Revisiting gender-fair language and stereotypes – A comparison of word pairs, capital I forms and the asterisk

Silke Schunack; Anja Binanzer

doi:10.1515/zfs-2022-2008

Article Open Access

Revisiting gender-fair language and stereotypes – A comparison of word pairs, capital I forms and the asterisk

Silke Schunack and Anja Binanzer

Published/Copyright: September 6, 2022

Published by

Become an author with De Gruyter Brill

Author Information Explore this Subject

From the journal Zeitschrift für Sprachwissenschaft Volume 41 Issue 2

Abstract

In this paper we replicated two influential studies on gender-fair language that investigated how gender-fair language influences stereotype perception and recall of exemplars. We also updated the original studies to assess new forms of gender-fair language. A first set of studies replicated Gabriel et al. (2008) by eliciting gender proportion ratings for occupational nouns from adult German native speakers for word pairs, capital I forms and the asterisk. Results were mixed with effects mainly for female-biased nouns. Only the non-binary asterisk form (Lehrer*innen) showed an increase of women for male-biased nouns. A third study replicated Stahlberg and Sczesny’s (2001) naming study. Here, the number of women answers was higher than in the original study and increased when using gender-fair language with a larger increase for capital I forms (LehrerInnen) than for word pairs (Lehrer und Lehrerinnen). Overall, the effects of word pair forms were weaker than those of other forms of gender-fair language.

Keywords: generic masculine; gender-fair language; stereotypes; German; feminine plural

1 Introduction

Gender-conscious^[1] language has received an increasing amount of attention in the last few years. However, it is far from a recent phenomenon with its roots in the feminist movement of the 1970s that attempted to increase women’s visibility in language and to challenge gender stereotypes. Some of the earliest studies in German on gender-conscious language and the cognitive representation of men and women date back more than 20 years (Braun et al. 1998; Stahlberg and Sczesny 2001). These early studies found mixed results regarding the effects of different types of gender-conscious language and the circumstances under which these effects appear. Other studies like Gabriel et al. (2008) were not originally designed to investigate effects of gender-conscious language, but nevertheless found them. These studies represent two directions of research: One is more concerned with research questions rooted in societal reality, while the other focuses more on general cognitive representations. In our paper, we combine these two strands of research which allows us to look at gender stereotypes and the recall of exemplars in the same person and to investigate how mental stereotypes and actual exemplars interact, i. e. can strong stereotypes be overcome by salient exemplars? In the years since the first psycholinguistic studies on gender-conscious language, societal and linguistic changes have also led to the continued development of new options of gender-conscious language and an increased awareness of questions surrounding gender in society. In our updated replications of the exemplar naming study by Stahlberg and Sczesny (2001) and the norming study by Gabriel et al. (2008), we included new forms of gender-fair language and addressed questions that had remained unanswered in the original studies.

1.1 Gender stereotypes and language

Stereotypes are expectations related to members of particular social groups which are not based on personal experience. Gender stereotypes in particular reflect what is (stereo)typical for a given gender and involve, for example, typical occupations, behaviors or traits. While general stereotypes are fluid and can depend on context, this is less true for gender stereotypes (Ellemers 2018). The description and prescription of stereotypical behavior often happens indirectly, for example in media portrayals or evaluations (Angelini et al. 2014; Maass 1999). Language is one piece of the puzzle how gender stereotypes are maintained, but does it also have the potential to help overcome them?

The call to look towards language in the fight against gender stereotypes comes from the idea that language and thought are intertwined as expressed in linguistic relativism (Whorf 1956). In this theory, it is assumed that what we express in language somehow influences our thoughts. Grammatical distinctions in a language supposedly motivate the speaker to pay attention to particular characteristics of objects, actions or concepts. Experimental evidence has been presented for different aspects of language, but continues to be a topic of debate (Samuel et al. 2019).

Linguistic forms that are used for person reference are under particular scrutiny for effects on the mental representation of gender. The focus here is on grammatically masculine forms that can either be interpreted specifically – exclusively referring to male persons, or generically – referring to a mixed-gender group or contexts in which gender is irrelevant. Opponents of gender-conscious language argue that the specific and the generic interpretation are equally accessible and using masculine forms is therefore not discriminating. However, multiple studies have shown that the generic masculine is often interpreted as specific leading to less cognitive representation of women (Heise 2000; Irmen and Linner 2005) and processing problems whenever a generic interpretation is required (e. g. Gygax et al. 2008; Irmen and Schumann 2011; Misersky et al. 2019).

Gender-conscious language seeks to weaken gender stereotypes and to increase the mental and linguistic representation of women and non-binary persons. The strategies employed to reach this goal are highly language-specific and include the re-interpretation of already existing words, e. g. singular they in English, the replacement of gender-specific forms, e. g. policeman replaced with police officer, and the creative development of new forms, e. g. the gender-neutral pronoun hen in Swedish. Due to the nature of grammatical gender in German, the focus of gender-conscious language is mostly on nouns and to a much lesser degree on articles and pronouns.

1.2 Linguistic forms of German gender-fair language

Recent legislative changes in Germany introducing a third non-binary gender option, have brought more attention to the topic of gender-conscious language and led to the publication of several handbooks on how to appropriately use gender-conscious language (e. g. Diewald and Steinhauer 2020). However, there is not a single form that has normative character. Instead, many parallel forms co-exist with different origins and degrees of acceptability and there is a continued development of new forms (Kotthoff 2020).

In this article, we will focus on gender-fair language options that can be divided into feminization strategies that stress the feminine form along the masculine form assuming a binary concept of gender, and non-binary forms that seek to represent additional options beyond female and male.

Table 1

Systematic overview of gender-fair language forms in German.

	Linguistic form	Example
1	Word pairs, binary forms, order of nouns random
1a	masculine and feminine forms, connected by a conjunction (and/or)	Lehrer und Lehrerinnen ‘male teachers and female teachers‘
1b	masculine and feminine forms, separated by a forward slash	Lehrer/Lehrerinnen
2	Shorthand version of word pairs, binary forms, masculine form always first
2a	round brackets separate feminine ending from masculine form	Lehrer(innen)
2b	forward slash and hyphen separate feminine ending from masculine form	Lehrer/-innen
2c	forward slash separates feminine ending from masculine form	Lehrer/innen
2d	capitalized I marking the feminine ending	LehrerInnen
3	Shorthand version of word pairs, non-binary forms, masculine form always first
3a	underscore between masculine form and feminine ending	Lehrer_innen
3b	asterisk between the masculine form and the feminine ending	Lehrerinnen*
3c	colon between the masculine form and the feminine ending	Lehrer:innen

For many role nouns the feminine form is derived from the masculine form by adding the ending -in, e. g. Lehrer (m. ‘teacher’) becomes Lehrerin (f.). This derivation also forms the base for the different gender-fair options presented in Table 1. The most straightforward solution is word pairs that explicitly list the masculine and the feminine form in any order (1a, 1b). An advantage of word pairs is the use of existing forms that allow easy production and recall while not violating the rules of German orthography. Using word pairs results in longer and often syntactically more complex sentences rendering this strategy less applicable to formats with limited space. It could also potentially impair sentence processing and comprehension by placing a burden on working memory. However, studies investigating text comprehension and content recall in German have not found any negative impact of word pairs (Blake and Klimmt 2010; Braun et al. 2007; Friedrich and Heise 2019).

