Using a cognitive load theory approach, we investigated the effects of speaker variability when individuals are learning to understand English as a foreign language (EFL) spoken by foreign-accented speakers. The use of multiple, Indian-accented speakers was compared to that of a single speaker for Chinese EFL learners with a higher or lower English language proficiency. Translation accuracy scores and cognitive load self-rating scales were used to measure differences between various acquisition conditions. After 3 consecutive days of acquisition, contrasting effects were found depending on levels of expertise. Analyses revealed that multiple-speaker acquisition conditions facilitated perceptual learning of foreign-accented English for higher expertise EFL learners, whereas single-speaker acquisition facilitated learning for lower expertise EFL learners. These results suggest that effective learning of foreign language in spoken form should take into account both the instructional format and learners' expertise levels.