2.1 “I always knew” narrative

The most familiar trans experience is one in which the individual has always known or known from a very young age that they were of a different gender than the one assigned to them at birth. This is summarised succinctly by Catherine’s and Taylor’s experiences:

2.2 Later in life narrative

Although the “I always knew” narrative is common, it is not the universal gender experience. One alternative is the Later in life narrative where the realization of one’s true gender identity occurs when one is a teenager or older. In such cases it is likely that at t₁ an individual can truthfully utter “I am a man” and the proposition be true, whilst at t₂ the same proposition is false. Here an individual is likely to take back their previously true utterance. Later in life narratives are supported by extracts from the following interviews:

Whilst we think that most semantic accounts have focused on the “I always knew” narrative in their motivations and considerations, we want to illuminate the Later in life narrative. As we show below, we take these types of narratives to be suitable for cases of retraction.

2.3 Retraction

To clarify what retraction is we begin with MacFarlane’s (Reference MacFarlane2014) retraction rule. In doing so we utilize examples from the literature on the semantics of predicates of personal taste (e.g., fun and tasty) and epistemic modals (e.g., must and might) before moving on to retraction and gender terms.Footnote ² MacFarlane takes the following rule to be a norm for retraction:

Consider two contexts: in context c₁ X is a small child and does not find olives tasty hence she utters (1); in context c₂ X has grown up and now find olives tasty hence she would not utter (1).

1. (1) Olives are not tasty.

Since X’s tastes have changed at c₂, (1) is no longer true. Thus, according to the retraction rule, X must take back her utterance of (1) as it is not true as used at c₁ and assessed from c₂. X should utter I take that back; what I said was false, olives are tasty or I retract that; olives are tasty. Importantly, X’s utterance at c₁ was sincere: X, at the time of c₁, did not find olives tasty. By retracting her assertion X is not claiming that she’s blameworthy for her earlier assertion (e.g., that she was lying at c₁). As MacFarlane writes: “Retraction is not admitting fault” (Reference MacFarlane2014, 110).

Before we move on to discuss how this applies to gender, a word of caution is needed. We, like many others, do not find the claim that one is required or obligated to retract an utterance that is no longer true very convincing. This seems to ask too much of an individual. If we were to demand X take back her earlier assertion of (1) there is no reason why she cannot protest and say I’m not taking it back, when I said it it was true. Doubts regarding obligatory retraction have been raised by von Fintel and Gilles (Reference von Fintel and Gilles2008). They focus on epistemic modals (in this case might) as illustrated below:

1. (2)

   1. a. Alex: The keys might be in the drawer.

   2. b. Billy: (Looks in the drawer, agitated.) They’re not. Why did you say that?

   3. c. Alex: Look, I didn’t say they were in the drawer. I said they might be there—and they might have been. Sheesh. (von Fintel and Gilles Reference von Fintel and Gilles2008, 81)

Were we to follow MacFarlane we would have to insist that Alex should retract her earlier claim once she finds out that the keys are not in the drawer. As such, (2c) should read as infelicitous at worst and unreasonable at best. It does not. Of course, more can be said about these examples and the nature of retraction itself. But, given that we are seeking a retraction rule that fits with the range of data spanning both the “I always knew” and the Later in life narratives, we clearly need a weakened rule compared to MacFarlane’s. One can retract an utterance that is no longer true (but it is not the case that one must retract).Footnote ³ Consequently, we amend the retraction rule:

It should be noted at the outset that this modification of the rule will require some moves to be made within the details of the semantic theory we endorse. MacFarlane’s retraction rule is a direct consequence of MacFarlane’s semantic theory, and our preferred semantic theory will have to be constructed so as to support our rule instead. We will return to this point when introducing relativist semantics in section 5. We take the possibility of retraction to be essential to a theory of gender terms since we want an account to allow for a wide range of gender experiences. Retraction provides the option to retract one’s earlier gender claim if one wishes. Perhaps after her transition, X might want to reassess her whole life and be able to take back her previous utterances. This sentiment is reflected by Bettcher who, in considering a self-identificatory account, writes:Footnote ⁴

To clarify what retraction looks like with gender terms consider two contexts: in context c₁ X is younger and identifies as a man; in c₂ X is older and has come to realize that her true gender identity is that of a woman. In such a scenario there is no reason why X cannot retract assertions made at c₁.

