This study investigates the interpretation of scopally ambiguous sentences involving a universally quantified direct object NP and negation (e.g., The boy didn’t eat every cookie) from a processing perspective. Using an online truth-value judgment task implemented with a self-paced reading technique, data were collected from native Korean speakers, native English speakers, and Korean L2 learners of English. The results indicate that native Korean speakers strongly preferred the full set interpretation (every > not). In contrast, native English speakers strongly preferred the partitioned set interpretation (not > every). L2 learners showed a developmental divergence according to the learners’ L2 proficiency; the low proficiency group showed a strong preference for the full set interpretation, whereas the advanced L2 learners showed no preference for either of the patterns. The main findings were examined within the framework outlined by O’Grady (2005), who proposes that the nature and acquisition of scopal contrasts are best understood with reference to the operation of an efficiency-based processor.