This is an open access article distributed under the terms of the Creative Commons Attribution License (The common traffic sign “Radfahrer bitte absteigen” (Cyclists [masc.] please dismount) illustrates the generic use of grammatically masculine role nouns in German: The feminine form Radfahrerinnen addresses female cyclists only, while the masculine form is supposed to address all cyclists. However, research has shown that using masculine forms whenever the gender of the referent(s) is unknown or irrelevant leads to a male bias (e.g., Gygax et al., 2021). A rather new, but popular gender-fair form intended to reduce this bias is the gender star (Genderstern) (Krome, 2020). Its proponents suggest that a sign displaying “Radfahrer*innen bitte absteigen” (Cyclists [star] please dismount) would address all people, that is, persons identifying beyond a female-male dichotomy, as well as women and men (Diewald & Steinhauer, 2017). In contrast, the Council for German Orthography1 disapproves of the star for not aligning with German orthography and impeding the readability of texts (Rat für deutsche Rechtschreibung, 2021). Experimental research on this claim is still scarce. Because word recognition is a crucial component of the reading process (e.g., Coltheart, 2006), we developed a lexical decision task to investigate the readability of role nouns in star form with an implicit measure. Data from two experiments will be presented: Experiment 1 was conducted with a sample of students (18–30 years). Experiment 2 was conducted with a more heterogenous group of older non-students (30–80 years).
Gender-Fair Language in German
In English, a natural gender language (Gygax et al., 2019), most personal nouns are gender neutral. In contrast, in German, a grammatical gender language, all nouns carry grammatical gender (masculine, feminine, neuter). Dependent forms, such as articles and pronouns, have to correspond. For the majority of role nouns, there are two forms: a masculine (e.g., der Radfahrer – the (male) cyclist [masc.]) and a feminine form. The latter is usually derived from the former by adding the feminine suffices -in [sing.] or -innen [plural] (e.g., die Radfahrerin – the female cyclist [fem.]) (Diewald & Steinhauer, 2017). As illustrated by the example above, feminine and masculine forms are used asymmetrically: Feminine forms only refer to women, while masculine forms can be used to refer to men (specific use) and to referents whose gender is unknown or irrelevant (generic use)2. Since the 1980s, feminist linguists have argued that this asymmetry leads to an underrepresentation of women in mental representations (e.g., Pusch, 1984). Their claims found support in psycholinguistic studies showing that the generic use of masculine forms leads to a male bias. In contrast, gender-fair alternatives can increase the mental inclusion of women (e.g., Braun et al., 1998; Gygax et al., 2008; Körner et al., 2022; Sato et al., 2016; Stahlberg & Sczesny, 2001).
Two strategies of gender-fair language (GFL) can be distinguished (cf. Gabriel et al., 2018): Feminization makes the inclusion of women explicit (e.g., binary pair forms: die Studentin oder der Student – the female student [fem.] or the male student [masc.], or their abbreviations, e.g., the capital-I: FlugbegleiterIn – stewardEss). Neutralizations remove any clues to the referent’s gender (e.g., nominalized participles: Studierende – those who study).
With the growing acknowledgement of the existence of gender identities beyond a female-male dichotomy, additional non-binary gender forms (GFs) have been introduced. They are intended to refer to all genders (Diewald & Steinhauer, 2017). The most popular non-binary GF and the one central to the public debate (Krome, 2020; Meuleneers, 2024) is the gender star. It has gained popularity particularly since 2018, when a third gender option (diverse) was introduced into the German personal status law in addition to the existing binary options (female and male). It is created by adding an asterisk between a role noun’s stem and its feminine suffix (e.g., Radfahrer*in – cyclist). It intends to go beyond being a mere abbreviation for binary pair forms, such as capital-I forms, to also address persons identifying as non-binary. Recent psycholinguistic studies investigating the functionality of the star form showed that its use can successfully reduce the male bias evoked by masculine generics (e.g., Keith et al., 2022; Körner et al., 2022). One study moreover showed that it evokes well-balanced mental representations of all genders (Zacharski & Ferstl, 2023).
Grammatical gender is not the only relevant factor influencing gendered representations of person referents in German: Many role nouns are associated with semantic gender stereotypes encoding the readers’ expectations of how likely it is that the referent is female or male (e.g., female: social workers, male: surgeons). This information from general world knowledge, which is related to the actual binary gender distribution within a specific group, might be altered when there is societal change (cf. Zacharski & Ferstl, 2023). Given that previous research has shown interactions between grammatical gender and stereotypical gender in German (e.g., Braun et al., 1998; Sato et al., 2016; Vervecken et al., 2013), it is crucial to take gender stereotypes into account when conducting research on GFL.
Comprehensibility and Readability of Gender-Fair Forms
In order to build an appropriate mental representation in the first place, gender-fair alternatives have to be readable and comprehensible (Friedrich & Heise, 2019). While opponents of GFL have argued that use of binary pair forms and neutralizations that are in line with German orthography (in the following: orthographic forms) makes texts harder to read and understand (e.g., Gesellschaft für deutsche Sprache, 2020), experimental studies using self-report measures have not confirmed this claim (e.g., Blake & Klimmt, 2010; Friedrich & Heise, 2019; Pöschko & Prieler, 2018). Using eye-tracking, Steiger-Loerbroks and Von Stockhausen (2014) moreover found that gender-fair neutralizations required more processing effort than masculine forms only at early reading stages, but not at later ones.
The case is different for binary forms that are not in line with German orthography (in the following: non-orthographic forms): Pöschko and Prieler (2018) found significantly lower subjective readability ratings for texts using slash forms (e.g., der/die Lehrer/in – the [masc.]/the [fem.] male teacher [masc.]/female teacher [fem. suffix]). For texts using capital-I forms (e.g., FlugbegleiterIn – stewardEss), Blake and Klimmt (2010) reported longer total reading times than for texts using orthographic forms (masculine forms, pair forms, and neutralizations). The authors suggested that this might be due to the more complex and unusual shape as well as the rare occurrence of the former. Another open question is whether, as seems likely, frequent exposure to non-orthographic forms can reduce these processing difficulties (cf. Gabriel et al., 2018).
So far, only Friedrich et al. (2021) have investigated the readability of the star form using subjective comprehensibility ratings: In two different experiments, students were presented with texts in one of two variants (masculine/star). The text for the first experiment contained predominantly plural star forms (e.g., die Spieler*innen – the players), while the second text contained more singular star forms. For the first experiment, the ratings did not yield any difficulties for the star. However, in the second experiment, it significantly decreased comprehensibility, particularly due to increased sentence difficulty ratings. Interpreting these results, it should be noted that there is only one definite plural article in German (die), which is used for feminine and masculine nouns. Hence, no adaptation of dependent articles is necessary when using plural star forms. In contrast, using singular star forms requires a more complex and very salient adaptation of the article (e.g., der*die Spieler*in – the[masc.]*the[fem.] player[star])—the actual use of these more complex constructions is, however, very uncommon.