There are several shorthand versions of word pairs which merge masculine and feminine forms into one (2a–d). The merger aims to display both forms at once while setting them apart at the same time. This type of gender-fair language works best when the feminine ending -in does not cause any changes to the stem, such as umlaut, leaving the male noun intact. A gender-fair noun such as KöchInnen (‘cooks’) is problematic as the masculine forms Koch (sg.) or Köche (pl.) are not recoverable.

There are also non-binary options that use non-letter graphemes like the underscore (3a), the asterisk (3b) or the colon (3c). These graphemes did not originally have a referential function, instead their new function was established by the LGBTQI+ community and is interpreted differently even within that community. Binary and non-binary forms of gender-fair language as well as the generic masculine are used by different communities of practice and represent different ideologies (Kotthoff 2020).

Several strategies for gender-neutral language also exist, like the use of nominalizations (Lehrende, m./f. ‘people who are teaching’) or composite nouns with suffixes such as -kraft (f.) (lit. ‘force’, Lehrkraft ‘teacher’) or -person (f.) (Lehrperson ‘teaching person’). These suffixes have a fixed grammatical gender, but the resulting compounds are considered semantically gender-neutral. These neutralization strategies are less standardized and less systematic than the gender-fair language examples in Table 1 which is why we decided against using them in our studies.

Despite the rise in psycholinguistic investigations in recent years, not all of the forms introduced above have received equal attention. The most frequently investigated option is word pairs (1a, e. g. Friedrich and Heise 2019; Vervecken et al. 2013), while alternative forms have received less attention (e. g. Blake and Klimmt 2010; Stahlberg and Sczesny 2001). Our studies compare both word pairs and the alternative forms capital I and asterisk.

2 Rating studies on role nouns

The original intention of the rating study by Gabriel et al. (2008) was to provide norms on the perception of gender in 126 role nouns in English, French, and German. The researchers presented the participants with a rating task in which the proportion of men and women was to be estimated. The rating scale consisted of eleven steps with a 10 % increase for one gender, while decreasing the proportion of the other gender at the same rate. The first experimental manipulation was the direction of the rating scale: the direction was either from 100 % male–0 % female to 0 % male–100 % female or vice versa in the direction of reading. The second experimental manipulation was the form in which the words were presented: word pairs (e. g. male teachers/female teachers) or only the generic masculine (e. g. teachers). For the word pairs condition, the order of the two nouns corresponded to the order of the anchors of the scale, i. e. feminine forms were shown below the 0 % male–100 % female anchor.

194 participants (76 German-speaking) completed the questionnaire with the word pair forms and 177 participants (67 German-speaking) completed the questionnaire with the generic masculine. Across all three languages, word pairs received a higher estimate of women than the generic form. However, this effect was modulated by the scale direction. When the left anchor of the scale was 100 % male–0 % female, there was no difference between the generic masculine and word pair forms. When the left anchor was 0 % male–100 % female, the gender-specific form with the feminine noun on the left received a higher percentage of female ratings than the generic form. The authors explained this finding with a higher amount of reflection on the role of women, when the feminine noun is named first.

A later study by Misersky et al. (2014) used only the word pair condition and the same rating task to provide norms on more than 400 nouns in seven languages. They did not find a consistent scale effect in all languages. In those cases where an effect of scale direction was detected, it was in the same direction as in the study by Gabriel et al. (2008). Effects of participant gender were also not consistent across studies and languages.

Both studies focused on the establishment of gender norms and not on how these norms can be affected by the language used in the task. The comparison of generic masculine and word pairs was also dropped from the 2014 study. The finding of an increase of women in the word pairs condition compared to generic masculine has not been replicated since. It is also still unclear whether the beneficial effect of word pairs is due to the explicit mention of a full feminine form or whether forms of gender-fair language that merge masculine and feminine forms, such as capital I forms or the asterisk, achieve the same outcome. This is what our replications investigated.

2.1 Binary forms of gender-fair language

In our first study, we used the gender rating method by Gabriel et al. (2008) to investigate the effects of capital I forms and the feminine plural on stereotypes. If the presence of a feminine ending is sufficient to cause more reflection on women, we should find a similar increase for the capital I form as for the word pairs. If, however, the explicit spell-out of a feminine form is needed, we should not find an increase for the capital I form. For the feminine plural, we expect participants to interpret it as such assigning only 100 % women ratings. The study was preregistered and additional materials can be found on OSF (https://osf.io/fzbkn/).

2.1.1 Method

2.1.1.1 Participants

Participants were recruited through university newsletters, social media channels and personal contacts. In accordance with our preregistered criteria, participants were excluded if they were a minor ( n = 2), reported a history of dyslexia ( n = 4), were non-native speakers of German ( n = 1), had grown up bilingually before the age of 3 ( n = 1) or if they had participated in previous studies on this topic ( n = 2). Participants were also excluded if they showed clear answer patterns ( n = 2). Answering exclusively with 100 % women in the conditions capital I and feminine plural was not treated as an answer pattern as those are valid answers. Students received course credit for their participation, all other participants were not reimbursed.

The final sample consisted of 135 participants: 55 participants were male, 80 were female. Age range was 18 to 72 years ( M = 27, S D = 11.25). 80 of our participants were students enrolled at a German university at the time of testing. Of the non-student group, 20 participants had a university degree and 23 had obtained a high school diploma (Abitur or Fachabitur).

We also asked participants about their familiarity with and attitude towards gender-conscious language. This part of the questionnaire was not completed by four participants.

The majority of the participants ( n = 102, 78 %) reported at least some familiarity with gender-conscious language. Familiarity was distributed unequally across educational backgrounds with higher familiarity rates in more highly educated groups. Despite relatively high familiarity rates, the majority of participants reported no use of gender-conscious language ( n = 92, 70 %). The attitude towards gender-conscious language was distributed almost equally between positive ( n = 41, 31 %), neutral ( n = 42, 32 %) and negative attitudes ( n = 48, 38 %).

2.1.1.2 Design

The study design was a between-subjects design with the single factor language form and the four levels generic masculine (GM, Lehrer ‘teacher’), word pairs (WP, Lehrer und Lehrerinnen), capital I (CI, LehrerInnen) and feminine plural (FP, Lehrerinnen). Unlike Gabriel et al. (2008) and Misersky et al. (2014), we did not manipulate scale direction and only used the scale with 100 % men–0 % women as the left anchor and 0 % men–100 % women as the right anchor. In the conditions generic masculine, capital I and feminine plural the to-be-rated noun was on the left side of the scale (see Figure 1A). In the word pairs condition, the male form was placed on the left side of the scale and the female form on the right side (see Figure 1B). Due to formatting reasons the spelled-out scale could not be put directly above the rating in this condition. Instead a reminder of the instructions repeating the rating scale was given at the beginning of each page together with a visual aid on top of the scale.

