This study investigated order-of-learning effects on the acquisition of classifier-noun associations in Chinese in two experiments modeled after Arnon and Ramscar's (2012) study of artificial language learning. In Experiment 1, learners with no prior exposure to Chinese showed better learning of classifier-noun associations when exposed to larger units (sentences) before smaller ones (words) than vice versa, replicating Arnon and Ramscar's findings in a natural language. In Experiment 2, learners with 5 to 7 weeks of classroom exposure to Chinese completed the same procedure. No order-of-learning effects emerged, suggesting that even basic prior knowledge of Chinese eliminated the advantage of initial exposure to larger units. These findings shed light on the extent to which blocking effects arising from carefully controlled lab-based training scale up, from artificial to natural language learning and to learning contexts involving relevant experience beyond a controlled training phase. Implications for language instruction and curriculum design are discussed.