Introduction
In many languages with grammatical gender, masculine forms are used as the default for generic reference, i.e., to refer to people regardless of their gender. Thus, it is theoretically and grammatically correct to use masculine personal nouns to denote mixed groups or people of unknown or irrelevant gender (Gygax et al., Reference Gygax, Elmiger, Zufferey, Garnham, Sczesny, von Stockhausen, Braun and Oakhill2019). It is also possible to use masculine pronouns generically (Paterson, Reference Paterson2014). In Example 1, he refers to a person of any gender: male, female, non-binary, or otherwise.
-
1. If a person wants to learn a new skill, he should practice regularly.
The focus of the current paper is on pronouns. The generic use of masculine pronouns constitutes a gender asymmetry in language, as feminine forms are not used generically and only refer to women (Sczesny et al., Reference Sczesny, Formanowicz and Moser2016). Masculine generics have increasingly been criticized for perpetuating an androcentric worldview and have been shown to cause a male bias, i.e., a mental overrepresentation of men (e.g., Gastil, Reference Gastil1990; Moulton et al., Reference Moulton, Robinson and Elias1978; Noll et al., Reference Noll, Lowry and Bryant2018; Redl, Reference Redl2021). Such a male bias can reinforce existing inequalities between genders in society (Pérez & Tavits, Reference Pérez and Tavits2019; Tavits & Pérez, Reference Tavits and Pérez2019), e.g., the perception of women’s lack of fit with leadership positions (Horvath & Sczesny, Reference Horvath and Sczesny2016). Stout and Dasgupta (Reference Stout and Dasgupta2011) found that when the masculine generic pronoun he is used in job descriptions during mock interviews, female participants experienced a decreased sense of belonging, lower motivation, and less identification with the job.
As a reaction to linguistic androcentrism and in an effort to counter a potential male bias, users of languages that use masculine pronouns as generics have turned to different gender-inclusive language strategies, such as replacing masculine forms with pair forms, which combine masculine and feminine forms (e.g., “he or she,” “s/he,” …) (Sczesny et al., Reference Sczesny, Formanowicz and Moser2016). Such differentiation strategies, also known as feminization or balancing strategies, have been shown to facilitate the cognitive inclusion of women and counteract the male bias (Hyde, Reference Hyde1984; Lindqvist et al., Reference Lindqvist, Renström and Gustafsson Sendén2019). However, masculine-feminine pair forms increasingly face criticism. Often, they follow the pattern of male firstness, in that the masculine form in a pair almost always precedes the feminine form (Hegarty & Parr, Reference Hegarty and Parr2024; Koster, Reference Koster2020; Willis & Jozkowski, Reference Willis and Jozkowski2018). Additionally, pair forms are increasingly seen as not truly gender-inclusive, since people who do not identify with the male/female dichotomy are not represented in the language. The German gender asterisk (Genderstern), such as in er*sie (“he*she”), is an innovative example of a differentiation strategy aiming to encompass all genders, and more specifically to make non-binary gender visible, by placing the asterisk as a placeholder between the masculine and the feminine form (Friedrich et al., Reference Friedrich, Drößler, Oberlehberg and Heise2021).
In recent years, gender-neutral pronouns, which in most cases have been introduced for specific reference to non-binary individuals, have also been proposed as a suitable strategy for generic reference, since they circumvent the masculine-feminine dichotomy. The most well-known examples of gender-neutral pronouns are currently English singular they and Swedish hen.
In this paper, we focus on Dutch, where the forms die and hen have been introduced as gender-neutral pronouns. As Table 1 shows, these pronouns complement masculine hij (“he”) and feminine zij (“she”) as third person singular pronouns (Taal En Gender: Verwijswoorden Voor Vrouwen, Mannen En Non-Binaire Personen (Algemeen) - Taaladvies.net, n.d.). Dutch is spoken by about 24 million people in the Netherlands and Belgium (Feiten & cijfers - Taalunie, n.d.). The region is characterized by the presence of two institutionalized varieties: Netherlandic Dutch in the Netherlands and Belgian Dutch in Flanders, the northern part of Belgium. Belgian Dutch is also commonly referred to as “Flemish.” Our study includes both Belgian and Dutch language users; we will refer to these varieties as BE Dutch and NL Dutch.Footnote 1 Even though mutual intelligibility is high, differences between the varieties exist at the morphological, phonetic, syntactic, lexical, and semantic levels (Impe et al., Reference Impe, Geeraerts and Speelman2008; Verhoeven, Reference Verhoeven2005).
Table 1. Dutch singular third-person personal and possessive pronouns (adapted from Taaladvies.net)
The Dutch Language Union, which provides official language advice for both Netherlandic and Belgian Dutch, acknowledges the use of gender-neutral pronouns die and hen for specific non-binary reference (as in Example 2) and die for generic reference (as in Example 3) (Taal En Gender: Verwijswoorden Voor Vrouwen, Mannen En Non-Binaire Personen (Algemeen) - Taaladvies.net, Reference Renström, Lindqvist, Klysing and Gustafsson Sendénn.d.).
-
2. Sasha ging uit gisteren. [Die/Hen] had een leuke tijd met [diens/hun] vrienden.
-
Sasha went out yesterday. [They] had a great time with [their] friends.
-
3. Als een persoon een nieuwe vaardigheid wil leren, dan moet [die] regelmatig oefenen.
-
If a person wants to learn a new skill, [they] should practice regularly.
Although the Language Union accepts the use of these newly emerged gender-neutral pronouns, their influence on readers’ perception of text quality has only been investigated for non-binary reference by Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024). Building on this work, our study now shifts its focus to the generic use of these pronouns. An often-heard criticism of gender-inclusive language forms is that they may hinder mindful reading and disrupt the reading flow, thereby reducing text comprehensibility and text appreciation (e.g., Vergoossen et al., Reference Vergoossen, Pärnamets, Renström and Gustafsson Sendén2020). This study posits the question of whether this concern holds true for the use of Dutch gender-neutral pronouns for generic reference. More specifically, we test whether the use of gender-neutral pronouns impairs the perceived text quality compared to more conventional generic referential strategies, such as the use of masculine-only forms or plural forms.
In addition to examining comprehensibility and appreciation, our study aims to assess familiarity with and attitudes toward gender-neutral pronouns in Flanders and in the Netherlands. By comparing these findings to earlier measures of familiarity and attitudes (Decock et al., Reference Decock, Van Hoof, Soens and Verhaegen2024), we can trace the evolution of gender-inclusive pronominal reform and observe changing attitudes toward this reform over time.
Dutch masculine generic pronouns
Research into whether masculine pronouns in Dutch, interpretable as both specific or generic, cause a male bias has yielded mixed results (Redl, Reference Redl2021; Redl et al., Reference Redl, Eerland and Sanders2018, Reference Redl, Frank, Swart and Hoop2021, Reference Redl, Szuba, De Swart, Frank and De Hoop2022). Masculine pronouns can induce male bias in specific circumstances, contingent on various contextual factors, including the genericity (episodic vs. truly generic statements), the stereotype context (male vs. female vs. neutral), the gender of the participant (male vs. female), the conceptual number of the antecedent (singular vs. plural), the method used in the experiment (e.g., eye-tracking vs. sentence evaluation), and grammatical case of the masculine pronoun (personal vs. possessive).
In a first eye-tracking experiment by Redl et al. (Reference Redl, Eerland and Sanders2018), participants read episodic sentences followed by a clause containing the name of an individual group member, thereby identifying them as either female or male (e.g., Iedereen was zijn tanden aan het poetsen. Zo was ook Daphne/Robert zich aan het klaarmaken om naar bed te gaan. “Everyone was brushing his teeth. Daphne/Robert, too, was getting ready to go to bed.”) No evidence of a male bias was found. In a conceptual replication of this experiment, Redl et al. (Reference Redl, Frank, Swart and Hoop2021) did find zijn (“his”) to cause male bias, however, only for male participants and in stereotypically neutral contexts. Note that the same stimuli did not reveal male bias when used in a sentence evaluation task, in which participants had to evaluate whether the continuation of the sentence, which identifies part of the group as female or male, was seen as a good continuation of the first part of the sentence (Redl et al., Reference Redl, Frank, Swart and Hoop2021).
In another eye-tracking study, zijn (“his”) was embedded in truly generic contexts (e.g., Iemand/Iedereen met een absoluut gehoor kan snel zijn instrument stemmen. “Someone/Everyone with perfect pitch can tune his instrument quickly.”) (Redl et al., Reference Redl, Szuba, De Swart, Frank and De Hoop2022). Generic statements differ from episodic ones in that they do not refer to a specific situation at a specific time but generalize across situations. The results show that the pronoun zijn causes male bias only among male participants and only when the antecedent is conceptually singular (i.e., iemand “someone” vs. iedereen “everyone”).
Employing a self-paced reading paradigm, Redl (Reference Redl2021) tested the effect of hij (“he”) in generic sentences (e.g., Iemand die steeds belooft dat hij echt op tijd zal komen, zoals mevrouw/meneer Knoop, zal alsnog soms te laat zijn. “Someone who always promises that he will really be on time, such as Ms/Mr Knoop, will sometimes be late anyway.”) The study confirmed the expectation that hij causes a male bias among both male and female participants.