Figure 1

Example of the first two items in the rating conditions generic masculine (A) and word pairs (B).

For our items, we selected a total of 80 role nouns from the larger norming study by Misersky et al. (2014). Predefined selection criteria included:

availability of gender-fair forms, (e. g. Lehrlinge ‘apprentices’ is gender-neutral)
no additional hyphenation as in LKW-Fahrer ‘lorry drivers’
spelling as a single word (exclusion of Betreiber eines Partyservice ‘caterers’)

As one of the goals of gender-fair language is to increase women’s visibility, we expect different effects depending on the baseline gender stereotype established by the norming studies. If reading a feminine form or ending leads to active thinking about women, this effect should be stronger for male-biased nouns with little female representation than for female-biased nouns that already have a strong underlying female representation. In order to test this, we selected 20 nouns with a male bias (ratings below .3 in Misersky et al. 2014, e. g. Bauarbeiter ‘construction workers’), 20 nouns with a female bias (ratings above .7, e. g. Nagelpfleger ‘manicurists’) and 40 neutral nouns (ratings between .45 and .55, e. g. Musiker ‘musicians’). Noun bias is a between-items factor.

2.1.1.3 Procedure

The study was conducted using an online questionnaire developed on the platform Sosci Survey (Leiner 2019) and distributed via a link. Participants first saw a welcome page and the consent form followed by a biographical questionnaire. This questionnaire collected data on participant gender, age, language background, potential dyslexia and highest education or student status. After this page, participants were randomly assigned to one of the four language conditions and completed the rating task. After the rating task, participants were again randomly assigned to one of the four language conditions for the naming task reported in Section 3. The experiment ended with three questions regarding familiarity with, attitudes towards and personal adoption of gender-conscious language. Participants could also provide examples of what kind of gender-conscious language they use. Complete debriefing could be requested from the first author. The completion of the questionnaire took approximately 30 minutes.

The instructions for the rating task were the same as in the study by Misersky et al. (2014) and asked the participants to give a rating on an 11-point scale to indicate the ratio of men and women in the social and occupational groups presented. The participants were also instructed to give their rating based on their assumption of the actual ratio and not how they think the ratio should be. We replaced the examples Schauspielerinnen (‘actresses’) and Schauspieler (‘actors’) with Kinder (‘children’), a neutral noun that has no gender-fair forms and could be used for all four conditions without biasing the participants. Following the procedure of Gabriel et al. (2008) and Misersky et al. (2014), we presented 20 role nouns per page in a fixed pseudorandomized order.

2.1.2 Results

The data was processed and analyzed using R 3.6.2 (R Core Team 2019) and the original 1–11 scale provided by the questionnaire software. In order to facilitate comparison to the Gabriel et al. (2008) and Misersky et al. (2014) data, the means and standard deviations reported below were transformed to a 0–1 scale (1 = 100 % women). The number of participants per condition was almost balanced. Mean ratings for the four conditions showed a numeric increase for the gender-fair conditions: M GM = 0.48 ( S D = 0.22), M WP = 0.49 ( S D = 0.25), M CI = 0.56 ( S D = 0.26) and M FP = 0.56 ( S D = 0.25). The much higher means in the capital I and feminine plural condition are due to seven participants that had a rating average of 1 indicating exclusively 100 % women answers.

As the dependent variable was ordinal in nature with equal numeric distance between the steps (i. e. a 10 % increase in women for each step), we ran a cumulative link mixed effects model using the ordinal package (v2019.12-10, Christensen 2019). Based on our design, the starting model contained an interaction between language form and noun bias as fixed effects and random intercepts for subject and item. The random intercept for subject can account for some of the variance introduced by those seven “outlier” participants mentioned above. From this starting model, we followed a stepwise procedure and added random slopes to the random effects structure, however, none of these models converged. There was no main effect of language form (see first three rows of Table 2) and an expected main effect of bias as neutral nouns and nouns with a female bias are supposed to have higher numbers of women than nouns with a male bias. Numerous significant interactions were driven by this bias and are therefore omitted from Table 2, but can be recovered from the analysis file on OSF.

Table 2

Coefficients of the cumulative link mixed effects model with generic masculine (GM) and male bias as reference levels.

	Estimate^a	Std. Error	z-value	p-value^b
Condition (WP^c)	−0.33	0.74	−0.44	n. s.
Condition (CI)	1.28	0.74	1.73	.08 .
Condition (FP)	1.47	0.75	1.95	.05 .
Bias (neutral)	3.68	0.21	17.83	<.001 ***
Bias (feminine)	7.59	0.25	30.73	<.001 ***

Formula: rating.scale ∼ condition*bias + (1 | subject) + (1| item), threshold = equidistant
^aPositive estimates indicate higher ratings for women compared to the reference level.
^bSignificance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘’ 1
^cWP = word pair, CI = capital I form, FP = feminine plural

In order to investigate effects between language form conditions within the same bias level, we calculated pairwise comparisons using the R package emmeans (v1.6.2-1, Lenth 2021). p-value adjustment to compensate for multiple comparisons was calculated using the Tukey method. Table 3 only shows the results of the relevant comparisons to the generic masculine as baseline. There were no differences between the generic masculine and word pairs for any of the noun bias levels. The capital I condition only differed for female-biased nouns and the feminine plural condition differed for both neutral and female-biased nouns.

Table 3

Results of pairwise comparisons between language forms within each noun bias level.

Comparison	Estimate^a	Std. Error	95 % confidence interval	z-value	p-value^b
Noun bias: male
GM^c – WP	0.33	0.75	−1.59 – 2.25	0.44	n. s.
GM – CI	−1.28	0.75	−3.19 – 0.63	−1.73	n. s.
GM – FP	−1.47	0.75	−3.41 – 0.47	−1.95	n. s.
Noun bias: neutral
GM – WP	−0.14	0.74	−2.05 – 1.77	−0.19	n. s.
GM – CI	−1.59	0.74	−3.49 – 0.31	−2.15	n. s.
GM – FP	−2.02	0.75	−3.95 – −0.09	−2.68	.037 *
Noun bias: female
GM – WP	−0.67	0.75	−2.59 – 1.25	−0.89	n. s.
GM – CI	−2.08	0.74	−4.00 – −0.17	−2.8	.026 *
GM – FP	−2.63	0.76	−4.57 – −0.68	−3.47	.003 **

^aPositive estimates indicate lower ratings for women in the second condition of the comparison.
^bSignificance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘’ 1
^cGM = generic masculine, WP = word pairs, CI = capital I form, FP = feminine plural

Figure 2

Comparison of by-item medians in the condition word pairs (WP), capital I (CI) and feminine plural (FP) with the generic masculine as baseline.

Figure 2 visualizes the by-item ratings in the word pair, capital I and feminine plural conditions on the y-axis. Instead of using item means that are strongly affected by the seven “outlier” participants, we used the item median that is less affected by extreme values.

