A short guide to analytic philosophy of gender

Thjs post is a series of notes I took while making my way through the philosophy of gender. Since writing it about 6 months ago I've done a lot more reading into epistemology and the philosophies of science and language, and there's a lot of revisions I'd like to make. In other words, I now think much of the content of this post is dubious but I'll keep it up for the time being as it might serve as a helpful resource for understanding the literature.

The post is about the philosophy of language, the question "what is a woman?" and taxonomizing positions in the philosophy of gender.

1. Metasemantics and Methodologies

1.1 The dictionary. A natural starting place for answering questions like “what is a woman?” is the dictionary. But what can the dictionary actually tell us? Quine writes:

The lexicographer is an empirical scientist, whose business is the recording of antecedent facts; and if he glosses 'bachelor' as 'unmarried man' it is because of his belief that there is a relation of synonymy between these forms, implicit in general or preferred usage prior to his own work. The notion of synonymy presupposed here has still to be clarified, presumably in terms relating to linguistic behavior. (Quine 2004, p. 35).

        Insofar as we think that usages may diverge from meanings and meanings from metaphysics, the dictionary's report of synonymies in usage will be of limited import for answering "what is" questions. In order to extract answers to “what is” questions from the dictionary’s reports of linguistic behavior, one must have some scheme for inferring semantic and metaphysical theses from them.

1.2 Lessons from Byrne and Dembroff. Sometimes answering the question “what is a woman” is hindered by philosophers not making these schemes clear. A particularly illuminating example is found in a disagreement between philosophers Robin Dembroff and Alex Byrne. Byrne uses the fact that dictionaries define ‘woman’ as ‘adult human female’ (hereafter abbreviated as AHF) in his case for the thesis that, necessarily, something is a woman iff it is an adult human female (Byrne 2020, p. 3786-7). Dembroff, critiquing Byrne, writes:

Words get their meanings and application conditions from their everyday use, and we have little to no reason to believe that such everyday use would glom onto the metaphysical structures that ultimately explain the phenomena that people refer to using natural language. (Dembroff 2021, p. 10.)

Dembroff argues that Byrne puts too much stock into ordinary language in determining the metaphysics of gender. I will put aside my general agreement with Dembroff’s sentiment here and instead use this dispute to dig up an underlying metasemantic conflict. I’ll make this disagreement clear by contrasting Dembroff’s above position on meanings with the positions Byrne gestures to in the introduction to his paper. Byrne writes:

It is a familiar point from Kripke and Putnam that conceptions can be radically mistaken.  Since some people’s conceptions of women include AHF, while others’ conceptions are inconsistent with it, some people’s conceptions of women will be in error no matter what … there is some connection between ordinary use and meaning, but it is highly indirect and not well understood. (Byrne 2020, p. 3784-5.)

Dembroff endorses an account of meaning that makes the leap from semantics to metaphysical structures look sort of preposterous. On Dembroff’s view, meanings–particularly those of gender terms–evolve constantly with culture and language. So the meaning of ‘woman’ is unstable, which could make it a poor guide to metaphysical structures which are often seen as less subject to the accidents of social development. Byrne doesn’t endorse such a theory, and in fact gestures towards certain semantic externalisms–possibly mitigating the strength of Dembroff’s concern. Semantic externalism here will be understood as the thesis that meanings are largely determined by features external to the minds of speakers—i.e. semantic externalism may be true of some term t if t’s meaning when uttered by a speaker is at least partially determined by features external to the speaker’s mental states.

