With regard to general gender differences in language use, a considerable body of empirical research has systematically compared male versus female speech, for a range of different linguistic phenomena. For instance, Newman et al. (2008), in a study comparing 14,000 text samples from male and female speakers of English, found profound gender differences, with texts from female speakers containing significantly more words related to social processes and male speakers instead producing significantly more words related to properties of inanimate objects and impersonal topics. Leaper and Ayres (2007), in a meta-analytic study comparing male and female speech, found small but significant gender effects for talkativeness, affiliative speech, and assertive speech.

A substantial number of studies have investigated Lakoff's (1975) claim that male language is relatively more assertive, while female language is more tentative and polite. Leaper and Robnett (2011), in a meta-analysis comprising a total of twenty-nine studies investigating Lakoff's claim, indeed found a small but significant gender effect, with more tentative language in female than in male speech. Specific linguistic phenomena aside, Labov (1990, and much subsequent work), based on a review of several empirical studies investigating the use of standard vs. non-standard linguistic forms by men and women, argued that men are less affected by negative social stigmas attached to the use of non-standard linguistic features than women. Men thus, at least in linguistically stable situations, produce a higher frequency of non-standard linguistic features than women.

The studies mentioned so far have all investigated general gender effects independent of the specific communicative situation. However, there is at least some evidence that properties of the communicative situation may modulate gender effects. For example, Reid et al. (2003) investigated gender differences in elicited conversations between male and female students, and experimentally manipulated the saliency of gender. When gender was highly salient, profound differences in language use emerged, with female speakers producing substantially more tentative language than male speakers. In contrast, when gender was not salient and a shared student identity was made salient instead, the observed gender effect disappeared, with male and female speakers producing equal amounts of tentative language.

How are gender effects in language use explained at a theoretical level? In variationist sociolinguistics, gender differences are considered to be related to the theoretical construct of social meaning (e.g., Gal, 2023; Labov, 1993, 2010; Ochs, 1992; Silverstein, 2003). Such theoretical accounts assume that meaning is socially situated²2

See, for example, Eckert (2012) for an in-depth review of the three waves of sociolinguistics.

With regard to the process of deriving social meaning, Labov (1993) suggests a distinction between lexical and phonological versus grammatical features, with only the former carrying social meaning and displaying social stratification. This is supported by research on diverse linguistic features and contexts, which showed the same distinction (e.g., Labov, 2001; Levin et al., 1994; Trudgill, 1986). Other research suggests a more nuanced picture of the influence of social meaning conveyed through grammatical features (e.g., Acton & Potts, 2014; Levon & Buchstaller, 2015; Levon & Fox, 2014; Moore, 2020; Sauermann et al., 2023).

Based on research on variation in language contact, several authors have proposed hierarchies of linguistic features (Matras, 2009; Wiese et al., 2022). Matras (2009, especially pp. 58–59 and 310–314) argues that the use of overt features (phonological or lexical features) is consciously controllable and influenced by sociolinguistic factors such as gender or language prestige, whereas the use of abstract features (grammatical features) is less consciously controlled and thus it is less easy to distinguish and to derive indexicality. The Social Salience Hierarchy proposed by Wiese et al. (2022) assumes that different types of noncanonical (for instance, Namibian German) linguistic features differ with regard to the salience of the social meaning they carry. The account distinguishes between several types of linguistic features, and puts them in a hierarchical order according to how salient the conveyed social meaning is. Crucially for our purposes, the account assumes that social meaning is relatively more salient for lexical features (such as the Namibian German borrowing trolley 'shopping cart' versus its Standard German equivalent Einkaufswagen 'shopping cart') than for grammatical features (such as noncanonical word order, e.g., nonstandard placement of negation).

Most studies have addressed gender effects in Western linguistic communities with a largely monolingual habitus. While gender effects also occur in multilingual communities and settings (e.g., recent migration), the picture is not as clear-cut. In a recent literature review, Meyerhoff and Birchfield (2019) highlight several points regarding language contact and gender stratification: Mainly, gender stratification in language contact shows unpredictable effects, often incompatible with principles of gender and language variation/change in monolingual speech communities. In multilingual settings with a high degree of language contact, individuals do not adopt gendered stances or attitudes as often or as consistently as in monolingual speech communities. However, when they do, they tend to rely on different sets of linguistic features, and the speech communities assign gendered labels to those features. Also, social factors other than gender are of greater practical importance in language contact contexts. Additionally, Meyerhoff and Birchfield (2019) also points towards a discrepancy between findings from qualitative versus quantitative studies in this field: Qualitative studies show ample evidence of gender effects or rather on individual speakers of speech communities attributing certain language use as gendered. However, quantitative studies, in contrast, struggle to find evidence that gender is a significant predictor of variation in language contact or show inconsistent effects.

