5) earlier, it can be argued that the relevant criteria of similarity have to do with having similar DNA. The claim is not that this is the criterion of similarity that speakers would probably have in mind (this is what attributor contextualism would say), but rather that this is the relevant criterion of similarity, given the moral and political considerations that have priority in that context. To repeat: according to subject contextualism, the relevant criteria of similarity that determine the reference of ‘woman’ at a context are not the beliefs that speakers have in mind, but rather the moral and political considerations that have the most weight in that context, regardless of what speakers have in mind. This is why Bettcher, Laskowski, and Zeman have all argued that the duty not to misgender would yield the same referent in all those contexts. In response, I believe that there could be additional normative considerations to weigh up. To mention one: Wendy Weisberger (Reference Weisberger2023) explains in an article at CNN that recent archaeological studies of a 90,000-year-old skeleton found in the Andes Mountains have shown that the skeleton is actually female, although it was first believed to be male (because the presence of weapons next to the skeleton led archaeologists to assume that the skeleton was male). This indicates that the assumption that men but not women were hunters is mistaken. This example is relevant because it is crucial for the purposes of explaining the misinterpretation of archaeological data, that the skeleton was first thought to be male but DNA testing has confirmed it was female. Hence, with this evidence, we have support for claims such as (7): ‘the myth of men as hunters and women as gatherers has been shattered’ (as explained in Weisberger’s article). In my view, subject contextualism would entail that the term ‘woman’ as it occurs in (7) refers to those who are biologically female, precisely because this is the relevant criterion of similarity in this context, namely, a context where the aim is to make salient the sexist biases in archaeological studies, and the debunking of sexist myths such as that women were not hunters.

Something similar could be argued about utterance (6) earlier. There is now a consensus that symptoms of a heart attack in women are more likely to be misinterpreted, because in medical school the standard symptoms associated with a heart attack are those that occur more often to men. Again, it could be argued that the relevant moral and political considerations that have the most weight for the purposes of this claim would select as the referent of ‘woman’ those who are biologically female, or at least those who tend to have symptoms of heart attack associated with those who are biologically female (those two groups do not have to be co-extensional). Crucially, the reason that women’s symptoms of heart attacks are not so commonly studied is because they are women.

However, to repeat once more, it would be misguided, as advocates of gender-critical feminism often do, to conclude from these examples that the term ‘woman’ always refers to those who are biologically female. Subject contextualism shows that there is a plethora of contexts where the relevant criteria of similarity do not have to do with being biologically female (for instance, (1–3) earlier, among many other cases). Indeed, we can plausibly say that in most ordinary contexts, the criterion of similarity that will have the most weight corresponds to self-identifying as a woman. For instance, when we talk about who should have access to women-only spaces such as bathrooms, lockers, and shelters, it is clear that the relevant criterion of similarity has to do with gender identity. However, it is an open question what the relevant criteria of similarity are in cases such as (4–7). Sentence (4) is particularly difficult. It could be argued that it can be time-efficient for the nurses at the clinic to use the term ‘woman’ to refer to those who are biological female, for the specific purpose of calling those who need to make appointments for a smear test (but not, for instance, when it concerns who should be allowed to use the women’s toilets at the clinic, where the relevant criterion is self-identification). But on the other hand, there are also good reasons to hold that in a medical context, clinical practitioners should always use gender terms in a way that is trans-inclusive, so the case regarding (4) is still open for further debate.Footnote ³²

Much more would need to be said to settle this matter. My main aim in this section is to argue that subject contextualism can illuminate what is at stake in those debates and can help us make progress. Questions about the referent of ‘woman’ in utterances in different contexts can be understood as questions about what criteria of similarity should have the most weight in different contexts, given the moral and political considerations regarding the subjects of those utterances that have priority in those different contexts.Footnote ³³ On this interpretation of the debate, it is clear that invariantism according to which ‘woman’ always picks out those who are biologically female is a non-starter.