For instance, suppose Byrne endorsed a causal theory of reference where (crudely) the term ‘woman’ referred to the kind it was applied to at its ‘baptism’ in a linguistic community. The application conditions of ‘woman’ would not exactly be determined by everyday use, and in fact, everyday use would be capable of being wrong. From this standpoint, the link between metaphysics and dictionary evidence may be that dictionaries give us prima facie reason to think AHF is the relevant kind picked out by applications of ‘woman’ (although, of course, the linguistic behaviors that dictionaries track might be systematically mistaken), and therefore people who are accurately referred to as women would have to be instances of the type AHF. This would make the bridge from semantics to metaphysics more straightforward, and would support Byrne’s claim that this bridge could be built entirely from “disquotational principles'' (Ibid., p. 3785, Byrne 2021 p. 14). Unfortunately, Byrne does not make his metasemantic views clear, so the appropriateness of his claim that dictionary definitions (and thus, linguistic behavior) support his modal thesis is hard to discern. In his subsequent reply to Dembroff, Byrne says little more to clarify the inferential machinery used to construe dictionary definitions as support for his thesis, leaving us only with the notion that the relationship between the two is “highly indirect and not well understood” and an offhand reference to Putnam and Kripke. Thus, even if one does not accept Dembroff’s position on how the meaning of ‘woman’ is determined, their objection brings to light one potential shortcoming of Byrne’s argument: the obscurity of the link between dictionary definitions and semantics. Prodding this link uncovers metaphilosophical disagreements.

Hopefully this example makes my concern more concrete: this is an instance where metaphilosophical ambiguity inhibits progress. Some philosophers do a better job of making metasemantic commitments explicit, however, providing a clearer picture of how our inquiry into womanhood might proceed.

1.3 Externalism and Haslanger. Byrne is not the first to remember the famous externalists in attempting to find out what a woman is: Sally Haslanger defends social constructionism about race and gender from charges of unintuitiveness by establishing a distinction between the conceptions often employed by language users and the kinds to which racial and gender terms actually might refer. (The next section will explain what exactly ‘social constructionism’ means.) Supposedly, the goal of such an approach is not revising our concepts but “understanding our original concept[s] better” (Haslanger 2006, p. 106). Put more explicitly in externalist language, people can be mistaken about the concepts they employ in everyday use, and while they may think that terms like ‘woman’ or ‘black’ refer to natural kinds (e.g. reproductive features or skin color, respectively), the kinds these terms refer to may instead be unified by an objective social type.

This line of thought creates the basis for Haslanger’s distinction between manifest and operative concepts: manifest concepts are the concepts language users think they are employing and operative concepts are the concepts they actually do employ. Haslanger’s descriptive methodology, as distinguished from her ameliorative project covered in 1.4, is an attempt to discern the operative concepts underpinning gender terms by searching their paradigmatic applications for unifying objective types. This methodology places secondary priority on intuitions or the reports of language users concerning the concepts they employ, as Haslanger thinks these evidences provide more insight into the manifest concepts at play. There is a lot of further discussion to be had regarding the implications and presuppositions of Haslanger’s externalism (see: Diaz-Leon 2012; Haslanger 2010), but this outline explains a popular angle for answering some “what is” questions. It is, however, not the angle Haslanger is famous for.

1.4 Stipulation and amelioration. Elizabeth Spelman and Judith Butler popularized what is called the normativity problem for definitions of ‘woman’. Haslanger explains that, “[t]he normativity problem raises the challenge that any effort to define women will problematically privilege some women and (theoretically) marginalize others, and will itself become normative. One worry is that bias inevitably occurs in deciding which experiences or social roles are definitive; a second worry is that if someone wants to be a “real” woman, she should conform to the definition of women provided, and this will reinforce rather than challenge male dominance” (Haslanger 2000, p. 46). This is often seen as a concern for methodologies like the sorts Byrne used in 1.2 and Haslanger outlined in 1.3. But “what is” questions sometimes aren’t answered with definitions that are merely intended to describe. There is

…a variant type of definitional activity which does not limit itself to the reporting of pre-existing synonymies. I have in mind what Carnap calls explication -- an activity to which philosophers are given, and scientists also in their more philosophical moments. In explication the purpose is not merely to paraphrase the definiendum into an outright synonym, but actually to improve upon the definiendum by refining or supplementing its meaning. (Quine 1951, p x.)