In conclusion, a male bias was found for subject form hij and possessive form zijn, though the latter was limited to male participants. Taken together, the results of these studies led Redl (Reference Redl2021) to conclude that Dutch masculine pronouns, even when intended as generic, are not always readily processed as gender-neutral.
Research on gender-neutral pronouns
English singular they and Swedish hen are currently the most well-known examples of gender-neutral pronouns (Gustafsson Sendén et al., Reference Gustafsson Sendén, Renström and Lindqvist2021; Hekanaho, Reference Hekanaho2020). These pronouns have officially been codified and are well-established within their respective language communities (Hekanaho, Reference Hekanaho2020; Renström et al., Reference Renström, Lindqvist and Gustafsson Sendén2022b). Furthermore, they have been shown to be easily processed (e.g., Foertsch & Gernsbacher, Reference Foertsch and Gernsbacher1997; Noll et al., Reference Noll, Lowry and Bryant2018; Speyer & Schleef, Reference Speyer and Schleef2019).
English
The first results of studies on the perception of singular they suggest that it may indeed be an inclusive alternative for generic reference. For English, in two self-paced reading experiments, Foertsch and Gernsbacher (Reference Foertsch and Gernsbacher1997) found that singular they was read at a similar pace and was thus processed as easily as the pair form he or she. Speyer and Schleef (Reference Speyer and Schleef2019) successfully replicated these results for advanced non-native speakers and showed that singular they did not lead to processing costs compared to gendered pronouns.
In two lexical decision task experiments conducted 15 years apart, Noll et al. (Reference Noll, Lowry and Bryant2018) documented a change in the interpretation of singular they and he as generic by comparing the processing of masculine and feminine words after seeing either he or they. The results suggest that the rise of generic singular they in more recent years at the expense of he as generic may have caused the latter to no longer function as truly generic anymore.
Furthermore, comparisons of generic and non-binary uses of they in several studies have shown that the former is more comprehensible and more acceptable than the latter (Bradley et al., Reference Bradley, Salkind, Moore and Teitsort2019; Conrod, Reference Conrod2019; Hekanaho, Reference Hekanaho2020). Bradley et al. (Reference Bradley, Salkind, Moore and Teitsort2019), for example, found that they is the preferred singular pronoun for a generic antecedent or hypothetical person in English. However, singular they is perceived as less grammatical and less acceptable when referring to a specific antecedent, especially when the person is named.
Swedish
Since approximately 2012, the Swedish language has adopted a neopronoun, hen, which is used for both specific and generic reference (Renström et al., Reference Renström, Lindqvist and Gustafsson Sendén2022b). The pronoun complements han (he) and hon (she) and is modelled after Finnish hän, a genderless pronoun. Since its introduction in 2012, the usage and appreciation of hen have quickly risen. Gustafsson Sendén et al. (Reference Gustafsson Sendén, Renström and Lindqvist2021) showed that even though the introduction of the pronoun was met with hostility and resistance, attitudes toward hen became more positive between 2015 and 2018, and the use of hen increased by 14%, to about half of the Swedish population actively using hen in 2018. Using eye-tracking, Vergoossen et al. (Reference Vergoossen, Pärnamets, Renström and Gustafsson Sendén2020) found that specific hen took slightly longer to process than gendered pronouns. However, they found no evidence that hen might compromise reading comprehension.
Similar to English singular they, attitudes toward the specific meaning of hen were found to be more negative compared to the generic meaning (Renström et al., Reference Renström, Lindqvist and Gustafsson Sendén2022b).
Dutch
As Dutch gender-neutral pronouns are relatively new, academic research on the topic is scarce and general awareness about gender-neutral pronouns is not as high for Dutch as it is for English and Swedish. Moreover, a number of potential gender-neutral pronouns are currently in use, notably hen and die (Vriesendorp, Reference Vriesendorp2024). Examining non-binary pronouns used by inclusivity-oriented NL Dutch speakers, Vriesendorp (Reference Vriesendorp2024) found that die was predominantly used as a subject, hen as an object, and both diens and hun as possessive pronouns. In their choice of non-binary pronouns, the respondents followed a pattern where they stayed relatively close to existing usage: most respondents used pronouns in syntactic functions consistent with their usage in contexts outside of non-binary reference. Die can occur in subject position as a personal pronoun in Dutch, while hen never occurs as a subject (Algemene Nederlandse Spraakkunst: Colleman et al., Reference Colleman, Caluwe, Haeseryn, Boogaart, Landsbergen and Hoorde2021: § 5.6.3.3.1). Vos and Nutters (Reference Vos and Nutters2022) compared the reading experiences of NL Dutch participants reading translations of an excerpt of Bernardine Evaristo’s novel Girl, Woman, Other, featuring a non-binary character. The translations used either hen or die as the gender-neutral pronoun to refer to that character. The study revealed no clear preference for either pronoun in literary texts. Consequently, Vos and Nutters (Reference Vos and Nutters2022) tentatively concluded that there is no distinct advantage to or preference for die or hen, and that both were equally suitable for translation of English non-binary they. Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) focused on both die and hen in BE Dutch for specific non-binary reference and compared the effect of these gender-neutral pronouns on perceived text quality to that of alternative referential strategies (i.e., binary pronouns and pronoun avoidance). Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) found no particular detrimental effect of either die or hen on text comprehensibility. However, although die showed no effect on text appreciation, hen significantly lowered text appreciation compared to binary pronouns. This effect was mediated by perceived awkwardness, which is defined as the extent to which the referential strategy is considered well-chosen, appropriate, irritating, strange, and/or surprising. Perceived awkwardness can be considered as an explanation of the negative effect of hen on text appreciation. The effect of referential strategy on perceived text quality mediated through perceived awkwardness was moderated by cueing, which made readers aware that the person reported identified as non-binary and preferred non-binary pronouns. More specifically, when cueing was absent, hen-hen-hun (subject-object-possessive) had a significant negative effect on perceived text quality through perceived awkwardness. When cueing was present, these indirect effects were not significant, leading the authors to conclude that cueing is a successful strategy to guide readers’ interpretation of non-binary pronouns.
In light of these findings and the findings on participants’ attitudes toward the different gender-neutral pronouns, which showed less aversion to die than hen, Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) concluded that die shows most potential to become accepted in Flanders as a Dutch gender-neutral pronoun. Furthermore, in a sentence evaluation experiment among BE Dutch speakers, Decock et al. (Reference Decock, Van Hoof, Vanrobaeys, Verhaegen and Vincent2025) showed that die was better appreciated for generic reference than for non-binary reference.
The present study
Overview of experiments and hypotheses
In this study, we examine the effect of Dutch gender-neutral pronouns used for generic reference on perceived text quality by means of two experiments conducted in Flanders and the Netherlands. We conceptualize perceived text quality as comprising two essential dimensions: Text Comprehensibility and Text Appreciation. Comprehensibility is one of the central criteria for evaluating gender-inclusive language forms and can be defined as the ease with which one can engage in mindful reading of a text (Friedrich et al., Reference Friedrich, Drößler, Oberlehberg and Heise2021). Appreciation refers to the aesthetic appeal of a text and the reading flow. The referential strategies under investigation are (1) gender-neutral pronouns in three different combinations, i.e., hen-hen-hun, die-hen-hun, and die-die-diens (subject-object-possessive), (2) slash forms combining a feminine, masculine, and gender-neutral pronoun, i.e., zij/hij/die (subject), haar/hem/hen (object), and haar/zijn/hun (possessive); (3) masculine forms, i.e., hij-hem-zijn (subject-object-possessive); and (4) plural forms, i.e., ze-hen-hun (subject-object-possessive). We based our selection of referential strategies on the advice by the Dutch Language Union, which suggests using gender-neutral pronouns, plural forms, or a combination of several pronouns as an alternative to masculine generics. We embedded these referential strategies in an instance of a text genre in which generic reference is common: employee guidelines.
Two between-subjects online survey experiments were conducted, in which participants assessed an excerpt from fictional employee guidelines on remote work. The first experiment tested the effect of Referential Strategy and Cueing (i.e., making readers aware of the fact that the pronouns used in the text are meant to be interpreted as gender-inclusive) on perceived text quality in a sample of speakers of BE Dutch. The second one compared the effect of Referential Strategy on perceived text quality in NL Dutch and BE Dutch.