Subject contextualism about the meaning of a term is a semantic theory about the term that posits two aspects of meaning, namely, the character and the content of the term. The character is the function or description that is conventionally associated with the term, and that states how the referent of the term is determined in different contexts, given the factors in the contexts that are responsible for determining the referent, which can vary from context to context. The content, on the other hand, is the referent that the term would have in each context. According to contextualism, the character associated with a term is fixed by convention, and the content can change from context to context (given that the relevant factors change from context to context). In this section I have explored subject contextualism as a theory about the ordinary meaning of ‘woman’. Some reasons in its favour include the idea that many ordinary concepts are defined in terms of some paradigmatic instances and some criteria of similarity with respect to the paradigms that can vary from context to context. However, the idea that the varying criteria of similarity depend on moral and political considerations that have the most weight in that context might seem too revisionary. That is, it might be argued that it is implausible that this is what the term ‘woman’ actually means. Remember, though, that we are interested in the operative concept of ‘woman’, not the manifest concept. It seems clear that the manifest concept associated with the term ‘woman’ does not correspond to the contextualist view developed in this section. However, there is still space to argue that this view captures the operative concept. Furthermore, it could be argued that even if this view is not plausible as an answer to the descriptive question of what gender terms actually mean, it could be a useful answer to the ameliorative question of what gender terms should mean.

A final clarification is in order: As we have said, contextualism is a semantic view according to which a term has two aspects of meaning, namely, a character and a content. For many semanticists, this makes contextualism a version of internalism, since on this view terms are associated with some information in our heads (i.e., the character) which, together with features of the corresponding context, will yield a content or extension. However, it is important to notice that this is compatible with the claim that the character, even if it is technically speaking ‘in our heads’, could come apart from the manifest concept (that is, the character would correspond to the operative concept, which can then in turn yield different extensions in different contexts).

In any case, in order to have a more exhaustive discussion of the different options, in the next section we are going to explore a different sort of view, namely, externalist views of the meaning of gender terms, according to which the operative concepts are not accessible to the subjects, not even in the way the character is associated with the term according to contextualism. It will be interesting to review these externalist approaches, in part to assess to what extent they are compatible with the metaphysical deflationist framework I have been assuming (spoiler: I think they are compatible), and second, to assess whether externalism would yield different results than contextualism with respect to the actual meaning of ‘woman’ (spoiler: I do not think so).

6.2 Externalism

Externalism is an approach in philosophy of language according to which the meaning of our terms is fixed by external factors that are not accessible to the speakers by introspection. Externalist views are concerned with both the semantics (i.e., the semantic content or meaning of terms) and the meta-semantics (i.e., the mechanisms that fix the semantic content of terms) of linguistic expressions. One of the best developed defences of externalism as applied to the aims of feminist philosophy has been offered by Haslanger (Reference Haslanger2006, Reference Haslanger, Marques and Wikforss2020). Haslanger’s characterization of externalism goes as follows:

Haslanger uses the label ‘objective-type externalism’ to emphasize that the insights of externalism can be applied not only to explain the meaning of natural kind terms, but also to other linguistic expressions including social kind terms and all general terms that pick out an objective type. Externalist views became very influential following the work of philosophers such as Kripke (Reference Kripke1972) and Putnam (Reference Putnam1973), who defended externalism mostly as applied to natural kind terms, which are terms posited in the natural sciences that purport to refer to underlying theoretical kinds ( ‘water’, ‘gold’, ‘tiger’, etc.). Haslanger’s claim is that these insights would also apply to other kind of terms that purport to track an underlying objective type, whether it has been posited by the natural sciences or not. A central idea of externalism is that the nature of the underlying referent can be discovered only empirically. And Haslanger’s point is that in many cases, especially concerning areas of political significance such as race, gender and the family, the empirical disciplines that are in a better position to discover empirically the nature of the corresponding referents are the social sciences. Hence, Haslanger’s idea is that when it comes to the philosophical project of asking about the nature of gender and race and other politically significant kinds, our intuitions about the corresponding terms are moot: what matters is what the best empirical theories in the social sciences say about the nature of those referents. And it is plausible that ultimately, the social sciences reveal that the core nature of gender corresponds to people’s social position in a social hierarchy (which corresponds to Haslanger’s account of gender in her (Reference Haslanger2000) piece).

