Literature review

Although it is not always known by this name, linguistic androcentrism has been a major concern for feminists in many countries, since at least the 12th century (Weatherall, Reference Weatherall2002). Psychological and psycholinguistic studies of masculine and inclusive forms date back to the 1970s for English (e.g. Kidd, Reference Kidd1971; Bem and Bem, Reference Bem and Bem1973; Schneider and Hacker, Reference Schneider and Hacker1973; Soto and Cole, Reference Soto and Cole1975; Pincus and Pincus, Reference Pincus and Pincus1980), where researchers showed that anglophones interpret words with masculine marking (chairman, fireman, policeman etc.) or masculine pronouns (he, him, his) as more likely to refer to men, even if the context favours a gender neutral reading (i.e. The chairman must arrive before his colleagues). Later studies such as Stout and Dasgupta (Reference Stout and Dasgupta2011) and those described in Ng (Reference Ng2007) have shown a link between these male-biased interpretations and women’s feelings of ostracism on the job market and other kinds of gender-based discrimination.

Psychological and psycholinguistic research on grammatical gender in French is more recent, yet substantial (and growing). In the past 15 years, many scholars have argued based on experimental results that grammatically masculine noun phrases and pronouns create male-biased mental representations (e.g. Brauer and Landry, Reference Brauer and Landry2008; Chatard et al., Reference Chatard, Guimond, Lorenzi-Cioldi and Désert2005; Gabriel et al., Reference Gabriel, Gygax, Sarrasin, Garnham and Oakhill2008; Garnham et al., Reference Garnham, Gabriel, Sarrasin, Gygax and Oakhill2012; Gygax et al., Reference Gygax, Gabriel, Sarrasin, Oakhill and Garnham2008, Reference Gygax, Gabriel, Lévy, Pool, Grivel and Pedrazzini2012, Reference Gygax, Schoenhals, Lévy, Luethold and Gabriel2019; Richy and Burnett, Reference Richy and Burnett2021; Pozniak and Burnett, Reference Pozniak and Burnett2021), providing extra support to qualitative arguments (e.g. Beauvoir, Reference Beauvoir1949, Yaguello, Reference Yaguello1979; Houdebine, Reference Houdebine1987; Michard, Reference Michard1996) that linguistic androcentrism is also a problem in French. Much more recently, there has been a fair amount of psycholinguistic work on the processing of proposed solutions to linguistic androcentrism in French: the various inclusive forms. This work has been partly inspired by current large societal debates in France and elsewhere in the francophone world on whether gender inclusive writing should be used and, if so, which type. As Abbou et al. (Reference Abbou, Arnold, Candea and Marignier2018), Burnett and Pozniak (Reference Burnett and Pozniak2021) and Abbou (Reference Abbou2022) describe, the most recent debates in France started in 2017, when members of the right wing press discovered an elementary school history textbook that had been written using a form of gender inclusive writing, and started writing articles claiming that such orthographic conventions were problematic for the French language and society. ^{Footnote 4} These complaints, which were primarily directed towards the point médian form, received much attention from major governmental institutions. The Académie Française declared that gender inclusive writing was a “mortal peril” for the French language, ^{Footnote 5} and the prime minister at the time, Édouard Philippe, even issued an official statement proscribing the point médian, among other forms, in official government documents. ^{Footnote 6} Interestingly, in this statement, Édouard Philippe specified that, while composite forms are prohibited, the double forms (étudiante ou étudiant) are allowed. Édouard Philippe’s distinction therefore raises the question of whether double form, which is sanctioned by the state, can play the same role as other composite forms, like the point médian, which are now outlawed in official documents.