A small number of recent studies have investigated loan word use in Namibian German (Bracke, 2021; Radke, 2021; Zimmer, in press). While these studies did not primarily focus on gender effects, they report additional supplementary analyses that are potentially relevant with respect to gender effects in Namibian German. Bracke (2021), in a study based on data from the free speech subcorpus of the DNam corpus (see section “Data” for further information on the subcorpora), found a general preference for English over Afrikaans loan words. Also, adolescents produced a higher frequency of loan words than adults. Interestingly, the results also revealed differences between particular male and female speaker groups: While female adolescents preferred English loan words, male adults instead showed a higher frequency of Afrikaans loan words than English loan words. A similar pattern was also found by Radke (2021) for computer-mediated communication in Namibian German expats: Male speakers were more likely to use Afrikaans loan words than female speakers. For English loan words, in contrast, no gender effects occurred. Additionally, other-language male nominal vocatives, also referred to in the literature as brocatives³3

The term brocative was coined by Urichuk and Loureiro-Rodríguez (2023) to describe vocatives such as bro (short for brother) often used by and when addressing male speakers. In the context of Namibian German, vocatives in other languages, such as Afrikaans or English only appear as brocatives and never as female nominal vocatives.

Unlike German as a minority language in the USA, where its use is typically limited to informal settings, German in Namibia is used in both formal and informal contexts. In formal contexts, German is present in a multitude of written and spoken situations in Namibia: signs, media, and schooling. Especially in Windhoek and Swakopmund, German is present in public spaces, for instance, in the form of store signs and official signs. German media presence in Namibia includes the German newspaper Allgemeine Zeitung, as well as German-speaking radio programs and television broadcasts (Shah & Zappen-Thomson, 2018). Church services in German are offered by several churches, and German-speaking societies offer a variety of activities. German also plays a major role in education, with fourteen public and private German-speaking schools. These schools offer courses in Deutsch als Muttersprache (DaM; German as a mother-tongue) and other subjects in German from Grade 1–4. The highest number of students⁴4

677 pupils with DaM attend the DHPS. The second highest number of students with 364 attend the Delta-Schule (Delta school) in Windhoek and 154 pupils attend the Privat Schule Swakopmund according to a statistic by the Arbeits- und Fördergemeinschaft Deutscher Schulvereine in Namibia 'Support Association of German School Societies in Namibia`.

Speakers of Namibian German show systematic register differentiation between formal and informal registers. Both Sauermann et al. (2023) and Wiese and Bracke (2021) examined loan words from English and Afrikaans, choices of existing words in new contexts, and variations in word order in two corpus studies of the DNam corpus and two correction experiments (Sauermann et al., 2023; Wiese et al., 2022). The most clear-cut register differentiation is seen in the lexicon in the two corpus studies, where speakers use loan words, such as oukie 'man, dude', tannie 'woman, aunt' or bra 'brother' almost exclusively in informal registers. Features at the interface of lexicon and grammar, such as weh kriegen 'to hurt' (3), and word order variation, such as the position of the negation particle nicht (4), show less differentiation.

The findings from the correction tasks in Sauermann et al. (2023) suggest that formal and informal registers are more notably defined by the use of lexical material of non-German origin, rather than by the selection of existing words or variations in word order. Notably, individual features within each category exhibit varying degrees of influence. The presence of Namibian German-specific traits in formal settings and the absence of corrections for certain characteristics support the notion of a Namibian German-specific formal register (Kellermeier-Rehbein, 2016).

The results align with aspects of the Social Salience Hierarchy (Wiese et al., 2022). However, the studies reported in Sauermann et al. (2023) reveal variations within the categories, suggesting that factors such as the degree of perceptual salience of a linguistic feature (for instance, how much a particular unusual word order sticks out) may interact with how salient the social meaning conveyed by this feature is.