Let’s examine Haslanger’s characterization of objective type externalism in more detail. Her central idea is that the reference of the corresponding terms (including gender and racial terms) is determined by ostension to paradigms plus the implicit assumption that we are aiming to pick out the group of individuals of the same kind, where the nature of this common kind (the kind that unifies all the paradigms) is discovered empirically. As Haslanger emphasizes, it is not necessary that competent speakers are able to state the kind membership conditions. Often, speakers are ignorant or mistaken about the membership conditions and this can be discovered only empirically. However, it might be the case that after empirical investigation, experts find out that the paradigms have more than one type in common. (After all, things are typically similar in many ways.) Haslanger is aware of this problem and deals with it as follows: ‘Sets of paradigms will typically fall within more than one type. To handle this, one may further specify the kind of type (type of liquid, type of artwork), or may (in the default?) count the common type with the highest degree of objectivity’ (Reference Haslanger2006: 110). As we have seen, the problem is that the set of paradigms alone will probably not select a unique type as the referent. Hence, if we want to identify the operative concept, we need to pose further constraints. The worry is that if we assume that the concept is already tracking a unique objective type, then the hypothesis that ostension to the paradigms alone is sufficient to fix a unique objective type is flawed. Haslanger is aware of this worry and posits further constraints. First, she points out that the concepts may include some sortal information about the kind of kind the referent belongs to, such as the type of liquid, type of artwork, type of animal, and type of plant. The idea, then, is that the meta-semantics would need to invoke this sort of information in addition to other external mechanisms in order to explain reference-fixing. (Haslanger would reject the view that this information is part of the semantics, though, but she would accept that it can sometimes be part of the meta-semantics. See Haslanger Reference Haslanger, Marques and Wikforss2020 for further discussion.) However, Haslanger is sceptical that we often have sortal information associated with the concept. In these cases, the referent will just be ‘the common type with the highest degree of objectivity’, in the absence of further constraints. But how can we figure out the type (among those that unify the paradigms) that has the highest degree of objectivity?

A crucial question for our purposes in this section is whether the notion of objective type externalism is compatible with the framework of metaphysical deflationism that I rehearsed at the beginning. I think they are clearly compatible. Recall that the version of metaphysical deflationism defended earlier has it that the debates in metaphysics of the form ‘What is X?’ or ‘Are Xs real?’ can be solved by means of two stages. First, we need to figure out the application conditions associated with the relevant term ‘X’. These application conditions can be pretty thin, and often, they will amount to only necessary conditions for something to fall under ‘X’, or only sufficient conditions (very rarely will we have necessary and sufficient conditions associated with the term by all competent speakers). Now, it is important to realize that Haslanger’s sortal information associated with the concept, or even the minimal assumption that the term will refer to ‘the common type with the highest degree of objectivity’ can be identified with the first stage of Thomasson’s two-stage meta-metaphysical framework. That is to say, in order to explain reference-fixing, Haslanger acknowledges that we need to posit some sortal information, or at least the assumption that the corresponding terms purport to pick out the most objective type unifying the paradigms. This is what Thomasson would call ‘application conditions’. One discussion that we can put aside is whether these application conditions are part of the semantics or only part of the meta-semantics. What is more crucial for our purposes is to realize that in order to fix the referent, we need to appeal to at least some sort of sortal information or another kind of ‘anchorage’ mechanism. And then, as Thomasson acknowledged, this information or anchorage can yield different referents in different scenarios. That is to say, with the same sortal information, given that we do not know what the external world is like (this can be discovered only empirically), we can conceive of several scenarios that are compatible with the same sortal information. And then, the empirical sciences can tell us which of those scenarios is real, and hence, what the objective type that the paradigms have in common actually is.

In Haslanger’s view, the sortal information (or application conditions, to use Thomasson’s term) would not be non-negotiable (as Glasgow Reference Glasgow2009 put it), but on the contrary, this is something we can and often do negotiate. Thomasson (Reference Thomasson2008) used to argue that we can discover the application conditions by means of engaging in conceptual analysis about the contours of our ordinary concepts. But in more recent work (e.g., Thomasson Reference Thomasson, Cappelen, Plunkett and Burgess2020, Reference Thomasson2021), she has changed her mind, and now she is more interested in what the application conditions should be (which we can discover by engaging in conceptual engineering), not what the application conditions actually are (which is difficult to ascertain and not the most important question anyway).