The notion of chiseling out of pre-existing synonymies, usages, and meanings an improved definition for a particular context or purpose explicitly allows for the admission of practical reason into one’s investigation. Often these reasons are pragmatic, meaning we

…fix on the particular functions of the unclear expression that make it worth troubling about, and then devise a substitute, clear and couched in terms to our liking, that fills those functions. (Quine 1960, p. 258-9.)

This idea of explication or stipulation or conceptual engineering has found a comfortable home in the philosophy of gender after being championed by Haslanger, who argued that we need not limit the constraints of our explication to just pragmatic concerns. That is, we may include ethical/political concerns in the list of things that make the definition of ‘woman’ “worth troubling about,” opening up our candidate conceptions to ones that are revisionary and meant to tackle political problems. This approach integrates normativity from the get-go, supposedly accounting for Butler’s worry by hoping for, not fearing, normative consequences. Contrasted with operative and manifest concepts, the products of Haslanger’s ameliorative inquiry are called target concepts. Haslanger lists several objectives that a target concept ought to accomplish–some normative and some theoretical, and later ameliorativists often add to or revise these constraints.

There have been many challenges to ameliorative inquiry (hereafter ‘AI’) and Haslanger’s particular ameliorative definitions. One such challenge is that Haslanger fails to overcome the normativity problem. That is, she ends up reinforcing oppressive gender norms with her particular social-positioning account or excluding people she ought to include (e.g. trans people). Others argue that Haslanger’s definitions, while potentially fruitful for some ameliorative projects, have less merit in clinical contexts (Toyoshima et al 2021). An objection to Haslanger from a different metaphilosophical standpoint is that AI looks problematic given certain views about meaning. This is another instance where metasemantics seep into the philosophy of gender. Koch explains how issues can arise:

If a given lexical item has its meaning determined by some condition c, then what conceptual engineers ought to do is to change c. (Koch 2021, p. 2.)

Metasemantic critiques of conceptual engineering (of which AI is a type) tend to argue that given certain popular metasemantic views, conceptual engineers overestimate the amount or type of control they have over c. This is intended as a problem for AI, because one of the supposed goals for AI is to amend or revise meanings. For example, Koch recites an externalist critique that might be amenable to philosophers like Byrne:

If (a certain branch of) [semantic externalism] is true, then changing the meaning of lexical items is either impossible or at least very hard to achieve. (Ibid, p. 4.)

Recall the causal theory outlined in section 1.2. If the reference for a term within a linguistic community is at least largely determined by the conditions of its baptism, then intentional meaning-change (and therefore conceptual engineering) may be difficult. Or suppose meanings are largely determined by patterns in usage within a linguistic community. Conceptual engineers, to effectively change meanings, would have to change patterns in common usage. Similar problems may be produced for a range of theories where meanings are not easily altered by mere changes in the verbal behaviors, attitudes, or intentions of language users.

Other philosophers worry about the authority problem, which is the hurdle, “of explaining, to the people who are urged to adopt an engineered concept, why they have reason to adopt this concept and structure their affairs in terms of it” (Queloz 2022, p. 1). Putting aside for a second the potential challenge of changing meaning, there is also the difficulty in producing justifications to the target language users for why they ought to adopt the target concept. For example, suppose we revise a concept C into the target concept T. How can we urge the adoption of T over C? Some philosophers gesture towards theoretical virtues: perhaps T offers better explanations and predictions at less theoretical cost than C. But for many concepts, their prevalence and usefulness is not because of their theoretical virtues at all—so why would we accept T as a proper substitute for C if it improves upon C in ways deemed unimportant to C. Wittgenstein gives an analogy:

…if your concern is to cut bread and you ask for a bread knife, you will hardly thank me if I give you a razor blade because it is sharper. (Wittgenstein 2000, MS 120, 142v; quoted from Queloz 2022.)

