English includes several personal pronouns to refer to humans, including she, he, and they. While they is often plural, it also has been used in the singular for centuries, for example “Every student does their homework” (Balhorn 2004; Baron 2020). Recently singular they has also emerged as a socially significant form for referring to individuals who take they as their personal pronoun, often people who identify as gender nonbinary. For example, an article on Demi Lovato reports “‘I’ve had the revelation that I identify as non-binary,’ they said ….” (Olson & Fuentes 2022). People vary in the degree to which they accept singular they in different linguistic contexts (Bradley et al. 2019; Camilliere et al. 2021; Conrod 2019; 2022; Conrod, Schultz & Ahn 2022; Konnelly & Cowper 2020). This raises questions about how singular they is cognitively represented and used.

In this squib I propose that it is theoretically and empirically useful to distinguish between two broad classes of singular they: “Underspecified singular they”, which is unrelated to the referent’s personal pronouns, and “Personal singular they”, used when the referent’s personal pronouns are they/them. This proposal makes three innovative contributions while building on current evidence. First, it fills a gap in current theoretical accounts by distinguishing between uses with different discourse and processing conditions. Second, it highlights the importance of the discourse representation in determining the appropriateness of both kinds of singular they. Third, it highlights the distinction between Personal and Underspecified uses of definite specific they.

Part of this proposal is that it is worth grouping together multiple uses of singular they under the umbrella term “underspecified singular they”. This is the “old” form, often defined as the use of singular they for a person whose gender is either not known or not important. The most widely accepted uses of singular they are with quantified antecedents (e.g., Everyone should make their bed), or epicene referents (those with unknown gender, e.g., A teacher should know their students; The clerk had their back to me). But some speakers also accept they with known referents, e.g. My neighbor said they would stop by (Bjorkman 2017; Konnelly & Cowper 2020). While these uses differ from each other in some ways, I argue that they share processing features that distinguish them from Personal they.

What are the conditions supporting underspecified singular they? Some scholars suggest it is defined by the absence of a need to mark gender on the pronoun (e.g., Haegeman 1981, as cited by Bjorkman 2017). This explains its use when the gender is not known because it is distributed (everyone, anyone), indefinite and not specified (a teacher), or known but not important to the story (my neighbor). In some cases they may be used to hide person’s identity for privacy.

Consistent with this idea, empirical evidence suggests that they is less common when the gender is salient or assumed. Camilliere et al. (2021) collected acceptance ratings for singular they in a variety of contexts, for example My dentist… they or My friend…they. The sentences were out of context and referred to unknown people, so the only information about gender came from assumptions based on the linguistic labels. They found that they was rated as less acceptable for antecedents with an assumed gender (My sister; Sophia) compared with antecedents that were not gender-biased (My friend, Taylor). A similar picture emerges from comprehension studies; people rate singular they as more natural if the referent does not have a strong stereotypical gender than if it does (Doherty & Conklin 2017), and read it just as quickly as he or she for gender-neutral antecedents (Foertsch & Gernsbacher 1997).

This evidence might suggest that singular they simply represents an ungendered pronoun. If so, we could follow a nice, neat, rule for choosing pronouns: she for feminine, he for masculine, and they for ungendered/neutral. But other evidence shows that things are not this simple.

First, numerous examples show that they is often used for gendered antecedents (Bjorkman, 2017; Newman, 1998); see also Table 1. Second, there appear to be additional pragmatic constraints that determine whether they is used instead of he or she. In particular, the pragmatic conditions for using they are constrained to contexts where the referent is not representationally detailed. Newman (1992) suggests that singular they is biased toward “nonsolid” antecedents, where solidity is defined as the “existence or not of a specific concrete referent.” Newman (1998)’s “individuation” measure contrasts “referents treated as individuals with their identity relevant” and generic referents. A related concept is Camilliere et al.’s (2021) manipulation of what they call “social distance”, also called “distal they” by Conrod (2019). Camilliere et al. found that they is accepted more for referents that are socially “far” from the speaker, e.g. for the dentist more than my friend, and the actress more than my sister. Social distance/distal they are described in terms of the social relation in the real world, suggesting that it is important to be more detailed about the people we know personally.