The next question that we have to face is: how can we discover what the type with the highest degree of objectivity is? How can we assess degrees of objectivity? With respect to this question, Haslanger (Reference Haslanger and Kendig2016) has offered a very plausible account. I want to rehearse this account of objectivity, also drawing from other philosophers such as Elizabeth Anderson (Reference Anderson1995) and Philip Kitcher (Reference Kitcher2001). The central idea is that objectivity is a matter of the kinds that are the most explanatorily useful, that is, the kinds that have inductive and predictive potential. This is a big question that we do not have the space to examine in detail, but we can at least sketch the sort of view defended in Haslanger (Reference Haslanger and Kendig2016), and which draws from feminist philosophers of science such as Anderson (Reference Anderson1995) and Kitcher (Reference Kitcher2001). The central idea is that when we compare the degree of objectivity (or ‘naturalness’) of several kinds, what matters is what kinds are the most explanatorily useful given the (legitimate) explanatory aims that we have, that is, given our explanatory needs and purposes. What counts as the most objective type is not a question that is independent from us (as Sider Reference Sider2011 claims). On the contrary, this view goes, there is no way of assessing degrees of objectivity of different kinds in a way that is mind independent. We need to appeal to our (legitimate) explanatory aims and goals in order to ascertain what kinds are the most explanatorily useful with respect to the explanatory needs that we have, as theorists.Footnote ³⁴

This is how Haslanger (Reference Haslanger2006) defends the claim that gender terms might refer to the social kinds she identified in Haslanger (Reference Haslanger2000). Indeed, one could argue that the common type with the highest degree of objectivity concerning gender would probably be a biological kind rather than a social kind. However, once we clarify the picture of objectivity that is relevant here, it is clear that what matters is the theoretical kind that has the most explanatory power with respect to the (legitimate) explanatory aims of our inquiry. And the explanatory aims of feminist inquiry include offering explanations of social reality that can help to explain and resist women’s oppression and being conducive to social justice. With respect to these explanatory aims, the kind that has the highest degree of objectivity is likely to amount to something akin to Haslanger’s social-hierarchical account of gender.

Our final question is the following: what is the connection between contextualism and externalism? Do these different semantic frameworks yield different conclusions with respect to the operative concept of gender? My answer is that, after reflection, they seem to yield similar results. On the one hand, as we saw in the previous section, subject contextualism has it that the referent of ‘woman’ at a context will be determined by the moral and political considerations regarding how the subjects of the utterance should be treated. On the other hand, in this section we have explored a version of externalism according to which the referent of ‘woman’ is determined by the most objective type that unifies a set of paradigms, where the most objective type amounts to the type that has the most predictive and explanatory power given our legitimate aims and goals in that inquiry. It is likely, though, that different theorists give priority to different aims and goals. Hence, it might be the case that there are different aims and goals that pull us in different directions, even when all those aims and goals are legitimate aims of inquiry. Hence, it is probable that our descriptive inquiries into the operative concept of gender (both from the perspectives of contextualism and externalism) will yield a pluralist conception of gender. To illustrate, let’s consider again some of the examples we examined in the previous section:

1. (1) Women are more likely to be subject to strict norms of beauty than men.

2. (2) Women are likely to be paid less than men for the same jobs.

3. (3) Women are more likely to suffer domestic violence than men.

Regarding these examples, the externalist account would imply that the term ‘women’ refers to the most objective type unifying the paradigms. But as we have seen, according to a plausible account of what the highest degree of objectivity consists in, those terms will refer to the common type that has the most explanatory and predictive power, given the explanatory aims and goals that we give priority to. And it is clear that the hypothesis that ‘women’ in those occurrences refers to people who are biologically female is a non-starter, on this account. For belonging to the type ‘being biologically female’ in itself does not explain why women are subject to norms of feminine beauty, or are paid less for the same jobs, or suffer domestic violence. A better hypothesis, with respect to these explananda, amounts to an objective type along the lines of Haslanger’s account of gender in terms of a social position in a social hierarchy.

