Grammatical gender in German: A case for linguistic relativity?

Abstract

The “principle of linguistic relativity” holds that, by way of grammatical categorization, language affects the conceptual representations of its speakers. Formal gender systems are a case in point, albeit a particularly controversial one: Previous studies obtained broadly diverging data, thus giving rise to conflicting conclusions. To a large extent, this incoherence is related to task differences and methodological problems. Here, a priming design is presented that avoids previous problems, as it prevents participants from employing gender information in a strategic manner. Four experiments with German native speakers show priming effects of the prime's grammatical gender on animate and nonanimate targets, an effect for the prime's biological gender on animate targets, but no effect for the prime's biological gender on nonanimate targets, and thus speak against an effect of language on thought for German gender.

Keywords:

• Grammatical gender
• Cognition
• Language
• Linguistic relativity
• Priming
• German

Acknowledgments

The research reported in this paper was supported by Grant Kl 614/31–1 from the Deutsche Forschungsgemeinschaft (DFG) to K. C. Klauer. We are grateful to Sarah Mannion and three anonymous reviewers for valuable comments on earlier versions of this article.

Notes

¹ To minimize confusion between the two, the term “sex” will be used in the following for the biological gender of animates, and the term “gender” for the grammatical category (cf. Comrie, 1999).

² We are primarily interested in whether the grammatical gender of nouns affects the conceptual representation of the respective objects; hence, we restricted our material to the two genders that might be linked to sex (neuter is not). This restriction should have no consequences for people's decisions in the tasks, and it should not be critical for generalizing the findings. As reported above, the observed differences in gender impact on thought between German and Romance languages have been explained through their differences in gender systems (Sera et al., 2002; Vigliocco et al., 2005). Accordingly, the two genders in Romance languages facilitate a sex-to-gender mapping during learning, whereas the three genders in German do not. In this sense, neuter is crucial only for the difficulty with which associations are acquired during the lifetime; it is not relevant for performance in our task.

³ While we took care to exclude nouns that mark gender morphologically in order to prevent the task from becoming too simple, it was not necessary to exclude any regularities. As we were not interested in how people identify gender (i.e., by way of retrieval or assignment), but rather in whether grammatical properties affect the semantic level, it would not be of critical importance if our participants were aware of the phonological regularities. Such an awareness would only have facilitated the task in general, independent of priming, and would hence have rendered our assessment of possible effects more conservative (N.B., in the case of grammatical priming only).

⁴ More precisely, pronouns are sensitive to the grammatical gender of the possessor. However, animates (and particularly persons) can be regarded as the “prototypical” possessors, which should activate a connotation of sex. This was supported by a post hoc test (n = 20), in which we checked how reliably the different types of primes were identified as male or female. The possessive pronouns ihr(e) and sein(e) proved to be only slightly weaker (88.75%) than the pictograms (90%) used in Experiment 4 and the nouns Frau/Mann (100%). This overlap is picked up in the discussion.