Grammatical Gender

Grammatical gender is a system of noun classes — commonly two or three in European languages, typically ten to twenty in Bantu languages — that every noun belongs to whether or not it denotes anything with sex at all. German's das Mädchen, "girl," is grammatically neuter; French la personne is feminine regardless of who is meant. The classes are a grammatical Taxonomy, not a semantic one, though a handful of core categories echo the graded membership Prototype Theory describes elsewhere in the lexicon: cognate nouns for the same referent routinely disagree in gender across related languages, evidence that assignment is fixed at Naming or borrowing by a noun's form and history, largely independent of what it currently means.

What the system buys is agreement machinery. A noun's class ripples to its articles, adjectives, and pronouns, and in some languages to past-tense verbs, marking gender redundantly at several points in a sentence — a rule-governed cousin of Collocation, forced rather than merely preferred. A listener who missed the noun can often recover it from the agreement pattern, and a non-native speaker's gender errors are among the most persistent Shibboleth markers of imperfect fluency, since misclassing an unfamiliar word is a mistake native speakers rarely make. Old English had the full three-gender system; English lost it over the Middle English period, region by region — essentially complete in the north by the early eleventh century, the Midlands by the early thirteenth, Kent into the mid-fourteenth. Nouns still pick up gendered pronouns informally — ships called "she" — a vestige of personification, not grammar, in the spirit of Folk Etymology.

Whether the surviving classes shape thought beyond grammar is contested, and the evidence is thinner than the popular claim. Studies asking speakers of gendered languages to describe objects have sometimes found gender-congruent adjectives — a bridge called elegant where its gender is feminine, sturdy where it is masculine — but effect sizes are small and inconsistent across replications, inside the broader, mixed record of Linguistic Relativity. The Ground Truth here is a weak, task-dependent nudge, not a difference in perception.