Building on these observations, I suggest that the key criterion for selecting Underspecified singular they is discourse specificity. Singular they is preferred when the speaker intends to evoke an underspecified mental representation for the addressee in the discourse. By contrast, a specific personal pronoun (e.g., she) is preferred when the referent representation is highly developed. This property is defined in terms of the treatment of the referent within the discourse. If I’m talking about my neighbor in passing, their identity is not important, which is signaled by singular they. But if I want to build a detailed picture of my neighbor’s personal qualities, I might say she. Audience knowledge also likely matters; anecdotally people use they to refer to a friend who uses he or she pronouns more often when the addressee doesn’t know the friend.

This criterion differs from some previous suggestions because it focuses on variation in the mental representation of the referent as portrayed in the discourse (extending Newman’s (1998) concept of individuation) and not the real-world relationship between the speaker and referent (cf. Camilliere et al. 2021; Conrod 2019). At the same time, real-world relationships are expected to correlate, such that speakers likely provide more elaborated descriptions of close friends than acquaintances.

This proposal also breaks with previous accounts (Konnelly & Conrod & Bradley 2023) by lumping definite specific referents (e.g., my neighbor…they) with generic and quantified referents in the underspecified category. Even though definite specific cases are less accepted than epicene uses of they, they also highlight the role of discourse specificity. On the proposed account, it’s not the lexical item that determines acceptability of they but rather the discourse context and how elaborated the referent is. This predicts that acceptability is modulated by specificity within a particular discourse context, on top of individual variation in overall willingness to accept the form.

The idea of discourse specification also breaks with the common assumption that singular they is driven by the lack of a salient gender.1 While gender matters too, two examples suggest that discourse specification may be stronger than gender. First, Camilliere et al. (2021)’s social distance effect reveals that referential properties affect acceptability over and above definiteness and gender. Similarly, Moulton et al. (2020) found that people accepted the use of singular they equally well whether the hearer knew the referent’s gender or not. Second, the examples in Table 1 show that they can be used for gender-specific antecedents. Hypothetically, the role of discourse specification may have developed because it co-occurs with gender uncertainty. Epicene and quantified contexts are almost always contexts where the details of the person’s identity are unspecified. Over time, the discourse conditions that trigger singular they may have shifted from “lack of gender” to “lack of representational detail”.

Unlike social distance or solidity, the concept of discourse underspecification predicts that they co-varies with other pronouns dynamically within and across discourses. Thus, I may know my neighbor well, but use they if she plays only a passing role in a story I am telling. Indeed, I once observed my friend tell a story about a student; at the beginning of the story the student was they, and then as the story progressed and more details about the student emerged, the speaker shifted to she.

The discourse specificity proposal highlights two open questions for future research. First, singular they is used in contexts that are relatively underspecified, but individuals vary in their acceptance of singular they (Camilliere et al., 2021; Conrod 2019; Conrod et al., 2022). If underspecification is the driving criterion, it focuses predictions on identifying how the threshold for they varies across situations and individuals. Second, it raises questions about whether gender is an independent constraint or not. It may be hard to distinguish between gender representation and discourse specificity as the key determinant because they are likely to be correlated. If I don’t know the gender (anyone, a child, the person who called and didn’t leave a message), I also probably don’t know enough to construct a detailed mental representation. An open question is whether gender specificity is as an independent constraint from discourse specificity. The importance of gender underspecification as a selectional constraint likely varies across individuals and may be diminishing over time.

A newer usage of singular they has emerged in the last decade or so as some people have adopted they as their personal pronoun, frequently people with nontraditional gender identities (see also Konnelly et al., 2023). This movement highlights the view that pronoun use is a matter of personal identity and has spurred the social practice of introducing pronouns in some communities. This has led to increased mainstream awareness that politeness requires learning and using someone’s personal pronoun, rather than assuming it. We term this usage “Personal they”, defined as the usage of they specifically because the referent’s personal pronouns are they/them.2

Personal they is critically different from both underspecified they and binary pronouns (she and he) in that the pragmatic conditions for using it are a linguistic fact — knowing that someone’s personal pronouns are they/them. This is a separate fact from gender identity, because some people identify as gender-queer3 but use other pronouns (e.g., neopronouns, hybrid pronouns like she/they, or binary pronouns), and some people use they as a personal pronoun without identifying as nonbinary (Sanders, 2019). Nevertheless, this usage still implicitly acknowledges the relevance of gender fluidity due to its association with nonbinary gender identity.