For the first experiment, we focus on gender-neutral combinations with die in subject position. Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) concluded that die-hen-hun showed more potential to be accepted as non-binary pronoun by Belgian participants, based on the finding that hen-hen-hun significantly lowered Text Appreciation in comparison to binary pronouns and resulted in significantly higher levels of Perceived Awkwardness compared to die-hen-hun. Die-hen-hun, on the other hand, did not differ significantly from binary pronouns or repetition of name and profession in terms of Text Comprehensibility and Text Appreciation. In addition, when tapping directly into respondents’ attitudes toward hen and die, Decock et al. noticed a preference for die compared to hen. However, Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) did not test die-die-diens, in spite of the fact that the Dutch Language Union also suggests this combination for generic reference. The Language Union states that a possible disadvantage of using a gender-neutral pronoun could be that different genders are not made visible, which may unintentionally lead to male bias. To address this, the Language Union recommends combining masculine, feminine, and non-binary pronouns. Although this approach may be seen as disadvantageous due to the increased length of the reference form, we included slash forms in our experimental design to align with the Language Union’s recommendation.
For the second experiment, we focus on the effects of gender-neutral pronouns on perceived text quality, comparing a BE Dutch and an NL Dutch speaking sample. The stimulus text from Experiment 1 was adapted to appear supraregional (i.e., the text contained no words specific to NL or BE Dutch) and we only manipulated the referential strategies, i.e., the versions with Cueing were deleted from the design. The second experiment thus contained five versions of the guidelines, only differing in the Referential Strategy that was used. We included hen-hen-hun in the design, rather than slash forms, as studies on the comprehensibility and appreciation of gender-neutral pronouns in the context of Netherlandic Dutch have not yet been conducted. Vriesendorp (Reference Vriesendorp2024) found that, even though die was the more popular choice, in 25% of cases, Dutch participants used hen to refer to a non-binary person in subject position. Note that language users might also occasionally opt for hen for generic reference.
Hypothesis building
No research has yet been conducted on the effect of generically used Dutch gender-neutral pronouns on perceived text quality. Therefore, we formulate the first hypothesis as follows:
(H1) Perceived text quality is affected by the choice of Referential Strategy: gender-neutral pronouns, masculine pronouns, and plural pronouns affect Text Comprehensibility and Text Appreciation differently.
Since different gender-neutral pronoun combinations are available to language users, but little is known about their use or perception, we also want to explore whether there are differences among the gender-neutral pronoun combinations to see which combination has the most potential to be accepted for generic reference.
In their research on Dutch gender-neutral pronouns for specific non-binary reference, Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) found that the Perceived Awkwardness of the gender-neutral pronouns explains the negative effect of hen on Text Appreciation. This study sets out to test whether the Perceived Awkwardness of gender-neutral pronouns plays a similar mediating role for generic reference. This leads to our second hypothesis:
(H2) Perceived Awkwardness of the gender-neutral pronouns in the text mediates the effect of Referential Strategy on perceived text quality.
Research on Dutch, English, and Swedish has shown that people’s awareness of gender-neutral pronouns influences their perception of such pronouns. For example, the use of Cueing has been found to reduce Perceived Awkwardness while reading (Decock et al., Reference Decock, Van Hoof, Soens and Verhaegen2024), facilitate the singular interpretation of they (Arnold et al., Reference Arnold, Mayo and Dong2021), and lead to more accurate representations of non-binary gender (Renström et al., Reference Renström, Lindqvist, Klysing and Gustafsson Sendén2023). Since Cueing provides additional contextual information, it is expected to help people in their interpretation of the Referential Strategy used in the text, which in the case of gender-neutral pronouns may be unfamiliar or awkward to them. A lack of familiarity with a word can reasonably result in difficulties in its interpretation (Renström, Reference Renström2025). Since Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) found that half of their sample was not familiar with gender-neutral pronouns, including Cueing conditions in the experimental design constitutes an interesting avenue for research. This leads us to formulate a third hypothesis as follows:
(H3) When Cueing is used (vs. not used), the effect of gender-neutral pronouns on perceived text quality, mediated through Perceived Awkwardness, will be more positive.
Figure 1 brings these hypotheses together and shows the hypothesized moderated mediation model. Table 2 gives an overview of the referential strategies and hypotheses that were tested in each experiment.
Figure 1. The hypothesized moderated mediation model (figure taken from Decock et al. Reference Decock, Van Hoof, Soens and Verhaegen2024).
Table 2. Overview of referential strategies and hypotheses tested per experiment
Experiment 1: gender-neutral pronouns and Cueing
Methods
Participants
Market research agency Bilendi provided us with 500 complete data sets from Dutch-speaking Belgian participants who completed the survey experiment online. Individuals who did not provide consent, those for whom Dutch was not their mother tongue, those under the age of 18, those who did not complete the survey, and those who did not pass the attention checks had been excluded. The agency also provided weights to be used in data analysis to ensure that the sample was representative in terms of gender, age, and education. All participants were native Dutch speakers. The sample comprised 252 men, 247 women, and one participant who selected the option “other.” The latter category also provided an open text box, where they wrote “X.” The participants’ mean age was 49.41 (SD = 17.64, range = 18–86). With regard to educational attainment, 205 participants (41%) had attended higher education, while 295 (59%) had not. The descriptive analysis of the distribution of gender, age, and education across the different conditions is summarized in Table 3.
Table 3. Participant distribution Experiment 1
Materials
The referential strategies under investigation were embedded in employee guidelines. The referential strategies occurred five times in the subject form, twice in the object form, and four times in the possessive form. Example 3 provides an excerpt from the die-hen-hun condition. In this example, the gender-neutral pronouns are written in bold.
-
4. Daarbij krijgt het personeelslid de autonomie om binnen afgesproken grenzen zelf te bepalen hoe, waar en wanneer die hun werk organiseert. … Het personeelslid moet zich er ook van vergewissen dat de door hen gekozen locatie voldoet aan de nodige vereisten op vlak van connectiviteit.
-
The staff member is given the autonomy to decide how, where, and when they organize their work within agreed limits. … The staff member must also ensure that the location chosen by them meets the necessary connectivity requirements.
In the Cueing conditions, the text started with a sentence explaining that the Referential Strategy used in the guidelines was meant to be interpreted as gender-inclusive. We thus created ten versions of the guidelines, identical in content but differing in the Referential Strategy that was used and the presence or absence of Cueing. The full survey, including various stimulus texts, is available in the OSF repository.
Perceived text quality and Perceived Awkwardness
Perceived text quality was measured on two dimensions: Text Comprehensibility and Text Appreciation. Text Comprehensibility was measured by means of four constructs: (1) Recall, (2) Word Difficulty, (3) Sentence Difficulty, and (4) Subjective Comprehensibility. To measure Recall, six multiple-choice questions were created to assess participants’ memory of the information presented in the employee guidelines (cf. Pöschko & Prieler, Reference Pöschko and Prieler2018). Word Difficulty (the perceived ease with which readers can ascribe meaning to the words of a text), Sentence Difficulty (the perceived ease with which readers can analyze the syntax of the sentences of a text), and Subjective Comprehensibility were measured on a subscale of three items each. All items were based on the validated comprehensibility questionnaire of Friedrich (Reference Friedrich2017) and Friedrich and Heise (Reference Friedrich and Heise2019). Text Appreciation and the mediator Perceived Awkwardness were both measured on subscales of five items taken from Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024). The perceived text quality measures and the mediator Perceived Awkwardness were rated on 5-point Likert scales, ranging from I disagree (1) to I agree (5). With internal consistencies of Cronbach’s α over .85, all scales proved to be good. The full list of items is available in the Appendix on OSF.
Procedure
The online experiment was drafted using Qualtrics. In the introduction, participants were informed that they were taking part in research on the assessment of text quality in written professional communication and informed consent was obtained. Participants were asked to read an excerpt from employee guidelines on the topic of remote work. After reading the text, participants completed the Recall questions and assessed the statements regarding Text Comprehensibility and Text Appreciation. In the second half of the experiment, participants were asked about the study’s intent, the Perceived Awkwardness of the referential strategy used in the guidelines, their familiarity with gender-neutral pronouns, genderism (7 items measuring participants’ belief that gender is binary, adapted from van Ditzhuijzen & Motmans, Reference van Ditzhuijzen and Motmans2020, by Decock et al., Reference Decock, Van Hoof, Soens and Verhaegen2024), and they provided demographic information. Additionally, we directly asked participants whether they felt there is a need for gender-neutral pronouns in Dutch and if so, which pronouns they would deem suitable. The experiment concluded with a manipulation check. For this, participants had to indicate which Referential Strategy had been used in the guidelines to refer to employees throughout the text. Participants were presented with all options tested in the experiment and were also given the option of responding “I don’t remember.” A second manipulation check was conducted for Cueing: Participants were required to indicate whether the guidelines had explicitly stated that gender-inclusive language had been used (with the response options “yes,” “no” and “I don’t remember”). In both manipulation checks, respondents could only pick one response option.