Lawford-Smith (Reference Lawford-Smith2022) argues that the group of people who suffer the patterns of discrimination associated with being a woman corresponds to the group of people that are biologically female. But this is unjustified. In fact, many individuals, including cis women, trans women, and intersex people perceived as female all suffer similar patterns of subordination and oppression, especially when it comes to norms of feminine beauty, being subject to domestic violence, being excluded from positions of authority in academia, politics, and business, or being paid less for the same jobs. And furthermore, what is relevant according to externalism to determine the referent of ‘women’ is the question of what common objective types are the most explanatorily useful. And it is clear that the type ‘being biologically female’ cannot explain why the members of that group suffer those patterns of discrimination, since there are possible worlds where people who are biologically female are not so oppressed. What is explanatorily useful is a matter of the social position in a social hierarchy. With respect to this debate, social kinds are clearly more explanatorily useful than biological kinds.

I also examined other examples earlier where, I submitted, biological kinds seem, prima facie, to be more relevant:

1. (4) ‘We have to call all women over 40 for a smear test’ (uttered in the context of the clinic described earlier)

2. (5) ‘This bone belonged to a woman’ (uttered by a forensic anthropologist)

3. (6) Women are more likely to have their symptoms of a heart attack ignored.

The externalist framework would yield a similar conclusion. For example, regarding (5), this is interesting because it is useful to posit a biological kind as the referent of the occurrence of ‘woman’ in this utterance, precisely in order to explain the sorts of biases and prejudices that were common among archaeologists. The relevant factor is that archaeologists used to assume that a skeleton that was buried next to weapons was male, and this is why it is useful, in order to explain and resist sexism, to emphasize that DNA tests have shown that the skeleton in the Andes Mountains was actually female. This is why the externalist framework entails (in my view) that the occurrence of ‘woman’ in (5) refers to those who are biologically female. But this is a claim about utterance (5), and in no way can we infer anything about other occurrences of ‘woman’. Indeed, a preliminary analysis of these examples points to a pluralist conception, according to which different occurrences of the term ‘woman’ pick out different objective types.Footnote ³⁵

7.1 Witt on Uniessentialism about Gender

Charlotte Witt (Reference Witt2011) begins her discussion by focusing on the question of whether a person would become a different person if they changed their gender, or if their gender had been a different one. Witt is interested in exploring to what extent gender is part of our individual identity. She develops a very clear framework in order to offer an answer to this question, namely, what she calls uniessentialism, drawing on Aristotelian metaphysics. The main idea, in a nutshell, is that for an individual to be an individual, over and above the set of parts that constitute the individual, the individual must have a certain property or feature that turns those scattered parts into a unified individual. We can then say that the unifying property is uniessential to the individual. For example, a house becomes an individual house, over and above the parts (which could be neatly arranged on the floor before construction without yet constituting an individual house), because it acquires the function of providing shelter. The house is an individual, over and above the parts, because it has the function of giving shelter. It can perform this function better or worse (for instance the windows might be malfunctioning, or the roof might be leaky), but it is a house insofar as it has the function of giving shelter to people. Likewise, a human being is an individual over and above the parts (heart, lungs, liver, etc.) because the human being has several unifying functions that the individual organs do not. That is to say, the human being qua living organism is organized through some functions or principles that cause the individual human being to exist, over and above the parts. Furthermore, a person is an individual over and above the parts because the person has a unifying perspective, a self. This self then unifies all the random thoughts and emotions the person might experience.

