This article presents results of two off-line comprehension tasks, investigating the acceptability of novel and regular metonymy by speakers of English, Korean, and Spanish. We are interested in uncovering regular–novel metonymy computation discrepancies, and whether they are treated differently in the different languages. The distinction between novel and regular metonymy is discussed by the existing theoretical treatments of metonymy as well as in psycholinguistic research. The findings of this study constitute further experimental support for the psychological reality of this distinction. In addition, it is demonstrated that the speakers of the three languages treat novel and regular metonymy differently. Significant findings are the acceptability of novel metonymy in Korean and the relative lack of conventionalization effect for regular metonymy in Korean and Spanish. We conclude that current theoretical approaches to metonymy should focus more on cross-linguistic differences and that further language comparisons are warranted and needed, in comprehension as well as in processing.