Gender-inclusive speaking: a quantitative study of noun phrases referring to humans in a corpus of spoken French

1 Introduction

In the past decade, how the gender of human beings is represented in language, and what effects the linguistic encoding of social gender might have on society, has been a major topic of inquiry and discussion for both lay people and specialists (linguists, psychologists, and sociologists). Of course, this is not new: the role of language in producing and reproducing gender inequality has been a major focus for scholars and activists in the anglophone world, in francophone Canada, and in Sweden since the 1970s (see Pauwels 1998 for English; Arbour et al. 2014; Vachon-L’Heureux 1992 for French Canada; and Hornscheidt 2003 for Swedish), and since the 1980s and 1990s in countries such as Germany (Bußmann and Hellinger 2003), Belgium (Arbour et al. 2014), and France (Burr 2003; Houdebine-Gravaud 1998; Yaguello 1979; among others). Nevertheless, the past ten years has seen renewed interest in how gender is encoded in language, and how languages might be changed to be more egalitarian. In some languages, like English, the primary interest is in how to include people whose gender does not fall into the binary categories male/female (for example, non-binary, gender fluid, or genderqueer individuals, see Baron 2020; Conrod 2018 for overviews). In other languages such as French, the recent attention to the inclusion of non-binary people is just a small part of a much larger revival of discussions around male bias in these languages and what should (not) be done about it. As Abbou et al. (2018) and Burnett and Pozniak (2021) describe, since 2017, there has been an explosion of discourse about French gender-inclusive language in traditional and social media, in academic publications, and in political, activist, and even legal circles. These debates concern not only how/whether to represent non-binary people in the language, but also whether/how so-called “generic” masculines should be replaced with non-masculine-marked forms. So-called generic masculines are masculine-marked expressions whose intended referents are not limited to men, e.g., using les étudiants _M to mean simply ‘the students’.^[1]

An important feature of the current intensity of “verbal hygiene” (Cameron 1995) on this topic in France and elsewhere in the francophone world is that it is almost exclusively focused on the written language. Indeed, the phrase écriture inclusive has become the dominant way of referring to a wide variety of feminist linguistic practices, replacing other older terms less tied to the written language such as langage/communication inclusive/epicene/non-genrée ‘inclusive/epicene/non-gendered language/communication’, or féminisation ‘feminization’ (see Abbou 2023; Elmiger 2021b; Toma 2021, for reviews), and guidelines about how to talk in an inclusive or gender-neutral way take a backseat to how to write in one (see Crémier 2023). Likewise, the recent scientific literature on the relationship between grammatical and social gender in French has focused on how francophones use (Burnett and Pozniak 2021; Diaz Colmenares 2021; Diaz Colmenares and Heap 2020; Flesch and de Beaumont 2023; Simon and Vanhal 2022; among others) and interpret (Liénardy et al. 2023; Pozniak et al. 2023; Spinelli et al. 2023; Xiao et al. 2023; among others) gender-inclusive forms in writing only. The main goal of this paper is therefore to center the spoken language in a way that is not generally done when describing gender-inclusive strategies, either in activist or in scientific discussions, and document the ways in which grammatical gender marking is expressed in speech.

Our investigation of spoken language referring to humans has the potential to yield surprising results that have consequences not only for language policy, but also for descriptive and theoretical linguistics. In the past 50 years (since at least Blanche-Benveniste 1986), the availability of large corpora of naturalistic speech has brought to light the ways in which varieties of French used on both sides of the Atlantic differ when they are written and spoken. Morpho-syntactic differences identified include the role of pronominal subjects (Auger 1994; Culbertson 2010; Liang 2023; Roberge 1990), the structure of negation (Barra-Jover 2004; Palasis 2013), the distribution of the subjunctive mood (Kastronic 2016; Poplack 1993), and the pronominal inventory (Blondeau 2001; Laberge and Sankoff 1979), among many others (Berrendonner 2004; Blanche-Benveniste 2007; Blanche-Benveniste et al. 1987; Deulofeu 2001). Indeed, some researchers even propose that francophones live in a situation of diglossia (Massot 2010; Palasis 2013; Rowlett 2011; among others), so great are the grammatical differences between the oral and written forms of the language. Therefore, this paper also investigates the descriptive question (which has theoretical implications) of whether there are important differences between spoken and written grammatical gender marking in French. In order to do so, we present a quantitative study of gender marking in noun phrases referring to humans in the Cartographie linguistique des féminismes (CaFé) corpus (Abbou and Burnett 2025), a corpus of 102 sociolinguistic interviews with feminist and queer activists in three large multicultural francophone cities: Paris, Marseille, and Montreal. We have chosen to study the speech of activists because these are people who are likely to be leading changes in the use of the grammatical gender system, which may then diffuse in francophone communities more generally.

The results of our study paint a picture of the use of the French grammatical gender system in oral language which contrasts with common assumptions made about written French, both in grammatical descriptions and in gender-inclusive language guides. We find that the overall rate of use of masculine-marked expressions whose intended referents are not necessarily men is extremely low: only at most around 5 % of all the noun phrases in the corpus (and 12.5 % of noun phrases with gender-neutral reference). This also contrasts with the proportion of generic masculines found in the rare studies which have quantified the proportion of various types of nouns in corpora of written data: 31 % of all nouns referring to gender-mixed groups in a corpus of Belgian news and political texts (Simon and Vanhal 2022), and 24.97 % of all nouns referring to humans in a corpus of German press texts (Müller-Spitzer et al. 2024). We show furthermore that the use of these so-called generic masculines is conditioned by social factors including age, geographical location, and education in such a way that, for younger, less educated speakers from Montreal, the rate of using a so-called generic masculine is even lower: 3.89 %. We argue that these surprising results can be understood to arise from three different, unrelated, sources which, together, conspire to almost eliminate so-called generic masculines from the spoken language of feminist and queer activists:

1. Differences between the spoken and standard written language such that a lot of gender marking present in writing is neutralized in speaking in Paris, Marseille, and Montreal (e.g., ami _M and amie _F ‘friend’ are both pronounced [ami] in these three locations).