Analysis
The data were prepared, transformed, and checked using SPSS (version 29) and analyzed using R (version 4.3.1) and RStudio (version 2024.12.0 + 467). The dependent variable Recall was a categorical outcome (right vs. wrong answers to the content questions) and was examined by fitting a generalized linear mixed-effects model using the glmer function from the lme4 package (Bates et al., Reference Bates, Mächler, Bolker and Walker2015). The dependent variables Word Difficulty, Sentence Difficulty, Subjective Comprehensibility, Text Appreciation, and the mediator Perceived Awkwardness were ordinal outcomes (5-point Likert scale ratings) and were examined by fitting ordinal mixed-effects models using the clmm function from the ordinal package (Christensen, Reference Christensen2023). The models included covariates age (standardized), education level (no higher education vs. higher education), gender (male vs. non-male)Footnote 2 , familiarity with gender-neutral pronouns (yes vs. no), and genderism (standardized). All binary categorical control variables were dummy coded using R’s default treatment coding, with the first level as the reference category: “male” (gender), “no higher education” (education), and “familiar with gender-neutral pronouns” (familiarity). Random intercepts were included for both participants and items, where items referred to the individual questions (in the Recall task) or Likert-scale statements (in the Word Difficulty, Sentence Difficulty, Subjective Comprehensibility, Text Appreciation, and Perceived Awkwardness measures). The independent variables of interest were Referential Strategy (die-die-diens, die-hen-hun, slash forms, masculine forms, plural forms) and Cueing (absent vs. present), and their interaction. The categorical predictors Cueing and Referential Strategy were effect-coded using sum-to-zero contrasts (contr.sum in R), such that each level’s effect is interpreted relative to the grand mean. P-values were computed using the Anova function from the car package (Fox & Weisberg, Reference Fox and Weisberg2019). Post-hoc tests were performed using the emmeans package (Lenth, Reference Lenth2025), using Bonferroni-adjusted pairwise comparisons to control for multiple comparisons. The results section reports all significant main and interaction effects; full model estimates for all predictors are presented in Table 4. Standardized beta coefficients were computed using the parameters package (Lüdecke et al., Reference Lüdecke, Ben-Shachar, Patil and Makowski2020) and are reported in Table 5. To evaluate sensitivity, a post-hoc power analysis was conducted using the simr package (Green & MacLeod, Reference Green and MacLeod2016).
Table 4. Analysis of deviance table (Type III tests) for outcome variables (Experiment 1)
Signif. codes: 0 ‘***’ .001 ‘**’ .01 ‘*’ .05 ‘.’ .1 ‘ ’ 1
Table 5. Effect sizes (Standardized Coefficients) of Referential Strategy, Cueing, and their interaction (Experiment 1)
Predictors were effect-coded (sum-to-zero contrasts), meaning that each coefficient represents the effect of a given level compared to the grand mean, not to a reference category. For predictors with more than two levels (e.g., Referential Strategy), only k-1 coefficients are estimated; the remaining level (plural forms) is implicitly represented as the negative sum of the others. For binary predictors (e.g., Cueing), the single coefficient reflects the deviation of one level from the grand mean, with the other level implied to have the opposite sign. Standardized beta coefficients are used to allow comparison across predictors.
Results
Effect of Referential Strategy and Cueing on perceived text quality
Manipulation checks
A chi-square test indicated a significant relationship between the Referential Strategy condition to which participants were assigned and the Referential Strategy they selected as the answer to the question, which Referential Strategy was used in the guidelines (χ 2 [20] = 185.748, p < .001). In the die-hen-hun condition, 21 of the 96 participants remembered the Referential Strategy correctly (21.9%). In the die-die-diens condition, 25 out of 100 participants remembered correctly (25%). In the masculine forms condition, 18 of the 95 participants remembered correctly (19%). In the slash forms condition, 41 of the 105 participants remembered correctly (39.1%). In the 3rd person plural form condition, 4 of the 106 participants remembered correctly (3.8%). These figures suggest that the slash forms combining feminine, masculine, and non-binary pronouns are the most salient.
Another chi-square test indicated a significant relationship between the Cueing condition to which participants were assigned and the Cueing strategy they indicated as the answer to the question whether the guidelines made explicit that gender-inclusive language was used (χ 2 [2] = 85.931, p < .001). In the condition without Cueing, 92 of the 257 participants remembered this correctly (35.8%). In the condition with Cueing, 142 of the 243 participants correctly indicated that Cueing was present (58.44%). Participants thus remembered better when they had read the sentence clarifying the gender-inclusive language strategy used in the guidelines, than when assigned to the condition without Cueing.
Analyses of variance
Referential Strategy did not have a significant effect on the dependent variables as main or interaction effect and Cueing only had a significant main effect on Word Difficulty (χ 2 [1] = 5.97, p = .014, all other p-values > .07). Pairwise comparisons of the estimated marginal means with Bonferroni correction indicated that Word Difficulty is significantly higher when Cueing is present versus absent (z = 2.361, p = .018).
The analysis revealed a statistically significant main effect of Referential Strategy on the mediator Perceived Awkwardness (χ 2 [4] = 10.42, p = .034). The effect of Cueing only approached significance (p = .09), but the interaction effect was significant (χ 2 [4] = 9.6, p = .048). Participants who had read the guidelines with die-hen-hun rated the text as significantly more awkward than those in the plural condition, but only when Cueing was present (z = 3.08, p = .020). This implies that, when Cueing is absent, Perceived Awkwardness remains more stable across referential strategies. Additionally, the post-hoc comparisons show that die-hen-hun is perceived as significantly more awkward when Cueing is present than when Cueing is absent (z = 2.57, p = .010). This means that for die-hen-hun, Cueing “activates” the Perceived Awkwardness of the pronoun combination, whereas for the other referential strategies, Cueing does not make a significant difference. This is visualized in Figure 2.
Figure 2. Adjusted means (mean.class) and standard errors of Perceived Awkwardness, grouped per Referential Strategy and colour-coded on the presence or absence of Cueing.
Since we found no main effect of Referential Strategy on perceived text quality, we did not proceed with the mediation analysis and the moderated mediation analysis.
Familiarity with and attitude toward gender-neutral pronouns
In addition to the perceived text quality experiment, our questionnaire also assessed familiarity with and attitudes toward gender-neutral pronouns. The majority of the participants (63.4%) indicated being familiar with gender-neutral pronouns in Dutch; of these, 42.4% were familiar with die-die-diens, 41.1% with die-hen-hun, and 30% with hen-hen-hun (participants could indicate that they were familiar with more than one combination). Of those who are familiar with gender-neutral pronouns, 12.5% indicated they have already used such pronouns themselves.
Among all participants, a slight majority (54%) indicated that they do not see the need for a gender-neutral pronoun for specific non-binary reference, 22% expressed a neutral stance, and 24% agreed that such a pronoun is needed. To allow comparison with the Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024), we calculated the mean agreement score on the 7-point Likert scale. The mean in our study (M = 3.13) is lower than that reported by Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024; M = 3.53), indicating a slightly stronger tendency toward disagreement that a non-binary pronoun is necessary in 2023 compared to 2022. For generic reference, 45% of respondents did not see the need for a gender-neutral pronoun, 24% are neutral, and 30% agreed that a gender-neutral pronoun is needed. Overall, the majority of our participants disagreed that the Dutch language needs gender-neutral pronouns, but the need for a gender-neutral generic pronoun was rated slightly higher than that for a non-binary pronoun (see Figure 3).
Figure 3. Likert scale responses (1 = helemaal niet akkoord, “I totally disagree”; 7 = helemaal akkoord, “I totally agree”) regarding the need for gender-neutral pronouns in Experiment 1.
Participants who agreed a gender-neutral pronoun was needed were asked to rate the combinations die-die-diens, die-hen-hun, and hen-hen-hun on a 7-point Likert scale for suitability. Die-die-diens was rated most suitable, with 66% of participants agreeing to its use for specific non-binary reference, and 63% agreeing with die-die-diens for generic reference. Ratings of die-hen-hun and hen-hen-hun were also more positive than negative (see Figures 4 and 5).
Figure 4. Likert scale responses (1 = helemaal niet akkoord, “I totally disagree”; 7 = helemaal akkoord, “I totally agree”) for the suitability of die-die-diens, hen-hen-hun, and die-hen-hun for non-binary use in Experiment 1.
Figure 5. Likert scale responses (1 = helemaal niet akkoord, “I totally disagree”; 7 = helemaal akkoord, “I totally agree”) for the suitability of die-die-diens, hen-hen-hun and die-hen-hun for generic use in Experiment 1.
Discussion
We had hypothesized that the referential strategies would have different effects on perceived text quality. However, we did not find evidence to support this hypothesis, since there were no significant effects on the dependent variables. For this reason, we did not continue with the mediation (H2) and moderated mediation (H3) analyses.
There was a significant main effect of Cueing on Word Difficulty, but not on the other outcome measures. Neither Referential Strategy nor the interaction between Cueing and Referential Strategy significantly influenced Word Difficulty. This suggests that the effect of Cueing on Word Difficulty operates independently of pronoun choice and is therefore not directly related to our hypotheses about gender-neutral pronouns and their impact on perceived text quality.
There was a significant effect of Referential Strategy on the mediator Perceived Awkwardness, and a significant interaction effect between Referential Strategy and Cueing. We found that die-hen-hun heightens Perceived Awkwardness in comparison to plural forms and that this is activated by Cueing. This challenges our reasoning behind Hypothesis 3 that Cueing would mitigate the awkwardness of the fairly unknown gender-neutral pronouns and thus facilitate Text Comprehensibility and Appreciation. The results suggest that Cueing had the opposite effect.