This question was recently investigated by Tibblin et al. (Reference Tibblin, van de Weijer, Granfeldt and Gygax2023a). These authors asked French L1 participants to estimate the distribution of men and women on a scale ranging from 0 to 100% in 22 non-stereotyped French role nouns. These nouns were either in the masculine form (joggeurs), or double form (either feminine-masculine joggeuses et joggeurs, or masculine-feminine joggeurs et joggeuses), the contracted double form (joggeur·euses) or the gender-neutral form (un groupe de jogging). They found that using inclusive writing, whatever the form (either gender-neutral or double forms), generated higher perceived percentages of women compared to the masculine form and no statistical difference was found between the different inclusive forms used.

Tibblin et al. studied nouns in isolation, but another study, Xiao et al. (Reference Xiao, Strickland and Peperkamp2023), looked at the interpretation of inclusive writing in context. In addition, they studied an expanded set of nouns, differentiating according to their stereotypicality (with male-stereotyped (diplomat), women-stereotyped (secretary) and neutral-stereotyped professions (journalist)), they also contextualised the roles nouns in a short text (three sentences) as in (2).

1. (2) Le rassemblement régional des PROFESSION NAME a eu lieu cette semaine à Amiens. La localisation centrale de cette ville a été particulièrement appréciée. Les PROFESSION NAME ont aussi adoré l’apéro offert à l’hôtel de ville le premier jour.

‘The regional gathering of PROFESSION NAME took place this week in Amiens. The central location of this city was particularly appreciated. The PROFESSION NAME also loved the aperitif offered at City Hall on the first day.’

Although Xiao et al.’s stimuli were more elaborate than Tibblin et al.’s, the question they asked participants was roughly the same: they asked them to estimate the percentages of men and women during the gathering mentioned in the short text (see (3)). Results showed that use of gender inclusive forms (either contracted double forms with the mid-dot or complete double forms) led to higher estimations of women compared to the use of masculine form. They also showed that inclusive writing’s effects changed depending on the stereotype class of the noun: with both inclusive forms, there was less male bias for neutral- and female-stereotyped professions, and more female bias for men-stereotyped professions.

1. (3) Selon vous, quels étaient les pourcentages d’hommes et de femmes dans ce rassemblement?

“In your opinion, what were the percentages of men and women in the gathering?”

In summary, these very important experimental investigations of French gender inclusive writing forms have shown that double form and the point médian do an equally good job at countering linguistic androcentrism, at least when they are presented in isolation or with short contextualization. This being said, we would like to question how directly relevant these results are to the issue of using gender inclusive writing in, for example, official documents. Our interrogation is based on some recent research in sociolinguistic perception on the relation between context and the interpretation of sociolinguistic variables.

Importance of context

It has been known for a long time that the context, especially the discourse context, plays a role in the processing of many linguistic phenomena, such as asymmetries in the processing of subject vs object relative clauses (see Roland et al., Reference Roland, Mauner, O’Meara and Yun2012). Hilton and Jeong (Reference Hilton and Jeong2019) also insist on the importance of context from a sociolinguistic perspective. People usually make social inferences about others based on how they speak. However, these inferences – or impressions on the speaker – may change depending on the context. As described in Hilton and Jeong (Reference Hilton and Jeong2019), one example is Campbell-Kibler’s (Reference Campbell-Kibler2009) study on the ING variant in English that shows that people using the velar variant (like smoking) will be perceived as being more educated than those using the alveolar variant (smokin’), unless they have a Southern US accent. Following that, Hilton and Jeong (Reference Hilton and Jeong2019) suggest that people will ignore the social information conveyed by the sociolinguistic variables if they have gathered sufficient initial impressions of the speaker from contextual enrichment (i.e. from the information about the speaker and the speaking situation). This implies that some sociolinguistic variables known to have an effect in contextually less detailed stimuli (i.e. without sufficient information about the speaker and the speaking situation) could have little or no influence on perception in more contextually detailed stimuli where contextual enrichment takes over the social information conveyed, except if the social meanings conveyed by the forms are highly differentiable within a language (i.e. enregistered, Agha, 2003). Through ten experiments, they investigate to what extent this contextual enrichment influences the sociolinguistic perception of three linguistic variables in English: number agreement in there constructions, intonation contours in declarative sentences and overlapping speech in conversation. ^{Footnote 7} They find that contextual enrichment overall diminishes the effect of the social meaning for the first two variables (number agreement and intonation contours in declarative sentences), but not for overlapping speech in conversation. They suggest that the difference found between the variables is due to the fact that speech overlaps are subject to metalinguistic commentary, which is why participants have strong attitudes about it, contrary to the other two variables. They therefore conclude that contextual enrichment can obscure the effects of the linguistic forms, except if the social meanings conveyed by the forms are enregistered. This effect of context has been found for other sociolinguistic variables such as the social indexicality of English numerical expressions in Beltrama et al. (Reference Beltrama, Solt and Burnett2022).