Furthermore, if T diverges greatly enough from C, how can conceptual engineers avoid the charge that they are changing the subject entirely? Conceptual engineers are thus faced with three primary challenges when offering revised conceptions: (i) providing a metasemantics and/or cost-benefit analysis that overcomes the apparent difficulty of meaning-change, (ii) identifying and preserving the functions that made the original concepts “worth troubling about,” and (iii) preserving the continuity of subject matter.

1.5 Different projects. To summarize: drawing from Haslanger, I think there are a few main projects that are undergone when philosophers give accounts of gender. The first is the conceptual project. Conceptual projects primarily investigate linguistic behaviors, manifest concepts, beliefs about gender, etc.. The conceptual projects are not so much concerned with who does and doesn’t actually belong to a given gender category, but rather who is popularly considered to belong to that category, what popular beliefs there are about the membership requirements for such a category, and what treatment these beliefs prompt. Dictionary definitions can perhaps be construed as results of a conceptual project, because they report linguistic behaviors but are disinterested in whether certain usages cohere with meanings or what metaphysical structures underpin gender discourse. Another example is in Kwame Anthony Appiah’s “ideational” method in the philosophy of race, which starts by surveying beliefs about race categories by investigating historical views.

Conceptual projects may be contrasted with descriptive projects which are more interested in metaphysics and semantics and operative concepts. Descriptive projects aim to tell us who actually is a woman and in virtue of what, the actual meanings of our terms, our operative concepts, and may be happy to call those beliefs and behaviors explored in the conceptual project inappropriate. The relationship between or separability of these two projects is determined by metaphilosophical inclinations and the conclusions of the projects. For example, those who are skeptical of externalist views of meaning or sympathetic to use-theories might think that it is hard to pull apart the two projects. Perhaps there is no meaning or metaphysics of ‘woman’ that can be separated from the way ‘woman’ is used, or its meaning largely depends upon internal features of speakers—thus the conceptual project more or less may exhaust our semantic facts. Or perhaps the conceptual project reveals to us certain important features that gender is thought to have, and our descriptive project reveals that nothing matches these descriptions. For others, like Haslanger, the conceptual project might reveal the beliefs of language users that help to construct systems of oppression and mark people for specific treatment, and these social factors will be utilized in the descriptive project.

The last kind of project is the ameliorative or engineering project that asks what our concepts ought to do for us and builds an account with these goals in mind. These projects can clash with the other projects because they do not fear revision.

2. Taxonomizing Views

Views in social ontology can be difficult to taxonomize, so I’ll be drawing heavily from Matthew J. Cull’s “Pluralism, Trans Identities, and Feminist Philosophy.” Cull separates positions in the philosophy of gender into categories peculiar to social ontology which will let me limit my use of terms like realism and nominalism, which is good because they are stupid. I find the resulting taxonomy to be uniquely helpful for understanding key disagreements.

2.1 Essentialism and Anti-essentialism. ‘Gender essentialism’ is a common term in the philosophy of gender. In the case of ‘woman,’ essentialists are often thought to say one or both of the following things:

Weak essentialism (WEss): there is some property that all women share in virtue of which they are women

Strong essentialism (SEss): there is some property that is essential to all women in virtue of which they are women. 

 

WEss means merely that there is some single property or group of properties that is a necessary and sufficient condition for being a member of the kind women. The difference between this claim and SEss is that SEss adds that this characteristic is essential to every woman (e.g. that it cannot be changed). SEss points more clearly in the direction of things like gendered souls or certain AHF accounts where the feature which grounds one’s femaleness cannot be changed, whereas WEss is arguably compatible even with Haslanger’s social positioning account of gender. That is, under Haslanger’s account, all women are women in virtue of their being systematically subordinated on the basis of perceived sex characteristics (though this subordination can differ in dimension and context). Thus within essentialism I will distinguish between social essentialism like that of Charlotte Witt and Haslanger, and a biological/natural essentialism that unifies all women through biological characteristics.