2. The introduction of person-centered language, following proposals by disability activists in anglophone North America, which replaces noun phrases referring to people headed by property nouns, which are often so-called generic masculines like les sourds ‘the deaf’ or les noirs ‘black people’, with a noun phrase headed by personne (i.e., les personnes sourdes; les personnes noires), which is grammatically feminine.

3. The replacement of isolated so-called generic masculines (e.g., les étudiants) with syntactic doublets (les étudiants et étudiantes), following proposals by influential gender-inclusive language guides.

Two of the three (socio)linguistic phenomena described above have no direct relation to feminist linguistic activism; however, as we will show in the paper, they combine with the doublets to render the spontaneous spoken language of feminist and queer activists highly gender inclusive. In fact, we will show that the contribution of the doublets, which currently occupy the most central place in the discourse around gender-inclusive language in French, to reducing the use of so-called generic masculines is far lower than the contribution of both person-centered language and, especially, the neutralizations created by written and spoken French grammar. We argue that our results have implications for gender-inclusive language policy not only for French but also for other languages: guidelines cross-linguistically are almost always formulated with the written language in mind; however, the spoken language can differ greatly from the written form, and is just as important when it comes to the production and reproduction of gender inequality. In order to be useful, language policies and guides should not neglect it. Our paper also makes a contribution to descriptive and theoretical linguistics: our results suggest that, in our corpus (and possibly also in other corpora of spoken French), masculines play a very minor role when it comes to accomplishing generic, gender-neutral, or gender-mixed reference, something that is at odds with many linguists’ descriptions of French grammatical gender, and on which they base morphological, syntactic, and semantic theories.

The structure of this article is as follows: in Section 2, we present a brief overview of the types of noun phrases in written and spoken French, as well as some of the major proposals that have been made to render the written language more gender inclusive. In Section 3, we describe the CaFé corpus, and in Section 4, we present the results of our quantitative study of noun phrases referring to humans in this corpus. Section 5 concludes with a discussion of the implications of our findings for language policy and linguistic descriptions of French.

2 French grammatical gender and écriture inclusive

In languages that typologists describe as having grammatical gender, such as French, there is often some relation between the meaning of at least some of the nouns and whether they give rise to feminine or masculine agreement marking patterns. Broadly speaking, for nouns referring to humans, there are four cases in French. The first case comprises nouns whose basic semantic meaning incorporates gender, and their grammatical gender lines up with the social gender of the referent. Following Hellinger and Bußmann (2003), we call these nouns gender-specific, and this class includes words like femme ‘woman/wife’, homme ‘man’, frère ‘brother’, sœur ‘sister’ and mari ‘husband’. The second case consists of nouns whose basic semantic meaning includes people of all social genders; however, the noun itself has a single fixed grammatical gender. Terminology describing this second class varies greatly according to author and grammatical tradition; however, in this paper, we will call these nouns, such as personne, parent, victime, and gens, epicene nouns. The vast majority of French nouns referring to humans come in masculine-feminine pairs of two types. The first type consists of noun pairs where the masculine and the feminine have a different form, for example, boulanger _M  – boulangère _F ‘baker’. The masculine form appears with masculine-marked dependents (le beau boulanger) and the feminine form appears with feminine-marked dependents (la belle boulangère). The second type consists of cases where there is a single form of the noun (e.g., journaliste ‘journalist’) but this form can appear with either masculine or feminine-marked dependents: le beau journaliste – la belle journaliste. In this paper, we will follow Corbett (1991) in calling nouns like journaliste “common gender” nouns. In almost all cases, the interpretations of nouns with feminine endings and common gender nouns in feminine-marked noun phrases are clear: feminine-marked noun phrases (almost) always refer to women. The interpretation of masculine-marked noun phrases, on the other hand, is more controversial.

In French grammars (see, for example, Grevisse and Goosse 2016), masculine-marked noun phrases referring to humans are described as allowing two types of reference, one referring specifically to men (specific masculine), the other to mixed-gender groups or persons whose gender is irrelevant or unknown (generic masculines). Many descriptive and theoretical linguists also describe French masculine noun phrases in this way (see Atkinson 2012; Dubois and Lagane 1973; Ihsane and Sleeman 2016; Riegel et al. 1994; Schafroth 2003; among many others). In other words, the claim is that, while la boulangère or la journaliste can only be interpreted as referring to a woman, le boulanger, les boulangers, or le journaliste can be used to refer to men, women, or people of other genders.