For Witt, the human being and the person are not the same individual, since they have different functions: the human being has the essential functions of a unified living organism, whereas the person has as an essential function, a unifying perspective or a self. The human being could exist without the person, for instance if the human being was in a deep coma, and the person could exist without the human being, for example if we could transplant the brain into another human being, keeping the same thoughts, memories, and so on, unified by the same self. Given this distinction, Witt wonders: is gender uniessential to either the human being or the person? And it seems that the answer is negative: the human being would be the same living organism even if they underwent gender reassignment treatment; and the person would also be the same person (the same self or conscious perspective) if they underwent gender reassignment treatment. The self or the unifying perspective (that is, the subject of experiences) would not change. How then can we capture the intuition that in some deep sense, the individual might have changed? For Witt, the answer is as follows: in addition to the human being and the person, we can distinguish a third individual, namely, the social individual. This corresponds to the individual agent that is responsive to social norms and is socially positioned given the social norms, practices, and expectations that are operative in their social milieu given their social position. In Witt’s view, the human being and the person are not subject to social normativity in the way that the social individual is (for instance, the human being is responsive to biological normativity or functions, which the human being can perform better or worse). And how is this social individual unified? Why is there a social individual, over and above the different social norms and practices? Because of gender, Witt answers. In the same way that providing shelter unifies the parts of the house to form an individual house, gender as a social feature unifies all the social positions and social roles that the social individual occupies. Why? Because all these different social roles that the individual occupies are shaped by gender. No other social role, Witt argues, has the same shaping force: gender is uniessential to the social individual because gender is what unifies all the social roles, and in this way determines the social individual that the person performs. For example, many social roles that we occupy are clearly gendered: being a parent is gendered since the social norms are different depending on the person’s gender. Being a professor or a lawyer or a doctor or almost any other profession are all shaped by gender, that is, the social norms that constitute those roles vary in virtue of gender. What counts as a good professor or a good doctor, and what is expected of them, varies a lot depending on the person’s gender.

Assuming this framework, Witt is able to clearly articulate the important intuition that many of our social roles are shaped by gender in a fundamental way. As Witt argues, this means that gender is uniessential to the social individual, that is, the individual exists as a social individual because of gender as a unifying factor that shapes all the other social roles that the person is expected to occupy. In this way, the claim that gender is uniessential to the social individual becomes a useful way of articulating the intuition that gender is central to our social lives.

Witt’s argument is debatable,Footnote ³⁷ but my main aim here is not to examine Witt’s argument in detail, but rather to assess the benefits of her account of gender, independently of the arguments she provides to support it. In Witt’s view, gender is a fundamental social role that shapes all other social roles and social norms that we are subject to, and in this sense, if our gender were different, all the other social roles and social norms that determine the social individuals we are would also change. As I have explained, Witt does not reach this view by means of analysing the meaning of gender terms and examining what they refer to. She starts from a very different approach. We could nonetheless wonder whether her account of gender as a social kind might be compatible with the meta-metaphysical approach we developed in the previous sections, and there is no reason to say that it is not. In particular, someone could wonder about the benefits of ameliorating gender terms so that they might come to express gender concepts that capture Witt’s account of gender as an essential feature of social individuals. In addition, someone could wonder whether either contextualism or externalism might be concerned with the sort of social feature of individuals that Witt identifies with gender. It might be the case that on a contextualist account, at least in some contexts, the normative considerations that on subject contextualism determine the referent of gender terms in a context, might yield a referent for gender terms along the lines of Witt’s uniessential account. Moreover, on an externalist framework, someone might wonder whether Witt’s characterization of gender might be explanatory useful, at least in some contexts, so that it helps to determine the referent of gender terms. At least, Witt gives us some reasons for why her conception of gender is explanatorily useful: it helps to make sense of the powerful intuition that our gender shapes many of the social roles and norms that we are subject to, and that in some sense, if we had had a different gender, we would be different social individuals.

These considerations seem to strengthen the case for pluralism about gender. And it also helps to make sense of the diversity of accounts in the metaphysics of gender, where many of these different accounts seem to do explanatory work that is useful at least for some goals and purposes that the feminist theorist might have.

7.2 Ásta on Conferred Kinds

Ásta (2018) provides a very useful account of the metaphysics of social reality, by means of offering a detailed account of the nature of social properties, and what makes a property social. In particular, she argues, social kinds are social in virtue of something about other people’s actions. In a nutshell, Ásta argues that a social property is a social status that is conferred upon an entity or individual (or a group of individuals) by a person or a group of persons who have a position of authority, and where the social status comes with constraints on and enablements to the behaviour of the individual who is so conferred (2018: 2).