This study provided valuable first insights in the readability of the star, but has two limitations: First, its results are based on subjective ratings of participants. Hence, it cannot be ruled out that responses were influenced by factors such as social desirability, political correctness, or emotions towards GFL. In contrast, implicit measures enable investigation of word recognition at its earliest, pre-conscious stages3 thus avoiding potential influences of these factors. Second, its results are based on a homogenous, mostly female student sample and participants are likely to have held a positive attitude towards GFL (e.g., Jäckle, 2022; Zacharski, 2024). However, previous research showed that attitudes towards GFL might affect the processing of GFs: In their eye-tracking study, Steiger-Loerbroks and Von Stockhausen (2014) reported that a more positive attitude towards the generically intended masculine led to increased reading times for gender-fair alternatives. In their word-picture matching task, Stahlberg and Sczesny (2001) (Experiment 4) found that participants with a more positive attitude towards GFL showed faster response times for images of women following the gender-fair capital-I form than the masculine form. In contrast, no such effect was found in participants with a negative attitude towards GFL. Further research on the readability of the star is thus needed that uses implicit measures and takes potential attitudes effects into account.
Visual Word Recognition: A Fundamental Component of Text Comprehensibility
If we want to know whether the gender star makes reading a text more difficult, investigating how it affects the recognition of role nouns on the word level is a crucial step towards it. According to van Dijk and Kintsch (1983), text comprehension consists of building a mental representation of what the text is about. The comprehension processes that contribute to its production occur at the word, sentence, and text level and interact with the reader’s world knowledge (Perfetti et al., 2005). Hence, “if we knew how people recognize whole words on the page, we [would] […] know part of what we need to know in order to understand how people comprehend whole printed sentences” (Coltheart, 2006, p. 6). Reflecting this, when investigating the readability of the star, Friedrich et al. (2021) did not only ask participants to rate the comprehensibility of the text as a whole, but moreover assessed word difficulty, that is, how easily participants grasped the meanings of the text’s words.
A well-established implicit measure to study word recognition is the lexical decision task (Coltheart et al., 2001; Rastle, 2016). In this paradigm, participants are presented with strings of letters. They have to decide as quickly as possible whether the stimulus shown is a word or not. Accuracy rates and reaction times (RTs) provide information on which factors influence the ease of word recognition. The most important lexical features are word length and word frequency: Shorter or more frequent words are recognized more quickly. Other factors include age of acquisition and familiarity: Words that are learned earlier in life and words that are more familiar are recognized faster (Coltheart et al., 2001; Rastle, 2016). Importantly, the paradigm even enables the detection of effects of within-word changes on word identification that are small in magnitude (e.g., case alternation: gArDeNeR, Coltheart & Freeman, 1974).
The Dual Route Cascaded model of visual word recognition (Coltheart et al., 2001) assumes that each word known by the reader is represented as an individual lexical entry in a mental lexicon. Words are recognized by mapping the identified letters on the correct entry (lexical access). This process allows the skilled reader to quickly and accurately read familiar words. Reading unknown words or pronounceable strings of letters (pseudowords), in contrast, requires the use of the non-lexical route (corresponding to spelling out the word) which is slower and less accurate.
For reading star nouns, there are two potential scenarios: First, owing to their non-orthographic form, they are not mapped on entries of the mental lexicon, but are processed via the non-lexical route. Second, star nouns are accessed via the lexical route, which is highly sensitive to word length, word frequency, and age of acquisition. Due to the insertion of the asterisk, star nouns are longer. Moreover, special character forms occur less frequently than feminine forms (Goldhahn et al., 2012), which are less frequent than masculine words (cf. Friedrich et al., 2021; Gabriel et al., 2018). Finally, owing to its relative recent introduction, participants will have first learnt the star form in adulthood, so that age of acquisition is higher for star nouns than for binary forms. Hence, in both scenarios, we expect slower reading times for star nouns. Interindividual differences, however, might influence the ease of recognition: First, as mentioned above, a positive attitude towards GFL might facilitate its processing. Second, owing to its recent introduction, younger participants will have become acquainted with it earlier in their life than older participants did. Moreover, many German universities recommend the use of the gender star in their guidelines for GFL (Schneider, 2022). Students might have thus encountered the gender star more frequently than non-students. Lexical access to star nouns might thus be easier for younger students than for older non-students.
The Present Study
The present study consists of two experiments employing the same lexical decision task to investigate whether the insertion of the asterisk impedes lexical access to role nouns. The first experiment was conducted with a sample of students (18–30 years). The second experiment was conducted with a more heterogenous group of older non-students (30–80 years). In each experiment, we assessed attitudes towards GFL. This and comparisons between the samples allowed us to evaluate the influence of interindividual differences.
In the lexical decision task, 72 role nouns were used as experimental items. GF was varied within subjects (star/feminine/masculine) so that every participant saw 24 role nouns in each of the forms. The semantic stereotype of the role noun was controlled (Misersky et al., 2014). To conceal the purpose of the experiment, a large number of filler items was used (48 regular words, 120 pseudowords), leading to 240 trials in total. Participants had to decide as quickly as possible whether the string presented was a German word or not, and respond via keypress. Acceptance rates and RTs for experimental items were used as dependent variables.
Our study was guided by two main research questions (RQs) for which we derived specific hypotheses (Hs). The first question addresses the status of the gender star as a German word. The second one addresses the question of its readability.
Even though the star form is, by now, frequently used by media channels and institutions (Krome, 2020; Schneider, 2022), previous surveys showed that only about 25% of the population in Germany are in favor of its use (Jäckle, 2022: 21%; Welt am Sonntag, 2020: 26%). We thus wanted to know:
(RQ1) Which participants accept role nouns in star form as German words?
We expected that the polarization of the public debate (Meuleneers, 2024) would manifest itself in the distribution of acceptance rates of star nouns: While some participants accept star nouns as words, others would always reject them. Moreover, we expected that participants with a more positive attitude towards GFL and participants of the younger, student sample would be more likely to accept the star, thus leading to the following hypotheses:
(H1.a) Participants’ mean acceptance rates for star nouns show a bimodal distribution.
(H1.b) Participants with a more positive attitude towards GFL are more likely to accept star nouns as words (within samples).
(H1.c) Participants of the younger, student sample are more likely to accept star nouns as German words than participants of the older, non-student sample (between samples).
Next, we were interested in whether and how the gender star affects word recognition for those participants who, in general, accept star nouns as German words:
(RQ2) Is lexical access to role nouns in star form more difficult than to feminine and masculine forms?