1. (3)

   1. a. X at c₁: I am a man.

   2. b. X at c₂: I take that back; what I said was false, I am not a man/I retract that; I am not a man.

To be clear the type of phenomenon we’re interested in is when one sincerely identifies as a different gender in an earlier context. Thus, cases described by the “I always knew” narrative seem less likely (although not impossible) to be of the kind to give rise to retraction. This is because if X has always known that she is a woman, then a plausible explanation of utterances of (3a) at an earlier context would be that those utterances were never true. Perhaps X felt pressured into uttering (3a) but she never really believed it.Footnote ⁵

The paradigm cases that we would consider of retraction would be those occurring in the Later in life narratives. If X only realizes later in life that she’s a woman (similar to Tracy or Robyn in the section above), it seems that her earlier utterances of (3a) would be sincere. She would have truthfully made those claims. Later in life however, realizing her true identity she is entitled to take back those earlier utterances (if she wishes to do so). If someone sincerely identifies as one gender at c₁ yet identifies as a different gender at c₂ we want to be able to account for the need to retract that statement without making the claim (as used and evaluated) at c₁ false. What we want is a semantic theory that allows for (3a) to be true at c₁ and (3b) to be true at c₂. Since retraction is not connected with being at fault, endorsing retraction will respect this intuition. Again, it’s worth reiterating that one need not retract. Retraction will be the main challenge we present against other theories as well as motivation for our account.

4.1 Indexical contextualism and retraction

In this section, we demonstrate how indexical contextualism struggles to deal with retraction claims. We’re going to stay neutral over exactly which standard is applied as we’re neutral to which version of contextualism we’re talking about. Recall our amended retraction rule from section 2.3:

Consider a context cu₁ where X sincerely utters:

1. (6) X at cu₁: I am a man.

   1. a. X is a man_{-according-to-standard-of-cu1.}

At cu₂, X no longer sincerely identifies as a man, instead she sincerely identifies as a woman:

1. (7) X at cu₂: I take that back; what I said was false, I am not a man.

   1. a. X takes it back; what X said was false, X is a man_{-according-to-standard-of-cu2.}

On the indexical contextualist’s framework, the proposition expressed by (6) is (6a). Thus (6a) will only be true if X meets the standard of cu₁. The content of each utterance is anchored to the context (and standard of that context) in which it is uttered. Since the standard for man changes in cu₂, man in (7) will mean something different from man in (6).

When X is retracting her earlier utterance, she is not actually retracting what she said, rather she’s saying that if we take the new standard into consideration, she no longer meets the standard of man. Thus, by uttering (7) X has not expressed that she’s not a man from the standard of cu₁, she’s only expressed that she no longer meets that standard. In other words, X cannot say that what she said (at cu₁) is false, for at cu₁ (6a) will always be true as X did meet the standard of man at cu₁. The only way she could utter I was wrong, I take (6) back is if at the time of the utterance the proposition was evaluated as false, i.e., the contextual standard was not met.Footnote ⁷ This is not the scenario that we’re considering. So, prima facie, indexical contextualism cannot deal with retraction claims.Footnote ⁸