The controversy revolves around whether the masculine marking on noun phrases such as tous les citoyens ‘all citizens’ or tout citoyen canadien ‘All Canadian citizens’ is biased towards a male interpretation; that is, are readers of these expressions more likely to think that they apply to men than women (or people of other genders)? In fact, many researchers working in a feminist and/or a psycholinguistic perspective have argued that these so-called “generic” uses of masculine noun phrases are in fact male-biased in this way. Some, for example Michard (1996, 1999), Michel (2016) among others, have highlighted naturally occurring written examples in which noun phrases that intuitively ought to have a gender-neutral interpretation are instead interpreted as male oriented. Others (Brauer and Landry 2008; Gygax et al. 2008; Richy and Burnett 2021) have conducted controlled experiments in which francophone participants read sentences containing gender-marked noun phrases, and show, using various response measures, that the masculine ones reliably trigger faster and easier male mental representations. Both of these lines of research have fueled feminist linguistic activism aiming to eliminate the so-called “generic” (yet male-biased) uses of masculine-marked noun phrases in favor of expressions that, depending on the political orientation of the activist, would either make women more visible or remove any inferences related to gender entirely.

In what follows, we present the major strategies present in francophone guidelines. As Elmiger, who has compiled the most comprehensive cross-linguistic database of gender-inclusive guidelines to date, notes, the proposals are as varied as they are numerous (Elmiger 2021a). For this reason, we will present only proposals that appear in the most influential guidelines in France and Quebec. For France, we follow Elmiger, who identified the 2015 (revised 2022) Guide pratique pour une communication publique sans stéréotype de sexe, published by the French government’s Haut conseil d’égalité entre les femmes et les hommes (Bousquet and Abily 2015/2022), and the 2016 (revised 2019) book Manuel de l’écriture inclusive, published by the communications agency Mots-Clés, which popularized the now ubiquitous term écriture inclusive (Haddad 2016/2019). For Quebec, we include the guidelines elaborated by the Université du Québec à Montréal (Lamothe et al. 1992), the guidelines written by the highly influential Office québécois de la langue française (OQLF) (2012, revised 2020), and the book Manuel de grammaire non sexiste et inclusive by Lessard and Zaccour (2018). Note that some of the Quebec references are older: this reflects the fact that feminist linguistic activism is older in Canada than in France, being a product of Second Wave feminist interest in language in the 1970s and 1980s (Vachon-L’Heureux 1992).

1. Composed forms

The idea is to replace a masculine that would be used to refer not only to men with an expression that combines aspects of the masculine and feminine forms of the noun, separated by some punctuation. Étudiant can thus be replaced by étudiant-e, étudiant·e, étudiant.e, or étudiant(e). These forms can be inclusive of all gender identities, including those beyond the binary.

1. Masculine – feminine doublets

The second strategy that is proposed by all the guides is to write both masculine and feminine forms of the nouns, either in a conjunction (les étudiants et étudiantes), disjunction (les étudiants ou étudiantes), or simply repeating them (les étudiants, étudiantes). In French, these are often described as doublets.

1. Neutralization

A third strategy found in all the guides is the use of expressions where grammatical gender is neutralized. Under certain morphophonological conditions, it is impossible to tell whether a noun phrase is masculine or feminine. For example, while the common gender noun journaliste’s membership in the masculine class is visible in le _M gentil _M journaliste ‘the nice journalist’, if we wanted to say ‘the incredible journalist’, we would use the vowel-initial adjective incroyable, and the vowel in the determiner would be elided, making l’incroyable journaliste. Likewise, if we wanted to pluralize: since the French definite plural determiner les is not gender marked, the noun phrase les journalistes is not either.

1. Epicene and collective nouns

Closely related to the neutralization strategy, the final strategy found in all the guides is to use an epicene noun, like personne or gens, or a noun referring to a collective. For example, instead of writing le directeur pense que…, one can write la personne en charge du projet pense que… or la direction pense que… ‘Management thinks that…’. Collective nouns (e.g., la foule ‘the crowd’ or le groupe ‘the group’) are like epicene nouns in that they do show gender marking, but their semantic meaning is gender neutral.

The guidelines only occasionally make reference to the spoken language; however, when they do it is to predominantly discuss what place the composed forms, which make special use of punctuation, might have in speech. Bousquet and Abily (2015/2022) simply considers the composed forms irrelevant for speaking and proposes to use doublets. Other guides give instructions about how to pronounce the composed forms: Lamothe et al. (1992: 16) propose that if one is reading a composed form, one should read it as a doublet in which the masculine is read before the feminine, and Haddad (2016/2019) propose that composed forms should also be read as doublets starting with the feminine form.

3 The Cartographie linguistique des féminismes (CaFé) corpus

The CaFé corpus comprises 102 sociolinguistic interviews (of approximately 90 min each) with people who are engaged in what we described as “feminism, women’s issues and/or activism for queer and sexual rights” in Paris (42 interviews), Montreal (40 interviews), and Marseille (20 interviews), and in which we collected their positions on issues related to gender and sexuality, and their link with language. Two interviews were done with multiple people: one in Montreal (two people) and one in Paris (three people). The corpus was collected between 2021 and 2022, and consists of 168 hours of speech, transcribed with an enriched orthographic transcription (approximately 2 million words). For more details about the content and technical aspects of the corpus, see Abbou and Burnett (2025); however, we highlight here that many of the topics discussed were political and theoretical; therefore, they were conducive to talking about people and things in general, and so conducive to the production of noun phrases with generic or gender-mixed reference.

As is to be expected in a corpus collected from feminist activists, the majority of participants are women (cis or trans); however, it also includes a small minority of cis and trans men and non-binary people. Paris, Montreal, and Marseille are all large multicultural cities, and the composition of feminist spaces reflects this. Many of our participants are bilingual (French-Arabic, French-Kabyle, French-Portuguese, French-English, French-Spanish, among others), and we have chosen not to exclude the participants whose first language is not French, provided French is the language of their feminist engagement. As a result, we have three speakers who learned French after childhood: one whose native language is Spanish, one whose native language is Nigerian English, and one whose native language is Swedish.

