l'appel du vide

ceci n'est pas une mise en abyme

(or: it’s fine to say that women are people who identify as women)
18 03 2026


‘Women are people who identify as women!’ goes the pro-trans mantra. On its face, this account seems to underwrite progressive views for cheap; it doesn’t incur contentious empirical commitments, does not (at least, not irreparably) exclude marginalized trans women, and isn’t a mouthful. In the face of this success, a contingent of social conservatives—and some progressive dissenters—have identified an apparent trick at play in this use of self-identification: it’s circular! It’s not informative or contentful or enlightening (or anything else that a definition is supposed to be), because it defines the word ‘woman’ using the word ‘woman.’ Self-id and its definitions for our gender concepts, then, are to be buried alongside brain sex and constructionism and long-winded jabber about schemata and desires. This conclusion, though, is too quick. The circularity objection relies on an oversimplification of circularity and an ignorance of ‘self-identification’’s logical behavior. Tidying up these concepts, I think, reveals how the objection from circularity plays a trick of its own.

semantic circularity

There is a prolific characterization of semantic circularity in terms of the inclusion of a word within its own definition. This, the supposition that

a definition is circular when (and only when) the definiendum appears within the definiens

can be called the ‘syntactic definition’ of circularity. This definition is often assumed, without argument, in the course of conversations about self-id. Anti-trans demagogue Matt Walsh, for example, uses this syntactic definition in his 2022 documentary ‘What is a Woman?’ when he urges Dr. Patrick Grzanka (one of the film’s progressive interviewees), to “define the word woman without using the word woman,” before accusing Grzanka of defining ‘woman’ “circular[ly].”

The syntactic definition of semantic circularity is wrong. There are definitions which feature their definienda which are nonetheless not circular. One need not do much legwork to find such counterexamples: Oxford defines the verb ‘to dust’ as ‘to sprinkle with dust or powder’; Merriam-Webster defines ‘to pin’ as ‘to fasten, join, or secure with a pin’; and Cambridge defines ‘to google’ as ‘to search for something on the internet using the Google search engine.’ Each of these definitions, jointly representing the top three English-language dictionaries, are circular in the syntactic sense, as the words they define appear within their definientia. But seemingly, they are not really circular. Supporting this intuition, the weaker argument is that these institutions are highly respected authorities on English-language lexicography, and that they all commit unawares the same obvious error is, on its face, highly unlikely. The stronger argument is that each of these definitions do what definitions are supposed to do! They are informative and enlightening, and anyone who failed to grasp what ‘dust,’ ‘pin,’ or ‘google’ meant as verbs would come to understand their meanings by reading the above definitions.

Of this, a minimal upshot is that a definition’s circularity (in the syntactic sense) is not itself a force for revision: there are successful and informative definitions which would nonetheless be classed by the syntactic definition as circular. More significantly, these arguments compel us to refine the landscape surrounding the circularity objection. A different, revised definition of semantic circularity is required, one which is able to (1) maintain semantic circularity’s status as something to be avoided, while (2) avoiding the above counterexamples, and, for the objection to work, (3) count self-id as circular.

Two points ought to guide this revision:

  1. The issue with paradigm instances of semantic circularity (like ‘something is large if and only if it takes up a large space’) is that they can only be understood by those who already understand the definiendum. This is where the normative force of accusations of circularity comes from: definitions are supposed to be informative (and enlightening and contentful etc.), but circular ones are not, as interpreting them requires the knowledge they are supposed to provide. If our revision captures what induces failure in these paradigm instances, then it will avoid the counterexamples to which the syntactic definition is prone while still tracking the term’s intuitive extension.
  2. The above counterexamples all involve cases of polysemy, where one word takes on multiple meanings (in these cases, one as a noun, and one as a verb). These counterexamples pose a problem for the syntactic definition because individual words can have multiple meanings, and one can understand one of these meanings without understanding them all. Cases like these allow us to generate counterexamples of the above sort, where the source of circularity’s normative force (that one must understand the definiendum to understand the definiens, preventing the definition from being enlightening) is absent even though its typical syntactic markers (the word’s appearance in its definition) are present. To avoid this, the revised definition can focus meanings as opposed to the words that bear them.