As a reviewer for this journal notes, there are options available for the indexical contextualist when it comes to retraction. They suggest two. The first strategy would be to deny that when X says what I said in cu ₁ is false X is referring to the proposition expressed in cu₁, but rather X is saying that the proposition that sentence would express in the new context is now false. Another possibility would involve metalinguistic negotiation. Metalinguistic negotiation occurs when a speaker or multiple speakers negotiate what a particular term means or should mean in a context. Hence, when X utters what I said in cu ₁ is false she is genuinely referring to the past proposition, but what she is metalinguistically implicating is that the use of man in that context should have been used to express a different proposition—one which would consider the standard of cu₂ and hence that proposition would be false. In other words, the speaker is negotiating (with her past self) over the correct meaning of man. Regarding the first avenue, we recognize that this is a possibility but it is not a case of genuine retraction—the most one can say is that I am no longer a man, the previous standard no longer applies, the proposition expressed by that sentence would now be false, but all of this has to do with a current context and not the one in which the utterance was first made. One has not taken anything back. We’re also sceptical about appealing to metalinguistic negotiation. There are no speech markers to indicate that what the speaker is engaged in is metalinguistic negotiation. Unless the indexical contextualist is prepared to say that every instance of retraction involves metalinguistic negotiation (a claim for which an argument would need to be made), it’s more likely that retraction markers like I take that back directly target the previous speech act and they target it for truth and not the meaning of the word. This is not to say that metalinguistic negotiation never occurs. For example, matters would be different if X was askedwhat counts as a “man” in your country?Footnote ⁹ Here there is space for X to negotiate the meaning of “man,” perhaps even consider her own previous utterances. Again, this does not seem to be a case for retraction.

Of course, an indexical contextualist might find our replies unsatisfactory or they might simply want to explain away retraction.Footnote ¹⁰ This might be sufficient for some; however, we find the indexical contextualist’s inability to give a straightforward account of retraction motivation enough to look to alternative avenues.

5.1 Assessment-sensitive relativism and retraction

One often cited benefit of assessment-sensitive relativism over contextualism is its ability to account for retraction.Footnote ¹² This is because one can assess a proposition uttered at cu₁ as false from ca_2, even though it would be true if assessed from the context in which it was uttered. For example, if X utters I am a man at cu₁, then this proposition (namely that X is a man) will only receive a truth value when evaluated from a particular context of assessment and importantly this context of assessment need not coincide with the context of use. The problem with indexical contextualism is that the content of a proposition is anchored to a particular context of use. For an assessment-sensitive relativist, this issue is non-existent since the content of a gender term is stable across different contexts (i.e., man and woman have the same meaning in all contexts) and the truth values are sensitive only to contexts of assessment. As such one can easily take back an utterance from the current context of assessment, as that proposition can be judged as false at the judge-time pair of the context of assessment (at ca₂) if the judge no longer deems it true, i.e. the judge has changed their mind. In section 6, we develop a version of assessment-sensitive relativism in such a way that restrictions on the relevant parameter are introduced. However, a word of caution about relativism and retraction regarding gender has been given by Zeman:

We take Zeman’s concern that retraction is not a universal experience of trans folk seriously and we would like to make two comments. First, as discussed in section 2, we recognize that there is no one universal experience of being trans, thus we wholeheartedly agree with Zeman’s observations.Footnote ¹³ Secondly, the existence of alternative narratives would only present a problem for relativism if we demanded retraction in every case. We considered why this is an unreasonable demand not just when considering gender terms, but any terms—hence the Amended Retraction Rule.

A concern one might have with our proposal is the ease with which we have substituted MacFarlane’s retraction rule with our amended rule. As an anonymous referee rightly pointed out to us, this is no minor modification to MacFarlane’s position as it follows from his semantic theory for predicates of personal taste that retraction is compulsory if the utterance is no longer true.Footnote ¹⁴ The reason why comes down to the choice of parameter that we relativize intuitively subjective judgements to. MacFarlane notes that the required truth-conditional clauses can be equipped to provide the coordinates required for assigning denotations to predicates of personal taste in at least two ways. One can posit a judge parameter, such that a taste expression like “fun” has a denotation, and “rollercoasters are fun” has a truth value, relative to a judge. Or one can simply posit a standard of taste that “fun” and the like are relativized to. His choice is to opt for the latter, precisely because he thinks that changes of taste over time by the judge would lead to cases where retraction of past assertions regarding taste would be unrequired even when one’s tastes had changed (Reference MacFarlane2014, 163–64). MacFarlane’s retraction rule has it that a sentence S uttered in c₁ should be retracted at c₂ just in case S (as used at c₁) is false with respect to parameters relevant to c₂. If S is false with respect to the standard of taste initialized in c₂ then the speaker should retract their earlier utterance of S. From a semantic point of view, the question is simply whether S is true relative to c₂. So, if the agent finds rollercoasters fun at c₁, but has had a drastic change of heart at c₂ (one year later, let’s say), the semantics predicts that the agent should say things like:

(R1) I was wrong about rollercoasters, they are not fun after all.

And not things like:

(R2) I wasn’t wrong/I spoke truly, rollercoasters were fun.

(where this is meant purely as an observation about a change in taste, not of any objective features of rollercoasters such as their design being less extravagant than a year ago). For others (MacFarlane cites Lasersohn Reference Lasersohn2005 and Stephenson Reference Stephenson2007), results are not the same. On Lasersohn’s approach, for example, the semantic theory does not contain a standard of taste coordinate, but a judge parameter. Judges can change their tastes over time and Lasersohn is accordingly less demanding than MacFarlane in his approach, allowing utterances like (R2) to be true. It is important to note that this gives the temporal parameter a different role in Lasersohn’s semantics to the role it has in MacFarlane’s because we can now have cases where two contexts of assessment ca₁ and ca₂ differ only with respect to the temporal element of a judge-time pair, yielding different judgments of taste. MacFarlane’s adoption of a standard of taste rather than a judge does not predict that in cases where the standard of taste remains constant, while the time parameter shifts, a difference in judgment regarding taste could occur.Footnote ¹⁵ Accordingly, we follow Lasersohn’s approach as a better model for our purposes and take gender identity statements to express propositions which are true or false relative to judge-time pairs, rather than standard of taste-time pairs.

Of course, if Lasersohn’s semantic theory predicted that retraction was forbidden (i.e., that the fact that rollercoasters used to be fun relative to the agent’s tastes at the time they asserted it blocked them from subsequently taking it back), this would be a disastrous result not least because it would mean that the theory would lose its power to accommodate various retraction data that are often taken to support relativism.Footnote ¹⁶ But this is not the case. As Lasersohn (Reference Lasersohn2017, 139–41) makes clear, what agents are faced with is a choice between which context of assessment to assess an earlier assertion from. Just as one can (and often should) assess a statement of taste from a different perspective to one’s own, so one can assess a statement of taste from a different time to the present. For example, when Amy (a vegan) asks George whether his beefburger is tasty, she is seeking assessment from George’s perspective, not her own. Just as such exocentric judgments are relevant in many uses of predicates of personal taste, so also can earlier times be relevant to assessments of taste, making true utterances of (R2) felicitous. By uttering (R2), one is “sticking to their guns” and refusing to retract a previously true utterance, since the non-retraction marker “I wasn’t wrong/I spoke truly,” makes it clear that the relevant time of the context of assessment is one where an earlier judge-time pair is under consideration. But, equally, one can adopt a perspective anchored to the current time, making retraction like in (R1) appropriate, i.e., it is the current judge-time pair that is relevant. Both contexts of assessment are available within the semantic theory, but it is a pragmatic matter as to what guides the selection (Lasersohn Reference Lasersohn2017, 139).

In terms of the retraction rule we have been working with, what this means is that the full statement of the rule can be expanded as follows:

In short, an agent can choose between assessment from the time of their current situation, or from their earlier perspective. The latter will yield cases where utterances like (R2) are perfectly felicitous and true.

