This study examined the effects of bottom-up instruction (BI) on a high-stakes listening test for second language (L2) listeners and compared these effects with those of strategy instruction (SI) in relation to their proficiency levels. Two intact classes (48 L2 listeners) of different L2 listening proficiency levels received five weeks of BI and two weeks of SI. BI offered explicit knowledge of spoken English combined with reproduction practice, whereas SI helped learners deploy metacognitive knowledge and test-taking strategies. The data were collected three times: before the intervention (pre-test) and after each instruction (mid- and post-test). A common-item design that permits Rasch analysis was used. The results showed that BI improved L2 ­listeners' test performances regardless of their proficiency levels, ­highlighting the usefulness of BI in boosting listening proficiency. Conversely, SI benefited only high proficiency L2 listeners, which supports the threshold hypothesis claiming that L2 listeners require a certain level of L2 proficiency to deploy metacognitive knowledge and test-taking strategies successfully.