When it comes to explaining the social categorization of people, I've been an advocate for performative theories since long before they became popular/mainstream. To be clear, I find the current mainstream notions of performativity deeply problematic because they overemphasize social constructivism and fail to highlight what I see to be the actual insight behind the original formulation of performativity. But all the same, I've long been a fan of (my understanding of) performativity.
However, in the tail end of chapter 8 of Whipping Girl, Julia Serano raises a major complaint against performative theories of sex/gender in particular— a complaint I agree with wholeheartedly, and which is not easily reconciled. Before getting into the problem she raises, I should probably explain what performativity is and why I've been such an advocate for it.
What does it mean to be human, or a woman, or an atheist, or a scientist? For any specific categorization the exact details will vary, of course. The question I'm asking is, once we abstract over the particular category, what does it mean to say that some person does or does not belong to that category? Many social categories are uninteresting in this regard. I am an IU student in virtue of the fact that I am registered here, pay tuition, attend classes, etc; there's a clear definition, and that definition is wholly uninteresting and uncontroversial. However, for many categories things aren't so cut and dried.
An ancient but still prevalent theory of these interesting social categories is the modernist theory of essential natures. That is, a human is anyone who has the essential nature of being human; a woman is anyone who has the essential nature of being a woman; etc. By relying on essential natures we can avoid problems with characteristic-based theories— such as being able to properly categorize birds that can't fly, humans who may not posses the requisite number of limbs, women who can't give birth, etc.
However, there are deep seated issues with the modernist theory. In particular, how do we identify this essential nature? As we move to different cultures or different time periods the idea of what it is to be X vary. Thus, there is no unitary essence of being X. Postmodernism resolves this problem by positing the existence of multiple different essences of being X. Thus, to be a woman is a context-dependent trait of possessing the particular essence of womanliness which is relevant to the given context. This pluralism is why postmodernist texts often pluralize everything in sight. We're not discussing "masculinity", we're discussing "masculinities"; etc.
But essentialist postmodernism still doesn't solve the question of how to define essential natures. If there are essential natures then we know they're not unitary, but we still don't know what they actually are, nor how to tell whether someone has them or not. More to the point, because one cannot —by definition— change their essential nature, the theory of essential natures (whether modernist or postmodernist) is reliably used to discredit, subjugate, and oppress minorities. People can never overcome the caste of their birth. Women can never overcome being born female. "Savages" can never overcome being born outside of "civilized" countries. And so on.
One response to this issue of oppression is, rather than asking whether someone possesses the essential nature of being X (and denying them category membership if they do not), instead ask them whether they identify as being a member of category X (and respect it when they demur). Thus, people who call themselves "gay" are in fact gay, whereas people who do not identify as "gay" are not gay regardless of who they sleep with. This provides an alternative solution to the problem of definitions varying across cultures and times since the particular terms describing categories will vary as well. That is, we do not attempt to translate the names of categories, because the categories themselves differ. From this perspective it does not make sense to ask if ancient Athenians were "gay" because the modern American conception of "gay" did not exist then, so clearly they were not.
Shortly after the rise of identity-based postmodernism the problems of this approach became apparent. The problem, of course, is that category membership becomes a political act. If I say I am a feminist, that's a political statement. If I say I am not a feminist, that's a political statement too. If I use a different term and say that I am a womanist (instead of a feminist), that too is a political act. Moreover, if anyone else says I am or am not a feminist or womanist, they are making a political accusation.
In addition to the social problems of these identity politics, there's another subtler problem. To satisfy the politics, when it comes to romantic issues, I should properly identify myself as a kinky polyamorous pansexual femme lesbian trans woman, or something similarly convoluted. But this bag of identifiers fails to express a holistic view of my identity, of who I am. By breaking my identity down into these separate parts, something is lost. There are interactions between these different facets of identity— for example, whether I am attracted to someone or not depends both on my pansexuality (I don't exclude any particular gender) and my lesbianism (I prefer women and female-of-center genders). And there are intersectionalities between these different facets— as a trans woman I get to deal with not only misogyny and transphobia but also transmisogyny.
A different alternative is to look at what it is that people actually do. Part of what we do is to declare (non)membership in various groups, but we also do other things as well— wear particular clothes, wear our hair a particular way, talk a particular way, pray a particular way, etc. Everyone does things, and everyone is something, and there tends to be a correlation between our actions and our identities. But correlation is not causation, so the question of causation is still open.
From the essentialist perspective our identity comes first, and our actions are merely expressions of that inherent identity. The performative trick is to consider that it might be the other way around: perhaps our actions come first, and our actions serve to construct our identities.
One of the other major problems with the essence- and identity-based theories is that they fail to account for the fact that the way we act varies across different contexts. Do Muslims stop being Muslim while they don't pray? Do women stop being women while they act in stereotypically masculine ways? Are people only queer or straight when they're in the bedroom? These questions sound absurd, but if all our actions are merely the expression of some inner nature then we need to be able to account for the fact that people do not always act in a coherent and non-contradictory manner. Changing the way we behave is not a symptom of having a fractured or poorly developed identity, it's a symptom of the fact that we have to interact with other people and people are messy.