The 105 recorded speakers were recruited on the basis of their engagement in feminist movements according to five categories:

1. Academics: gender studies scholars, but also people who promote a feminist reading of science (scientists, doctors) or promote women in STEM (science, technology, engineering, and mathematics).

2. Professionals: diversity practitioners, salaried community activists, lawyers, and instructors.

3. Associative: activists, spokespeople, and others from associations, collectives, political parties, or unions.

4. Media: authors, editors, librarians, translators, and journalists.

5. Collective: volunteers in online activism, grassroots collectives, informal networks, etc.

The speakers in the corpus range from 19 to 83 years old, with a median age of 35 (SD = 22). In the analyses below, we divide the speakers into three age groups: participants under 35; participants over 60; and participants between 35 and 59. The speakers in CaFé are also highly educated, with all but nine having at least an undergraduate degree, and around a quarter of the corpus (27) having a PhD. The breakdown of participants by corpus according to age, education, and feminist engagement is shown in Table 1.

4 Quantitative studies of gender-inclusive speaking

This section presents a quantitative study of grammatical gender marking on noun phrases referring to humans in the CaFé oral corpus. Of course, grammatical gender marking is present on other linguistic expressions in the language, particularly pronouns, adjectives, and participles; however, we have chosen to study noun phrases for practical reasons: first of all, studying gender marking on pronouns requires that, for each occurrence of the masculine and feminine pronouns il and elle, we identify whether its antecedent is human or non-human and, if human, what the social gender is. Given that there are 17,408 occurrences of il(s) and 7,017 occurrences of elle(s) in CaFé, it would be extremely time-consuming to process these data manually, and automatic identification of coreference chains in French is still a work in progress, despite recent advances (see e.g., Wilkens et al. 2020). The same observation holds for predicative adjectives and participles in morpho-syntactic agreement configurations.

Neologisms, like iel ‘they’ and toustes ‘all’, are certainly relevant to our central research question concerning gender marking and inclusive speaking. Non-metalinguistic uses of these expressions are found in our corpus; however, they are very infrequent: only 17 non-metalinguistic occurrences of iel, 13 of which are uttered by speaker Paris 24, the youngest speaker in the corpus (19 years old), and nine non-metalinguistic occurrences of toustes. We therefore limit our study to noun phrases referring to humans. The data and code (R notebook) used for the analyses are available on the OSF platform (https://osf.io/dc3v7/?view_only=2a294f3347764a89a04fcd937b07330a).

4.2 Overall descriptive results

We extracted 29,990 noun phrases referring to humans from the corpus. We found 942 metalinguistic uses and 158 noun phrases in languages other than French, that were excluded from the statistical analyses. Of the 28,890 remaining noun phrases, 10,858 (37.58 %) are feminine only, 6,367 (22.04 %) are epicenes, 4,921 (17.03 %) are gender-neutralized, 5,647 (19.55 %) are masculine, 762 are collectives (2.64 %), and 335 (1.16 %) are doublets. Crucially, it turns out that most of the masculine-marked noun phrases (n = 3,985 or 70.58 % of all masculine noun phrases) are used to refer to a man or to men only. This distribution gives us the first main result of our paper: so-called generic masculine noun phrases account for only 5.75 % (n = 1,661) of all noun phrases referring to humans in the corpus. If we add up all the relevant expressions that can be used to accomplish generic, gender-neutral, or gender-mixed reference (so-called generic masculines, epicene noun phrases, neutralizations, and doublets), the percentage rises only to 12.5 %. For information, Table 2 shows the ten most frequent uses of masculine-marked noun phrases in the corpus to possibly refer to people other than men. Note that our coding scheme is very generous to so-called generic masculines: as soon as non-male reference did not create a contradiction in the context, we coded a masculine as “so-called generic”. It seems probable to us that most of the occurrences of violeur and agresseur are made with men in mind; however, since the context in which they occurred did not clearly specify this, we coded them as “so-called generic”. In this way, the result that only 12.5 % of noun phrases with generic or gender-neutral/mixed reference are masculine should be taken as an upper-bound: in reality, there are probably even fewer truly generic or gender-neutral/mixed masculines in the corpus.

Table 2:

Ten most frequent nouns in generic masculine noun phrases.

Montreal			Paris			Marseille
Noun	Freq.	%	Noun	Freq.	%	Noun	Freq.	%
étudiant ‘student’	42	8.96	étudiant ‘student’	63	7.38	étudiant ‘student’	54	15.98
enfant ‘child’	21	4.48	violeur ‘rapist’	29	3.40	sourd ‘deaf person’	15	4.44
expert ‘expert’	20	4.26	militant ‘activist’	28	3.28	enfant ‘child’	12	3.55
policier ‘police officer’	18	3.84	enfant ‘child’	25	2.93	expert ‘expert’	10	2.96
ennemi ‘enemy’	16	3.41	travailleur ‘worker’	23	2.69	ennemi ‘enemy’	9	2.66
blanc ‘white person’	14	2.99	expert ‘expert’	19	2.22	agresseur ‘assaulter’	8	2.37
enseignant ‘teacher’	14	2.99	agresseur ‘assaulter’	18	2.11	militant ‘activist’	8	2.37
québecois ‘quebecker’	13	2.77	avocat ‘lawyer’	17	1.99	bourgeois ‘bourgeois’	7	2.07
imposteur ‘imposter’	12	2.56	acteur ‘actor’	13	1.52	copain ‘friend’	7	2.07
militant ‘activist’	10	2.13	client ‘client’	13	1.52	blanc ‘white person’	6	1.78

This result is noteworthy because, as discussed in Section 2, the question of what to do with masculines is generally treated as one of the fundamental, hotly debated issues in French feminist linguistic activism. To better understand why the proportion of so-called generic masculines is so small in our spoken corpus, we will now examine how the distribution of gender-marked expressions referring to humans varies according to city.

