This article reports findings from 3 experiments examining whether 2nd language (L2) words are represented in episodic memory, as originally proposed by Jiang and Forster (2001). Experiment 1 was a direct replication of Jiang and Forster, testing highly proficient Chinese-English bilinguals. Masked translation priming was obtained in an episodic recognition task from L2 to the 1st language (L1) for studied "old" L1 targets but not for unstudied "new" targets. This experiment also confirmed the translation asymmetry generally found in lexical decision tasks, namely, priming in the L1-L2 direction but not in the L2-L1 direction. Experiment 2 showed that recently learned words in an unfamiliar language (therefore, words that are obviously represented episodically) could also prime their L1 translations in an episodic recognition task but not in a lexical decision task. Finally, in Experiment 3, masked repetition priming was used with an episodic recognition memory task. For native speakers of English, repetition (L1-L1) priming is obtained only for old words, because there is no episodic representation for new words. However, Chinese-English bilinguals tested with the same items showed repetition (L2-L2) priming for both old and new words, indicating that the new L2 words were represented episodically as well. Overall, the results from these 3 experiments support the hypothesis that L2 words are represented in episodic memory. Finally, the mechanisms behind why L2-L1 translation priming can be obtained in episodic recognition and not in lexical decision are discussed. (Contains 2 figures, 1 footnote and 5 tables.)