Ásta is explicit about the methodology that she uses to arrive at her account of social properties, to wit:

Ásta makes clear that she is mainly interested in giving an account of the nature of social properties, not an account of the meaning of our terms or the content of our concepts. However, given the meta-metaphysical framework that I developed in the first part of this Element, in order to give an account of what a social property is, the first step would be to figure out the application conditions of terms like ‘social property’ and related terms, and then to examine what entities, if any, satisfy those application conditions. (There is also room to ameliorate or revise the application conditions of the terms, if these are unclear, indeterminate, or flawed for any other reason.)

What should we make of this methodological difference? One could take an approach along the lines of Plunkett (Reference Plunkett2015) and Thomasson (Reference Thomasson2016), and claim that indeed, many debates on the metaphysics of the social are at least in part about what certain terms should mean (or what concepts those terms should express), even if this was not explicit in those projects. Plunkett (Reference Plunkett2015) and Thomasson (Reference Thomasson2016) revisit several traditional debates in mainstream metaphysics and argue that those debates could be seen at least in part as debates about what certain central terms should mean, or what concepts should be employed. I am sympathetic to this strategy, but I will not press it further here. Rather, I want to make two different points.

First, Ásta’s approach starts from paradigm cases of social properties. This is somewhat similar to Haslanger’s externalist approach in Section 6.2, that is, an account for figuring out which kinds or properties our general terms pick out (and therefore, an account to figure out the referent of certain terms). And as we explained earlier, this is also somewhat similar to Thomasson’s two-stage metaphysical inquiry, because Haslanger also emphasizes the need to include some additional information in the form of sortal descriptions (which arguably correspond to what Thomasson calls application conditions, even if minimal). Second, the versions of contextualism about gender terms that we examined in Section 6.1 also appeals to paradigm cases of ‘woman’, that is, those who have all the markers associated with being biologically female (and then the referent will be fixed by a relation of similarity to those paradigm cases, a relation of similarity which can change from context to context). Hence, there are some methodological similarities among these different approaches, even if one were to reject some of the theoretical assumptions of metaphysical deflationism.

Let’s put methodological issues aside and let’s have a look at Ásta’s conferralist account of social properties, and in particular her account of sex and gender as conferred social properties. Ásta argues that social properties are social statuses that are conferred to someone by others, where those who do the conferring are attempting to track a base property that (they believe) the subject instantiates, but what determines that the subject has the social property is not that base property, but rather the act of conferring the social status upon the subject. Ásta provides a useful schema to capture the different elements of the conferral of a social property, as follows:

Let’s now look at how Ásta applies this schema to the categories of sex and gender. Ásta makes a distinction between two kinds of social properties: institutional and communal (2018: 17–18). Institutional categories are those that are conferred by individuals in a position of authority in virtue of their place in an institution, that is, those who do the conferring have institutional authority. Some examples include, for instance, being the Prime Minister and being legally married. On the other hand, communal properties are also conferred properties, but they are conferred by people who do not have institutional authority, although some kind of social standing is necessary. One example is the property of being cool or being popular.

Ásta then argues that whereas sex is an institutional property, gender is a communal property. Her account of sex as an institutional property goes as follows:

And her account of gender is the following:

As we can see, in Ásta’s view, someone’s gender as a communal property can vary from context to context, because what property is conferred upon someone by the people who happen to have social standing in a context can vary. (As I understand it: both the people who have social standing, and what status they confer, can vary from context to context.) Ásta mentions several base properties that the people who do the conferring might be attempting to track. The first one sounds similar to Haslanger’s account of gender, because it has to do with the perception (right or not) of someone’s role in reproduction. But this is not the only ‘marker’ that matters for the conferral of gender, in Ásta’s view.

