In languages with grammatical gender, such as French, the use of masculine forms to refer to both males and females, as well as unspecified or mixed-gender groups, is a common linguistic convention (Gygax et al., 2019). As a result, while the feminine form always has a specific interpretation, the masculine form can be used in both a specific and a generic sense. For instance, in French, the word étudiantes [students_FEM] unambiguously refers to a group of female students, while the word étudiants [students_MASC] can refer to a group of only male students, both male and female students, or even students whose gender is unknown. This ambiguity poses challenges for language processing as the cognitive system needs to resolve the reference of the noun. This is evident in the following example, which is particularly challenging for French speakers due to the alternation of generic and specific senses of the same masculine form within a single sentence¹:

The resolution of this kind of ambiguity relies on various cognitive mechanisms, including world knowledge, context, and pragmatic cues. However, previous research suggests that the ambiguity is typically resolved to the disadvantage of the feminine gender, with the masculine generic favoring the activation of masculine-specific representations. For instance, in Experiment 1 of Gygax and Gabriel (2008), French-speaking participants had to decide, as fast as possible, whether the person referred to by a kinship term (e.g., un frère [a brother] or une sœur [a sister]) could be part of a group described by a role noun in the masculine generic form (e.g., musiciens [musicians_MASC]). The results revealed a strong bias toward male kinship terms, indicating that male kinship terms were more readily associated with role nouns written in the masculine form. Similarly, when asked to estimate the percentage of women in a group described by a generic masculine term, participants consistently selected gender ratios below 50%, despite the role nouns being non-stereotyped in terms of gender (Tibblin et al., 2023). Moreover, studies have shown that the use of generic masculine language in questionnaires can impact the self-reported efficacy of female respondents (Chatard-Pannetier et al., 2005; Vainapel et al., 2015). Wilson and Ng (1988) used a priming paradigm to manipulate the perception of gender in ambiguous photographs through short sentences presented beforehand. The results indicated that the use of masculine generic language in the sentences biased perception toward male faces. Together, the findings from these experiments, as well as other studies mentioned below, provide robust evidence supporting the notion that the use of masculine-generic forms automatically activates masculine-specific representations.

One effective approach to reduce the influence of the masculine bias caused by masculine-generic forms is to employ gender-fair forms. Two main strategies can be distinguished regarding gender-fair language (see, e.g., Tibblin et al., 2023, Figure 1): neutralization on the one hand, which can be achieved, for instance, through the use of gender-unmarked words, and re-feminisation on the other hand, exemplified by the contracted double form. Gender-unmarked words have the same form for both male and for female referents,² and are only differentiated by the preceding article. For instance, in French, le ministre vs. la ministre [the_MASC minister_unmarked vs. the_FEM minister_unmarked] refer to a male or female minister respectively. In certain cases, gender information can be entirely removed: when using plural forms (les ministres [the_{plur,unmarked} ministers_{plur,unmarked}]) or when an elision occurs on the article due to the phonological context (le or la replaced by l' if the following word begins with a vowel). For instance, l'acrobate [the_unmarked acrobat_unmarked] can refer to either a female or a male acrobat. On the other hand, the re-feminisation strategy consists in explicitly listing the masculine and feminine forms, such as un ou une ministre [a_MASC or a_FEM minister_unmarked]. This expression can be sometimes shortened into a contracted double form using a mid-dot (e.g., un·e ministre).³

There are two important distinctions between the gender-unmarked form and the contracted double form. First, while the gender-unmarked form is truly neutral in the sense that it lacks any explicit gender markings, the contracted double form explicitly includes both the feminine and the masculine forms. Secondly, these two forms hold different sociolinguistic status. Gender-unmarked forms are relatively common in standard French, even in a generic masculine context, whereas contracted double forms are still socially marked: although their use has been promoted in France since 2015 by the Haut Conseil à l'Égalité entre les femmes et les hommes (Guide pratique pour une communication publique sans stéréotypes de sexe, 2022), and is occasionally found in official documents, they are still considered as a radical linguistic innovation and have been the subject of intense public and political debates in the French society (see Candea and Véron, 2019; Loison-Leruste et al., 2020 for more sociocultural context).