The majority of participants indicated prior familiarity with Dutch gender-neutral pronouns, but a minority agreed that there is an actual need for such pronouns. Among those in favor of gender-neutral pronouns, die-die-diens seems to be the preferred pronoun combination. As both Decock et al.’s study and the current Experiment 1 were conducted among a representative sample of Flemings, this measure allows us to map the evolution of familiarity with gender-neutral pronouns in BE Dutch. Overall, familiarity rose from 53.7% to 63.4% between January 2022 (Decock et al., Reference Decock, Van Hoof, Soens and Verhaegen2024) and January 2023. However, despite the increase in familiarity, attitudes toward gender-neutral pronouns remain negative. A majority of participants in Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) and our study express disagreement with the necessity for a specific non-binary pronoun, and in our study, this disagreement was slightly more pronounced than in Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024). Participants were slightly more open to a gender-neutral pronoun for generic reference than for non-binary reference, suggesting that resistance may be stronger when pronouns are explicitly linked to non-binary identities rather than used more broadly.
Based on the findings by Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) and advice by the Dutch Language Union, this experiment only focused on gender-neutral referential strategies with die in subject position among a BE Dutch readership. Whereas Decock et al. compared two referential strategies die-hen-hun and hen-hen-hun, we tested die-hen-hun, die-die-diens, and slash forms combining feminine, masculine, and gender-neutral pronouns. The absence of a negative impact of the gender-neutral pronoun combinations on perceived text quality in our study is in line with the results of Decock et al., who only found a negative impact on Text Appreciation for the combination hen-hen-hun compared to masculine forms, and none for die-hen-hun. For our second experiment, in which we compare BE Dutch and NL Dutch, we deemed it important to again include the combination hen-hen-hun, as there are currently no studies on the comprehensibility and appreciation of gender-neutral pronouns in the Netherlandic Dutch context.
Experiment 2: gender-neutral pronouns in Flanders and the Netherlands
Methods
Participants
The survey experiment was distributed through Prolific and Facebook. Participants recruited via Prolific were selected based on their location (either the Netherlands or Belgium) and their fluency in Dutch. A total of 250 Dutch participants completed the survey through Prolific, and 258 Belgian participants were recruited through a combination of Prolific and convenience sampling via the Facebook network of the first author. Among the Belgian sample, two participants had to be excluded because they failed to correctly respond to two attention checks. From the 250 participants from the Netherlands, 235 were native speakers of Dutch, as were 235 out of 256 participants from Belgium. The Belgian sample comprised 97 men, 152 women, and 7 non-binary people. Their mean age was 28.25 (SD = 8.49, range = 18–65). The Dutch sample comprised 140 men, 104 women, and 6 non-binary people. Their mean age was 30.24 (SD = 9.36, range = 18–71). The majority of participants from both the Belgian and Dutch samples had a high level of education: 231 Belgian participants (90%) and 214 Dutch participants (85.6%) attended higher education. Table 6 summarizes the descriptive analysis of gender, age, and education distribution across the different conditions.
Table 6. Participant distribution Experiment 2
Materials and procedure
The same materials as in Experiment 1 were used to measure perceived text quality and Perceived Awkwardness. The stimuli and procedure were similar, except that there were no Cueing conditions. Cueing was excluded from the experiment as it was not identified as a moderator of perceived text quality in Experiment 1. Consequently, this experiment tests Hypothesis 1 and Hypothesis 2 (cf. hypothesis building section).
The stimulus text was adapted to appear supraregional, i.e., it contained no words specific to NL or BE Dutch. Internal consistency was sufficient (Cronbach’s α > .74) for the scales Word Difficulty, Sentence Difficulty, Subjective Comprehensibility, Text Appreciation, and Perceived Awkwardness.
To measure the perceived suitability of different gender-neutral forms, we did not present participants with set combinations of pronouns (i.e., comparing die-hen-hun vs. hen-hen-hun vs. die-die-diens) as in Experiment 1 but let participants rate each of these forms separately in their respective grammatical conditions (clarified by an example sentence). The full surveys, including various stimulus texts, can be found in the OSF repository.
Analysis
Data preparation and coding procedures were identical to those used in Experiment 1. The analysis followed the same modeling approach, with the exception that Experiment 2 did not include any Cueing. Instead, the Sample (Belgian vs. the Netherlands) and its interaction with Referential Strategy (die-die-diens, die-hen-hun, hen-hen-hun, masculine forms, plural forms) were included in the models. The mediation analysis was performed using Hayes’ (Reference Hayes2022) PROCESS macro (Model 4; 5.000 bootstraps; 95% confidence intervals [CIs]). Genderism, age, education, familiarity, and participant gender were included as control variables, and the mean rating of Perceived Awkwardness was included as mediator. The results section reports all significant main and interaction effects; full model estimates for all predictors are provided in Table 7. Table 8 presents the standardized coefficients of Referential Strategy, Cueing, and their interaction.
Table 7. Analysis of deviance table (Type III tests) for outcome variables (Experiment 2)
Signif. codes: 0 ‘***’ .001 ‘**’ .01 ‘*’ .05 ‘.’ .1 ‘ ’ 1
Table 8. Effect sizes (Standardized Coefficients) of Referential Strategy, Cueing, and their interaction (Experiment 2)
Predictors were effect-coded (sum-to-zero contrasts), meaning that each coefficient represents the effect of a given level compared to the grand mean, not to a reference category. For predictors with more than two levels (e.g., Referential Strategy), only k-1 coefficients are estimated; the remaining level (plural forms) is implicitly represented as the negative sum of the others. For binary predictors (e.g., Cueing), the single coefficient reflects the deviation of one level from the grand mean, with the other level implied to have the opposite sign. Standardized beta coefficients are used to allow comparison across predictors.
Results
Effect of Referential Strategy and Sample on perceived text quality
Manipulation check
A chi-square test indicated a significant relationship between the Referential Strategy condition to which participants were assigned and the Referential Strategy they selected as the answer to the question, which Referential Strategy they remembered reading in the guidelines (χ 2 [20] = 283.4, p < .001). Among the Belgian participants, 19 of the 51 participants (37.3%) correctly remembered being in the masculine forms condition. In the die-hen-hun condition, 17 of the 49 participants remembered the Referential Strategy correctly (34.7%). In the hen-hen-hun condition, 18 of the 51 participants remembered correctly (35.3%). In the die-die-diens condition, 17 of the 51 participants remembered correctly (33.3%). In the 3rd person plural pronouns condition, 5 of the 54 participants remembered correctly (9.3%). These figures suggest that the 3rd person plural pronouns condition is the least salient. Among the Dutch participants, 12 of the 51 participants (23.5%) correctly remembered being in the masculine forms condition. In the die-hen-hun-condition, 8 of the 49 participants remembered the Referential Strategy correctly (16.3%). In the hen-hen-hun condition, 1 of the 51 participants remembered correctly (2%). In the die-die-diens condition, 23 of the 50 participants remembered correctly (46%). In the 3rd person plural pronouns condition, 10 of the 49 participants remembered correctly (20.4%). In the hen-hen-hun condition, 10 participants of the 51 indicated having read a text in the die-hen-hun condition (19.6%), while in the die-hen-hun condition, 15 participants of the 49 indicated having read a text in the hen-hen-hun condition (30.6%). This suggests that there is confusion between having seen die-hen-hun or hen-hen-hun.
Analyses of variance
There were no significant main effects of Referential Strategy on Text Comprehensibility, i.e., Recall, Word Difficulty, Sentence Difficulty, or Subjective Comprehensibility. However, Referential Strategy did have a significant effect on perceived Text Appreciation (χ 2 [4] = 13.040, p = .011). Pairwise comparisons of the estimated marginal means with Bonferroni correction indicated that die-die-diens was appreciated significantly less than masculine forms (z = 3.139, p = .017). More specifically, die-die-diens received a mean rating of 2.67 (SE = .257), while masculine forms were rated 3.04 (SE = .262).
Although no hypothesis was formulated regarding Sample, there was a significant main effect of Sample on Recall (χ 2 [1] = 4.840, p = .028), Word Difficulty (χ 2 [1] = 8.76, p = .003), and Subjective Comprehensibility (χ 2 [1] = 12.696, p < .001). More specifically, Recall was higher for Belgian participants (p = .027, OR = 1.39), with an estimated probability of 90.4% (SE = 3.02%) compared to 87.1% (SE = 3.75%) for Dutch participants. Subjective Comprehensibility was also higher for Belgian participants (z = 3.38, p < .001), with a mean rating of 4.06 (SE = .193) compared to 3.79 (SE = .213) for Dutch participants. In contrast, Word Difficulty was higher for Dutch participants (z = −2.76, p = .005), with a mean rating of 1.55 (SE = .174) compared to 1.35 (SE = .150) for Belgian participants. There were no significant main effects of Sample on Sentence Difficulty or Text Appreciation. That means that Belgian participants remembered the content of the text better and reported high comprehensibility, while Dutch participants reported that they struggled with some of the words.
