The process of word translation evaluated in this study manipulated the word learning method and concreteness of words. The participants were fluent English (L2) speakers whose native language was Japanese (LI). The participants learned new L2 words by rote rehearsal or imagery method. Then the participants were tested on word translation. The translation asymmetry effect (L2 to L1 translation being faster than L1 to L2 translation) predicted by the model of bilingual memory was not demonstrated. The present findings were not consistent with previous studies. It suggested that a modified version of the asymmetry model should take the word retrieval time into account.