Figure 1 represents the proportion of noun phrases per grammatical gender category in the three subcorpora: Montreal (10,466 noun phrases), Paris (12,959 noun phrases), and Marseille (5,465 noun phrases). It shows that the proportion of masculine and feminine noun phrases is lower in Montreal than in other cities, but that the proportion of epicene noun phrases is the highest in Montreal (27.77 % of noun phrases, versus 19.03 % in Marseille, and 18.68 % in Paris).

Figure 1:

Distribution of the types of noun phrases referring to humans in CaFé.

As described in Section 3, the main topic of the interviews is feminist and queer activism, so it is not surprising that all the speakers frequently refer to women and therefore have a high rate of feminine-marked noun phrases. What is a bit more unexpected is that both epicene and gender-neutralized noun phrases appear more frequently than both gender-specific masculines and so-called generic ones. In the next sections, we examine both of these categories in detail, starting with the neutralizations.

4.4 The role of person-centered language in epicenes

As discussed in Section 4.1, speakers in the Montreal corpus have both the lowest rates of so-called generic masculines and the highest rates of epicenes. Table 3, which shows the list of epicene nouns in each subcorpus, gives some insight into how speakers in Montreal differ from their French counterparts.

Table 3:

Five most frequent epicene nouns, by city (personne ‘person’, gens ‘people’, parent ‘parent’, personnage ‘character’, victime ‘victim’, grand-parent ‘grandparent’).

Montreal			Paris			Marseille
Noun	Freq.	%	Noun	Freq.	%	Noun	Freq.	%
personne	1,693	58.26	gens	1,213	50.1	personne	485	46.63
gens	1,012	34.82	personne	990	40.89	gens	454	43.65
parent	152	5.23	parent	147	6.07	parent	67	6.44
personnage	12	0.41	victime	35	1.45	personnage	18	1.73
victime	11	0.38	grand-parent	13	0.54	grand-parent	6	0.58

Although there is much overlap between the lists from Paris, Montreal, and Marseille, one difference immediately stands out: speakers in Montreal use the noun personne much more often (58.26 % of epicene nouns) than speakers in Paris and Marseille, who use personne about as often as gens. Rather than being a coincidence, we hypothesize that part of the reason that the Montreal speakers are using fewer so-called generic masculines than the speakers from France is that they have a higher rate of use of a construction that is a variant of a different sociolinguistic variable, unrelated to gender: person nouns versus property nouns. In particular, we find examples in which, to describe a person or a group of people, speakers in Montreal construct a NP composed of the noun personne (and to a lesser extent gens) and a modifier describing a property, while speakers in France use noun phrases which directly describe the property. Some examples of person noun versus property noun pairs in the corpus are shown in (11)–(13).

(11)

a.

homo ça devrait être pour qualifier et les personnes gays et les personnes lesbiennes (Montreal 09)

‘“Homo” should be to qualify gay people and lesbian people’

b.

enfin le truc c’est que les lesbiennes j’ai l’impression que c’est toujours les meilleures alliées des gays et les gays ils nous laissent tomber tout le temps (Paris 17)

‘Actually the thing is I have the impression that lesbians are always the best allies of gay people, and gay people let us down all the time’

(12)

a.

en plus des personnes juives ou identifiées comme telles c’est euh les personnes tsiganes et roms et les personnes musulmanes ou identifiées (Paris 32)

‘More and more Jewish or identified people, it’s uh Romani and Gypsy people and Muslim people’

b.

je compare euh les chrétiens euh les musulmans euh les juifs euh les con- enfin Confucius, Bouddha, euh voilà (Marseille 10)

‘I compare Christians, Muslims and Jews, euh the Con- well Confucious, Buddha, there you go’

(13)

a.

la police n’est pas une réponse pour les surtout les personnes immigrantes et racisées (Montreal 39)

‘The police is not an answer, especially for immigrant and racialized people’

b.

de plus en plus les collectifs de féministes qui sont pas nécessairement racisés ou qui ou où les racisés sont minoritaires (Paris 36)

‘More and more feminist collectives that are not necessarily racialized or where racialized people are in the minority’

The opposition between a description of a person (or a group of people) based around a property that they have and one which describes them first as a person and then states the property is reminiscent of the oppositions that are at play in the Person-centered (or person-first) language movement. In the 1990s, in the wake of the passage of the Americans with Disabilities Act (1990) in the United States, a number of important publications, including one from the American Psychological Association (1992), argued that “using labels [property nouns] to define people, which had long been used and was widely accepted, resulted in increased stigma in the medical, legal, and social realms” (Granello and Gibbs 2016: 33). The proposed solution was to form an NP around person or people, and describe their disability with a post nominal modifier, such as the disabled instead of people with disabilities. In English, there is an additional variant which contains a person noun: disabled people. This variant is known as the identity-centered language option because, according to its advocates, the prenominal adjective communicates that the property described in the adjective is part of the person’s identity in a way that the “person first” variant (people with disabilities) does not (see Andrews 2019 for a review). Proponents of the identity-centered variant draw a parallel between noun phrases describing people with disabilities and noun phrases describing people with other cultural or social properties: in the same way that no one finds French people offensive, the same holds for disabled people, under the disability-as-cultural-experience view. Although the property noun variant (the disabled) is universally disfavored by disability activists, whether the person first (people with disabilities) or identity first (disabled people) variants are favored depends on a variety of factors including age, gender, country of residence, and, above all, which particular disability community is being studied (Sharif et al. 2022; Vivanti 2020; among others). The person- and identity-centered language movements have not been very influential in France; however, they have become mainstream in French Canada. For example, the OQLF has both langage centré sur la personne and langage centré sur l’identité in its Grand dictionnaire terminologique,^[3] and the use of one of these variants (or both) is frequently recommended in guidelines for institutions and organizations, such as the Centrale des syndicats du Québec.

