This paper challenges a widely held assumption in linguistics – that all normal speakers master the basic constructions of their language – and argues that proficiency with a particular structure depends on individual speakers’ linguistic experience. Our argument is based on an experimental study testing speakers’ ability to interpret passive sentences. Since full passives are used predominantly in written texts, more educated speakers have more experience with the construction, and hence might be expected to perform better. In order to determine whether the type of linguistic experience matters as well as the sheer amount, we also tested non-native speakers. Highly educated adult second language learners have the benefits of schooling, but quantitatively less experience with passives than native English speakers, and hence should perform worse than native speakers if proficiency is merely a function of the amount of exposure. We tested sentence comprehension using a modified version of a task developed by Ferreira [Ferreira, F., 2003. The misinterpretation of non-canonical sentences. Cognitive Psychology 47, 164–203]. Participants were asked to identify the agent in four types of sentences: plausible active, implausible active, plausible passive, and implausible passive. We found that both of the highly educated groups and the less-educated non-native group performed at ceiling in all conditions. The less-educated native group performed at ceiling on the plausible sentences, but had difficulty with implausible actives (65% correct) and especially implausible passives (36% correct). These results suggest considerable (possibly education-related) differences in level of attainment among native speakers. However, the performance of the less-educated non-native group indicates that this effect is not solely attributable to the number of passives in the speakers’ experience. We suggest that processing implausible non-canonical sentences depends to some extent on metalinguistic skills, which may be enhanced by explicit L2 instruction.