The aim of this investigation was to observe both children's and adults' knowledge of grammatical gender in Welsh, and assess the role of exposure, structure complexity and form-function mapping in acquisition. Four different elements of the gender system were examined: local gender marking via SM of nouns and adjectives, local gender marking via production of gendered numeral forms, and distant gender marking via production of gendered prepositions and their pronominal objects. Participants' knowledge was assessed with various controlled elicitation tasks. One additional task tested the productivity of one gender construct by assessing gender marking in relation to novel nouns. Participants (aged four, five, seven, nine and adult) came from one of three home language backgrounds, based on level of exposure to Welsh and English in the home: only Welsh at Home, Welsh and English at home or only English at home. Patterns of responding for individual tasks revealed piecemeal acquisition that seemed to be influenced by frequency effects, resulting in higher levels of productivity in relation to more frequent nouns and lower productivity for less frequent ones. Further probing of the productivity of gender marking across the different gender constructs also revealed piecemeal acquisition patterns, as shown by a lack of uniformity in gender marking in relation to the same lexical items (nouns). The results suggest effects of the varied complexity of the different gender constructs, and the role of semantic or natural gender information in relation to grammatical gender assignment. Overall, a role of relative language exposure seemed evident, as marked differences were observed across home language groups, with OWH children showing the most productive 2 command ofthe gender system. The adult data revealed incomplete acquisition of some aspects of the system, suggesting that there are limits to bilinguals' acquisition of highly complex grammatical features.