The scope of this paper does not extend to an evaluation of the competing arguments for MacFarlane’s position compared to Lasersohn’s with respect to the semantics of predicates of personal taste, and their ensuing predictions regarding retraction.Footnote ¹⁷ But for our application of relativism to gender terms we will clearly be better off adopting Lasersohn’s approach. On this view, retraction of an earlier assertion of gender self-identity is permitted, but not obligatory.Footnote ¹⁸

6.1 Gender relativism and retraction

To complete our proposal, we explicate how gender relativism deals with retraction cases. We begin by reminding the reader of the Later in life testimonial data from section 2 and then move on to explain how those experiences can be accounted for within our semantics.

What we want from a semantic theory for gender terms is to allow for the possibility that Kortney, Gerry, Tracy, or Robyn, in light of what they say here, might want to retract some earlier utterances regarding their gender identities at the times they are describing even if those utterances were true at the time of their use. For example, if Robyn had asserted “I am a man” during the time she is here recalling, she may choose to retract that claim now in light of her current gender identity. Following the amended (extended) retraction rule we can see that if any of the individuals who subscribe to the Later in life narrative want to retract a previous utterance they can. We represent such a situation with X as an example below:

1. (9)

   1. a. X: I am a man (true as evaluated from ca₁).

   2. b. X: I take that back; what I said was false, I am not a man (true as evaluated from ca₂)/I retract that; I am not a man (true as evaluated from ca₂).

Under gender relativism, as we present here, we can do this quite easily. Propositions are not anchored to any context since the contents of gender terms are constant in each context of use. Consequently, one can evaluate utterance (9a) from a judge-time pair separate from context of use. Thus, since X is still the subject of the sentence from the new context of assessment (ca₂), it is still her who has the ability to evaluate the proposition. Since she no longer identifies as a man, her utterance of (9a) is false as evaluated from the new context of assessment (ca₂).

We can also see how non-retraction is possible. Consider someone asking if X was wrong to utter “I am a man”. On our view X can respond:

1. (10) X: I wasn’t wrong. At that time, I was happy and contented to be a man (true as evaluated from ca₁).

Here X makes it clear that she is taking the time of her original utterance to be the relevant one for evaluation (as marked by the non-retraction marker “I wasn’t wrong”) and at that time X did identify as a man. This satisfies the second clause of the expanded retraction rule, i.e., “an agent can choose not to retract that assertion of p by assessing p relative to c₁.” Hence, she can explicitly refuse to retract that earlier assertion.

This also provides an explanation for why Robyn could, from the present context of assessment in which she identities as a woman, truly utter the past tense sentence “I was a man.” Given the past tense of “was,” the relevant judge-time pair for the context of assessment is at the time where Robyn was a man. Hence as uttered today the proposition—Robyn was a man—will be evaluated relevant to an earlier time rendering it true.Footnote ²⁶

We take the possibility of retraction to provide a new argument for the context sensitivity of gender terms which is wholly independent of existing arguments. Furthermore, we take retraction to be demanded by the testimony considered in this paper. Because of the ease with which relativism can account for retraction, we take relativism to be the most promising approach to the semantics of gender terms. When amended, as we have done here, gender relativism provides a semantic theory which best accommodates the full range of gender experiences. A crucial feature of gender relativism as we have developed it is the appeal to self-identification. We therefore now turn to a major criticism against self-identificatory views.

6.2 Absence of self-identification

Elizabeth Barnes (Reference Barnes2020, Reference Barnes2022) has presented a serious question for self-identificatory views of gender: what do we say about those individuals without the capacity for self-identification? Barnes argues that, although self-identificatory accounts provide a way to include all of the individuals who identify as their chosen gender, at the same time such accounts manage to exclude those individuals without a capacity for self-identification. Barnes notes that “denying womanhood to cognitively disabled women seems like a gross injustice, especially given that cognitively disabled women are particularly vulnerable to gendered abuse” (Barnes Reference Barnes2020, 7). She attributes this point to Haslanger:

This objection has a particular bite against our account as self-identification of the subject is a necessary component for assessing gendered propositions. We take this point seriously since to have a theory that harms marginalized and vulnerable individuals would go against our aims and feminist goals more broadly. As such, we’ll provide a possible answer that we take to give a way of assigning truth values to gender sentences involving those who cannot self-identify whilst at the same time granting gender sentences non-trivial truth values.