It thus seems necessary to take the context into account when looking at different forms to see which forms are really differentiable and still convey social meanings beyond the contextual enrichment. As explained earlier, studies from Xiao et al. (Reference Xiao, Strickland and Peperkamp2023) and Tibblin et al. (Reference Tibblin, van de Weijer, Granfeldt and Gygax2023a) on the interpretation of inclusive writing in French showed that forms like the double form or the point médian both succeeded in reducing linguistic androcentrism to the same degree. Still, one may wonder whether the social meaning of these forms would still be conveyed in a more detailed context, or whether it would be diluted by extra information. Moreover, previous studies asked participants directly to estimate the proportions of men and women. In this way, the question encouraged focusing on the form used (either inclusive or masculine). Semantically speaking, the interpretation of inclusive forms should also be tested in a more indirect way to see to what extent the social meaning of the inclusive forms is robust.

For these reasons, we decided to do three experiments testing whether the point médian and the double forms still reduce linguistic androcentrism in a more detailed context like the university brochures and in a more indirect way.

Experiment 1. Interpretation of point médian

Materials and design

We created 12 items that consisted of undergraduate brochures which were inspired by Burnett and Pozniak (Reference Burnett and Pozniak2021). As illustrated in Figure 1, each brochure was a 100 words long text with four occurrences of the word étudiant ^{Footnote 8} (student). The first variable we manipulated was a within-item variable: the use of linguistic forms on the word étudiant (use of point médian or use of the generic masculine), leading to two conditions per item. The second variable was a between-item variable: the gender balance (parity) of the academic discipline described by the brochure, with six brochures referring to a rather masculine discipline (ex. mathematics) and six brochures referring to a less masculine discipline (ex. literature). Discipline parity was decided according to the Conseil National des Universités (CNU) numbers for faculty parity used in the corpus study of Burnett and Pozniak (Reference Burnett and Pozniak2021) (i.e. proportions of women in disciplines). Table 1 presents the disciplines selected in the two categories: male-dominated disciplines vs less male-dominated disciplines. All materials are available on https://osf.io/kg2sr.

Figure 1. Example of an item used in experiment 1, in the point médian condition, for a less male-dominated discipline.

Table 1. Proportion of women in the disciplines used in the experiments

Three continuous scales ranging from 0% to 100% were added: one asking how many women will apply, one asking how many international students will apply, and the last one asking how many people from the region of Paris will apply to the degree (see Figure 2). The last two scales were actually added as distractors. The order of the three scales was counterbalanced through the participants. As said in the introduction, the question used in previous studies focused on a direct comparison between the proportion of women and men that are in each named occupation. As such, it might have increased the saliency of both stereotypes and language form attached to the occupation. In our study, by asking “how many women will subscribe”, we want to have a general question associated with the whole description, and not just the activity name.

Figure 2. Example of an order of the three scales used in all experiments.

Twenty-four fillers were created with a structure really similar to the critical items, except that the word étudiant was never mentioned, and any gendered words (nouns, adjectives) were avoided. Five were brochures of undergraduate degrees, three were brochures of majors, nine were brochures of academic classes and nine were brochures of sport classes. Two practice items (similar to the fillers) were added. An example is given in Figure 3.