Although person/identity-centered language comes from disability activism, the strategy of avoiding using a property noun to describe people in minorities or oppressed groups has become very general, albeit variable in the CaFé corpus, as shown in (11)–(13).

In order to test the hypothesis that Montreal speakers use more person/identity-centered language, and that this affects the distribution of gender marking in their speech, we conducted a quantitative study of variation in the use of person nouns versus property nouns. In the corpus, we found 785 occurrences of epicene nouns followed by an adjective or by a prepositional phrase, which could be considered direct alternatives to nouns (regardless of their type: feminine, masculine, or neutralized). Les personnes blanches ‘white people’ can be considered a direct alternative to les blancs or les blanches, for example. Likewise, une personne en situation d’itinérance ‘person experiencing homelessness’ can be considered an alternative to un itinerant ‘a homeless person’. The vast majority of these people noun phrases are formed with personne(s) (n = 751); others are formed with gens (n = 32) and individu(s) (n = 2). These people noun NPs are distributed into 103 different types; the 20 most frequent are displayed in Table 4, with their frequency in the corpus and their proportion in all people noun NPs. Many of these people noun NPs are related to gender and sexuality (personne trans, non-binaire, queer, lgbtq, hétéro, etc.), to race (personne racisée, blanche, noire, autochtone), and to disability (personne sourde, handicapée, autiste, malentendante). Most people nouns (78.73 %, n = 618) are in the plural form; 21.27 % (n = 167) are in the singular form.

Table 4:

Twenty most frequent “people noun” NPs in the corpus.

NP	Freq.	%	NP	Freq.	%
personne(s) trans	183	23.31	personne(s) handicapée(s)	10	1.27
personne(s) non-binaire(s)	115	14.65	personne(s) immigrante(s)	10	1.27
personne(s) racisée(s)	99	12.61	personne(s) autochtone(s)	9	1.15
personne(s) queer	45	5.73	personne(s) homosexuelle(s)	9	1.15
personne(s) blanche(s)	37	4.71	personne(s) militante(s)	9	1.15
personne(s) noire(s)	33	4.2	personne(s) cis	8	1.02
personne(s) lgbtq	14	1.78	personne(s) lesbienne(s)	8	1.02
personne(s) féministe(s)	13	1.66	personne(s) intersexe(s)	7	0.89
personne(s) hétéro	13	1.66	personne(s) marginalisée(s)	7	0.89
personne(s) bisexuelle(s)	10	1.27	personne(s) gay	6	0.76

In order to explore social variation in the use of people nouns versus property nouns, we created a subset based on all “properties” (trans, non-binaire, racisé, etc.) expressed by both people nouns and property nouns in the corpus. It contains 64 properties and 4,714 noun phrases: 703 people nouns and 4,011 property nouns (including 2,020 neutralizations, 1,433 feminine noun phrases, 517 masculine noun phrases, and 41 doublets). Most nouns are in the plural form (2,920, with 2,368 property nouns and 552 people nouns). There are 1,794 nouns in the singular form (1,643 property nouns and 151 people nouns). Figure 2 shows the observed probabilities of people nouns versus property nouns in each subcorpus and for each age group. It suggests an effect of both factors. The probability of producing people nouns is higher in Montreal than in Marseille and Paris, especially for the two younger groups. Moreover, age seems to be negatively correlated with the probability of people nouns, with a decrease in probability from one age group to another.

Figure 2:

Observed probabilities of people nouns versus property nouns, by city and age.

We created a mixed-effects logistic regression model with the lme4 R package (Bates et al. 2015), with speaker and property as random effects, and city, age, education, engagement, and number as fixed effects. The model predicts the production of people nouns. The reference levels of predictors are Montreal, older speakers, high school graduates, speakers with associative activities, and the singular form of nouns. The results are displayed in Table 5. The model shows a significant effect of city: the probability of using people nouns is lower in Marseille (OR = 0.48, p = 0.018) and Paris (OR = 0.43, p = 0.001) than in Montreal. Age also has a partial effect: the two youngest groups use more people nouns versus property nouns than older speakers (OR = 7.21 and p < 0.001 for 19–34 y.o.; OR = 4.94 and p < 0.001 for 35–59 y.o.). For the engagement factor, there is a significant difference between the associative groups and the collective group (OR = 0.46, p = 0.032). Finally, the model shows that the probability of using people nouns is higher for plural noun phrases than for singular noun phrases (OR = 2.53, p < 0.001).

Table 5:

Mixed-effects logistic regression model 1 (boldface indicates statistically significant p-values).