The x-axis displays the item medians in the generic masculine condition. Lower values on the x- and y-axis indicate a male bias, whereas high values indicate a female bias. The black diagonal is calculated from the generic masculine condition and serves as a reference. Any points in the scatter plot above the diagonal indicate medians that are higher than in the generic masculine. Any points below indicate a lower median compared to the generic masculine. The blue line shows the modeled values based on a linear regression with a 95 % confidence band. The overlap between the black diagonal and the blue model lines in all three gender-fair language conditions reflects the non-significant results in Table 3. An overview of all item means and standard deviations by condition can be found on OSF.

2.1.3 Discussion

The results of this rating study using binary gender-fair language forms did not support our hypotheses. While we did find an overall numeric increase in the proportion of women for all three conditions compared to the generic masculine, this increase was not significant.

For the word pair condition, this is in line with the original study by Gabriel et al. (2008) that had found an overall increase for the word pair condition, but not for the scale direction that we used. Using the same task exclusively for word pairs, Misersky et al. (2014) had not found an effect of scale direction for the German group. The role of scale direction and order within the word pairs remains unclear as studies that found an effect used the female-first order (e. g. Vervecken et al. 2013).

We did find a significant effect for the capital I condition, but only for female-biased nouns. The risk of confusing the capital I form with the feminine plural was low and only occurred in three cases that also reported complete unfamiliarity with the concept of gender-conscious language.

We had introduced the feminine plural as a control condition with the expectation that participants would not treat it as a generic form, but instead interpret it as referring exclusively to women. However, only four participants in this condition actually showed this rating behavior. One explanation could be the instructions as we asked the participants to report their estimation of the proportion of men and women in the respective group. With a noun like Bauarbeiterinnen (‘female construction workers’), the number of female construction workers in real life is indeed very low and the stereotype could be so strong that it overrules the grammatical information provided by the feminine plural. Participants also could have treated the feminine plural as an actual generic feminine that has been suggested by some feminist linguists to include women and men in the same way as the generic masculine does. Finally, it is also possible that participants felt that repeating the same answer 80 times was not the expected response and altered their rating behavior after the first few ratings. As participants were free to change their answers as many times as they wanted and 20 ratings were presented on one page, we cannot say for sure whether and how often this behavior happened. A possible solution to this could be the presentation of single items for ratings without the possibility of correcting one’s judgment afterwards.

Figure 3

Reanalysis of the data by Gabriel et al. (2008), comparison of by-items means in the word pair condition with the generic masculine.

Our second hypothesis regarding more impact of gender-fair language for male-biased nouns as opposed to female-biased nouns was not supported by the data. Instead, we found no effects for male-biased nouns. It seems that gender-fair language has a weaker impact on male stereotypes than on female stereotypes, i. e. it increases the female ratings of nouns that are already strongly associated with women more than of nouns that are strongly associated with men. Therefore, our data does not fully support the argument that gender-fair language increases female visibility.

In their original article, Gabriel and colleagues (2008) only reported the average ratings across all items. It is therefore possible, that their 1.4 % increase in female ratings for word pairs in German is also unevenly distributed. Fortunately, the authors made some datasets available with the article that we could analyze in a similar fashion as our own data. The data that we reanalyzed consists of the by-item means and standard deviations for the role nouns in the generic masculine and the word pair condition in German. The data reported was aggregated over both scale directions, so a direct comparison with the scale direction used in our study was not possible. In Figure 3, we used the same plotting technique as in Figure 2 with the black diagonal representing the item means in the generic masculine condition. We see the same trend as for our own data, with no difference for male-biased nouns and a small difference for female-biased nouns.

2.2 A non-binary form of gender-fair language

So far, we investigated the ‘traditional’ forms of gender-fair language that have been investigated in multiple studies. However, there are newer, non-binary forms that have received little to no attention in research so far. In order to see whether a non-binary gender-fair option fares better at increasing the visibility of non-stereotypical genders, we repeated the rating task with asterisk forms (Lehrer*innen) as a new category and dropped word pairs and the feminine plural. Additional material for this study can again be found on OSF (https://osf.io/w3bqv/).

2.2.1 Method

2.2.1.1 Participants

Participants were again recruited through university newsletters, social media channels and personal contacts. The following exclusion criteria applied: participants were a minor ( n = 1), reported a history of dyslexia ( n = 5), were non-native speakers of German ( n = 8) or had grown up bilingually before the age of 3 ( n = 5). Participants with clear answer patterns were also excluded ( n = 5). Students could receive course credit for their participation, non-students were not reimbursed.

The final sample consisted of 157 participants: 51 participants were male, 106 were female. Age range was 18 to 62 years ( M = 27, S D = 11.14). 58 participants were students enrolled at a German university at the time of testing. The majority of the sample had a high school diploma (Abitur or Fachabitur, n = 96) as their highest degree.

Familiarity with gender-conscious language was assessed differently in the student and non-student sample. In the student sample, nine participants (15 %) reported no familiarity with gender-conscious language. 27 participants (46.5 %) reported to use gender-conscious language in official documents, like course work, and nine participants reported using it in everyday situations. 27 participants reported to not use gender-conscious language at all. In the non-student sample, 36 participants (36 %) reported no knowledge of gender-conscious language. 70 participants (71 %) reported no use of gender-conscious language. Most participants in the non-student sample had a neutral attitude towards this topic ( n = 39, 39 %), followed by positive attitudes ( n = 31, 31 %), and negative attitudes ( n = 24, 24 %). The larger non-student sample can therefore be characterized by less familiarity with gender-conscious language and less use when compared to the student sample. When we compare these attitudes with the preceding questionnaire (see Section 2.1.1.1), we see that familiarity does not necessarily lead to adoption of gender-conscious language and that it is a topic with very divided opinions.

2.2.1.2 Design

The study design was again a simple between-subjects design with the single factor language form and three levels: the generic masculine (GM, Lehrer) as a baseline, the capital I form (CI, LehrerInnen) as a binary form, and the asterisk (AS, Lehrer*innen) as a non-binary form. Gender bias of the selected nouns again served as a between-items factor.

The scale was the same as in the previous study with 100 % men–0 % women as the left anchor and the critical noun always to the left of the scale. We did not provide participants with an answer option for a third gender in order to keep the answer design constant across studies and also not to bias the participants towards overestimating the amount of third gender persons.

The same 80 items from the previous study were used in the same pseudorandomized order.

2.2.1.3 Procedure

The study was again conducted using the Sosci Survey questionnaire platform. After participants had given consent to their participation, they received the same instructions as in the previous study. Participants then completed the rating task that consisted of four pages of ratings with 20 role nouns per page. The questionnaire ended with questions on biographical data and attitudes towards gender-conscious language. Complete debriefing could be requested from the first author. The duration of the whole questionnaire was less than 15 minutes.

