Learners acquire grammatical constraints (e.g., the knowledge that giggle’s use in The joke giggled me is ungrammatical) in part through statistical learning. The entrenchment and preemption hypotheses claim that correlated statistics are relevant. This makes it difficult to find unambiguous evidence in favor of one or the other. The present work circumvents this issue by orthogonalizing effects of entrenchment and preemption in a learning task with a novel verb. We find evidence that both entrenchment and preemption have significant independent effects in adult learners.