Crossing grammar and biology for gender categorisations: investigating the gender congruency effect in generic nouns for animates

ABSTRACT

In formal gender languages, all nouns have grammatical gender, regardless of whether their referents have a biological sex. The question of whether the grammatical gender affects how the denoted entities are conceptualised is subject to ongoing debate. Here, we investigate the extent to which a gender congruency effect emerges for three categories of nouns, with a particular focus on generic nouns (or epicenes). In two experiments and two replications with native speakers of German, we used an implicit measure to test possible associations of nouns with biological sex. These experiments revealed a stable gender congruency effect for congruent animates, and a weaker effect for generic animates and non-animates. In a fifth experiment, we combined the implicit measure with an explicit measure and contrasted items that have strong versus weak associations with biological sex. The results indicate that the congruency effect is driven by item-specific associations rather than by grammatical gender.

KEYWORDS:

• Cognition
• language
• grammatical gender
• Extrinsic Affective Simon Task
• linguistic relativity

Acknowledgements

We are grateful to Susanne Bubser and Rainer Leonhart for assistance with material and data collection, to Dirk Wentura for advice on data analysis, to Henrik Saalbach for valuable comments on an earlier version of this manuscript, and to Sarah Mannion de Hernandez for proofreading.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹To distinguish between these two sources of gender and keep terminology brief and concise, in the remainder of this paper the term “sex” will be used for the biological gender of animates, and the term “gender” will be reserved for the grammatical category (see also Comrie, 1999; Foundalis, 2002). The former relates to what Hellinger and Bußmann (2001) denote as “referential gender,” and the latter as “grammatical gender,” while what they label “lexical gender” relates to items for which sex is part of the core meaning such as “aunt” or “stallion,” the grammatical gender of which is typically congruent with the sex of the referent. “Gender,” as used here, should not be confused with the social construct, as distinct from the biological sex of a person. Likewise, the two instantiations of sex will be labelled “female” and “male” (abbreviated in figures and tables as “f” and “m,” respectively), while the two (grammatical) gender classes will be labelled “feminine” and “masculine” (abbreviated as “F” and “M,” respectively; accordingly, “neuter” is abbreviated as “N”).

²Grammatical gender in German is most often not transparent, and the only substantial regularity – a greater likelihood of feminine terms (as compared to masculine ones) to end on e – does not hold for animates (hence der Löwe or der Rabe), nor is it commonly noted by native speakers. Even simple assignment rules do not operate on a conscious level (Corbett, 1991, p. 92); their identification took years of meticulous research (e.g., Köpcke & Zubin, 1983, 1984; Zubin & Köpcke, 1986); and when native speakers access gender information, they normally do so via the lexical route (Hohlfeld, 2006).

³After having conducted the experiments, we were made aware of a more topical database for word frequencies, SUBTLEX-DE (Brysbaert et al., 2011). Overall, the frequency values from the SUBTLEX-DE database were higher, but this general increase had no effect on the matching of feminine and masculine items, neither for generic animates (log frequencies feminine: 1.917 vs. masculine: 1.826; t(38) = −.449; p = .656), nor for non-animates (feminine: 2.378 vs. masculine: 2.416; t(38) = .220; p = .827).

⁴As indicated in Appendix B, the final model in most of the reported analyses includes random slopes for response association as a function of participants. This indicates inter-individual differences with regard to the preference for the “female” or “male” response key, perhaps due to an interaction between the participant's handedness and the balanced assignment of the response keys.

⁵From the results reported by Imai et al. (2014), our findings differ in that we always did (or did not) find a gender congruency effect for generic animates and non-animates alike, while they found such an effect only for animates, but not for non-animates. The reason for this difference may be that their task requested inferences on biological properties, which were highly relevant for one category, but not the other, while the requirements for our task were equally (ir)relevant for both categories alike (for a discussion of how relevant the specificity of tasks may be for effects of linguistic relativity, see Saalbach & Imai, 2007).

⁶We also reasoned that a perfectly balanced pattern of sex associations in the pilot study would not guarantee the same pattern in the study at stake in any case. As noted above, individuals differ in their associations of nouns for non-animates with a specific biological sex (Bender et al., 2016), and these differences may be amplified for whole samples. The subsequent rating task therefore also served as a means to control for this potential variation.

⁷Post hoc (see footnote 3), we ran analogous analyses for the log frequencies reported in the SUBTLEX-DE database (Brysbaert et al., 2011). Again, the overall frequencies were higher than those reported in the Celex database. The re-analysis of the generic animates indicated again no significant differences between the four sets of nouns (average log frequency: 1.723; largest F(1, 36) = 2.269; p > .140), whereas the re-analysis of the non-animates indicated that strongly associated nouns were less frequent on average (1.795) than weakly associated nouns (2.413); F(1, 36) = 16.535; p < .001. It should be noted, however, that the occurrence of one such difference may reflect the inflation of alpha errors implied by the large number of individual significance tests conducted in checking item matching across the many conditions in this study.

⁸As mentioned above, the intermixing of generic animates only with congruent animates in a between-participants design may have blurred the generic nature of their gender, thus facilitating a carry-over in gender–sex matching from the congruent animates category to all stimuli.

⁹Given the findings reported by Imai et al. (2014) and Saalbach et al. (2012), which indicate that the gender congruency effect in a deductive reasoning task among German speakers hinges on the presence of the gender-marking article, this might sound counterintuitive. However, the article presented in these previous studies was always the correct one and therefore provided reliable information on grammatical gender. In contrast, the articles in the current study were assigned randomly and requested speeded colour/sex decisions, thus rendering it rather unlikely that the article may have paved the way for the observed effect.

¹⁰This is at least the case in Experiment 3, whereas in the first set of experiments (1, 2, 1R, and 2R), we do not have clear evidence that the effect was really driven by stereotypical associations.

Additional information

Funding

This work was supported by the Deutsche Forschungsgemeinschaft [grant number Kl 614/31-1] to Karl Christoph Klauer and [grant number Be 2451/8-2] to Andrea Bender.