Four experiments are reported which were designed to test hypotheses concerning the asymmetry of masked translation priming. Experiment 1 confirmed the presence of L2-L1 priming with a semantic categorization task and demonstrated that this effect was restricted to exemplars. Experiment 2 showed that the translation priming effect was not due to response congruence. Experiment 3 replicated this finding, and demonstrated that the 150 ms backward mask that had been used in earlier translation priming experiments was not essential. Finally, it was demonstrated in Experiment 4 that L2-L1 priming was not obtained for an ad hoc category, indicating that priming was not obtained merely because the task required semantic interpretation. These results provide further support for the Sense Model proposed by Finkbeiner et al. (2004).