I think that anti-essentialism ought to be construed as the denial of both WEss and SEss: anti-essentialist theories of gender reject that there is some characteristic of all women that serves as the necessary and sufficient condition for their membership in the category women. Examples include Butler’s performativism, error theory, deflationism, contextualism, dispositionalism, property-cluster views, and family resemblance. There are four main anti-essentialist arguments, summarized briefly by Theodore Bach:

1. Inseparability: Gender is not a feature that exists and develops independently of other (social) features such as race, class, and religion.

2. Universality [or commonality]: There is simply no feature that all women of all times and places have in common.

3. Immutability: By defining women according to property P it follows that (i) the elimination of P entails the elimination of women, (ii) if an individual possesses P at time 1 and loses P at time 2, then that individual is no longer a woman.

4. Normativity: Defining women according to an essential property privileges those who possess this property, or who possess more of it, and marginalizes those who do not possess this property, or who do not possess enough of it. (Bach 2012, p. 235.)

Unsurprisingly, I think the strength of these arguments depends on metaphilosophical commitments and what methodology one uses to give their account of gender. For example, arguments 1, 3, and 4 seem pressing for conceptual engineers, whereas arguments 1 and 2 seem pressing for those who think women form a natural kind. I’ll hopefully investigate the significance of these arguments and the views they mesh the best in a later post. (I’m thinking that if I say enough shit like “I’ll talk about this in the next post” I’ll start to understand what I’m doing here.)

2.2 Structuralism and anti-structuralism. Cull describes structuralism as the position that gender categories can be given a logical structure. That is, that there are principles that govern membership and the boundaries of the category, and these principles and boundaries follow some structure. For example, a family-resemblance category or stable cluster of properties seem to have a structure to them. Anti-structuralists deny that any such logical structure can or should be produced. Cull distinguishes between three anti-structuralist claims:

Metaphysical Anti-Structuralism (MAS): The category woman does not have a unifying structure.

Epistemic Anti-Structuralism (EAS): We do not have epistemic access to unifying structure of the category woman (if it has one).

Ameliorative Anti-Structuralism (AAS): In putting forward a target concept or concepts for the category woman, we ought not to stipulate or give a unifying structure for that category. (Cull 2020, p. 74).

Accepting any of the above claims will constitute dissent towards structuralist accounts, so each is sufficient for anti-structuralism. The primary example of anti-structuralism is Butler’s performativism and its relatives, but error theory entails MAS and EAS as well. I lean towards a view like MAS and hopefully I’ll flesh that out later on. (Or maybe I won’t. I wrote most of this post while clocked out on Nyquil. I don’t know what externalism is.)

2.3 Constructionism. In section 1.3 I referenced constructionism, the view that gender is socially constructed. I’ll briefly explain a few different interpretations of this claim, starting with Haslanger’s famous distinction between causal and constitutive social construction. According to Haslanger,

X is socially constructed causally as an F iff social factors (i.e. X’s participation in a social matrix) play a significant role in causing X to have those features by virtue of which it counts as an F (Haslanger 2003, p. 317).

In other words, to say that X is socially causally constructed as an F is to say that there are some features in virtue of which X is an F and that social factors will play a large role in an explanation of why it is that X has those features. Haslanger argues that there are contexts where this definition (drawn from Hacking 1999’s analysis of social construction) does not capture the intentions of social constructionists. She argues for the inadequacy of causal social construction with the warning that 

we should avoid conflating social kinds with things that have social causes. Sociobiologists claim that some social phenomena have biological causes; some feminists claim that some anatomical phenomena have social causes, for example, that height and strength differences between the sexes are caused by a long history of gender norms concerning food and exercise. (Ibid.)

Haslanger uses concerns like this to distinguish causal social construction from constitutive social construction:

X is socially constructed constitutively as an F iff X is of a kind or sort F such that in defining what it is to be F we must make reference to social factors (or: such that in order for X to be F, X must exist within a social matrix that constitutes Fs). (Ibid., p. 318.)

So, to say that X is socially constructed constitutively as an F is to say that F itself cannot be defined without reference to social factors. This switches the focus from the causes of a particular individual being F to the category F itself. Because of this difference in scope, constitutive social construction is less permissive than causal social construction. For example, one would not need to reference social factors in defining what it is to be short, but sometimes shortness might be caused by social factors. 