Extensive empirical research has been conducted to investigate whether the masculine bias associated with masculine generic forms can be eliminated or reduced through the use of gender-fair forms. Brauer and Landry (2008) conducted five studies comparing the activation of masculine representations in a masculine-generic context and in a double-form context in French. The results of all five experiments consistently demonstrated that double forms promote more gender-balanced representations. For instance, in the first experiment, participants cited nearly three times as many woman as a potential Prime Minister when the question was asked using a double form (candidat/candidate [candidate_MASC/candidate_FEM]), as compared to a generic masculine form (candidat [candidate_MASC]) (Brauer and Landry 2008, exp. 1; see also Stahlberg and Sczesny, 2001). In a subsequent experiment, the use of masculine generics, as opposed to double forms, decreased the likelihood of participants naming a woman when asked to list their favorite heroes and artists (Brauer and Landry 2008, exp. 2; see also Stahlberg et al. 2001). The experimenters then conducted two experiments on adults and children, showing that grammatical form also influenced the gender of the imagined prototypical person when envisioning an occupational group (Brauer and Landry, 2008, exp. 3 and exp. 4). In a final experiment, participants read a short text about a professional gathering and estimated the percentages of men and women present. The findings indicated that the double form resulted in a more gender-balanced representation of the group when no information about its gender composition was provided (Brauer and Landry 2008, exp. 5; see also Braun et al., 1998, 2005; Xiao et al., 2023).

Studies investigating the efficiency of the neutralization strategy in reducing masculine bias are more seldom. Kim et al. (2023) conducted a study in which participants had to judge whether an individual, identified solely by a gendered first name, could be part of a group described either in a plural generic masculine form or a gender-neutral form. The results demonstrated that the use of gender-neutral forms dramatically reduced the male bias observed with masculine forms, and that both grammatical forms used equal cognitive resources, as indicated by comparable response times. Richy and Burnett (2021) conducted a gender judgement task that yielded similar findings. Non-gender-stereotyped role names introduced with a masculine article (e.g., le journaliste [the_MASC journalist]) were rated as masculine to a much higher degree than those introduced with a feminine article (e.g., la journaliste [the_FEM journalist]). In contrast, when the gender marking was completely eliminated, such as in l'imperturbable journaliste [the_unmarked unflappable journalist], the rating was close to balance (46%). These findings provide further support for the efficacy of gender-neutral forms in reducing masculine bias.

This raises the question of whether one of the two strategies (neutralization or re-feminization) is more effective in promoting gender-balanced representations. Only a few studies have been designed to address this question, and the conclusions drawn from them are generally inconsistent. Lindqvist et al. (2019) explored whether job candidates described in short paragraphs were perceived as more masculine or more feminine when the summaries were written using gender-neutral words (such as “the applicant” or the singular “they”) or a double pronoun (“he/she”). The results indicate that double forms produced more balanced gender representations compared to gender-neutral forms, both in English and in Swedish. These findings align with other experiments conducted in French and German, suggesting that the specific type of gender-fair form may have a differential effects, particularly when they interact with stereotypical information (Irmen and Roßberg, 2004; Braun et al., 2005; Irmen, 2007; Brauer and Landry, 2008). However, these conclusions contradict those of other researchers. In a recent study conducted in French, Tibblin et al. (2023) compared the perceived gender ratio within groups described using a single role noun in the masculine generic form or four different gender-fair forms. The results indicated that all tested gender-fair forms were equally successful in countering the male bias induced by the masculine form, resulting in estimated percentages of women ranging between 49.1 to 50.0%. Similar results have also been reported in a study by Stahlberg et al. (2001) investigating neutralization and re-feminization strategies in German. Furthermore, several studies have found that gender-neutral forms were able to largely eliminate the masculine bias (see, in particular, Richy and Burnett, 2021; Kim et al., 2023, in French). These discrepancies may be attributed to methodological issues, such as inadequate control for response biases or insufficiently precise measures that might have hindered the detection of potential effects. For example, Tibblin et al. (2023) demonstrated that the direction of the response scale in a gender ratio estimation paradigm has a systematic effect on the results.

In summary, both gender-unmarked forms (neutralization strategy) and contracted double forms (re-feminization strategy) have potential in reducing gender biases in language. However, gender-unmarked forms seem to yield intermediate results between double forms and masculine generic forms, suggesting a more nuanced impact on mental representations. Thus, the objective of the present study was twofold: firstly, to provide confirmation, using French sentences, that gender-unmarked forms are not fully effective in neutralizing the inherent masculine bias (Experiment 1); second, to compare this condition with a re-feminization strategy, specifically the contracted double form using a mid-dot (Experiment 2). Our hypothesis for the second experiment was that contracted double forms would yield more balanced representations than gender-unmarked forms.