Figure 3. Example of a filler used in all experiments.

In order to check whether participants were reading carefully the brochures, we included 18 reading comprehension questions (six on the experimental items and twelve on the fillers), each time appearing on a subsequent screen.

Results

Violin plots in Figure 4 show the mean proportions of women estimated who would apply for the bachelor’s degree depending on the linguistic forms (use of point médian, “Étudiant·e·s” or masculine, “Étudiants”). They also show the quartiles (white horizontal lines) giving more detailed information about the distribution of the data.

Figure 4. Proportions of women depending on the linguistic forms.

Violin plots in Figure 5 also show the mean proportions of women, but depending on linguistic forms (use of point médian or masculine, “Étudiants”) and discipline parity (fewer women in the discipline vs more women in the discipline).

Figure 5. Proportions of women depending on the linguistic forms and discipline parity.

As shown in Figure 4, the estimations do not really vary whether the brochure is written with point médian or masculine (51.82 vs 51.66). Indeed, no robust evidence for an effect of linguistic forms was found in the statistical analysis: Est. = 0.18, CrI = [−2.65, 2.96], P(Est. > 0) = .55. Figure 5 shows an effect of discipline parity confirmed in the analysis (strong evidence for an effect, Est. = 25.9, CrI = [12.38, 39.55], P(Est. > 0) = 1), with higher proportions of women in less male-dominated disciplines. No robust evidence for an interaction between parity and linguistic forms (point médian vs masculine) was found (Est. = 0.43, CrI = [−5.22, 6.12], P(Est. > 0) = .56).

Experiment 2. Interpretation of the double form

Materials and design

Materials were the same as in the first experiment, except that the double form (les étudiantes et les étudiants) was used instead of the point médian, as shown in Figure 6.

Figure 6. Example of an item used in experiment 2, in the double form condition, for a less male-dominated discipline.

Results

Violin plots in Figure 7 show the mean proportions of women estimated who would apply for the bachelor’s degree depending on the linguistic forms (use of double form “Étudiantes et étudiants” or masculine, “Étudiants”).

Figure 7. Proportions of women depending on the linguistic forms.

Violin plots in Figure 8 also show the mean proportions of women, but depending on linguistic forms (use of double form “Étudiantes et étudiants”, or masculine, “Étudiants”) and discipline parity (fewer women in the discipline vs more women in the discipline).

Figure 8. Proportions of women depending on the linguistic forms and discipline parity.

Figure 7 shows that proportions of women estimated were higher when double form was used compared to masculine (53.78 vs 50. 41), which was confirmed in the statistical analysis (strong evidence for an effect, Est. = 2.65, CrI = [−0.33, 5.60], P(Est. > 0) = .96). As shown in Figure 8, we again found strong evidence for an effect of discipline parity, with higher proportions of women estimated in less male-dominated disciplines (Est. = 22.27, CrI = [9.24,35.53], P(Est. > 0) = 1). We can also see an interaction between parity and type of inclusive form in Figure 8, for which we found weak evidence for an effect (Est. = −4.13, CrI = [−10.10,1.80], P(Est.<0) = .92): using double form leads to more proportions of women, especially in more male-dominated disciplines (strong evidence for an effect, Est. = 4.62, CrI = [−0.35,9.58], P(Est. > 0) = .97) and not so much in less male-dominated disciplines (Est. = 0.60, CrI = [−5.89,7.23], P(Est. > 0) = .59).