The DNam corpus is accessible online through the 'Datenbank für Gesprochenes Deutsch' (DGD, database for spoken German). The corpus consists of four sub-corpora: interviews, free speech, elicited formal speech (LangSit formal), and elicited informal speech (LangSit informal). It contains a total of 211,761 words from 115 speakers of Namibian German. Each data point consists of a single word annotated for a variety of linguistic properties, such as its phonological transcription (e.g., gekricht 'gotten'), its transcription in standard spelling (e.g., gekriegt [ɡəˈkriːkt] instead of gekricht [ɡəkɾˈɪçt]), the corresponding uninflected base form of the word (e.g., for gekricht, the infinitive form kriegen '(to) get', also referred to as a lemma) and part-of-speech (e.g., for gekricht, past participle of a verb). The corpus also includes metalinguistic information about each speaker, such as biographical data or the speaker's language background.

The LangSit subcorpora consist of 49,235 tokens in 4 hours, 41 minutes, and 33 seconds of transcribed audio data from the same 103 speakers in both subcorpora, and an additional eight speakers who appear in only one of the subcorpora. Out of the 111 speakers from Namibia, 49 identified as female and 62 as male. The age range of the speakers was between 14 and 75 with a mean age of 23.92 years and a standard deviation of 15.58. The majority of the speakers (N = 77) were adolescents aged between 14 and 18. Additionally, the corpus contains data from four other speakers who grew up in Germany; these were not included in the analyses.

The data points for the analysis were extracted from the DNam corpus in DGD version 2.13. To obtain data for the 'lexical features' category, all adjectives, adverbs, nouns, particles, discourse markers and verbs were extracted from the corpus. Each data point was subsequently coded as either Namibian German or Standard German. In cases of doubt, the annotators checked whether the word was listed in a Standard German dictionary (such as Duden online). For words existing in both Standard German and English (which could either constitute a Standard German word or a specific Namibian German word which entered Namibian German as an English loan word), such as 'hospital', the annotators examined the phonetic properties of the produced utterance, for instance whether the pronunciation of 'hospital' followed Standard German ([hɔspiˈtaːl]) or English ([ˈhɒspɪtl̩] or [ˈhɑːspɪt̬l̩]) pronunciation standards. In addition, all data points which were coded as Namibian German were also annotated for the word's language of origin (Afrikaans, English, or other).

For the 'grammatical features' category, a total of six different Namibian German grammatical features that occurred in the corpus were identified:

For each of these grammatical features, a 'Namibian German' variant and a set of corresponding 'Standard German' equivalents (as mentioned above) were defined. For instance, the Namibian German variant 'weh kriegen' was paired with its Standard German equivalents '(sich) verletzen' or 'weh tun'. Similarly, the Namibian German variant 'sicher machen'⁵5

The Duden contains verb sicher machen with the meaning etwas absichern ('to secure something'). However, sicher machen in this specific sense was not attested in the subcorpora used for the analysis, only the Namibian German homonym was used by speakers.

The data for the 'grammatical features' category were obtained through a corpus search for these six Namibian German features and their Standard German equivalents. To identify V3 clauses, a corpus search was conducted for adverbs followed by nouns or personal pronouns. A second query was then used to narrow down the matches by searching for finite verb forms that follow the first query. Long pauses (> 0.5 seconds) between adverbs and nouns were removed as they can be considered a restart of the utterance (Couper-Kuhlen & Selting, 2018). Main clauses with verb-second clauses were extracted by searching for a finite verb followed by a noun or personal pronoun and sorted manually.

To identify word order variation in subordinate clauses after matrix clauses with negation, I searched for the embedding verbs described in Freywald (2016a, 2016b) followed by the negation particle nicht. I found hits for the embedding verbs sagen 'to say', hoffen 'to hope', glauben 'to believe', and erinnern 'to remember', and sorted them manually into both variants (verb-second and verb-last).

The negation particle nicht is a crucial element for another extracted grammatical feature. It is placed after the finite verb but before the object (e.g., mag nicht seine Musik 'does not like his music') or after the verb and the object (e.g., mag seine Musik nicht 'does not like his music'). In German, negation can have a wide or narrow focus, which is realised through intonational patterns. Utterances with a narrow focus, specifically on the object, and utterances with verb-negation-object word order were excluded as they are standard in German (Helbig & Buscha, 2001, p. 545).