2.2.2 Results

Processing and modeling of the data was again done on the 0–11 scale provided by the questionnaire software, while visualization was done on the transformed 0–1 scale. The number of participants per language form was not fully balanced. There were slightly fewer participants in the capital I condition ( n = 48) as compared to the generic masculine ( n = 54) and the asterisk ( n = 55). Mean ratings for the three conditions showed a small numeric increase for the gender-fair conditions: M GM = 0.48 ( S D = 0.24), M CI = 0.52 ( S D = 0.24) and M AS = 0.51 ( S D = 0.24).

We again ran a cumulative link mixed effects model using the ordinal package (Christensen 2019). The modeling procedure was the same as in the previous study leading to the same final model. This time we did find a main effect of language form and the expected main effect of noun bias. The interactions are again excluded from Table 4.

Table 4

Coefficients of the cumulative link mixed effects model with generic masculine (GM) and male bias as reference levels.

	Estimate^a	Std. Error	z-value	p-value^b
Condition (CI^c)	0.23	0.1	2.27	.023 *
Condition (AS)	0.35	0.1	3.63	<.001 ***
Bias (neutral)	3.1	0.18	17.3	<.001 ***
Bias (feminine)	6.42	0.21	30.23	<.001 ***

Formula: rating.scale ∼ condition* bias + (1 | subject) + (1| item), threshold = equidistant
^aPositive estimates indicate higher proportions of women compared to the reference levels.
^bSignificance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘’ 1
^cCI = capital I form, AS = asterisk

Table 5

Results of pairwise comparisons between language forms within each noun bias level.

Comparison	Estimate^a	Std. Error	95 % confidence interval	z-value	p-value^b
Noun bias: male
GM^c–CI	−0.23	0.1	−0.46 – 0.01	−2.27	.06
GM–AS	−0.35	0.1	−0.58 – −0.12	−3.63	<.001 ***
Noun bias: neutral
GM – CI	−0.34	0.08	−0.52 – −0.15	−4.26	<.001 ***
GM – AS	−0.27	0.07	−0.45 – −0.09	−3.52	.0013 **
Noun bias: female
GM – CI	−0.67	0.1	−0.91 – −0.44	−6.69	<.001 ***
GM – AS	−0.29	0.1	−0.52 – −0.05	2.89	.011 *

^aPositive estimates indicate lower ratings for women in the second condition of the comparison.
^bSignificance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘’ 1
^cGM = generic masculine, CI = capital I form, AS = asterisk

We again calculated pairwise comparisons for each language form condition within each bias level. Table 5 again shows only the relevant comparisons to the generic masculine. For the male-biased nouns, there is a significant effect only for the asterisk condition. For neutral and female-biased nouns both gender-fair conditions differ from the generic masculine.

Figure 4 visualizes the mean by-item ratings in the capital I and asterisk conditions on the y-axis. It follows the same rationale as Figures 2 and 3 with the black diagonal indicating the means in the generic masculine. For the capital I condition, we see a highly similar pattern as in Figure 2, but with a larger deviation from the generic masculine baseline in the female-biased nouns. In the asterisk condition, we see a different pattern with a larger deviation from the generic masculine in the male-biased nouns than in female-biased nouns. An overview of all item means and standard deviations by condition can be found on OSF.

Figure 4

Comparison of by-item means in the capital I condition (CI) and the asterisk condition (AS) with the generic masculine.

2.2.3 Discussion

Unlike in the previous study with four levels of the factor language form, this time we did find an overall significant increase for both gender-fair language options. This suggests that also single-word forms can increase the mental representation of women in a rating task. It seems that the explicit spell-out of the feminine form as in the word pairs is not necessary for this effect. The larger sample size per condition could also be responsible for finding the effect.

The visualization of the ratings in Figure 4 shows that the effect goes in opposite directions for the capital I form and the asterisk. Just like in the previous study, the capital I form does not weaken existing stereotypes, but actually exacerbates them by increasing the number of women estimates in already female-biased nouns. The asterisk form on the other hand shows the biggest increase in women estimates for male-biased nouns and a much smaller increase for neutral and female-biased nouns. This increase for male-biased nouns is the only evidence that supports the idea that gender-fair language can increase the mental activation of non-stereotypical genders. However, while the increase for female-biased nouns was much smaller in the asterisk condition than in the other gender-fair options, we did not find an increased representation of the non-biased male gender. This suggests that the generic masculine can be considered the baseline for the proportion of men in any given role noun independent of gender bias.

Why did we find an increased proportion of women for male-biased nouns only in the asterisk condition? There are several potential explanations. First, compared to the capital I form and word pairs, the asterisk is orthographically unique as it introduces a non-letter grapheme into the word. This non-letter grapheme is visually more salient than the capital I form that could potentially be interpreted as a lowercase l. Readers are also more familiar with capital letters inside words as they are also used in abbreviations and acronyms. Second, the asterisk is a comparatively recent form that could attract more attention purely based on its novelty. Third, the asterisk form was the only form investigated that has a non-binary interpretation. Assuming this non-binary interpretation is known to the participants, the asterisk could signal diversity more strongly than binary options. Future studies could tease apart these three explanations by testing additional non-binary options like the gender gap (Lehrer_innen) or the colon form (Lehrer:innen) that both use non-letters, but differ from the asterisk with regard to how established they are. Additionally, a version of the rating task that allows for estimates of non-binary persons could shed light onto the different effects of binary and non-binary forms of gender-fair language.

3 Naming study involving free recall of exemplars

While rating studies investigate rather abstract stereotypical concepts, there are also studies on whether gender-conscious language can influence the recall and naming of specific exemplars. Stahlberg and Sczesny (2001) report altogether four studies that are among the earliest studies to investigate options of gender-conscious language beyond word pairs, among them neutralizations and the capital I form. All four studies either asked participants to name specific persons in response to a cued category or measured the response time in a category matching task. The results across all four studies pointed in the same direction: less mental representation of women in the generic masculine compared to gender-conscious forms. For our replication, we chose the naming study of famous persons (Study 3 in Stahlberg and Sczesny 2001). In this study, 90 participants had to fill in 16 questions with the four critical questions asking them to name three exemplars of the categories Sportler (‘athlete’), Sänger (‘singer’), Politiker (‘politician’) and TV-Moderator (‘tv show host’). The categories were either presented in the generic masculine (Sportler), as word pairs (Sportlerinnen und Sportler) or in the capital I form (SportlerInnen). The authors found a main effect of gender-fair language forms on the number of women named compared to the generic masculine. However, this effect was carried by the capital I forms as there was no statistical difference between the generic masculine and the word pairs. The authors attribute this stronger effect for the capital I form to its novelty and its orthographic markedness that makes it particularly salient. Overall, the number of women named in the study was fairly low at M = 2.37 out of 12 total responses for the generic masculine, M = 2.67 for the word pairs, and M = 4.72 for the capital I forms. The authors also found gender differences as female participants named more women than male participants in each of the language conditions. The authors do not provide separate analyses for the different categories which is unfortunate because when considering the norms on gender bias provided by Gabriel et al. (2008) and Misersky et al. (2014), we can see that the categories athlete and singer are neutral, whereas politician is a male-biased noun. This underlying gender bias can potentially influence the mental availability of specific exemplars.

