This study investigated the use of segmental and suprasegmental cues to lexical stress in word recognition by Mandarin-speaking English learners, Korean-speaking English learners, and native English listeners. Unlike English and Mandarin, Korean does not have lexical stress. Participants completed a visual-world eye-tracking experiment that examined whether listeners' word recognition is constrained by suprasegmental cues to stress alone or by a combination of segmental and suprasegmental cues. Results showed that English listeners used both suprasegmental cues alone and segmental and suprasegmental cues together to recognize English words, with the effect of stress being greater for combined cues. Conversely, Mandarin listeners used stress in lexical access only when stress was signaled by suprasegmental cues alone, and Korean listeners did so only when stress was signaled by segmental and suprasegmental cues together. These results highlight the importance of a cue-based approach to the study of stress in word recognition.