Discussion

Again, male-dominated disciplines led to lower proportions of women by the participants. However, contrary to the point médian in experiment 1, the use of double form led to higher proportions of women, especially in more male-dominated disciplines. Our results do not align with the results found in Xiao et al. (Reference Xiao, Strickland and Peperkamp2023) or Tibblin et al. (Reference Tibblin, van de Weijer, Granfeldt and Gygax2023a) who did not find a difference between the two inclusive forms. It seems that in a detailed context and when the question of the interpretation is more indirect, the effect of point médian is diluted by extra information, while the double form still succeeds in reducing linguistic androcentrism, especially in male-dominated disciplines. Context thus does not appear to have the same effect on the two sociolinguistic forms, with the double form appearing more salient than the point médian (we will come back to this point in the general discussion).

We now consider the question of how general this “contextual dilution” effect is. When looking at the corpus data from Burnett and Pozniak (Reference Burnett and Pozniak2021), qualitative results showed that inclusive forms appear to be rather used at the beginning of the brochures than at the end (27.2% vs 10.6%, see Figure 9). This raises the question of whether it is necessary to systematically use gender inclusive forms (particularly double form) for all relevant noun phrases in the discourse. Perhaps having inclusive forms only at the beginning is enough to signal that the noun phrases are meant to refer to both women and men, even in a detailed context. We decided to test this hypothesis in another experiment to see whether the double form remains salient and continues to convey social information even when it is not systematically used.

Figure 9. Use of inclusive writing depending on the position in the text (Burnett and Pozniak, Reference Burnett and Pozniak2021).

Experiment 3. Systematic use of the double form

Materials and design

Materials were the same as in the second experiment, except that the double form (les étudiantes et les étudiants) was only used in the first two occurrences, with the last two being in masculine, as shown in Figure 10. When constructing the stimuli, we considered the question of where to remove two occurrences of double form. We decided to remove the last two because, again, we were interested in constructing the most realistic stimuli possible.

Figure 10. Example of an item used in experiment 3, in the double form condition, for a less male-dominated discipline.

Results

Violin plots show the mean proportions of women estimated depending on linguistic forms (use of double form, “étudiantes et étudiants” in the first two occurrences or masculine, “étudiants”, in the four occurrences) in Figure 11, and according to discipline parity (fewer women in the discipline vs more women in the discipline) in Figure 12.

Figure 11. Proportions of women depending on the linguistic form.

Figure 12. Proportions of women depending on the form and discipline parity.

Figure 11 shows that participants estimated that more women would apply for the degree when the brochure was written with the double form in the first two occurrences compared to when it was written exclusively using the masculine (49.64 vs 48.42), for which we found weak evidence for an effect (Est. = 1.29, CrI = [−1.39, 4.01], P(Est. > 0) = .83). As seen in Figure 12, participants estimated that more women would apply for a degree in which parity was higher, which was confirmed in the analysis (strong evidence for an effect, Est. = 24.4, CrI = [11.60, 37.31], P(Est. > 0) = 1). No robust evidence for an interaction between linguistic forms was found (Est. = 0.24, CrI = [−5.4, 5.8], P(Est. > 0) = .53).

General discussion

The first experiment showed different results from those found in Tibblin et al. (Reference Tibblin, van de Weijer, Granfeldt and Gygax2023a) and Xiao et al. (Reference Xiao, Strickland and Peperkamp2023) in that the point médian did not appear to increase the proportion of women that participants thought were going to apply in the various disciplines described by the brochure. In the second experiment, the double form did somewhat increase the proportion of women in disciplines across the board, but mainly in male-dominated ones. In this way, double form does appear to favour a ‘gender neutral’ reading, more than a strictly feminized one. ^{Footnote 9}