Some philosophers have gripes with constitutive social construction too, and would rather cast the positions of gender theorists in terms of grounding–(briefly) a primitive, acausal, asymmetric, irreflexive, “in virtue of” explanatory relation. For example, Aaron M. Griffith amends the permissiveness of causal social construction with

Social Construction as Grounding (SCG): S is non-causally socially constructed as a K iff [S is a K] is at least partly grounded in particular features of social reality. (Griffith 2017, p. 3.)

SCG  is meant to account for Haslanger’s worry about the permissiveness of causal social construction while remaining faithful to its important implication that, “the grounds for social identities are, in some sense, within our control. They may be transformed or eliminated by specific kinds of social action” (Ibid.). Whether the metaphysical tool of grounding is needed as a supplement to Haslanger’s constitutive social construction basically hinges on whether the notions employed in Haslanger’s definition are adequate for a satisfying account of social construction. Griffith also argues that the introduction of grounding into the debates over theories of gender provides dialectical clarity: disagreements between views may be helpfully cashed out in terms of what they think grounds facts like [S is a woman].

While there are several undiscussed understandings of social construction, this section has listed a few prevalent ones that capture the heart of social constructionist accounts of gender. Each account will likely deviate from Hacking, Haslanger, or Griffith’s portrayals of social construction in minor ways, but hopefully this section provides insight into roughly what it is that philosophers of gender mean when they claim that “gender is a social construct.”

Works Cited

Bach, Theodore (2012). “Gender Is a Natural Kind with a Historical Essence.” Ethics 122 

(2):231-272.

Butler, Judith (1990). “Performative Acts and Gender Constitution”, in Performing Feminisms, 

S-E. Case (ed.), Baltimore: John Hopkins University.

Byrne, Alex

(2020). “Are women adult human females?” Philosophical Studies 177 (12):3783-3803.

(2021). “Gender muddle: reply to Dembroff.” Journal of Controversial Ideas 1 (1).

Cull, Matthew (2020) Engineering Genders: Pluralism, Trans Identities, and Feminist 

Philosophy. PhD thesis, University of Sheffield.

Dembroff, Robin (2021). “Escaping the Natural Attitude About Gender.” Philosophical Studies 

178 (3):983-1003.

Diaz-Leon, Esa 

(2012). Social Kinds, Conceptual Analysis, and the Operative Concept: A Reply to Haslanger. Humana Mente 5 (22):57-74.

(2016). “Woman as a Politically Significant Term: A Solution to the Puzzle.” Hypatia 31 (2):245-258.

Griffith, Aaron M. (2017). “Social Construction and Grounding.” Philosophy and 

Phenomenological Research 97 (2):393-409.

Haslanger, Sally 

(2000). “Gender and race: (What) are they? (What) do we want them to be?” Noûs 34 (1):31–55.

(2006). “What good are our intuitions: Philosophical analysis and social kinds.” Aristotelian Society Supplementary Volume 80 (1):89-118. Edited by Sally Haslanger.

(2010). “Language, Politics, and “The Folk”: Looking for “The Meaning” of ‘Race’.” The Monist 93 (2):169-187.

Koch, Steffen (2021). “The externalist challenge to conceptual engineering.” Synthese 198 

(1):327–348.

Queloz, Matthieu (2022). “Function-Based Conceptual Engineering and the Authority Problem.” 

Mind 131 (524):1247-1278.

Quine, Willard V. O.

(2004). “Two Dogmas of Empiricism.” In Quintessence. Harvard University Press. Edited by Roger F. Gibson Jr.

(1960). Word & Object. MIT Press.

Toyoshima et al (2021). “Representing Gender in Ontologies: A Dispositional Perspective.” 12th 

International Conference on Biomedical Ontologies (ICBO 2021), Bozen-Bolzano, Italy. 

pp.1-12.