For this purpose, we had a closer look at RTs for accepted experimental items. Specifically, we formulated the following hypothesis:
(H2.a) The insertion of the asterisk impedes lexical access thus leading to slower RTs for star nouns compared to binary forms (controlling for word length).
Because more exposure to the star form is likely to facilitate lexical access, we expected the processing of star nouns to get easier over the course of the experiment:
(H2.b) RTs to star nouns decrease across time.
Moreover, we formulated the following hypotheses with regard to interindividual differences:
(H2.c) For participants with a more positive attitude towards GFL, recognition of star nouns is easier than for participants with a less positive attitude (within samples).
(H2.d) For participants of the younger, student sample, recognition of star nouns is easier than for participants of the older, non-student sample (between samples).
The wide range of participants’ ages in the non-student sample furthermore allowed us to exploratorily test the influence of age on the ease of lexical access within this sample. Finally, it was an exploratory question whether gender stereotypes would influence the lexical decision process.
Experiment 1: Student Sample
In the first experiment, we tested university students to make it comparable with previous research (e.g., Friedrich et al., 2021). We focused on students younger than 30 years thus representing a group of typical students in Germany (Davies, 2023).
Method
Participants
Recruitment took place at the University of Freiburg, Germany. Only students between 18 and 29 years with sufficient knowledge of German (L1, or L24 with > 10 years of experience) were included in the dataset for analysis. Of the 124 participants who initially started the experiment, 27 were excluded (early drop-out: 5, insufficient German skills: 6, > 29 years: 6, non-students: 9). Moreover, one participant was excluded in the course of an outlier analysis (see Supplementary Materials, Zacharski, 2024S). Our final sample consisted of 97 students of psychology (74), cognitive science (21), and linguistics (2) (M = 21.1 years, SD = 2.50; non-binary: 2, female: 76, male: 18, unspecified: 1). All participants received course credit as compensation. An a priori power analysis using G*Power (Faul et al., 2007) showed that a sample of 73 participants would have been sufficient to yield a power of .80 (with α = .05) for finding even small differences between GFs (f = .15).
Design and Materials
240 items were presented to each participant: 120 words and 120 pseudowords. 72 role nouns were used as experimental items. For the latter, GF was varied within subjects (star/feminine/masculine), so that every participant saw 24 role nouns in each form. The remaining items were fillers to distract participants from the goal of the study. Item types will be described in the following.
Words: Experimental Items
72 role nouns that allowed feminine inflection were selected from Misersky et al. (2014), so that 24 nouns each were stereotypically female, neutral, and male. Word frequencies of masculine forms (Goldhahn et al., 2012) and word lengths of masculine forms were balanced over the three stereotypicality categories, one-way ANOVAs: word length: F(2, 69) = 0.19, p = 0.83; word frequency: F (2, 69) = 1.67, p = 0.20 (Table 1). For each role noun, three GFs were created: masculine, feminine (-in), star (-*in).
Table 1
Descriptives of all Items Used as Stimuli
| Stimulus type | Stimulus category | Gender stereotypicality | n (Stimuli) | Word length (masc. for exp. item) | Word frequency (masc. for exp. items) | Example |
|---|---|---|---|---|---|---|
| WORD | experimental item | female | 24 | 10.71 (2.88), 7-17 | 15.71 (2.74), 10-23 | Kosmetiker/Kosmetikerin/Kosmetiker*in (beautician [masc./fem./star]) |
| WORD | experimental item | neutral | 24 | 10.50 (3.44), 6-17 | 14.58 (3.05), 9-21 | Biologe/Biologin/Biolog*in (biologist [masc./fem./star]) |
| WORD | experimental item | male | 24 | 10.17 (2.94), 5-17 | 14.50 (1.69), 12-18 | Chirurg/Chirurgin/Chirurg*in (surgeon [masc./fem./star]) |
| WORD | filler, no special character | — | 24 | 10.75 (2.85), 4-16 | 14.42 (3.02), 7-19 | Menschenrechte (human rights) |
| WORD | filler, with special character | — | 24 | 8.50 (3.11), 3-16 | 14.58 (3.43), 10-22 | iPhone |
| PSEUDOWORD | ending on -er | — | 24 | 10.04 (2.93), 5-17 | — | Welchzieter |
| PSEUDOWORD | ending on -in | — | 24 | 12.29 (3.53), 7-19 | — | Schrommerin |
| PSEUDOWORD | ending on -in, with * (asterisk) at random position | — | 24 | 13.58 (2.96), 10-20 | — | Flise*rmin |
| PSEUDOWORD | no specific ending, no special character | — | 24 | 10.71 (2.82), 4-16 | — | Witschenbechte |
| PSEUDOWORD | no specific ending, with special character | — | 24 | 8.42 (3.17), 3-16 | — | Kassem-gof:nalge |
Note. Words: Experimental Items and Filler Items: Means, SDs, and ranges of word lengths and word frequencies by Stimulus Type. Word frequencies and word lengths of experimental are based on masculine forms. Pseudowords, all categories: Means, SDs and ranges of pseudoword lengths by Stimulus Type. Examples are given for both words and pseudowords.
Words: Filler Items
48 words (no role nouns; e.g., Menschenrechte – human rights) were used. 24 of these contained at least one special character (e.g., H&M) or majuscule (e.g., iPhone). Filler words (with and without special characters) were selected so that their word frequencies matched those of the experimental items (One-way ANOVA: F (2, 117) = 0.35, p = 0.71). Experimental items and filler items without special characters were equally long. However, words with special characters were significantly shorter, because longer ones are rare in German (One-way ANOVA: F (2, 117) = 4.40, p = 0.01) (Table 1).
Pseudowords
Based on every word (experimental and filler items), one pseudoword was generated using the multilingual pseudoword generator Wuggy (Version 0.1.7). The tool creates equally long pseudowords, i.e., nonwords that are in line with orthographic and phonological patterns of German (Keuleers & Brysbaert, 2010). Pseudowords based on experimental items carried the suffices that are typically associated with feminine (-in) or masculine gender (-er), or carried the feminine suffix -in and contained an asterisk at a random position. Pseudowords generated from fillers either contained no or at least one special character other than the asterisk (Table 1).
Experimental Lists
GF of experimental items was varied within participants, such that each participant saw 24 role nouns in each GF. The same 48 filler words and 120 pseudowords were used in each of the three lists. The entire experiment consisted of 240 trials.
To guarantee pseudo-randomization during presentation, 12 sub-lists of 20 items were created for each list. Each sub-list consisted of 10 words (2 of each type) and 10 pseudowords (2 of each type) and was presented in a randomized order. No list contained the pseudoword and the word used for its generation.