This study is roughly 20 years old and societal changes have since lead to more representation of women in the categories in question and more familiarity with the concept of gender-conscious language in the general public. We therefore expect a general increase in the number of women named across all conditions and a less extreme result for the capital I forms as the novelty effect should have worn off by now. Assuming that we can elicit more woman answers, we can also run separate analyses for the different categories with their respective biases. Additional information, data and materials can be found on OSF (https://osf.io/g3wx8/).

3.1 Method

3.1.1 Participants

The participants were the same as reported in Section 2.1.1.1, except for four participants who did not complete the naming task. The final sample then consisted of 131 participants: 53 were male, 78 were female. Age range was 19 to 72 years ( M = 27, S D = 11.36).

3.1.2 Design

The design followed the original study as reported in Stahlberg and Sczesny (2001) by using generic masculine (Sportler ‘athlete’), word pairs (Sportler oder Sportlerinnen) and capital I forms (SportlerInnen) as levels of the independent variable language form. In the word pairs condition, we used the same order as in the rating study with the masculine form named first. We added a feminine plural condition (Sportlerinnen) in order to check whether participants confused capital I forms with the feminine plural and to check for possible ceiling effects. The variable language form was a between-subjects variable.

As in the original study, participants had to name three exemplars of each category. We excluded the category TV-Moderator (‘tv show host’) and added the categories Schauspieler (‘actor’) and Superheld (‘superhero’). We felt that tv show host might be a slightly outdated category and could cause problems regarding the fit of certain exemplars, e. g. presenters on video channels and other non-linear forms of media. We replaced it with the category actor that we assumed to be neutral regarding its bias and accessible to participants of all ages. In order to check the effect of gender-fair language on a category that is predominantly male and has been criticized for its lack of portrayal of women, we included the category superhero.

Drawing upon the bias values from the norming study by Misersky et al. (2014), three categories are neutral (athlete, singer, and potentially actor), and the other two categories have a male bias (politician, superhero). We did not include a female-biased category as they were either not suitable for the task (e. g. clairvoyant) or did not have a gender-fair form (e. g. model being neuter in German).

The dependent variable was the number of women per answer given, i. e. the odds of a woman answer, instead of the average of absolute numbers of woman answers as in the original study.

3.1.3 Procedure

The naming study followed after the rating task reported in Section 2.1. The instructions for the naming task asked for the participant’s help in constructing materials for a future study that included well-known personalities. Three exemplars for each of the five categories should be named. National as well as international personalities were explicitly allowed, just as living or dead personalities and those who had already ended their careers. This was done to ensure that participants of all ages could contribute equally well and did not feel pressured to name a current personality.

All five categories were presented on the same page in the order: actor, singer, athlete, politician and superhero. Each category was introduced with a prompt Nennen Sie bitte drei… (‘Please name three...’) followed by three slots into which answers could be typed. To make sure that participants did not accidentally skip a category, one answer in each category was mandatory.

3.2 Results

3.2.1 Data preprocessing and coding

Before entering into the analysis, the naming data was checked for answer eligibility. 66 data points were empty answers or other obvious placeholders and were excluded. The remaining 1899 answers were then coded for eligibility and whether the answer given was a woman. Eligible answers were defined as a real or fictional person with a clear gender identity who is uniquely identifiable by the name given in the answer. 12 data points were ineligible answers. Data analysis was conducted on a dataset with 1887 answers, 905 of which were woman answers. There were no non-binary answers. Data loss due to empty or ineligible answers was about 4 %.

3.2.2 Data analysis

The number of participants per language condition was fairly balanced. Using the lme4 package (v1.1-27.1, Bates et al. 2015), we calculated a binomial generalized linear mixed effects model with treatment contrast coding for language form. The dependent variable was the log-odds ratio of a woman answer for any eligible answer, i. e. the odds of providing a woman as an answer against a man transformed on the logarithmic scale. If the odds of a woman and a man answer were equal, the log-odds ratio would be 0.

When building the model, we started out with the basic model with language form as fixed effect and a random intercept for subject. Adding random slopes for subjects did not lead to an improved model fit.

In Table 6, the estimate for the intercept, i. e. the condition generic masculine, is a log-odds ratio of −1.14. If we transform the log-odds back to regular odds, the odds of getting a woman answer in the generic masculine are 0.32, meaning one woman answer for every three man answers. The positive values of the estimates for the other three conditions indicate that the odds of a woman answer are higher than in the generic masculine.

Table 6

Results of the generalized linear mixed effects model with generic masculine (GM) as the reference level.

	Estimate	Std. Error	z-value	p-value^a
Intercept (GM)	−1.14	0.25	−4.51	<0.001 ***
Condition (WP^b)	0.49	0.35	1.42	0.16
Condition (CI)	0.91	0.35	2.57	<0.01 **
Condition (FP)	3.86	0.43	9.01	<0.001 ***

Formula: N_female ∼ 1 + condition + (1 | subject), family = binomial
^aSignificance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘’ 1
^bWP = word pairs, CI = capital I form, FP = feminine plural

Table 7

Results of the pairwise comparisons of the glm model given on the log-odds ratio scale.

Comparison	Estimate	Std. Error	95 % confidence interval	z-value	p-value^a
GM^b – WP	−0.49	0.35	−1.39 – 0.4	−1.41	n. s.
GM – CI	−0.91	0.35	−1.82 – −0.003	−2.58	.049 *
GM – FP	−3.88	0.43	−4.98 – −2.77	−9.01	<.001 ***
WP – CI	−0.42	0.35	−1.31 – 0.47	−1.22	n. s.
WP – FP	−3.38	0.42	−4.45 – −2.32	−8.15	<.001 ***
CI – FP	−2.96	0.41	−4.03 – −1.9	−7.16	<.001 ***

^asignificance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘’ 1
^bGM = generic masculine, WP = word pairs, CI = capital I form, FP = feminine plural

Figure 5

Proportion of woman answers per language form with 95 % bootstrapped confidence intervals.

As in the previous analyses, we then used the emmeans package (Lenth 2021) to calculate pairwise comparisons between the different conditions (see Table 7). As expected, the feminine plural is significantly different from all other three conditions with a large increase in odds of receiving a woman as an answer. The capital I form differs significantly from the generic masculine. The word pair form does not differ from neither the generic masculine nor the capital I form, instead it seems to occupy a mid-way position. This can also be seen in Figure 5 that uses the proportion of woman answers for visualization.

3.2.3 Exploratory analysis

Additional exploratory analyses that separately investigated age, gender, student status, and data from the gender-conscious language questionnaire as fixed effects did not reveal a significant effect for any of these factors.

As the answer categories differed with regard to their gender bias, we ran an exploratory analysis to check for differences between the categories. For the categories athlete and superhero, the number of answers dropped markedly in the feminine plural condition to 74 and 87 out of 99 possible answers suggesting difficulties with these categories.