Noun phrases were extracted by searching for nouns in the subcorpora and manually sorted into non-canonical bare noun phrases and canonical noun phrases, which include noun phrases with determiners as well as canonical bare nouns.

All data points from both the lexical and grammatical categories were subsequently annotated for 'variant (Namibian German vs. Standard German)', which served as the dependent variable in the analysis, 'category (lexical vs. grammatical)' and 'register (formal vs. informal)', which were included in the analysis as within-speaker independent variables, and 'speaker gender' as a between-speaker independent variable.

The results were analysed with binomial generalised linear mixed effects regression models (GLMER) with lme4 Version 1.1-31 (Bates et al., 2015) in R Studio (RStudio Team, 2020). For each data point, the model predicted 'variant (Namibian German vs. Standard German)' as a binary dependent variable. The model included 'register' (formal vs. informal), 'category' (lexical features vs. grammatical features), and 'gender' (female vs. male), and their interactions as centred fixed predictors. As suggested by Barr et al. (2013), the initial model contained a maximal random-effects structure, with random intercepts and random slopes by speaker for each predictor included in the model. The random-effects structure was subsequently simplified as part of a stepwise procedure, to determine the most complex model which reached convergence. The final model contained register, gender, category, and their interactions as fixed effects, and a random intercept as well as random slopes for register and category per speaker. Any significant interactions were explored further through additional post-hoc models (Table 2 and Table 3). All plots were done with ggplot2 (Wickham, 2016).

The post-hoc analysis for lexical features contained a total of 19,464 data points, 4.40% of which (864 data points) were coded as Namibian German variants. Of these, 787 tokens were from the informal condition and 77 from the formal condition. The data points were relatively evenly distributed across male (9,369 tokens, 48.14%) and female speakers (10,095 tokens, 51.86%), as well as across the formal (9,615 tokens, 49.40%) and informal (9,849 tokens, 50.60%) registers. The model again predicted 'variant' (Namibian German versus Standard German) as a dependent variable, and contained 'register', 'gender', and the interaction between the two as centred fixed predictors. The results from the post-hoc model are shown in Table 2.

The model showed a significant main effect of gender (p < .001***, Z = 4.54, SD = 2.54), with male speakers (N = 601, 6.41%), producing significantly more Namibian variants than female speakers (N = 263, 2.60%). Additionally, the model also revealed a significant main effect of register (p < .001***, Z = -9.84, SD = 0.12), with significantly more lexical Namibian German variants in informal (N = 787, 7.99%) than in formal speech (N = 77, 0.80%) and a two-way interaction between register and gender (p < .014*, Z = -2.46, SD = 0.15), with male speakers producing more Namibian variants than female speakers in the informal (N = 552, 12.74% vs. N = 234, 4.26%), but not in the formal condition (N = 49, 0.97% vs. N = 28, 0.61%), again supporting Hypothesis 2.

The post-hoc analysis for grammatical features was based on 5,406 data points. 3.87% of these (209 data points) were Namibian German variants. The model analysis was conducted in the same way as for lexical features, with the model predicting 'variant' (Namibian German versus Standard German) as a dependent variable, and 'register', 'gender', and the interaction between the two included as centred fixed predictors. The results from this analysis are shown in Table 3.

Unlike the post-hoc analysis for lexical features, both the main effect of gender (p = .912, z = 0.11, SD = 0.18) and the interaction between gender and register (p =.277, z = -1.09, SD = 0.23) were non-significant. This confirmed that the significant two-way interaction observed in the main omnibus model reflects a significant effect of gender for lexical features (with male speakers producing significantly more Namibian German variants than female speakers), but no gender effect for grammatical features (see omnibus model above).

The main results from the present study is that lexical choices in language production between specific Namibian German variants and their Standard German equivalents are affected by the gender of the speaker in informal, but not in formal communicative situations. This finding indicates that Labov's (1990) classical claim that male speakers produce more non-standard forms than female speakers can be generalised to linguistic choices in language-contact situations. The results also suggest that gender effects in language use are at least considerably more substantial in the lexical than in the grammatical domain and that such differences can be overwritten by the linguistic demands of formal communicative situations.