Questionnaire
A questionnaire with five items was used to assess attitudes towards GFL (Dietsche, 2020) (e.g., In my opinion, more texts should be written in gender-fair language; for all items see Supplementary Materials, Table A, Zacharski, 2024S). Items were rated on a Likert-scale from 1–5. Two items were reverse coded so that a higher score suggests a more positive attitude towards GFL. Before filling out the questionnaire, participants read a brief definition of the term ‘GFL’. The mean score was used for statistical analysis. With an internal consistency of Cronbach’s α = 0.81, the scale proved to be reliable.
Presentation and Procedure
The experiment was implemented on lab.js (Henninger et al., 2023); the JATOS Server (Lange et al., 2015) was used for data collection. Participants received a link and completed the study at home on their computer. They were asked not to use smartphones/tablets, sit in a silent room, and make sure not to be disturbed. Although compliance to instructions cannot be directly measured in online-studies, their reliability has been confirmed (Schnoebelen & Kuperman, 2010).
Participants were told that the study investigates word comprehension, but were not informed about the goal of the study until after the experiment. They were instructed that they would be presented with letter strings and to decide as quickly as possible whether these are German words or not.
After giving consent and reading the instructions, all participants completed one practice block (8 trials). Then, each participant was presented with one of the experimental lists. Per trial, one letter string was presented. Participants gave answers using their keyboard (D/yes, K/no or K/yes, D/no, counterbalanced across participants). Each stimulus was displayed until keypress, but not longer than 2,000ms, after which the next stimulus automatically appeared. Two 30s breaks were allowed (after trials 80 and 160). After finishing the experiment, participants filled in the attitudes-questionnaire and gave demographic information. Completing the study took about 20 minutes.
Data Analysis
All steps of the statistical analysis were conducted with R (Version 4.3.1) run on R studio (Posit Team, 2023; Version 2023.06.0).
Before the main statistical analysis, an outlier analysis was conducted to identify participants who did not perform the task appropriately, and an item check to identify words and pseudowords that elicited unpredicted responses (for details see Supplementary Materials, Zacharski, 2024S).
(RQ1): The statistical analysis for acceptance rates was based on experimental items only (6,566 datapoints). To test whether mean acceptance rates for star nouns show the expected bimodal distribution, Hartigan’s Dip Test for Unimodality (Hartigan & Hartigan, 1985) was calculated using the diptest-package.
Next, a generalized mixed model (M1.1) was fitted using the glmer-function (lme4-package, Bates, Maechler, et al., 2015) with response type as the dependent variable. Yes-responses (acceptance) were coded as 1, no-responses (rejection) as 0. Based on our hypotheses, we chose a four-way interaction term for fixed effects including GF, gender stereotypicality, scaled attitude scores and scaled trial number. The parsimonious random effect structure (Bates, Kliegl, et al., 2015) including potential item effects (1|Item) and interindividual differences (1|ID) was controlled with the rePCA()-function (lme4-package) showing that these dimensions were sufficient to account for 100% of the explained variability. The Anova()-function (car-package) was used to produce type-III-Anova tables for fixed effects (for further details on contrast coding, post-hoc analyses, data visualization, and versions of R-packages see Supplementary Materials, Zacharski, 2024S).
(RQ2): In order to check how acceptance differed amongst participants generally accepting star nouns as words, we fitted the same glmer-Model (M1.2) for the sample reduced by participants who generally rejected star nouns (< 10% mean acceptance of the star; from now on: rejectors). For the analysis of RTs and to investigate lexical access, logarithmized yes-responses of all participants were used as the dependent variable (6,249 datapoints). We fitted a linear mixed effect model (M2) using the lmer-function (lme4-package). In addition to the fixed effect structure used for the glmer-Models (M1), we added scaled word lengths as an independent variable. We used the more complex, but still parsimonious random effect structure (1|Item) + (1+Trial Nr+Word Length|ID). The remaining procedures of statistical analysis were the same as for M1.
Results
Attitude Questionnaire
The mean attitude score M = 3.79 (SD = 0.77) suggested that, overall, participants held a positive attitude towards GFL (range = 1.80–5.00; scale: 1[negative] to 5[positive]). There were no significant differences in attitudes between female (M = 3.81, SD = 0.74, range 1.80–5.00) and male participants (M = 3.72, SD = 0.83, range = 2.00–4.8) (Wilcoxon rank sum test: W = 649.5, n1 = 76, n1 = 18, p = 0.743).
(RQ1) Acceptance of Role Nouns in Star Form as German Words
The statistical analysis for the complete sample (Table 2, M.1.1) yielded a significant main effect of GF. A priori contrast coding and post-hoc tests (Supplementary Materials, Tables C.1 & C.2, Zacharski, 2024S) showed that acceptance rates for star nouns are significantly lower than those for binary forms. Visualizing predicted probabilities for all participants (Figure 1, Plot A.1) showed a comparably large variance for the star. Having a closer look at the participants’ mean acceptance rates for star nouns (Figure 1, Plot A.3), we found, in line with H1.a, a bimodal distribution (Hartigan’s Dip Test for Unimodality: D = 0.14, p < .001). However, contrary to our expectations, there was only one rejector (100% rejection). The great majority of participants (99%) accepted star nouns as words (> 81% mean acceptance of the star).