The proportion of woman answers across language forms showed three subgroups of categories: actor as a neutral category ( M = 0.5), singer and politician with comparatively high numbers of woman answers ( M singer = 0.59, M politician = 0.57), and athlete and superhero with low numbers of woman answers ( M athlete = 0.34, M superhero = 0.37).

A binomial generalized linear effects model with category as a fixed effect supported this same grouping (see also Figure 6).

Figure 6

Proportion of woman answers per language form per category.

3.3 Discussion

Our hypothesis that twenty years after the original study we would find a generally higher proportion of woman answers was supported by the data. To better compare the means reported by Stahlberg and Sczesny (2001) with the proportions used in our study, we transformed their means to proportions. The resulting mean proportions are M GM = 0.2, M WP = 0.22, and M CI = 0.39 compared to M GM = 0.29, M WP = 0.36, and M CI = 0.44 in our study. Our hypothesis regarding the capital I form was also supported by the data. It was not confused with the feminine plural, as the feminine plural yielded nearly twice as many woman answers as the capital I form ( M FP = 0.85).

Just as in the original study, we do not find a significant difference in proportion of woman answers between the generic masculine and the word pairs. This could be due to the order in which the nouns appeared with the masculine form mentioned in the more prominent first position. We do, however, see a numerical advantage for the word pairs that could potentially increase with the reversed order of mentioning.

Judging by the amount of null answers, the feminine plural seems to be a more difficult condition compared to the other three. This can be explained by the expectations regarding eligible answers. The feminine plural suggests that three woman answers are expected, whereas in the remaining three conditions man and woman answers are equally valid answers. However, there were still man answers given in the feminine plural condition possibly to avoid defaulting to a null answer.

Answer patterns in the categories were as expected for three out of the five categories. Actor and singer were overall neutral categories and superhero showed a clear male bias. Based on the norming data by Misersky et al. (2014), we would have expected the category politician to also show a male bias. However, in our data we found it to behave like a neutral noun. This can be explained by the social circumstances. The study was run in Germany in 2019 with Angela Merkel as chancellor for the past fifteen years. Unsurprisingly, Angela Merkel was not only the most frequent answer in the category politician, but also out of all answers (107 mentions). This suggests that a comparatively strong gender stereotype can be overruled by a strong exemplar of the non-stereotypical gender, given sufficient exposure.

The opposite pattern emerged for the category athlete, which is a neutral noun in Misersky et al.’s (2014) norming data, but behaved like the male-biased superhero in our study. The two most frequently named athletes are emblematic of the whole category: Thomas Müller, an active male soccer player (24 mentions), and Steffi Graf, a long-retired female tennis player (23 mentions). Men received multiple mentions in a diverse set of sports, whereas this was restricted to tennis and swimming for women. This result probably has two explanations: First, it reflects a gender (and sports) bias in the German sports media coverage that is comparable to findings for other countries (Coche 2015) making male exemplars cognitively more easily available. Second, it reflects the gender biases of particular sports, also in line with previous studies (Plaza et al. 2017). Athlete is therefore the opposite example to politician in that the actual exemplars are much more stereotyped than the underlying representation.

4 Overall conclusions

The first aim of the studies reported in this article was to replicate the findings by Gabriel et al. (2008) and Stahlberg and Sczesny (2001). In the rating study reported in Section 2.1, we were not able to replicate the significant increase in proportion of women for the word pair condition compared to the generic masculine. We did, however, find a numerical trend pointing in the same direction. A reanalysis of the original data from Gabriel et al. (2008) also showed the same trend of a larger difference between ratings for female-biased nouns that we found in our data. A possible explanation for the absence of a significant difference could be the order of nouns in the word pairs. We had opted for the order with the male noun as the left anchor for which there had not been an effect in the original study. However, no effect of scale direction was found in the Misersky et al. (2014) study and its role when using word pairs in the context of gender-conscious language remains unclear.

We did replicate the findings of the naming study by Stahlberg and Sczesny (2001) as we found more women answers in the word pairs condition and the capital I condition compared to the generic masculine. Just as in the original study, this increase was significant only in the capital I condition. The general increase of women answers across all language forms is potentially related to a greater societal visibility of women compared to 20 years ago. This higher salience of women also reduces the gap between the generic masculine and the word pairs on the one side and the capital I forms on the other side making the difference between the three conditions less extreme than in the original study.

When bringing together the results from the gender stereotype ratings and exemplar naming task, we found that strong stereotypes can be overruled by highly salient exemplars, but originally neutral nouns can also become highly stereotyped under biased input that reduces the salience of women and other minorities.

Our second aim was to evaluate different forms of gender-fair language. The three forms, word pairs, capital I and the asterisk, do indeed behave differently in the two tasks. The popular option of word pairs seems to be the weakest choice when it comes to increasing the visibility of women, because while there was a numerical advantage for the word pairs, this trend was not statistically reliable in either study. This is in contrast to other studies that have found effects for word pairs (e. g. Vervecken et al. 2013) and suggests that the noun order within the word pairs and the type of task play a decisive role. Further research is therefore needed as the dependence of the effect on specific tasks and orders would weaken the overall applicability of word pairs as a means of gender-fair language expression.

The capital I option has shown a beneficial effect in both tasks and was rarely confused with the feminine plural. However, it had only weak, if any effects on strong male stereotypes, instead it increased the proportion of women for occupations in which women already are highly visible. The only option that showed some effect for male stereotypes was the non-binary asterisk option. Due to its orthographic markedness, the asterisk has the potential to disrupt language processing more strongly than other options. It has to be noted, however, that the asterisk was used in tasks that only required binary judgments and future studies will have to find ways to assess the cognitive availability and inclusion of non-binary persons without biasing participants.

Given the multitude of gender-conscious options in German, more research is needed to investigate their strengths and weaknesses across different tasks and even more so across different populations as the data from our questionnaires suggest lower familiarity and adoption rates and more critical attitudes in non-academic populations. In general, gender-conscious language can be one tool to raise awareness and increase the visibility of women and other genders, but in order to overcome or at least weaken gender stereotypes, enforcement through positive, non-stereotypical exemplars is needed as well.

Acknowledgment

The authors wish to thank the students of the class “Empirisches Praktikum” who conducted the studies reported in this article as part of their class requirement.