These results seem surprising compared to those found in other previous studies, but can be explained by the differences in the materials and the questions asked to the participants. First of all, as for the materials, we studied the university context and only looked at one noun: étudiant·e·s (students). It is possible that, were we to investigate a wider set of nouns with different associated stereotypes, we might find some effect with the point médian, as found in Xiao et al. (Reference Xiao, Strickland and Peperkamp2023). Moreover, previous researchers’ questions were more explicit and direct with respect to the proportion of women in groups. In Xiao et al. (Reference Xiao, Strickland and Peperkamp2023), there was a single item, therefore no fillers were needed. In contrast, our study had distractors (two scales) as well as fillers. Indeed, the absence of fillers may have induced some metalinguistic strategies from participants, especially in Tibblin et al. (Reference Tibblin, van de Weijer, Granfeldt and Gygax2023a). Furthermore, the actual semantics of the question we asked participants is very different: we asked them to put themselves in the shoes of a university administrator making decisions for their programs, some of which involved gender parity. In this way, our response question created a much more realistic interactional context, but then at the same time, the relationship between the question and gender parity was less explicit. Given all these differences, it is not surprising that we do not find as strong effects as previous researchers have done.

As we mentioned, our results are reminiscent of those found by Hilton and Jeong (Reference Hilton and Jeong2019), and Beltrama et al. (Reference Beltrama, Solt and Burnett2022): results involving the interpretation of sociolinguistic phenomena obtained in experiments with minimal context weaken or even disappear when the context becomes more detailed. We have named this phenomenon contextual dilution, and we suggest that it arises from how people process the information in the texts that they read. ^{Footnote 10} In shorter contexts, aspects of the form of the language may be more salient to listeners and readers than they would be in more detailed contexts where the listener/reader is required to take more information into account. A further argument for this analysis came from the results of the third experiment, focusing on how often double form is used. Indeed, our results could be interpreted in the following way: only using two occurrences at the beginning of the text may weaken the evidence for a contribution of double form to the creation of a gender neutral interpretation, especially in male-dominated disciplines. However, for this experiment, as mentioned by a reviewer, it could also be argued that adding the masculine forms in the same text with the double form created a stronger male bias. As said earlier, the choice of having the first two occurrences in the double form and the last two in the masculine form was inspired by what we found (Figure 9 above) in the corpus study from Burnett and Pozniak (Reference Burnett and Pozniak2021). Still, as is often the case for corpus studies, several factors may be confounded, ^{Footnote 11} and it would be interesting to test whether the results stay the same when only two double forms are used in the beginning of the text without the masculine forms. We leave this issue for future research.

The differences found between the point médian and the double form could also be imputed to the recent use of the point médian. Indeed, this form has become more frequent since 2017; whereas the double form was already quite widespread. It therefore seems reasonable to suppose that most people have not quite interiorized the social meaning of point médian, all the more in a very detailed context. Another explanation would be that contracted forms such as point médian ressemble masculine forms contrary to the double form that makes the feminine form much more salient. It is thus possible that the contracted form has been processed as masculine forms.

We argue that our results, joint with the results from previous studies such as Tibblin et al. (Reference Tibblin, van de Weijer, Granfeldt and Gygax2023a) and Xiao et al. (Reference Xiao, Strickland and Peperkamp2023) have important implications for language policy or guidelines. In particular, they suggest that using either the point médian or double form in a book title, on social media, in a headline, in a short sign (such as one would find in a feminist protest), or activist graffiti would have the desired effect, but it would be optimal to switch to double form (which is allowed even in government documents), and use it systematically in a more detailed context (such as brochures, as was tested in our experiments). Also, as said before, frequency may be a crucial factor at play for the difference in interpretation between the point médian and the double form. Future research should look at the frequency of the forms used by participants (on a daily basis) to see whether a frequent use of inclusive forms – and mainly the point médian – affects the interpretation of these inclusive forms in a more detailed context, through habituation experiments for example. Finally, our results come from offline experiments, and a few studies with online techniques are emerging like in Friedrich et al. (Reference Friedrich, Drößler, Oberlehberg and Heise2021) for the German star form, or in Gygax and Gesto (Reference Gygax and Gesto2007), and more recently Tibblin et al. (Reference Tibblin, Granfeldt and van de Weijer2023b) for French. Still, based on what we found, more research focussing on the influence of context with online measures are necessary to really capture how people process the inclusive forms.