Predictors	Odds ratios	CI	p
(Intercept)	0.05	0.02–0.17	<0.001
City: Marseille	0.48	0.26–0.88	0.018
City: Paris	0.43	0.26–0.72	0.001
Age: 19 to 34 y.o.	7.21	3.71–14.02	<0.001
Age: 35 to 59 y.o.	4.94	2.51–9.71	<0.001
Education: Undergraduate	1.05	0.45–2.44	0.916
Education: Master’s	0.85	0.37–1.96	0.703
Education: PhD	0.41	0.15–1.14	0.087
Engagement: Collective	0.46	0.23–0.93	0.032
Engagement: Media	0.56	0.29–1.08	0.086
Engagement: Professional	0.95	0.48–1.90	0.895
Engagement: Academic	0.92	0.45–1.87	0.815
Number: plural	2.53	1.89–3.40	<0.001

A priori, politically, the use of a person noun instead of a property noun in order to emphasize the referent’s “personhood” is orthogonal to gender-inclusive language; however, grammatically, the fact that personne is feminine means that switching to a noun phrase built around this noun from a non-common gender property noun will have the effect of decreasing the so-called generic masculines. If all epicene people nouns in the subset (n = 702) were replaced by generic masculines, the overall proportion of generic masculines in the corpus would be 8.18 % (n = 2,364), instead of 5.75 %.

We therefore conclude that one of the main reasons that Montreal speakers use fewer so-called generic masculines than their French counterparts is that they are greater users of person- and identity-centered language.

4.5 The role of feminist engagement in doublets

As mentioned in Section 2, much of the discourse surrounding current feminist linguistic practices focuses on doublets and composed forms, which are pronounced as doublets in speech. Although integrating doublets into their speech is perceived as requiring little effort for some speakers, other speakers in CaFé describe having to train themselves to use these forms systematically (14).

(14)

a.

ça vient vite il y a quand même forcément effectivement au départ à faire un tout p -un petit effort c’est-à-dire quel que s- même si vous êtes convaincue à cent mille pour cent qu’il faut utiliser le féminin et le masculin ce qui est mon cas euh il y a des telles habitudes voire habitus langagiers euh qu’il faut cognitivement euh défaire certaines euh réflexes de de parole pour en installer euh pour en installer d’autres mais ça vient très moi je trouve que ça vient très vite (Paris 14)

‘It comes quickly. There is indeed necessarily at the beginning a small effort, that is, even if you’re 100 000% convinced that you need to use the feminine and the masculine, which is my case, there are so many linguistic habits, or habitus, that we must cognitively deconstruct certain speech reflexes to put in other, but it comes very – I find that it comes very quickly’

b.

du coup je m’efforce de le faire à l’oral je le fais pas tout le temps parce que j’oublie (Paris 24)

‘So I force myself to do it in speech. I don’t do it all the time because I forget’

In order to investigate whether these impressions translate into production patterns, we conducted a quantitative study of variation between so-called generic masculines and doublets. The 335 doublets found in the corpus are formed with 86 different pairs of nouns (or in some case a single noun). The ten most frequent pairs are shown in Table 6. Most doublets are formed with nouns that have a different form in the masculine and in the feminine in speech (travailleurs et travailleuses ‘workers’, des spectatrices et des spectateurs ‘spectators’), but some are formed with neutralized nouns, with a determiner or adjective (un ou une alliée ‘ally’, certain certaines amies ‘certain friends’). Since the doublets are presented as ways to directly replace so-called generic masculines, we study the probability of using doublets versus these masculines.

Table 6:

Ten most frequent pairs in doublets in CaFé.

Nouns	Frequency	%
étudiant/étudiante ‘student’	58	17.31
travailleur/travailleuse ‘worker’	37	11.04
auteur/autrice ‘author’	22	6.57
copain/copine ‘buddy’	20	5.97
chercheur/chercheuse ‘researcher’	17	5.07
militant/militante ‘activist’	17	5.07
politicien/politicienne ‘politician’	9	2.69
colleur/colleuse ‘street activist’	6	1.79
expert/experte ‘expert’	6	1.79
interlocuteur/interlocutrice ‘interlocutor’	6	1.79

The dataset used in this analysis contains all generic masculine NPs (n = 1,661) and all doublets found in the corpus (n = 335). The probabilities of using a doublet compared to a so-called generic masculine according feminist engagement is shown in Figure 3. It suggests that academics use more doublets than other people in other categories of feminist engagement.

Figure 3:

Observed probabilities of doublets versus generic masculine, per type of engagement.

We created the model with the same random effects and predictors as model 1. Reference levels of predictors are Montreal, older speakers, high school graduates, academics, and singular NPs. The model (Table 7) predicts the probability of doublets. There is an effect of city, with the probability of producing doublets versus generic masculine NPs being lower in Marseille than in Montreal (OR = 0.30, p = 0.007). The effect of education is partial, with people with only undergraduate diplomas producing more doublets than those with only high-school diplomas (OR = 5.90 p = 0.023). The type of engagement is also significantly linked to the probability of doublets, which is lower in three groups (association, media and professional) than in the academic group. The effect of age is partially significant: younger speakers (19–34 y.o.) use more doublets than the oldest group (OR = 2.38, p = 0.44). There is no significant effect of grammatical number.