Thus, the pragmatic condition for using personal they is knowing that the referent’s personal pronouns are they/them. That means you have to know enough about the person to know what their pronouns are, so unlike underspecified they, it requires a mental representation that is specific enough to include knowledge of their pronouns. Using personal they also requires remembering a person’s pronouns (Gardner, 2023). Communication is facilitated when the referent’s pronouns are in common ground, so it is not uncommon to see written articles explicitly introduce pronouns,4 although some writers use personal they in passing.5 While the criterion is narrowly focused on knowing the personal pronouns, this detail tends to co-occur with other information about the referent. Indeed, using they/them pronouns probabilistically co-occurs with holding a nonbinary gender identity, to the extent that mis-using “he” for someone with they/them pronouns is termed an instance of misgendering (Conrod 2020). In sum, Personal they probabilistically co-occurs with referents that have highly specific discourse representations.

There is also significant individual variation in acceptance and ease of using personal they. It is a new form and is only relevant within communities where there are individuals who use it as a personal pronoun and share this information publicly. Arnold et al. (under review; see also Arnold et al. 2026) found that people who conceptualize gender as fluid and not linked to biological sex tended to be more accepting and more accurate at using personal they than those who view gender as fixed and binary. Both fluidity of gender concepts and linguistic facility were also correlated with increased exposure to personal they. This suggests a usage-based mechanism for acquiring personal singular they: people get better at it when it’s conceptually and communicatively relevant. Some people reject the use of personal they for ideological reasons (Hillard, 2019); they likely also avoid participation in communities where they might gain exposure to the linguistic form.

Despite all their differences, personal and underspecified singular they are also similar in that they share a phonological form (they) and the ungendered morphosyntactic feature. Two recent grammatical accounts (Bjorkman 2017; Konnelly & Cowper 2020) aim to explain how people vary in their acceptability of singular they in terms of grammatical change. Here we focus on Konnelly & Cowper’s account, which builds on Bjorkman (2017). They suggest there are three categories of grammar that explain the set of antecedents that can grammatically co-occur with singular they. Pronouns carry morphosyntactic features; for example the pronoun they has the feature [animate] but lacks the features [masculine] and [feminine], so it is used as a default in cases where the referent is animate but not masculine or feminine.

On their account, in the Stage 1 grammar these features are fully contrastive and obligatory. Critically, at this stage words automatically acquire a contrastive gender feature if you know the referent’s gender. So a child has the feature [female] when it refers to a known girl but not to an unknown person. This means that singular they can only be used for quantified referents or when the referent has a gender that is unknown or epicene. At Stage 2, gender features are still contrastive, but the gender feature is marked variably across lexical items. That means that some nouns like a friend can be unmarked for gender and occur as the antecedent of they. At the “super-innovative” Stage 3, they can refer to any referent, including gendered nouns like my sister. They suggest that this occurs by a shift in the system such that gender features on the nouns themselves are entirely underspecified.

Thus, under their view there is no grammatical distinction between using they in the personal and underspecified senses, because both are captured by the absence of gender marking on the noun. The only difference occurs at stage 2, where some (but not all) nouns may be ungendered. That means that using they for a nonbinary friend requires the speaker to add an ad-hoc new lexical entry for their name, e.g. “Mary-ungendered”, as opposed to “Mary-gendered” for other Marys. At stage 3 this distinction goes away because all gender marking is underspecified.

In sum, this grammatical account focuses on the linguistic properties of antecedents to explain when singular they is accepted and how people differ in their patterns of acceptance. The authors acknowledge that using they for a nonbinary individual is a specific case (see also Konnelly et al. 2023), but the grammatical features of underspecified and personal they are not different.

Distinguishing underspecified and personal they is theoretically beneficial because 1) it can explain a range of existing data, and 2) it raises useful questions for future research.