There was a significant interaction effect between Referential Strategy and Sample on Subjective Comprehensibility (χ 2 [4] = 10.404, p = .034), but not on the other dependent variables. Pairwise comparisons of the estimated marginal means with Bonferroni correction showed that Dutch participants scored significantly lower on Subjective Comprehensibility than Belgian participants for the strategies die-hen-hun (z = 2.96, p = .003) and die-die-diens (z = 2.51, p = .12). More specifically, Dutch participants rated the text with die-hen-hun as 3.61 (SE = .277) compared to 4.16 (SE = .215) for Belgian participants. Similarly, die-die-diens received a mean rating of 3.57 (SE = 0.286) from Dutch participants, while Belgian participants gave it 4.05 (SE = 0.207). This is visualized in Figure 6.
Figure 6. Adjusted means (mean.class) and standard errors of Subjective Comprehensibility, grouped per Referential Strategy and color-coded by Belgian or Dutch sample.
We found a significant effect of Referential Strategy on Perceived Awkwardness (χ 2 [4] = 11.37, p = .023) that was qualified by a significant interaction effect between Referential Strategy and Sample (χ 2 [4] = 20.21, p <.001). Pairwise comparisons of the estimated marginal means with Bonferroni correction show that die-hen-hun (z = −2.78, p = .005) and die-die-diens (z = −2.60, p = .009) were perceived as more awkward in the Netherlands than in Belgium. Hen-hen-hun, conversely, was considered more awkward in Belgium than in the Netherlands (z = 2.47, p = .013). In the Belgian sample, hen-hen-hun was perceived as significantly more awkward than die-hen-hun (z = −3.34, p = .008), and plural forms (z = 3.20, p = .014). In the Dutch sample, die-die-diens was perceived as significantly more awkward than plural forms (z = 3.18, p = .015). This is visualized in Figure 7.
Figure 7. Adjusted means (mean.class) and standard errors of Subjective Comprehensibility, grouped per Referential Strategy and color-coded by Belgian or Dutch sample.
Mediation analysis
We ran Hayes’ (Reference Hayes2022) PROCESS macro (Model 4; 5.000 bootstraps; 95% confidence intervals [CIs]) to test whether Perceived Awkwardness functions as a mediator for the effect of Referential Strategy on Text Appreciation. Given that the independent variable (i.e., Referential Strategy) was a multicategorical variable with five categories, four dummy variables were created, as recommended by Hayes (Reference Hayes2022). Based on the results from the pairwise comparisons, which revealed that die-die-diens has a significant negative effect on Text Appreciation compared to masculine generics, the die-die-diens condition served as the baseline category. Therefore, X1 represents differences between the die-die-diens condition and plural forms, X2 represents differences between the die-die-diens condition and masculine generics, X3 represents differences between the die-die-diens condition and hen-hen-hun, and X4 represents differences between the die-die-diens condition and die-hen-hun. The results are summarized in Table 9.
Table 9. Results of mediation analysis (Experiment 2)
There were significant relative indirect effects of Referential Strategy on Text Appreciation through Perceived Awkwardness. More specifically, using die-die-diens resulted in significantly higher levels of Perceived Awkwardness compared to using plural forms (b X1 = .115, 95% CI [.036, .196) and die-hen-hun (b X4 = .100, 95% CI [.017, .185]). Perceived Awkwardness, in turn, significantly affected Text Appreciation (b = −.370, p < .001), indicating that higher Perceived Awkwardness was associated with lower Text Appreciation. However, the difference between die-die-diens and masculine generics was not explained by Perceived Awkwardness, as the indirect effect was not significant, suggesting that other mechanisms might be driving this effect.
Familiarity with and attitude toward gender-neutral pronouns
The majority of participants were familiar with Dutch gender-neutral pronouns (98.1% of Belgian participants and 91.6% of Dutch participants). Of the Belgian participants, 82% of respondents indicated prior familiarity with die-hen-hun, 51.1% with die-die-diens, and 53.5% had already heard of hen-hen-hun. Among the Dutch participants, 74% of respondents indicated prior familiarity with die-hen-hun, 50% with die-die-diens, and 61% had heard of hen-hen-hun. 36.7% of the Belgian sample and 26.8% of the Dutch sample had previously used gender-neutral pronouns themselves.
As illustrated in Figure 8, 56% of the Belgian participants agreed with the statement that the Dutch language needs a gender-neutral pronoun for specific reference, 21% expressed a neutral stance, and 20% disagreed. For generic reference, 71% agreed that such a pronoun is needed, 17% were neutral, and 12% disagreed.
Figure 8. Likert scale responses (1 = helemaal niet akkoord, “I totally disagree”; 7 = helemaal akkoord, “I totally agree”) for the need for gender-neutral pronouns among Belgian participants of Experiment 2.
As illustrated in Figure 9, 46% of the Dutch participants agreed with the statement that the Dutch language needs a gender-neutral pronoun for specific reference, 19% expressed a neutral stance, and 35% disagreed. For generic reference, 61% of respondents agreed that such a pronoun is needed, 18% were neutral, and 21% disagreed.
Figure 9. Likert scale responses (1 = helemaal niet akkoord, “I totally disagree”; 7 = helemaal akkoord, “I totally agree”) for the need for gender-neutral pronouns among Dutch participants of Experiment 2.
Irrespective of the reference type (generic vs. specific), Belgian and Dutch participants who are in favor of gender-neutral pronouns clearly prefer die in subject position (see Figures 10 and 11). Hen in subject position is rejected by the majority of participants. As for the object form, both die and hen receive positive ratings. Similarly, for the possessive form, both diens and hun are rated as suitable. After a preposition, die is clearly rejected and hen is deemed more suitable.
Figure 10. Likert scale responses (1 = helemaal niet akkoord, “I totally disagree”; 7 = helemaal akkoord, “I totally agree”) for the suitability of gender-neutral pronouns among Belgian participants of Experiment 2.
Figure 11. Likert scale responses (1 = helemaal niet akkoord, “I totally disagree”; 7 = helemaal akkoord, “I totally agree”) for the suitability of gender-neutral pronouns among Dutch participants of Experiment 2.
Discussion
We found partial support for Hypothesis 1 and Hypothesis 2, in that Referential Strategy had a significant effect on Text Appreciation, mediated through Perceived Awkwardness, but not on Text Comprehensibility. The pronoun combination die-die-diens was rated significantly lower on Text Appreciation than masculine forms.
We found a significant interaction effect between Referential Strategy and Sample on Subjective Comprehensibility. More specifically, Dutch participants scored lower on Subjective Comprehensibility than Belgian participants for both combinations with die in subject position. Overall, Subjective Comprehensibility was relatively high, suggesting that the use of die-hen-hun or die-die-diens did not constitute an insurmountable hurdle for Dutch participants. This is substantiated by the absence of effects on the other Text Comprehensibility measures.
Still, die-hen-hun and die-die-diens were perceived as more awkward in the Netherlands than in Belgium. Conversely, the combination hen-hen-hun was perceived as more awkward by Belgian participants. In the Belgian sample, hen-hen-hun was rated significantly more awkward than die-hen-hun and plural forms. The significant difference between hen-hen-hun and die-hen-hun suggests that hen in subject position activates some higher level of awkwardness, as the rest of the combinations are quite similar. In the Dutch sample, die-die-diens was rated as significantly more awkward than plural forms.
While no hypotheses were formulated regarding the effect of Sample on perceived text quality, the results suggest that participants from the Netherlands may have experienced lower levels of comprehensibility compared to participants from Belgium. The text was written by authors from Belgium and proofread by colleagues from the Netherlands in an effort to eliminate features (such as vocabulary) specific to BE Dutch. Despite these efforts, the text may still have been “too Flemish.” Some Dutch participants made this explicit in the comment box at the end of the survey, which might explain the sample differences regarding Recall, Word Difficulty, and Subjective Comprehensibility.
The familiarity, need, and suitability measures suggest similar trends for Belgian and Dutch participants. The vast majority of the participants were familiar with gender-neutral pronouns in Dutch, and the most well-known combination was die-hen-hun. More people agree that the Dutch language needs gender-neutral pronouns for generic reference than for non-binary reference. The suitability ratings of the pronouns suggest that combinations die-die-diens and die-hen-hun are accepted. Perhaps die-hen-diens or die-die-hun could also be used, since Vriesendorp (Reference Vriesendorp2024) has shown that people who use gender-neutral pronouns do not necessarily stick to forms from one grammatical system (i.e., die-die-diens vs. hen-hen-hun). Unacceptable forms are hen as a subject or die after a preposition, regardless of reference type.
General discussion
The aims of this study were twofold: to investigate the influence of Dutch gender-neutral pronouns used for generic reference on perceived text quality, and to map familiarity with and attitudes toward gender-neutral pronouns in Flanders and the Netherlands. In two survey experiments, we tested and compared six generic referential strategies: the gender-neutral combinations die-die-diens, die-hen-hun, and hen-hen-hun; slash forms (e.g., zij/hij/die); masculine pronouns; and plural pronouns. Experiment 1 additionally tested the effect of Cueing. For Experiment 1, a representative sample of Belgian participants was surveyed, and for Experiment 2, the survey was distributed through Prolific and social media in the Netherlands and Belgium to allow for a comparison between the two varieties of Dutch.