Keeping these considerations in mind, we can state our revision of ‘semantic circularity’ thus:

A definition is semantically circular if and only if understanding the definiens requires knowing the meaning of the definiendum.

This definition provides the correct verdict on the above cases: understanding ‘Google’’s meaning as a noun does not require understanding its meaning as a verb, and, because the definition states the meaning of its verbal sense, using it in its noun sense therein induces no semantic circularity. Clear cases of semantic circularity, like the above ‘something is large when (and only when) it takes up a large space’ are also classed correctly by the revised definition: knowing what it means for a space to be large is parasitic on understanding what ‘large’ means, which is the expression being defined.

This much, so far, seems relatively intuitive. This understanding of circularity is not something I take to be novel or difficult, rather, my overlaboring thereof serves only a base-covering and pedagogical purpose. Specifically, I aim to make salient the narrow gap that opens up for self-id: the possibility that, despite its unproblematic syntactic circularity, it could escape the charge of circularity in the important, revised sense.

mention and use

Every word has a dual meaning: it can serve the function it normally serves, referring to the objects it normally refers to, or it can refer to itself, the word. ‘Tokyo is massive’ is true on the first of these meanings (the city is huge), but false on the second (the word is average-sized). It’s standard philosophical jargon to call first of these meanings, where the word takes on its typical meaning, a ‘use,’ and the second, where it refers to itself, a ‘mention.’ To consider a new example with this terminology, the sentence ‘Harper is of English origin’ is true when ‘Harper’ is read as a mention (the name comes from the Middle English ‘harper,’ for a player of the harp), but false when read as a use (I am from the United States).

To signal that a word or phrase is being mentioned, we employ a series of (what can be called) mention-forming devices. In written language, these are fairly obvious: quotation marks, block quotes, italics, and so on. In spoken language, there is a more varied list of mention-forming devices: one may make bunny ears with one’s hands, prepend the word or phrase with ‘the word,’ or ‘the sentence,’ use tonal emphasis, and so on. There are, further, some (spoken or written) words and phrases which less obviously serve as mention-forming devices. The word ‘said’ in ‘she said we went to the store,’ or the phrase ‘my name is’ in ‘my name is Harper,’ are both instances of these sorts of implicit mention-forming devices. These mention-forming devices are not as obvious as the others, like wrapping the word in quotes, but they nonetheless suffice to mention their complements.

On my analysis, the verb ‘to identify as’ is an instance of such an implicit mention-forming device, and to say that ‘S identifies as a Y’ is not necessarily to talk about Ys (or employ the concept ‘Y’) at all, but only to mention the word ‘Y’. This makes intuitive sense: one is not talking about dogs when one says that some therians identify as dogs, for example. This is a result of a straightforward analysis of ‘identify as,’ in terms of what one says about oneself, that is:

One identifies as an X iff one says ‘I am an X.’’

The suspicion that this analysis of ‘identify as’ is too thin, and fails to include how some may nonverbally identify as some X by performing a certain social role or having certain deeply held convictions (etc.), to me, seem to conflate ‘identify as’ with the (unfortunately similar-sounding) ‘identity.’ One’s social identity (say, as middle-class, as black, or whatever else.) is a matter of these complex psychological and social characteristics which outstrip mere self-ascription, but this is not what it means to identify as these things. These thicker analyses of ‘identify as,’ ones in terms of belief or complex social/psychological profiles, also exclude the phenomenon of deceptive self-identification. It seems well possible that one could insincerely self-identify as a woman in pursuit of social rewards, (say, to get discounted drinks during girls night at a bar) but this possibility is excluded by these more robust analyses.

