Masculine generics in German have long been considered to be gender-neutral (Doleschal, 2002). However, despite their neutrally intended usage, research of the last decades has repeatedly shown that masculine generics apparently are not neutral but biased towards a masculine reading (e.g. Koch, 2021; Misersky et al., 2019). How can this discrepancy between usage and meaning be explained?
Using a method novel to this area of research, we explored this question using linear discriminative learning (LDL; Baayen et al., 2019). LDL follows a discriminative perspective on language, arguing that the relation between form and meaning is fundamentally discriminative (Rescorla & Wagner, 1972). Thus, a word’s semantics emerges by its resonance with the entire lexicon. Semantics entered our LDL implementation via semantic vectors computed by naive discriminative learning (Baayen & Ramscar, 2015) based on a corpus of German news websites. The final LDL implementation allowed for extraction of measures on words’ semantics: comprehension quality, semantic neighbourhood density, and semantic activation diversity.
Taking these measures as well as stereotypicality judgements (Gabriel et al., 2008) to account for potential influences of stereotypicality, a multinomial regression analysis was conducted. The dependent variable was the type of word (masculine generic, masculine explicit, feminine explicit), while the measures and judgements were introduced as predictors. For stereotypicality, no significant effect was found. The LDL measures, however, showed
significant effects. Both masculine forms come with significantly higher comprehension quality and denser semantic neighbourhoods. Feminine forms showed significantly higher levels of semantic activation diversity in the singular and significantly lower levels of semantic activation diversity in the plural. Overall, masculine and feminine forms were significantly different in their semantic features, while masculine generics and explicits were highly similar.
The present results not only confirm the masculine bias of masculine generics, but puts forward an explanation to its source. That is, masculine generics are semantically highly similar to masculine explicits, which in turn is a result of their resonance with the entire lexicon. This then results in an overall biased reading towards the masculine.