Key findings
Experiment 1 revealed no significant main effects of Referential Strategy on perceived text quality, providing no support for Hypothesis 1. Cueing only had a significant effect on Word Difficulty, but not on the other dependent variables, and the interaction effects were not significant. Therefore, we did not continue the mediation and moderated mediation analysis (H2 and H3) in Experiment 1. We did, however, find a significant interaction effect between Referential Strategy and Cueing on Perceived Awkwardness: die-hen-hun was perceived as significantly more awkward than plural forms, but only when Cueing was present. The majority of the representative Dutch-speaking Belgian sample was familiar with gender-neutral pronouns. However, the majority did not see the need for non-binary or generic gender-neutral pronouns. Among those who were in favor of gender-neutral pronouns, die-die-diens was the preferred pronoun combination for both non-binary and generic reference.
Experiment 2, which was conducted among a convenience sample of Belgian and Dutch participants, showed a significant effect of Referential Strategy on Text Appreciation, mediated through Perceived Awkwardness, but not on Text Comprehensibility (partial support for H1 and H2). The combination die-die-diens significantly lowered Text Appreciation compared to masculine forms. Although we had formulated no hypotheses regarding Sample, we did find a significant interaction effect between Referential Strategy and Sample on Subjective Comprehensibility, which showed that combinations with die in subject position resulted in higher Subjective Comprehensibility for Belgian than for Dutch participants. Die-die-diens must have been perceived as distinctive by Dutch participants: it was perceived as most awkward and was the most salient Referential Strategy among Dutch participants.
Almost all of the participants, who were on average relatively young and highly educated, were familiar with gender-neutral pronouns. The majority also acknowledged the need for such pronouns. Suitability ratings show that the combinations die-die-diens and die-hen-hun are accepted, but hen-hen-hun is rejected.
Perceived text quality
Overall, we found no evidence that Dutch gender-neutral pronouns used for generic reference negatively impact Text Comprehensibility. The impact of die-hen-hun or die-die-diens on Subjective Comprehensibility was found to be more negative for Dutch participants than Belgian participants, but the overall rating remained high and other Text Comprehensibility outcomes (Recall, Word Difficulty and Sentence Difficulty) were unaffected. We would therefore reject the notion that die-hen-hun or die-die-diens constitute a hurdle for perceived text quality by Dutch participants.
Our findings do indicate that die-die-diens may negatively influence Text Appreciation. Other gender-neutral pronoun combinations had no such effect. The finding that gender-neutral pronouns did not impact Text Comprehensibility, despite negatively affecting Text Appreciation, aligns with the results of Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) for non-binary reference. They found no main effect of Referential Strategy on Text Comprehensibility, but did find a significant negative effect of gender-neutral hen-hen-hun on Text Appreciation. In Experiment 1, the combination with subject hen was excluded from the design based on the conclusion by Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) that subject die showed the most potential to be successfully implemented as gender-neutral pronoun. This decision was consistent with the guidelines from the Dutch Language Union, which put forward subject die but not subject hen as an option for generic reference. In Experiment 2, however, we decided to try and replicate the negative effect of subject hen on Text Appreciation, as found by Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024), in the context of generic reference. Given the scarcity of research on gender-neutral pronouns in Dutch, we also expanded our sample to include participants from the Netherlands. The negative effect of subject hen on Text Appreciation was, however, not replicated. Yet, consistent with Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024), Belgian participants perceived hen-hen-hun as significantly more awkward than die-hen-hun.
We are surprised to see a negative impact of die-die-diens in Experiment 2 but not in Experiment 1. On average, the participants of Experiment 2 are younger, more highly educated, and more likely to be already familiar with gender-neutral pronouns in Dutch than in Experiment 1. Based on previous research, it could be expected that this sample would be more accepting of gender-neutral pronouns (Gustafsson Sendén et al., Reference Gustafsson Sendén, Renström and Lindqvist2021; Hekanaho, Reference Hekanaho2020; Vergoossen et al., Reference Vergoossen, Pärnamets, Renström and Gustafsson Sendén2020). Conversely, our study indicates that gender-neutral pronouns may still negatively impact Text Appreciation, even within a young, highly educated sample, most of whom were already familiar with gender-neutral pronouns. We reflect on this in the section “Familiarity and attitudes.”
Across both experiments, Referential Strategy mainly showed null or very small effects on perceived text quality. To assess whether the studies had sufficient power to detect such effects, we conducted post hoc power analyses for Recall using the simr package in R (Green & MacLeod, Reference Green and MacLeod2016). We found that at the sample size of Experiment 1 (N = 500), power to detect the largest observed effect of Referential Strategy (β = 0.26) was approximately 40% (95% CI [37.1%, 43.2%]). Similarly, in Experiment 2 (N = 506), power to detect the largest observed effect (β = −0.31) was 41.3% (95% CI [38.2%, 44.4%]). This suggests that some small effects may have gone undetected due to limited statistical power. Recommendations for increasing power in future studies are discussed in the “Limitations and directions for further research” section.
Perceived Awkwardness
In Experiment 2, the pronoun combination die-die-diens was rated significantly lower on Text Appreciation than masculine forms. However, this difference was not mediated by Perceived Awkwardness, whereas the mediation analysis identified significant indirect effects for other referential strategies. The findings suggest that while Perceived Awkwardness is a key mediator—whereby lower awkwardness is associated with higher appreciation—additional cognitive or linguistic mechanisms may underlie the effects of certain referential strategies. Future research should further investigate these underlying processes and examine potential mediating factors (see section Limitations and directions for further research).
Cueing
In the study of Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024), the negative effect of hen-hen-hun on Text Appreciation was alleviated when Cueing was present in the text. By incorporating a statement clarifying the gender-inclusive interpretation of the Referential Strategy used in the guidelines (Experiment 1), we also expected the perception of text quality to be better for gender-neutral pronoun conditions compared to when Cueing was absent (H3). We cannot draw conclusions regarding Hypothesis 3, because the requirements for our hypothesized moderated mediation analysis were not met. The significant interaction effect of Referential Strategy and Cueing on Perceived Awkwardness, however, suggested that Cueing can heighten the Perceived Awkwardness of the gender-neutral referential strategies. More specifically, this was the case for die-hen-hun. This observation contradicts the conclusion by Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) and our reasoning behind Hypothesis 3, namely that Cueing can be an effective strategy for guiding readers’ interpretation of gender-neutral pronouns. However, it is important to note that the current study focuses on gender-neutral pronouns used for generic reference, whereas Decock et al.’s study was concerned with specific non-binary reference. It is plausible that our participants formed a plural mental representation because the generic use of “employee” and gender-neutral pronouns represents a workforce consisting of multiple people. The object form hen and possessive form hun are traditionally known as plural forms in Dutch, so they might go unnoticed as gender-neutral pronouns in this generic context because they fit the mental representation of multiple employees. Since Cueing serves as a reminder of the gender-inclusive meaning of gender-neutral pronouns, and gender-neutral pronouns are known to challenge an individual’s personal belief that gender is binary (Renström et al., Reference Renström, Lindqvist and Gustafsson Sendén2022b; Hekanaho, Reference Hekanaho2020), it is conceivable that Cueing might spark Perceived Awkwardness.
Familiarity and attitudes
In the period between the study by Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) and our first experiment, a marked increase in familiarity with gender-neutral pronouns was observed among Belgian participants (from 53.7% to 63.4%). However, the attitudes of the participants in our first experiment were rather negative and even went down slightly compared to Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024): The majority of participants did not see the need for gender-neutral pronouns (non-binary or generic) in Dutch or expressed a neutral stance. The sample of Experiment 2, which was on average younger and more highly educated than the representative Belgian sample from Experiment 1, was highly familiar with gender-neutral pronouns. The high level of familiarity translated into more positive—though not overwhelmingly positive—attitudes toward gender-neutral pronouns. Just over half of the participants agree that a non-binary pronoun is needed in Dutch. A slightly bigger majority agrees that a gender-neutral pronoun for generic reference is needed. This pattern mirrors existing literature on other languages, which has shown that gender-neutral pronouns are better accepted for generic use than for specific reference (e.g., Renström et al., Reference Renström, Lindqvist and Gustafsson Sendén2022b; Bradley et al., Reference Bradley, Salkind, Moore and Teitsort2019; Decock et al., Reference Decock, Van Hoof, Vanrobaeys, Verhaegen and Vincent2025).
Participants who are in favor of gender-neutral pronouns express a preference for the combination die-die-diens. Die-hen-hun is also acceptable, but hen-hen-hun is clearly rejected. These results are in line with the attitude measurement by Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) and insights from production data by Vriesendorp (Reference Vriesendorp2024). Taken together, it seems that we can conclude that hen has little potential to be accepted as gender-neutral subject pronoun. Taking into account the results from the statistical analyses, the combination die-hen-hun seems to show the most potential, given that die-die-diens impaired Perceived Awkwardness and Subjective Comprehensibility among the Dutch participants and lowered Text Appreciation for all.