In response to the above challenge, we take a similar line to Zeman (Reference Zeman2020).Footnote ²⁷ Zeman argues that in the case where someone is unable to have a sense of gender identity, we should “bite the bullet” and say that the truth value of a sentence like WOMAN is indeterminate (Reference Zeman2020, 775). The reason is that we do not know the element (namely self-identification) that establishes truth values, that is, we are in an epistemically deficient state to say whether WOMAN is true or false. Zeman goes on to say that such a strategy is “compatible with ascriptions of gender to the cognitively impaired for practical purposes” (Zeman Reference Zeman2020, 775, added emphasis), thus we might assign WOMAN a value if there is a practical reason to do so. The resources available to gender relativism accommodate this proposal, as we now show.Footnote ²⁸

As noted in the previous section, assessment-sensitive relativism allows exocentric judges or perspectives,Footnote ²⁹ i.e. perspectives other than the speaker’s. For example, in trying to convince a child to eat some broccoli X might utter:

1. (11) Broccoli is tasty.

By uttering (11) X need not think that broccoli is tasty, she might hate broccoli. Instead, X is adopting the child’s perspective. When evaluating (11) we do not look for X’s perspective in the context of assessment but rather the child’s and evaluate the truth of the proposition according to the tastes of the child. Before we find out whether the child does or does not find broccoli tasty, we may operate as if they do for practical purposes—we act as if (11) is true because broccoli is healthy. Importantly, (11) will not automatically be true just because we take it to be true, it will be true iff the child finds broccoli tasty. The matter concerning the truth value of (11) will only be settled once we know if the child likes broccoli. We propose that adopting something akin to an exocentric perspective for gender terms could help with Barnes’s concern.

Where an individual cannot self-identify yet there is a need to assign a gender to them (perhaps to address some of the worries Haslanger and Barnes bring up) something like exocentric perspectives could be adopted. For practical purposes only, WOMAN might be classed as true. However, this does not mean that we render WOMAN as actually true since there is no way to know the individual’s gender. As pointed out by a referee, this is where a clear difference between exocentric perspectives in (11) and individuals who cannot self-identify lies. Whilst there is a matter of fact whether the child finds broccoli tasty (hence the truth value could, in principle, be settled), there is no such fact when it comes to cognitively disabled individuals under consideration. In other words, the actual truth value of WOMAN is not settled and could never be settled. So, acting as if WOMAN is true might be useful for practical purposes but given that the truth value is indeterminate, we refrain from actually assigning a gender to someone on the basis of their sex.

A further possible worry concerns dead individuals who no longer have the capacity to self-identify. Should X pass away, would WOMAN be true or false? Given that we knew how X identified prior to her death, we could confidently assign the truth value of true to WOMAN (although “X was a woman” would be more appropriate given that “is” implies existence). However, if we had no clue how X identified, we should refrain from assigning WOMAN a definite truth value as the matter would be beyond our power to settle. This idea is captured nicely by Jaques’s discussion of D.’s piece she heard at the 2008 Transfabulous festival:

In some cases, we will never know the gender identities of the dead. But, of course, this is an observation about our epistemic state, not those individuals’ actual genders.

Whilst adopting something like an exocentric perspective might be useful for practical purposes, this should only be licenced in cases where the subject of the sentence does not have the means to self-identify. We are not proposing a system similar to Díaz-León’s where the best political/moral considerations would decide the meaning (or truth values in our case). For us, self-identification is the default, and it trumps all of the other considerations. We may adopt exocentric perspectives only when there’s no way for the subject to self-identify but the actual truth value of an utterance will always depend on the subject and in such cases, they will be indeterminate.