Many languages involve grammatical gender, in which each noun fits in a category such as masculine, feminine, or neuter. In Latin, for instance, puella “girl” was feminine, equus “horse” was masculine, and ferrum “iron” was neuter. Like many other languages, Latin also had case marking, such that words changed to indicate their role in the sentence (e.g., nominative for subject vs. accusative for object). A very common pattern in Indo-European languages is for neuter nouns to not distinguish nominative and accusative case, while masculine and feminine nouns do. This has been argued to be a consequence of the fact that neuter nouns are more likely to refer to inanimate nouns. Animate nouns can often be subjects or objects; inanimate nouns are more often objects: If you hear a sentence about a horse kicking a wall, you’re unlikely to be in any doubt as to what kicked what even without grammatical clues. It has been argued that the marking of gender and the marking case may be closely related. Here we propose an experiment designed to investigate how the same grammatical morphemes can be interpreted as marking different categories given different contexts. Participants will be exposed to an artificial language with animate and inanimate nouns and grammatical markers whose role participants have to infer. We will manipulate co-occurrence patterns between animacy and subject- or objecthood. Participants will then be asked to produce novel sentences in response to stimuli. We predict that the distribution of animacy and grammatical role in the training data will influence whether the grammatical marker is perceived as marking gender, case, or both.