The important thing about performativity, IMO, is this question about causality. Thus, it is important to distinguish the claims of performativity from the claims of social constructivism. Social constructivism says that categories like "birds", or "women", or "language" are things which are defined by social convention rather than things which arise from the natural state of the world. For example, "a language is a dialect with an army and a navy"; that is, whether we call something a language or not has more to do with political power and our prejudices than it does with the internal structure of that form of communication and how it relates to other forms of communication. Weak forms of social constructivism are self-evidently true: I am wren because I'm called "wren", I'm tall because I'm called "tall" (but I may not be "tall" in another country). The bigger question is the extent to which stronger claims of social constructivism are true.
Because performativity says that our identities arise from the actions we perform and how those acts are interpreted by others, it necessarily requires that notions of identity are at least to some extent socially constructed. But it does not, IMO, require that identity is wholly socially constructed. First, we can perform actions in the absence of other people, and these actions (or our choice to perform them) may still be driven by the desire maintain an identity. Second, performativity says nothing about what precisely motivates us to choose certain actions over others. We may choose to do certain things because they're required for admission to a group, or because they help us avoid bad consequences, but we may do other things for no reason other than some personal inclination. The point is, once we start to act those actions will be interpreted in particular ways and it is through that interpretation that people (including ourselves) will come to understand "who we are". Thus, whatever that motivating force is, it should therefore be considered separately from the notion of "identity" itself. Sometimes we do things because they "feel right", but other times we do things because we want to think of ourselves in certain ways or we want others to think of us in certain ways. Whenever the latter obtains, we are consciously using our actions to constitute our identity. And when the former obtains, nothing about that invalidates the claim that identity is constructed by performance.
The issue Serano brings up arises when performance-based theories are applied to sex/gender in particular. (Though similar problems may be evident for other social categories.)
As Serano discusses, all people upon encountering another person compulsively gender them; that is, we decide based on only a few cues whether the person is a "man" or a "woman", even if those cues are not at all sufficient to actually decide the question. What cues are most relevant vary by culture, but the compulsive gendering is universal. (Serano 2007:162–164)
In the US (and many other Western nations) the most important cues are secondary sexual characteristics: height, build, beard, breasts, voice, etc; cues like whether someone acts masculine or feminine or how they carry themselves only matter when the secondary sexual characteristics are too ambiguous to decide (Serano 2007:102,190–193). These secondary cues are important for deciding how "well" someone meets our expectations after gendering them, but they are not of particular import in the characterization process. (This is not the case in all cultures.) The problem is, secondary sexual characteristics are not actions. We often choose to exaggerate secondary sexual characteristics by wearing particular styles of clothing —and the choice to wear those clothes is most definitely a performance—, but if we see a woman in men's clothing we have no hesitation about calling her a woman. There is no confusion about whether she's a woman or not.
As evidence for the claim that gendering in the US is based on secondary sexual characteristics Serano cites her own transition. By her account, she doesn't dress up and yet noone has any hesitation in identifying her as a woman. I don't know whether she means she wears clothes which aren't tailored for women or what, but in any case I have no particular reason to question her account. In my own experience the effects of hormone therapy have been by far the deciding factor in how people gender me. So while I do prefer to dress in a way that shows off my femininity, doing so is part of "performing femme"; it is not part of "performing woman".
Indeed, I can think of no particular action which is relevant to influencing or constructing or otherwise affecting the fact that I am a woman. It took some time to realize that my innate feelings and tendencies meant that I was a woman, but those feelings and tendencies have always been there, and upon having that realization I've never had any justification for altering my conclusion. And, like most trans people, after having the realization, I spent a lot of time and energy trying to deny it or change it— all of which without any effect whatsoever. Wishing not to be trans cannot bring one's physiological sex and psychological sex into alignment; and wishing not to experience dysphoria cannot correct the incompatibility between one's hormonal sex and neurological sex. Biology just doesn't work that way, at least not in humans.
My decision to transition certainly is an action, so we could try to hang performativity on that. But I do not feel that transitioning has altered my gender. I've always identified as a woman— in spite of 32 years of being misgendered by everyone around me. That people have finally begun correctly gendering me over the past year doesn't change a thing. Sure, my understanding of women's lived experiences has changed as people now treat me differently, and that refined understanding has caused me to perceive the fine details of my identity somewhat differently, but it doesn't change any of the gross details— which are what's at stake here.
I'm still not entirely sure how to reconcile this data with my overall preference for performativity over identity politics and essentialism. The problem is, at the end of the day, certain facts are just anchored in the physical world and sometimes we don't do anything to demonstrate those facts to others. The facts are my psychological and neurological sex are female, my personal inclinations tend towards femininity, and I'm predominantly attracted to women. Given the cultural conceptions available to me, these facts entail that I'm a woman, I'm femme/feminine, and I'm a dyke/lesbian. But beyond the facts, these terms ("woman", "femme", "feminine", "dyke", "lesbian") have cultural interpretations and assumptions layered upon them. I do things to show off my lesbianism so that interested parties will know, and this can be easily understood as performing lesbianism. I also do things to show off my femininity, though this has more to do with enjoying the activities myself than it does with advertising my femininity to others; whether this counts as performative or not is far less clear. But I don't especially do anything to show off my womanhood itself, as distinct from femininity; and therein lies the challenge. Any decent theory of social categorization will need to handle all three of these examples without running afoul of the problems encountered by modernism, essentialism, or identity-ism.