Table 2
Results of Type-III Anova of glmer-Models of Acceptance Rates for Experimental Items for the Student Sample (M1.1, M1.2 – Experiment 1) and the Non-Student Sample (M3.1, M3.2 – Experiment 2)
| Predictor | Student sample | Non-student sample | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M1.1: Acceptance (97 participants) | M1.2: Acceptance (96 participants) | M3.1: Acceptance (80 participants) | M3.2: Acceptance (76 participants) | |||||||||
| Chisq | df | p | Chisq | df | p | Chisq | df | p | Chisq | df | p | |
| Intercept | 401.49 | 1 | < .001*** | 412.139 | 1 | <. 001*** | 340.82 | 1 | < .001*** | 352.733 | 1 | < .01*** |
| Gender Form | 13.361 | 2 | .001** | 3.391 | 2 | .184 | 62.536 | 2 | < .001*** | 14.784 | 2 | .001*** |
| Gender Stereotypicality | 2.938 | 2 | .230 | 3.805 | 2 | .149 | 2.226 | 2 | .329 | 1.501 | 2 | .472 |
| Trial Nr | 0.794 | 1 | .373 | 0.901 | 1 | .342 | 3.678 | 1 | .055 | 1.773 | 1 | .183 |
| Attitude | 0.002 | 1 | .960 | 0.544 | 1 | .461 | 0.131 | 1 | .718 | 0.019 | 1 | .889 |
| Gender Form:Gender Stereotypicality | 5.648 | 4 | .227 | 6.710 | 4 | .152 | 2.277 | 4 | .685 | 2.739 | 4 | .602 |
| Gender Form:Trial Nr | 6.633 | 2 | .036* | 5.650 | 2 | .059^ | 2.809 | 2 | .245 | 2.045 | 2 | .360 |
| Gender Stereotypicality:Trial Nr | 0.288 | 2 | .866 | 0.737 | 2 | .692 | 1.464 | 2 | .481 | 1.064 | 2 | .588 |
| Gender Form:Attitude | 11.112 | 2 | .004** | 3.902 | 2 | .142 | 35.409 | 2 | < .001*** | 0.153 | 2 | .926 |
| Gender Stereotypicality:Attitude | 2.024 | 2 | .364 | 1.198 | 2 | .549 | 1.863 | 2 | .394 | 2.910 | 2 | .233 |
| Trial Nr:Attitude | 0.448 | 1 | .503 | 0.592 | 1 | .441 | 0.009 | 1 | .926 | 0.182 | 1 | .670 |
| Gender Form:Gender Stereotypicality:Trial Nr | 0.898 | 4 | .925 | 0.375 | 4 | .984 | 11.673 | 4 | .020* | 10.053 | 4 | .040* |
| Gender Form:Gender Stereotypicality:Attitude | 2.960 | 4 | .565 | 7.196 | 4 | .126 | 7.183 | 4 | .127 | 6.183 | 4 | .186 |
| Gender Form:Trial Nr:Attitude | 0.877 | 2 | .645 | 0.504 | 2 | .777 | 0.850 | 2 | .654 | 0.594 | 2 | .743 |
| Gender Stereotypicality:Trial Nr:Attitude | 2.715 | 2 | .257 | 3.223 | 2 | .200 | 0.521 | 2 | .771 | 0.271 | 2 | .873 |
| Gender Form:Gender Stereotypicality:Trial Nr:Attitude | 5.940 | 4 | .204 | 4.972 | 4 | .290 | 4.665 | 4 | .323 | 4.950 | 4 | .293 |
***p < .001. **p < .01. *p < .05. ^p < .1.
Figure 1
Acceptance Rates by Gender Form and Distribution of Acceptance Rates of the Star Form for the Student and the Non-Student Sample
Note. Predicted Probability of Acceptance based on the Generalized Mixed Effect Models for the Student Sample (Plot A.1 & A.2) and the Non-Student Sample (Plot B.1 & B.2): Comparison of Complete (left) and Reduced Sample (right) within each of the samples. Error bars show SEs; SEs and significance levels are taken from post-hoc analysis. Histogram of the expected bimodal distribution of mean acceptance rates for the participants of the Student (Plot A.3) and the Non-Student Sample (Plot B.3). Abbreviations: fem = feminine; masc = masculine; nb = non-binary gender star.
***p < .001. **p < .01. *p < .05.
In line with H1.b, the main effect of GF (Table 2, M.1.1) was qualified by a significant interaction between GF and attitude: Participants with a more positive attitude towards GFL were more likely to accept star nouns as words. This effect was due only to the one rejector, whose attitude (score = 2.40) was less positive than the sample’s mean.
Moreover, the interaction between GF and trial number was significant. A priori contrast coding (Supplementary Materials, Tables C.1, Zacharski, 2024S) showed that this effect was driven by binary GFs only: While acceptance of feminine forms decreased across time, acceptance of masculine forms increased (see Supplementary Materials, Figure A, Zacharski, 2024S).
(RQ2) Lexical Access to Role Nouns in Star Form
Acceptance Rates
The statistical analysis of acceptance rates for the sample reduced by the rejector (Table 2, M1.2; Figure 1, Plot A.2) showed no main effect of GF.
Reaction Times
Predicted RTs of yes-responses are visualized in Figure 2 (left panel). As expected, the statistical analysis (Table 3, M2) yielded significant main effects of trial number and word length: Participants got faster across time and RTs to shorter words were faster. Importantly, and not in line with H2.a, there was no significant main effect of GF: RTs for star nouns were as fast as RTs for binary forms and do thus not suggest any difficulties in lexical access.
Figure 2
Reaction Times by Gender Form for the Student and the Non-Student Sample
Note. Predicted Reaction Times of yes-Responses of Experimental Item based on the Linear Mixed Effect Models for the Student Sample (left) and the Non-Student Sample (right). Error bars show SEs; SEs and significance levels are taken from post-hoc analysis. Significance levels for within-group models: red; significance levels for inter-group comparisons: black. fem = feminine; masc = masculine; nb = non-binary gender star.
***p < .001. **p < .01.
Table 3
Results of Type-III Anova of lmer-Model of Reaction Times (Yes-Responses) for Experimental Items for the Student Sample (M2 – Experiment 1) and the Non-Student Sample (M4 – Experiment 2)
| Predictor | Student sample | Non-student sample | ||||
|---|---|---|---|---|---|---|
| M2: Reaction times (yes-responses) (97 participants) | M4: Reaction times (yes-responses) (80 participants) | |||||
| Chisq | df | p | Chisq | df | p | |
| Intercept | 89,516.429 | 1 | < .001*** | 134,282.059 | 1 | < .001*** |
| Gender Form | 3.626 | 2 | .163 | 113.831 | 2 | < .001*** |
| Gender Stereotypicality | 1.444 | 2 | .486 | 0.979 | 2 | .613 |
| Trial Nr | 18.256 | 1 | < .001*** | 5.133 | 1 | .023 * |
| Attitude | 1.619 | 1 | .203 | 0.540 | 1 | .462 |
| Word Length | 9.968 | 1 | .002** | 12.088 | 1 | .001*** |
| Gender Form:Gender Stereotypicality | 3.354 | 4 | .500 | 9.664 | 4 | .046* |
| Gender Form:Trial Nr | 0.659 | 2 | .719 | 6.559 | 2 | .038* |
| Gender Stereotypicality:Trial Nr | 2.197 | 2 | .333 | 0.324 | 2 | .851 |
| Gender Form:Attitude | 3.648 | 2 | .161 | 2.430 | 2 | .297 |
| Gender Stereotypicality:Attitude | 3.742 | 2 | .154 | 1.200 | 2 | .549 |
| Trial Nr:Attitude | 2.487 | 1 | .115 | 0.001 | 1 | .976 |
| Gender Form:Gender Stereotypicality:Trial Nr | 5.922 | 4 | .205 | 2.452 | 4 | .653 |
| Gender Form:Gender Stereotypicality:Attitude | 11.589 | 4 | .021* | 2.246 | 4 | .691 |
| Gender Form:Trial Nr:Attitude | 0.249 | 2 | .883 | 0.364 | 2 | .833 |
| Gender Stereotypicality:Trial Nr:Attitude | 0.542 | 2 | .763 | 2.784 | 2 | .249 |
| Gender Form:Gender Stereotypicality:Trial Nr:Attitude | 0.432 | 4 | .98 | 2.193 | 4 | .700 |
***p < .001. **p < .01. *p < .05. ^p < .1.
However, we found a significant interaction between GF, gender stereotypicality, and attitude towards GFL: Participants with a more positive attitude responded faster to stereotypically female nouns in star form than in binary forms. In contrast, they showed slower RTs to stereotypically male nouns in star form than in binary forms (see Supplementary Materials, Figure B, Zacharski, 2024S).