What should we make of this diversity of genders? I want to make two points. First, Ásta’s view also makes someone’s gender context-dependent (like subject contextualism in Section 6.1), but for different reasons. Ásta’s view is not about the meaning of any particular terms, but rather, a claim about the nature of the conferral mechanisms: she claims that people conferring the social status of gender can be attempting to track different base properties in different contexts. This is not a claim about the meaning of gender terms. However, if one is sympathetic to contextualism about the meaning of gender terms, then one could appeal to Ásta’s view, in order to offer further support for context sensitivity. For if there really is such diversity of conferred social statuses as part of our diverse social practices, then it would make sense to appeal to different criteria of similarity in different contexts, when trying to figure out the referent of gender terms, as I explained in Section 6.1 earlier. It is possible to combine a contextualist view about the meaning of gender terms and Ásta’s conferralist account of gender as a communal conferred property that can change from context to context.

In Section 6.2 we talked about how different kinds can be explanatorily useful for different purposes, and how this is relevant in order to determine the referent of general terms on an externalist framework (since the kinds that are more explanatorily useful in the vicinity have more chances of being the referent of the term that was introduced by ostension). Given this, and on the face of the diversity of gender kinds that our discussion so far illustrates, we can ask which kinds have the potential of being more explanatorily useful. This is one of the questions that Jenkins (Reference Jenkins2023) undertakes, as I will explain in the following section

7.3 Jenkins on Gender Kinds for Emancipatory Purposes

Jenkins (Reference Jenkins2023) is interested in giving an account of social reality that helps to illuminate which gender (and race) kinds are most useful for emancipatory purposes. As a result of her useful exploration of this question, she arrives at three different sorts of gender kinds that are useful for emancipatory purposes. In a nutshell, these are the following: (i) what she calls hegemonic kinds (which correspond more or less to Haslangerian social structures such as Haslanger’s characterization of gender in terms of the social position of privilege or subordination in virtue of someone’s perceived or imagined role in reproduction); (ii) what Jenkins calls interpersonal kinds (which correspond roughly to Ásta’s communal conferred properties, such as her account of gender as a communal property); and (iii) what Jenkins calls identity kinds, which include identity kinds in two different senses: First, identity in the sense of norm-relevancy, which corresponds to Jenkins’ (Reference Jenkins2016, Reference Jenkins2018) previous account of gender identity in terms of norm-relevancy (see Section 5), and second, identity in the sense of self-identification, which is different from identity as norm-relevancy. (See Jenkins (Reference Jenkins2023: 158–65) for further discussion.)

Jenkins is concerned with giving a detailed explanation of these different gender kinds because of her aim of offering accounts of social reality that help with emancipatory purposes. In particular, she introduces the concept of ontic injustice, and shows how the different gender kinds that she describes can be helpfully illuminated by means of exploring to what extent they exhibit ontic injustice or not. She characterizes ontic injustice as follows:

The notion of ontic injustice is useful, among other reasons, because it can help us understand what is the kind of wronging that is done to a person by a society that constructs social kinds of that sort. And in turn, she explains how the different sorts of gender kinds that she identifies in her tripartite classification relate to the notion of ontic injustice.

Once again, the question for us, at the end of this arch, is whether the framework that I have developed in this Element can help to guide us around this territory. And I think it has potential. As I have painstakingly emphasized, if we want to answer the question of what gender terms actually mean or should mean, we need to investigate the application conditions that the terms have or should have, and then, what entities or kinds, if any, better satisfy those application conditions. Now we have explored (even if briefly) a rich diversity of accounts about what gender might be (or what gender kinds there might be). In my view, the deflationist framework helps to make sense of this diversity, and additionally, it helps to understand where the disagreement might lie. For example, there might be disagreement at the level of which paradigm cases are selected, or at the level of what application conditions are ascribed to the terms, or at the level of which contexts are selected, or at the level of empirical claims about which entities satisfy certain application conditions, or which kinds are more explanatorily useful. Moreover, there might be disagreement at the level of the normative considerations that we give priority to, at the different contexts, that is, in order to select the paradigms, or to ameliorate the meaning of the terms, or to select the similarity criteria that are relevant in a contextualist account, or to identify the kinds that are the most explanatorily useful in an externalist account. All these choice points lead to a pluralist account of gender kinds.