References

Angelini, James R., Paul J. MacArthur & Andrew C. Billings. 2014. Spiraling into or out of stereotypes? NBC’s primetime coverage of male figure skaters at the 2010 Olympic Games. Journal of Language and Social Psychology 33(2). 226–235. doi:10.1177/0261927X13508559.Search in Google Scholar

Bates, Douglas, Martin Mächler, Ben Bolker & Steve Walker. 2015. Fitting linear mixed-effects models using lme4. Journal of Statistical Software 67(1). doi:10.18637/jss.v067.i01.Search in Google Scholar

Blake, Christopher & Christoph Klimmt. 2010. Geschlechtergerechte Formulierungen in Nachrichtentexten. Publizistik 55(3). 289–304. doi:10.1007/s11616-010-0093-2.Search in Google Scholar

Braun, Friederike, Anja Gottburgsen, Sabine Sczesny & Dagmar Stahlberg. 1998. Können Geophysiker Frauen sein? Generische Personenbezeichnungen im Deutschen. Zeitschrift für Germanistische Linguistik 26(3). 265–283. doi:10.1515/zfgl.1998.26.3.265.Search in Google Scholar

Braun, Friederike, Susanne Oelkers, Karin Rogalski, Janine Bosak & Sabine Sczesny. 2007. „Aus Gründen der Verständlichkeit...“: Der Einfluss generisch maskuliner und alternativer Personenbezeichnungen auf die kognitive Verarbeitung von Texten. Psychologische Rundschau 58(3). 183–189. doi:10.1026/0033-3042.58.3.183.Search in Google Scholar

Christensen, R. H. B. 2019. ordinal – regression models for ordinal data. R package version 2019.12-10. https://CRAN.R-project.org/package=ordinal.Search in Google Scholar

Coche, Roxane. 2015. The amount of women’s sports coverage on international sports news websites’ home pages. Electronic News 9(4). 223–241. doi:10.1177/1931243115604882.Search in Google Scholar

Diewald, Gabriele & Anja Steinhauer. 2020. Handbuch geschlechtergerechte Sprache. Wie Sie angemessen und verständlich gendern. Berlin: Dudenverlag.Search in Google Scholar

Ellemers, Naomi. 2018. Gender stereotypes. Annual Review of Psychology 69(1). 275–298. doi:10.1146/annurev-psych-122216-011719.Search in Google Scholar

Friedrich, Marcus C. G. & Elke Heise. 2019. Does the use of gender-fair language influence the comprehensibility of texts? Swiss Journal of Psychology 78(1–2). 51–60. doi:10.1024/1421-0185/a000223.Search in Google Scholar

Gabriel, Ute, Pascal M. Gygax, Oriane Sarrasin, Alan Garnham & Jane Oakhill. 2008. Au pairs are rarely male: Norms on the gender perception of role names across English, French, and German. Behavior Research Methods 40(1). 206–212. doi:10.3758/BRM.40.1.206.Search in Google Scholar

Gygax, Pascal M., Ute Gabriel, Oriane Sarrasin, Jane Oakhill & Alan Garnham. 2008. Generically intended, but specifically interpreted: When beauticians, musicians, and mechanics are all men. Language and Cognitive Processes 23(3). 464–485. doi:10.1080/01690960701702035.Search in Google Scholar

Heise, Elke. 2000. Sind Frauen mitgemeint? Eine empirische Untersuchung zum Verständnis des generischen Maskulinums und seiner Alternativen. Sprache & Kognition 19(1/2). 3–13. doi:10.1024/0253-4533/.19.12.3.Search in Google Scholar

Irmen, Lisa & Ute Linner. 2005. Die Repräsentation generisch maskuliner Personenbezeichnungen. Eine theoretische Integration bisheriger Befunde. Zeitschrift fur Psychologie 213(3). 167–175. doi:10.1026/0044-3409.213.3.167.Search in Google Scholar

Irmen, Lisa & Eva Schumann. 2011. Processing grammatical gender of role nouns: Further evidence from eye movements. Journal of Cognitive Psychology 23(8). 998–1014.10.1080/20445911.2011.596824Search in Google Scholar

Ivanov, Christine, Maria B. Lange, Tabea Tiemeyer & Martin Ptok (eds.). 2019. Geschlechtergerechte Sprache in der Wissenschaft, Gebrauch und Motivation (Gender(ed) Thoughts, Working Paper Series, Vol. 2). doi:10.3249/2509-8179-gtg-9.Search in Google Scholar

Kotthoff, Helga. 2020. Gender-Sternchen, Binnen-I oder generisches Maskulinum, ... (Akademische) Textstile der Personenreferenz als Registrierungen? Linguistik online 103(3). 105–127.10.13092/lo.103.7181Search in Google Scholar

Leiner, D.J. 2019. SoSci Survey (Version 3.1.06). https://www.soscisurvey.de.Search in Google Scholar

Lenth, Russell V. 2021. emmeans: Estimated Marginal Means, aka Least-Squares Means. R package version 1.6.2-1. https://CRAN.R-project.org/package=emmeans.Search in Google Scholar

Maass, Anne. 1999. Linguistic intergroup bias: Stereotype perpetuation through language. Advances in experimental and social psychology 31. 79–121.10.1016/S0065-2601(08)60272-5Search in Google Scholar

Misersky, Julia, Asifa Majid & Tineke M. Snijders. 2019. Grammatical gender in German influences how role-nouns are interpreted: Evidence from ERPs. Discourse Processes 56(8). 643–654. doi:10.1080/0163853X.2018.1541382.Search in Google Scholar

Misersky, Julia, Pascal M. Gygax, Paolo Canal, Ute Gabriel, Alan Garnham, Friederike Braun, Tania Chiarini, et al.2014. Norms on the gender perception of role nouns in Czech, English, French, German, Italian, Norwegian, and Slovak. Behavior Research Methods 46(3). 841–871. doi:10.3758/s13428-013-0409-z.Search in Google Scholar

Plaza, Mélissa, Julie Boiché, Lionel Brunel & François Ruchaud. 2017. Sport = male … but not all sports: Investigating the gender stereotypes of sport activities at the explicit and implicit levels. Sex Roles 76(3–4). 202–217. doi:10.1007/s11199-016-0650-x.Search in Google Scholar

R Core Team. 2019. R: A language and environment for statistical computing. Vienna: R Foundation for Statistical Computing. https://www.R-project.org/.Search in Google Scholar

Samuel, Steven, Geoff Cole & Madeline J. Eacott. 2019. Grammatical gender and linguistic relativity: A systematic review. Psychonomic Bulletin and Review 26(6). 1767–1786. doi:10.3758/s13423-019-01652-3.Search in Google Scholar

Stahlberg, Dagmar & Sabine Sczesny. 2001. Effekte des generischen Maskulinums und alternativer Sprachformen auf den gedanklichen Einbezug von Frauen. Psychologische Rundschau 52(3). 131–140. doi:10.1026//0033-3042.52.3.131.Search in Google Scholar

Vervecken, Dries, Bettina Hannover & Ilka Wolter. 2013. Changing (s)expectations: How gender fair job descriptions impact children’s perceptions and interest regarding traditionally male occupations. Journal of Vocational Behavior 82(3). 208–220. doi:10.1016/j.jvb.2013.01.008.Search in Google Scholar

Whorf, Benjamin Lee. 1956. Language, thought, and reality: Selected writings of Benjamin Lee Whorf. Ed. John B. Carroll. Cambridge, MA: The MIT Press.Search in Google Scholar

Received: 2021-03-18

Accepted: 2022-02-11

Published Online: 2022-09-06

Published in Print: 2022-11-30

This work is licensed under the Creative Commons Attribution 4.0 International License.

Articles in the same Issue

https://doi.org/10.1515/zfs-2022-2008

Keywords for this article

generic masculine; gender-fair language; stereotypes; German; feminine plural

Creative Commons

BY 4.0