Summary
In line with hypothesis H1.a, we found the expected bimodal distribution of acceptance rates. Notably, however, only one of 97 students generally rejected star nouns as German words. In line with H1.b, we found that the general acceptance of the gender star might be influenced by attitudes towards GFL—however, as there was only one rejector, this finding is not reliable. For participants who generally accepted star nouns as words, there was no significant effect of GF on RTs. Hypothesis H2.a could thus not be confirmed. Consequently, H2.b and H2.c were not relevant.
Experiment 1 is subject to two limitations: First, we tested a homogenous sample of young students, who held a rather positive attitude towards GFL. Second, the majority of participants identified as female. Thus, Experiment 2 was conducted with a sample of older non-students, that was balanced with regard to binary gender identities.
Experiment 2: Non-Student Sample
In Experiment 2, we replicated the study described with a sample of non-students varying in age (30–80 years) and academic background. We stuck as closely to the initial design as possible, but used a more detailed questionnaire to assess attitudes towards GFL. All aspects in which Experiment 2 differed from the first one will be described in the following.
Method
Participants
84 non-student participants over the age of 30 were recruited via Prolific. Two participants currently enrolled at a university and one participant with insufficient knowledge of German were excluded. One participant was excluded in the course of an outlier analysis (see Supplementary Materials, Zacharski, 2024S). Our final sample consisted of 80 non-students (non-binary: 1, female: 32, male: 47) between 30–80 years (age distribution: Supplementary Materials, Table B, Zacharski, 2024S). 45 participants had an academic background, while 35 had not studied at a university. All participants received financial compensation in line with the recommendations given by Prolific.
Design and Materials
To avoid technical difficulties that occurred during the first experiment, two filler items were added after the start of the experiment and each of the two breaks.
Questionnaire
The 32-item ABNBL-questionnaire (Zacharski, 2024) designed to assess attitudes towards non-binary GFL in German was used. All items were rated on a Likert-scale from 0–9 and a higher score indicated a more positive attitude towards GFL. Following Zacharski (2024), participants read a brief definition of the term ‘non-binary gender identity’ before the questionnaire. The mean item score was used for statistical analysis.
Presentation and Procedure
The experiment was implemented on PCIbex (Zehr & Schwarz, 2022). Participants were redirected to the study via Prolific. They replied via keypress (Y/yes, N/no5). The procedure for the experiment was the same as in the first experiment, but participants went through two practice blocks (6 trials each)—one with, and one without feedback.
Data Analysis
(RQ1): The acceptance rates (M3.1) for experimental items (5,664 datapoints) were analyzed as in M1.1 (Experiment 1).
(RQ2): The model fit for RTs (M4) was based on yes-responses to experimental items of all participants (5,423 datapoints), as in M2 (Experiment 1).
We fit an additional exploratory model to test the influence of age on lexical access within this sample.6 The model specification (M4_age, Supplementary Materials, Tables K.1 & K.2, Zacharski, 2024S) was the same as for M4, except that Age was added to the fixed effects term.
Finally, for intergroup comparisons and to test H2.d, the processed datasets for both age groups were combined. The two models fitted (acceptance rates for reduced samples: M5; RTs of yes-responses for full samples: M6) were the same as for within-group analyses, except that, because different attitude-scales were used in the experiments, the attitude-variable was replaced by the factor Group (Students vs. Non-Students) (see Supplementary Materials, Tables F-K, Zacharski, 2024S).
Results
Attitude Questionnaire
The mean attitude score was M = 4.26 (SD = 1.73, range = 0.94–8.16; scale: 0 [negative] to 9 [positive]), with female participants having significantly more positive attitudes (M = 4.97, SD = 1.63, range = 2.38–8.16) than male participants (M = 3.75, SD = 1.62, range = 0.94–7.41) (Wilcoxon rank sum test: W = 1056.50, n1 = 32, n1 = 47, p = 0.002).
(RQ1) Acceptance of Role Nouns in Star Form as German Words
The statistical analysis of acceptance rates for the complete sample (Table 2, M3.1) yielded a significant main effect of GF. A priori contrast coding and post hoc analysis (Supplementary Materials, Tables F.1 & F.2, Zacharski, 2024S) showed that acceptance of star nouns was significantly lower than for binary forms. For the complete sample (Figure 1, Plot B.1), there was a comparably large variance for the star form. The participants’ mean acceptance rates for star nouns (Figure 1, Plot B.3) showed, in line with H1.a, the expected bimodal distribution (Hartigan’s Dip Test for Unimodality: D = 0.11, p < .001). However, in line with H1.c, compared to just one rejector in the student sample, more people, four in total, rejected star nouns in the majority of cases (less than 10% accepted). Still, as the major mode of the distribution shows, the great majority of participants (95%) accepted star nouns as German words (> 79% mean acceptance of star).
As in the student sample, and in line with H1.b, the main effect of GF in the complete sample was qualified by a significant interaction between GF and attitude towards GFL: All four rejectors held a rather negative attitude towards GFL (M = 2.0, SD = 0.72, range = 1.47–3.06). Interestingly, they all identified as male.
Moreover, we found a significant three-way interaction between GF, gender stereotypicality, and trial number. A priori contrast-coding (Supplementary Materials, Table F.1, Zacharski, 2024S) showed that this effect was driven by binary GFs: For stereotypically female role nouns, acceptance of feminine forms increased while acceptance of masculine forms decreased over time. Thus, over the course of the experiment, a consistency effect between stereotypes and GF for stereotypically female nouns emerged. For stereotypically neutral nouns, acceptance of feminine forms decreased while acceptance of masculine forms increased (see Supplementary Materials, Figure C, Zacharski, 2024S). However, overall, the acceptance of binary forms was very high and the differences that emerged were rather small.