As noted above, the choice of a particular methodology is crucial for detecting potential differences among gender-fair forms. To ensure precise measurements capable of capturing these distinctions, we modified the sentence evaluation paradigm employed in prior studies (Gygax et al., 2008; Garnham et al., 2012). Compared to the more commonly-used gender-ratio estimation experiment, the sentence evaluation paradigm offers several advantages. It is less prone to ceiling effects as it relies on response time measurements. Additionaly, it is less susceptible to criterion effects, by avoiding subjective scaling task. Finally, stereotypes associated with nouns and contexts are known to greatly contribute to gendered interpretations. Thus, it is crucial to carefully control for this factor in experiments aimed at assessing the influence of grammatical gender. In practice, experimenters usually evaluate the level of stereotype associated with their stimuli prior to testing, in order to establish a stereotype-neutral condition (Irmen and Roßberg, 2004; Xiao et al., 2023). We adopted a similar approach in the present study.

The objective of the present study was to examine the efficiency of gender-unmarked forms and contracted double forms in mitigating gender biases in language. Two experiments were conducted, in which participants had to judge the consistency of a pair of French sentences: a context sentence where the gender of the character was intentionally left unspecified, using gender-fair forms, and a continuation sentence beginning with a feminine or a masculine singular pronoun. All nouns included in the experiment were pre-tested for the absence of strong gender stereotypes. In the first experiment, we confirmed that gender-unmarked forms in French were not fully successful in neutralizing the masculine bias. Specifically, sentences beginning with a feminine pronoun exhibited longer processing times and higher error rates compared to those beginning with a masculine pronoun, indicating the persistence of a masculine bias even in a gender-neutral context. For instance, following the context sentence L'élève commençait à réviser [The_unmarked student_unmarked began to revise], the feminine continuation sentence Elle passait l'examen dans 15 jours [She was taking the exam in 15 days] required more time to process and resulted in more errors compared to the masculine continuation sentence Elle passait l'examen dans 15 jours [She was taking the exam in 15 days]. In the second experiment, we contrasted gender-unmarked forms (such as L'élève) with contracted double forms using a mid-dot (such as Un·e élève). The results revealed that the use of contracted double forms, which reintroduce grammatical gender markers associated with the feminine gender, eliminated the gender bias observed in the gender-unmarked condition. In the contracted double form condition, there was no significant difference in processing time between feminine and masculine continuation sentences, suggesting that the re-feminization strategy successfully counteracted the masculine bias.

The results of Experiment 1 are consistent with our hypothesis: gender-unmarked forms in French are not fully effective in neutralizing masculine bias. This finding is in line with previous research demonstrating that neutral wordings elicit a male bias, in English and Swedish (Lindqvist et al., 2019) as well as in German (Irmen and Roßberg, 2004). However, it contrasts with several studies that found no bias induced by gender-neutral forms (e.g., Tibblin et al. 2023, Richy and Burnett 2021 and Kim et al. 2023 in French, Stahlberg et al. 2001 in German). It is possible that these studies failed to capture the difference in masculine bias due to methodological limitations. More specifically, some estimates approached ceiling levels, making it difficult to detect potential effects (e.g., Tibblin et al., 2023). Furthermore, in some experiments, participants may have been able to understand the study's objective and adjusted their responses accordingly. In contrast, we believe that the sentence evaluation paradigm used in the present study is less susceptible to ceiling effects and less prone to the development of conscious strategies by participants. Finally, differences in the test population demographics may also explain these discrepancies. This point will be discussed in section 4.3.

The observation of a masculine bias in the absence of any grammatical marker may seem surprising. At this point we can only speculate about the causes of this effect. Two possible, non-exclusive explanations are considered in the following paragraphs: a first one based on lexical frequency effects, and another one based on semantics.

All the codes and data supporting this study are openly available on Github at https://github.com/LeoVarnet/GenderFair/.

The studies involving humans were approved by Comité d'Ethique de la Recherche Paris-Descartes. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

ES: Writing—review and editing, Data curation, Methodology, Resources. J-PC: Methodology, Writing—review and editing, Conceptualization. LV: Writing— review and editing, Formal analysis, Investigation, Writing—original draft.