The relationship between familiarity with and attitudes toward gender-neutral pronouns appears to be complex. The negative effect of gender-neutral pronouns on Text Appreciation in Experiment 2 may seem surprising, particularly given the characteristics of the sample, which was on average younger, highly educated, and highly familiar with gender-neutral pronouns. However, a close examination of familiarity and attitude rates suggests that acceptance may not increase linearly with exposure. This finding is at odds with research on Swedish hen that found that attitudes toward hen quickly shifted from majority-negative to majority-positive in a few years (Gustafsson Sendén et al., Reference Gustafsson Sendén, Renström and Lindqvist2021). Concurrently, Gustafsson Sendén et al. (Reference Gustafsson Sendén, Renström and Lindqvist2021) noted that the introduction of gender-neutral pronouns can trigger strong initial backlash. This suggests that early exposure in a charged context can produce resistance, rather than simple habituation. Ideological factors, such as a strong binary gender identity, appear to drive much of this resistance (Bailey et al., Reference Bailey, Dembroff, Wodak, Ikizer and Cimpian2024). Longitudinal research into the familiarity with and attitudes toward Dutch gender-neutral pronouns will be paramount, as will be discussed in the following section.
Limitations and directions for further research
A strength of our first experiment is that it was conducted among a representative BE Dutch-speaking sample. However, the second experiment is non-representative and was conducted using Prolific and convenience sampling, leaving us with a sample that was imbalanced in terms of gender, age, education level, and familiarity with gender-neutral pronouns. It is imperative to exercise caution when interpreting the descriptive results regarding familiarity and attitudes. These results should not be carelessly generalized to the broader Belgian or Dutch populations.
A few Dutch participants remarked that the stimulus text sounded somewhat Flemish. This may have had an effect on the results obtained, and it may help to explain the main effects of Sample on Recall, Word Difficulty, and Subjective Comprehensibility. We hereby call for further research in the Netherlands to be conducted, with optimized materials and among representative samples.
Since we were unable to conduct the moderated mediation analysis and Cueing did not establish itself as a moderator in Experiment 1, we excluded Cueing from the design of Experiment 2. Future research on the generic use of Dutch gender-neutral pronouns may want to include Cueing in the design. Based on its effect on Perceived Awkwardness, we emphasize the importance of recognizing the potentially different effects of Cueing in different contexts: while Cueing may assist readers in identifying non-binary references, its impact may differ when applied to generic references. A better understanding of the role of Cueing will be beneficial to policymakers, journalists, and others interested in implementing Dutch gender-neutral pronouns.
Perceived Awkwardness mediated the effect of Referential Strategy on Text Appreciation in Experiment 2, consistent with our hypothesis based on the study by Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024). However, this mediation appeared to depend on the Referential Strategy: Perceived Awkwardness did not explain the difference in Text Appreciation between die-die-diens and masculine forms. Other factors may contribute to the reduced Text Appreciation associated with the use of gender-neutral pronouns. These may include linguistic factors, such as perceived ungrammaticality of the forms (see Decock et al., Reference Decock, Van Hoof, Vanrobaeys, Verhaegen and Vincent2025) or prior exposure to different referential strategies. Additionally, personal beliefs and ideological convictions may also play a role. Future research should explicitly investigate these correlations for Dutch gender-neutral pronouns.
Importantly, the present study did not test whether the gender-neutral pronouns have the desired effect from a gender-inclusive language point of view: attenuating or even eliminating the male bias elicited by masculine singular 3rd person pronouns. Our results suggest that plural pronouns are a suitable alternative to masculine generic pronouns as well, but their effect on the male bias should be investigated further. Redl et al. (Reference Redl, Szuba, De Swart, Frank and De Hoop2022) found no male bias for plural pronouns, at least in neutral contexts and when the antecedent is not itself a masculine generic form. However, it remains unclear whether the absence of male bias holds in gender-stereotyped contexts or when the antecedent is explicitly masculine generic (e.g., werknemer rather than personeelslid). Mortelmans (Reference Mortelmans2008) has argued for personal nouns that neutralization strategies show great potential in the Dutch context, as grammatical gender is losing its salience and neutral forms include gender-diverse people better than differentiation strategies. However, in many languages, a male bias has been documented to the extent that even gender-neutral words are associated with male representations and masculinity. Renström et al. (Reference Renström, Lindqvist, Akbas, Hekanaho and Gustafsson Sendén2022a) found evidence for a male bias elicited by gender-neutral pronouns in Finnish and Turkish, which are genderless languages. This raises the question: can linguistic neutrality in the Dutch language override societal androcentrism? Future research should therefore explore the mental gender representations associated with both Dutch masculine singular pronouns and plural pronouns to better understand their effects on gender biases.
Decock et al.’s (Reference Decock, Van Hoof, Soens and Verhaegen2024) study was intended as the first in a row to track shifts in familiarity with and attitudes toward Dutch gender-neutral pronouns. Building on their findings, the current study documents a rise in familiarity with gender-neutral pronouns between 2022 and 2023 in Flanders, while attitudes remain rather negative, as reflected in low agreement scores for the necessity of non-binary or generic gender-neutral pronouns. Additionally, the current study is the first to explore familiarity with and attitudes toward gender-neutral pronouns in the Netherlands. We consider a longitudinal approach for Dutch of great value because it will allow us to see how attitudes evolve over time. A replication study among a representative sample of Dutch participants would be interesting to confirm the potential of die and the rejection of hen as gender-neutral subject. Additional follow-up studies in both Belgium and the Netherlands should track whether exposure to the gender-neutral pronouns breeds acceptance or backlash over the coming years.
A key strength of our study is the use of longer, continuous text, offering a more authentic reading experience of the gender-neutral pronouns in different syntactic functions. However, the between-subjects design and the limited number of pronoun occurrences per participant may have reduced the sensitivity to detect small effects, despite relatively large sample sizes. To enhance statistical power and sensitivity, future research should consider using within-subjects designs and increasing the number of pronoun instances presented to each participant. In general, further research on gender-neutral pronouns in their diverse usage contexts is vital. Future studies would benefit from incorporating indirect methods such as self-paced reading tasks or eye-tracking studies, which often yield continuous data and can detect subtle effects more reliably than binary or subjective outcomes.
Conclusion
The present study makes a meaningful contribution to the societal debate on gender-inclusive language. Redl (Reference Redl2021; Redl et al., Reference Redl, Eerland and Sanders2018, Reference Redl, Frank, Swart and Hoop2021, Reference Redl, Szuba, De Swart, Frank and De Hoop2022) has demonstrated that the generic use of Dutch masculine third-person pronouns can lead to male bias, which can hinder gender equality. In light of this, gender-neutral pronouns emerged as a promising alternative to masculine pronouns. Our empirical evidence suggests that Dutch gender-neutral pronouns, when used for generic reference, do not hinder Text Comprehensibility. However, both Decock et al. (Reference Decock, Van Hoof, Soens and Verhaegen2024) and the present study reveal that gender-neutral pronouns may negatively impact Text Appreciation. In line with Friedrich and Heise (Reference Friedrich and Heise2019), we would argue that any impairment of appreciation is a reasonable trade-off for greater inclusivity. We acknowledge that inclusivity extends beyond the use of gender-inclusive language and includes ensuring that texts are accessible to people from all backgrounds. In light of the finding that Dutch gender-neutral pronouns do not compromise comprehensibility, we align with Renström’s (Reference Renström2025) practice recommendation that governments, organisations, and communication professionals committed to gender-inclusive communication can incorporate gender-neutral pronouns whenever possible. We advise using die-hen-hun, as it seems to be accepted in both Belgium and the Netherlands. Alternatively, our study confirms that plural forms are an appropriate referential strategy for those who seek to avoid masculine generics, as it impairs neither comprehensibility nor appreciation.
Replication package
All materials are available at https://osf.io/zdfvu/.
Author contributions
H.V: Data curation, Formal analysis, Investigation, Methodology, Project administration, Validation, Visualization, Writing—original draft, and Writing—review & editing. S.V.H: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Supervision, and Writing—review & editing. R.V.H: Funding acquisition, Methodology, and Writing—review & editing. P.G. & U.G: Conceptualization, Funding acquisition, Project administration, Supervision, and Writing—review & editing. S.D. (Principal Investigator): Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Supervision, and Writing—review & editing.
Financial support
This research was conducted within a WEAVE project jointly funded by the Research Foundation - Flanders (FWO) and the Swiss National Science Foundation (SNF). Experiment 1 was funded by the Research Foundation - Flanders (FWO)(G097623N). Experiment 2 was funded by the Centre for Language Studies Radboud University (Small Research Grant RG2023-36).
Competing interests
The author(s) declare none.
Ethical statement
The online survey experiments received positive advice from the Ethics Committee of Ghent University. The participants provided their informed consent to participate in the study.
Open access