(RQ2) Lexical Access to Role Nouns in Star Form
Acceptance Rates
In contrast to the student sample, the statistical analysis of acceptance rates for the sample reduced by rejectors (Table 2, M3.2; Figure 1, Plot B.2) yielded a main effect of GF. A priori contrast coding and post hoc analysis (Supplementary Materials, Tables G.1 & G.2, Zacharski, 2024S) showed that this effect was due to higher rejection rates for star nouns compared to binary forms. This effect was, however, not qualified by an interaction with attitudes towards GFL as in model M3.1. Only the interaction effect of GF, gender stereotypicality, and trial number driven by binary forms described above remained significant.
Reaction Times
Predicted RTs of yes-responses for the different GFs are visualized in Figure 2 (right panel). Analogous to the student sample, the statistical analysis (Table 3, M4) yielded a main effect of trial number and of word length. In contrast to the student sample, however, it moreover yielded a significant main effect of GF: A priori contrasts and post-hoc analysis (Supplementary Materials, Tables H.1 & H.2, Zacharski, 2024S; Figure 2, right panel, significance levels in red) showed that RTs for star nouns were significantly slower than for binary forms. This finding is in line with H2.a and suggests difficulties in lexical access for star nouns.
In line with H2.b, this effect was qualified by a significant interaction of GF and trial number: RTs to star nouns decreased faster than those for binary forms. That is, there was an adaptation effect for star nouns over the course of the experiment—up to the point where initial differences between the three GFs disappeared (Figure 3, Plot A).
Figure 3
Reaction Times: Interaction Plots for Gender Form and A: Trial Nr; B: Participant’s Age for the Non-Student Sample
Note. Plot A: Predicted Reaction Times (yes-Responses) by Gender Form and Trial Number, Model M4.1 (Table 3) for the Non-Student Sample. Plot B: Predicted Reaction Times (yes-Responses) by Gender Form and Age, Model M4.2 (Table 3) for the Non-Student Sample. fem = feminine; masc = masculine; nb = non-binary gender star.
Contrary to H.2c, there was no significant interaction between GF and attitudes towards GFL. Hence, for the subgroup of people who generally accepted the gender star, the attitude score was unrelated to the performance in the lexical decision task.
A priori contrast coding (Supplementary Materials, Tables H.1, Zacharski, 2024S) showed that the interaction of GF and gender stereotypicality was due to binary GFs only, suggesting, again, consistency effects: For stereotypically female words, RTs for feminine forms were faster than for masculine forms, and vice versa for stereotypically male words (Supplementary Materials, Figure D, Zacharski, 2024S).
Finally, note that in the exploratory model M4_age (Supplementary Materials, Table K.1, Zacharski, 2024S), the main effect of GF was qualified by an interaction with age. A priori contrast coding (Supplementary Materials, Table K.2, Zacharski, 2024S) showed that, while RTs generally increased with age, this increase was stronger for the star compared to binary forms (Figure 3, Plot B).
Inter-Group Comparisons
The description of the results for the two samples confirmed the expectation that students have less difficulties processing the rather recently introduced gender star. Although the experiments were conducted separately, a direct statistical comparison was carried out to uncover group differences (Table 4, M5 & M6): In line with the observations from M1.1 and M.3.1, for acceptance rates of the reduced samples without rejectors (Table 4, M5), there was a main effect of Group qualified by a significant interaction with gender form. A priori contrast coding and post-hoc analysis (Supplementary Materials, Table I.1 & I.2, Zacharski, 2024S) showed that, while for binary GFs, non-students had significantly higher acceptance rates than students (feminine: non-students: M = 97.73, SD = 3.36, students: M = 95.96%, SD = 4.15, masculine: non-students: M = 97.66, SD = 3.10, students: M = 95.48%, SD = 5.90), no significant difference was found for star forms (non-students: M = 96.22, SD = 5.03, students: M = 95.07%, SD = 4.92). The statistical analysis for RTs (Table 4, M6) yielded a significant main effect of Group: Independent of GF, students responded faster than non-students (Figure 2, significance levels in black). A priori contrast coding for the significant interaction of GF and Group (Supplementary Materials, Table J, Zacharski, 2024S) confirmed the observation that, while there was no significant effect of GF on RTs in the student sample, GF had a significant effect on lexical access in the non-student sample.
Table 4
Inter-Group Comparison: Results of Type-III Anova of glmer-Models of Acceptance Rates (M5) and of lmer-Model of Reaction Times (Yes-Responses) (M6) for Experimental Items
| Predictor | Inter-Group comparisons | |||||
|---|---|---|---|---|---|---|
| M5: Acceptance rates (students and non-students, reduced sample, 172 participants) | M6: Reaction times (yes-responses) (students and non-students, complete sample, 177 participants) | |||||
| Chisq | df | p | Chisq | df | p | |
| Gender Form | 18.268 | 2 | < .001*** | 57.167 | 2 | < .001*** |
| Gender Stereotypicality | 2.680 | 2 | .262 | 1.292 | 2 | .524 |
| Trial Nr | 5.595 | 1 | .018* | 56.679 | 1 | < .001*** |
| Group | 17.449 | 1 | < .001*** | 22.983 | 1 | < .001*** |
| Word length | — | — | — | 13.520 | 1 | < .001*** |
| Gender Form:Gender Stereotypicality | 7.251 | 4 | .123 | 8.388 | 4 | .078^ |
| Gender Form:Trial Nr | 6.332 | 2 | .042* | 4.937 | 2 | .085^ |
| Gender Stereotypicality:Trial Nr | 1.504 | 2 | .472 | 1.199 | 2 | .549 |
| Gender Form:Group | 5.390 | 2 | .068^ | 55.057 | 2 | < .001*** |
| Gender Stereotypicality:Group | 1.616 | 2 | .446 | 2.209 | 2 | .331 |
| Trial Nr:Group | 1.926 | 1 | .165 | 2.955 | 1 | .086^ |
| Gender Form:Gender Stereotypicality:Trial Nr | 7.767 | 4 | .100 | 4.517 | 4 | .340 |
| Gender Form:Gender Stereotypicality:Group | 2.480 | 4 | .648 | 4.802 | 4 | .308 |
| Gender Form:Trial Nr:Group | 1.182 | 2 | .554 | 1.355 | 2 | .508 |
| Gender Stereotypicality:Trial Nr:Group | 2.302 | 2 | .316 | 2.421 | 2 | .298 |
| Gender Form:Gender Stereotypicality:Trial Nr:Group | 9.405 | 4 | .052^ | 3.657 | 4 | .454 |
***p < .001. **p < .01. *p < .05. ^p < .1.
Summary
With regard to (RQ1) and in line with hypothesis H1.a, we found the expected bimodal distribution of acceptance rates: Four of the 80 non-students generally rejected star nouns as German words. In line with H1.b, they held a rather negative attitude towards GFL. In line with H1.c, the percentage of rejectors was higher in the older, non-student sample than in the younger, student sample.
