This study investigated the relative effects of input-based versus output-based activities on the learning of English unaccusative constructions. A total of 73 high school students were randomly assigned into two experimental groups and one control group. Of the two experimental groups, the input-based input enhancement group (IE group) experienced a reading passage with the learning targets visually enhanced. The output-based dictogloss group (DG group) performed the dictogloss task with the identical passage given to IE group. The control group (CG group) did not experience the learning targets at all. Results showed that IE group statistically significantly outperformed CG group both at the immediate and the delayed posttests, while there was no significant difference between IE group and DG group. In terms of the generalizability of the learned knowledge, however, it was DG group that outperformed CG group at the immediate posttest. Again, there was no difference between IE group and DG group. More detailed findings are provided, along with some implications for English classrooms in Korea.