Trial by trial subjective measures indicated that exposure to a form–animacy regularity led to unconscious knowledge of that regularity. Under the same conditions, people did not learn about another form–meaning regularity when a linguistically arbitrary variable was used instead of animacy.Experiments 1 and 2 showed that people could learn to use an appropriate determiner based on the animacy of the noun. when modeling implicit learning using networks with randomly assigned pre-training weights is unlikely to accurately model real people.