With regard to (RQ2) and in line with H2.a, participants of the non-student sample who generally accepted star nouns as words, that is, the majority of participants (95%), showed significantly slower RTs for star nouns than for feminine and masculine forms. However, in line with H2.b, RTs for star nouns decreased more strongly over the course of the experiment than for binary forms suggesting that the processing of the star became less effortful over time. Contrary to H2.c, attitudes towards GFL did not predict the performance in the lexical decision task. However, in line with H2.d, comparisons between the two samples showed that processing was easier for the younger, student sample than for the older, non-student sample. In line with that, within the non-student sample, a higher age was accompanied by slower RTs to star nouns compared to binary forms. This suggests that, in addition to the (non-)student status, one factor driving the processing differences between the student and the non-student sample might be age.
General Discussion
The non-binary gender star in German (e.g., Radfahrer*in – cyclist) is the most popular gender-fair alternative for generically intended masculine role nouns to refer to persons of all genders, that is, persons identifying beyond a female-male dichotomy, as well as women and men (Krome, 2020). Opponents of its use argue that its non-orthographic form impedes the readability of texts (e.g., Rat für deutsche Rechtschreibung, 2021). Experimental research on this claim is rare. Because visual word recognition is a crucial component of the reading process (Coltheart, 2006), we developed a lexical decision task to test whether lexical access to singular role nouns is more difficult to star nouns compared to the orthographic and more common feminine and masculine forms. To the best of our knowledge, this is the first study investigating the readability of the star with implicit measures and, therefore, an important contribution to the current debate on GFL in Germany. In order to account for interindividual differences, we tested not only a homogenous student sample (Experiment 1), but also a heterogenous sample of non-students varying in age as well as academic background, which is a further strength of our study (cf. Jones, 2010).
Previous surveys on the acceptance of the gender star in Germany showed that only one quarter of the German population is in favor of its use (e.g., Jäckle, 2022). Based on these findings, we expected a bimodal distribution of the acceptance of star nouns as German words. While this hypothesis was confirmed, the number of rejectors was surprisingly small: only one student and four non-students, that is, merely 2.8% of all participants rejected star nouns as words in almost all of the cases. As expected, participants of the younger, student sample were more likely to generally accept star nouns as German words. Moreover, general rejection of star nouns as words was predicted by attitudes towards GFL: Participants with a less positive attitude towards GFL were more likely to be amongst the rejectors.
A closer inspection of the influence of GF on RTs for those participants who generally accept star nouns as words, sheds light on whether lexical access to star nouns is more difficult than to the more common binary forms. For the student sample, we found no evidence for difficulties in the recognition of role nouns in star form. These findings are in line with previous findings by Friedrich et al. (2021), who reported that the use of the gender star did not increase subjectively perceived word difficulty in a student sample. The case was different for the non-student sample: Participants showed significantly slower RTs to star nouns than to binary forms. These findings suggest difficulties in lexical access due to the insertion of the non-orthographic asterisk. However, processing of the star became easier over the course of the experiment: While RTs generally decreased across time, that is, independent of GF, the decrease was significantly steeper for star nouns—up to the point where RTs were comparably fast for all GFs. This suggests an adaptation effect for the non-student sample. Thus, while the initial processing of star nouns seems to be harder, their processing becomes less effortful with time (cf. Gabriel et al., 2018)—even over the short course of the experiment. Interestingly, and not in line with our expectations, attitudes towards GFL did not predict the success of lexical access to star nouns. However, participants’ age might be a predictor: First, participants of the younger, student sample had significantly less difficulties than participants of the older, non-student sample. Second, an additional exploratory analysis within the non-student sample showed that RTs increased with age, and this increase was largest for star nouns. A possible explanation for the crucial role participants’ age seems to play for successful lexical access—in line with theories on visual word recognition (Coltheart et al., 2001; Rastle, 2016)—is the fact that the gender star has been in the public for less than a decade. Consequently, older people first encountered star nouns as adults and age of acquisition is higher. Moreover, in its guidelines for GFL, the University of Freiburg explicitly suggests replacing the generically intended masculine with gender-fair alternatives such as the gender star (cf. Schneider, 2022). Thus, students are likely to be more familiar with the star than non-students: Even though the use of GFL in the media and other public domains also increases, the usage of GFL is, in general, less common and less systematic in informal contexts (cf. Gabriel et al., 2018). A gender-fair form useful to further investigate the role of age of acquisition and familiarity is the capital-I form, as this non-orthographic abbreviation of the pair form has already been in use since the 1980s. Comparing word recognition of the capital-I and the gender star form in older readers might allow us to differentiate more thoroughly between the influences of orthography, familiarity, and age of acquisition.
Another interesting finding is that gender stereotypes are activated already on the word level. In line with previous studies (e.g., Sato et al., 2016; Zacharski & Ferstl, 2023), we found consistency effects for the binary forms in the non-student sample: RTs were faster when the semantic stereotype matched the grammatical gender of the role noun. Thus, semantics are activated in the processing of different GFs—even if the meaning of role nouns is irrelevant for the participants’ task. A more subtle semantic influence was found for students with a positive attitude towards GFL. Here, RTs to star nouns decreased more slowly over the duration of the experiment when the role noun was stereotypically male, as compared to stereotypically female and neutral nouns. Further research is needed to investigate potential interactions between attitudes, gender stereotypicality, and non-binary forms such as the gender star.
Even though one strength of our study is that a lexical decision task allows processes on the word level to be tested early on and without potential influences of text context, word recognition is only one component of the reading process. Future studies should investigate the processing of the gender star on the text level using implicit measures such as eye-tracking. It will be moreover interesting to consider further interindividual differences. In particular, in order to find out more about the influence of participants’ gender identity on the processing of GFL, we have to take all genders into account—particularly those identifying beyond the binary. Furthermore, future studies should test how different forms of GFL affect persons with reading disorders, or L2-learners.
Conclusions
The experimental design was well-suited to test the influence of the within-word insertion of a special character in non-binary gender-fair forms on lexical access to role nouns, thus providing a highly valuable addition to studies based on self-report questionnaires. RTs showed differences in lexical access to the gender star compared to binary forms for the older, non-student sample, but not for the younger, student sample. While age—in addition to the (non-)student status of participants—predicted the ease of lexical access to star nouns, subjective attitude ratings did not. The worry that the gender star is more difficult to read is thus not completely unwarranted. However, our findings showed that initial difficulties can be overcome the more a person is confronted with gender-fair alternatives. In our opinion, the results presented are thus promising for proponents of the star: When the gender star becomes more established, traffic signs displaying “Radfahrer*innen bitte absteigen” (Cyclists [star] please dismount) might decrease the male bias—while still guaranteeing readability.