Table 7:

Mixed-effects logistic regression model 2. (boldface indicates statistically significant p-values)

Predictors	Odds ratios	CI	p
(Intercept)	0.03	0.00–0.18	<0.001
City: Marseille	0.30	0.12–0.72	0.007
City: Paris	0.50	0.25–1.01	0.053
Age: 19 to 34 y.o.	2.38	1.02–5.55	0.044
Age: 35 to 59 y.o.	2.03	0.87–4.72	0.101
Education: Undergraduate	5.90	1.28–27.31	0.023
Education: Master’s	3.05	0.67–14.00	0.151
Education: PhD	1.90	0.35–10.29	0.458
Engagement: Association	0.26	0.10–0.71	0.008
Engagement: Collective	0.33	0.10–1.05	0.061
Engagement: Media	0.30	0.11–0.86	0.025
Engagement: Professional	0.24	0.08–0.71	0.010
Number: Plural	1.00	0.64–1.57	0.992

Once again, we find that Montreal activists and younger speakers are leaders when it comes to avoiding so-called generic masculines (although there is no significant difference between speakers aged 35–59 and the oldest speakers).^[4] Feminist engagement also has a significant effect, with speakers engaged in academic feminism showing a higher rate of doublets than everyone else. We hypothesize that this effect is the result of the “technical” nature of the doublets. As described above, many of the speakers in CaFé describe having to “train” themselves over time to use them, and some mention still failing, despite their firmly held conviction that this feminist linguistic practice is important. Public speaking in classes (educational contexts) and conferences (scientific contexts) is part of an academic’s job, and so people in this profession have many opportunities to practice using doublets, which ends up making them more proficient. They are also engaged in training and knowledge transfer, so they may have a greater desire to model the linguistic behavior that they believe will have a positive effect on gender equality. Indeed, a number of speakers in CaFé bring up actively teaching their students how to use doublets (15), or having university professors model this linguistic behavior for them (16).

(15)

c’est ce que j’expliquais aux aux étudiants et aux étudiantes que j’avais en face de moi […] il faut au départ en prendre conscience qu’on ne parle qu’au masculin et et de se dire, “là, allez masculin et féminin” puis une habitude va en chasser une autre (Paris 14)

‘It’s what I was explaining to the students I had in front of me or the people that I might have in training […] one first needs to become aware that we only speak in the masculine and to tell oneself “now, let’s go masculine, feminine”, and then one habit will chase away another’

(16)

on a une professeure qui à chaque début de cours quand elle commence un nouveau cours elle conseille à ses étudiants étudiantes d’utiliser l’écriture inclusive (Montreal 20)

‘At university, […] we have a teacher that, at the beginning of each class when she started a new class, she advised her students to use inclusive writing’

In this vein, we also consider it significant that the most commonly doubled noun is étudiant(e) ‘student’, a noun that has been commonly featured in written doublets in university publications for many years (see Burnett and Pozniak 2021).

5 Conclusions

In this paper, we presented a quantitative study of noun phrases in the spoken French corpus Cartographie linguistique des féminismes (CaFé). We argued that CaFé is a particularly interesting corpus for studying reference to human beings because of how big a part grammatical gender plays in French feminist linguistic activism and how intensely the question of how/whether to avoid so-called generic masculines is debated among activists and the general public. We show that, although many of the CaFé speakers express the idea that gender-inclusive communication is easier in writing than in speech, reflecting the overwhelming focus on written French in linguistic guides and public debates, their spontaneous use of masculines to refer to people other than men is extremely low. Crucially, however, the rate of so-called generic masculines is not so low because speakers are following the recommendations of many guides and using masculine-feminine doublets systematically. In fact, the doublets represent only 1.4 % of the noun phrases in the corpus, and are predominantly used by academics who need to train themselves and others to use them. Rather, one of the reasons so-called generic masculines are so low in the CaFé corpus is the fact that the spoken language generates far many more morpho-phonological contexts in which grammatical gender is neutralized than the written one. With our corpus study, we were able to quantify these contexts and showed that, were the corpus written, the percentage of so-called generic masculines would double, going from around 5–10 % of the noun phrases. Further research is needed to see if this gap between the spoken French language and the written one is also present to that extent in other oral corpora; however, since this gap has nothing to do with feminism or LGBT activism, we hypothesize that, were we to replicate this study on an oral corpus with non-activist speakers, we would find a similar result in which neutralizations are much more frequent than so-called generic masculines.

We did, however, identify another sociolinguistic phenomenon that interacts with gender-inclusive speaking and that we do expect to be more common in activist speech: person/identity-centered language. We argued that linguistic practices originating in 1990s North American anglophone disability activism have come into French and been expanded to cover a wide range of property nouns, not just ones referring to disabilities. French Canada being more aligned with English Canada and the United States on these topics than France, it is not surprising that we find a much higher rate of person noun phrases in Montreal than in Paris or Marseille. We showed that the more widespread use of person/identity-centered language in Montreal significantly decreases so-called generic masculines in this city’s subcorpus, since the person noun personne is grammatically feminine, while many of the property nouns are gender marked even in speech.

Overall, we have argued that activist and scientific descriptions of French should give the spoken language more consideration. From the point of view of descriptive and theoretical linguistics, our paper makes a novel contribution to the growing literature on how spoken French looks very different from written French. Other researchers have highlighted how the grammatical systems of varieties of spoken French differ greatly in the (pro)nominal and verbal domains, and the conclusion that emerges from our study is that this is also the case when it comes to the types of noun phrases used: while masculines are one of the main ways to accomplish generic and gender-mixed reference in writing, our results suggest that so-called generic masculines may be marginal in speech, at least compared to neutralizations. From the point of view of gender-inclusive language policy, rather than emphasizing how complicated doublets are to spontaneously produce, guides could be emphasizing how French phonology and morphophonology make gender-inclusive speaking easier, and the synergies that exist with person-centered language. More generally, our study shows that francophones’ speech can be highly gender inclusive, even if little effort is made to double masculines with feminines. We therefore conclude that, at least for some activists, the problem of how to avoid so-called generic masculine noun phrases has already been mostly solved.