Cross-language priming effects using masked primes tend to be asymmetrical. Priming from L1 to L2 is strong, but priming from L2 to L1 is inconsistent and weak. Two kinds of explanations may explain this asymmetry. The representation hypothesis attributes the asymmetry to the different strength of connections between lexical items in the two languages. The processing hypothesis explains the asymmetry by emphasizing differences in processing speed or the general activation level of the two languages. In this study, three versions of the processing hypothesis were examined. Chinese–English bilinguals were tested with a masked priming paradigm on Chinese–English translation pairs in five experiments. Experiments 1 and 2 replicated the asymmetrical pattern of cross-language priming. Experiments 3 to 5 examined three processing-related hypotheses by varying prime and target presentation conditions. The results showed that none of the three processing accounts provides a satisfactory explanation for the asymmetry. The findings are discussed in the context of current models of bilingual memory organization.