When the meaning of ‘to identify as’ is clarified, its object (originally outside of quotes in the expression ‘…identifies as an X’) appears inside of quotes. But, since the expression ‘…identifies as an X’ is the same in meaning as ‘…says ‘I am an X,’ wherein ‘X’ is mentioned, ‘X’ is mentioned in the original expression as well. In light of this, we can conclude, minimally, that the meaning of ‘woman’ in self-id’s proffered definition is distinct from the sense of ‘woman’ it is defining. ‘Woman,’ as defined, refers to the people out there in the world who identify as women, while ‘woman,’ within the definition, refers to the word woman, because it is the object of ‘to identify as,’ an implicit mention forming device.

The worst is out of the way, it seems, for self-id. The force of the traditional circularity objection rests on the two instances of ‘woman,’ within the definiendum and definiens, sharing a meaning. Being able to separate the meanings of these two words, once again, advances the finish line for the circularity objection. At the very least, the opponent of self-id must offer an argument that understanding ‘woman’ as mentioned requires us to understand ‘woman’ as used. But this battle seems uphill, as this would make an exception of ‘woman’: in general, understanding words as mentioned does not require understanding their meaning as used. Understanding the French ‘bois’ as mentioned, for example, does not require understanding what it means as used. The non-francophone can say ‘I do not know what ‘bois’ means,’ or ‘‘bois’ starts with a b,’ even though they do not appreciate the meaning of ‘bois’ as used in French. They can produce and understand mentions of ‘bois’ despite not appreciating the meaning of ‘bois’ or speaking French. It’s easy to see how this case generalizes.

Moreover, there is positive reason to suppose that mentions of a word can be understood independently of its uses. Consider what it would take to get one to understand mentions of ‘bachelor’. If someone were ignorant of the word, I could exhaust the word’s orthographic or phonological properties to them: I could say, ‘it is the word spelled, ‘b’, ‘a’, ‘c’, ‘h’, ‘e’, ‘l’, ‘o’, and ‘r’, or pronounced /ˈbætʃ.ə.lɚ/.’ But this explanation, which is sufficient to single out and convey the meaning of ‘bachelor’ as mentioned, does not at all include (or require) explaining what ‘bachelor’ means as used. So goes for other mentions: once one grasps the spoken or written constituents of the word being mentioned, then mentions of any word making use thereof become comprehensible to them. One could even mention meaningless strings, like ‘asjdsafdg,’ and elucidate it as ‘a keysmash spelled, ‘a’, ‘s’ …’

Combining these two truths, that understanding a word as mentioned does not require understanding the word as used, and that ‘woman’ is mentioned in self-id’s definition for ‘woman,’ entails that understanding ‘woman’ in the definition proffered by self-id (as a mention) does not require understanding ‘woman’ as defined by self-id (as a use).

is this a mise en abyme?

The theses of the prior two sections combine to yield the failure of the circularity objection. For a definition to be circular, it must be the case that understanding the definiens requires knowing the meaning of the definiendum, and knowing the meaning of a phrase as mentioned does not require knowing the meaning of the phrase as used. But ‘woman’ is mentioned (because it is the object of ‘identify as’) within self-id’s definition for ‘woman.’ So knowing what ‘woman’ means within self-id’s definition for ‘woman’ does not require antecedently knowing what the word ‘woman’ means.

No significant moves, here, I take to be incredibly controversial, though I’ve never seen each move combined to ward off the circularity objection in this way. This, further, is why I take the circularity objection to be tricky: it relies on glosses and oversimplification of concepts whose character we all know (even if not immediately) to be just somewhat more complex.

There, of course, remain a myriad of other issues for self-id. This is all future grist for the mill, confidence and motivation permitting. All I aimed to do here was to assuage self-id’s most prominent critique, tabling the more complex offshoots (which would further pad this post’s length) for some other time.