We report three reaction time (RT)/event-related brain potential (ERP) semantic priming lexical decision experiments that explore the following in relation to L1 activation during L2 processing: (1) the role of L2 proficiency, (2) the role of sentence context, and (3) the locus of L1 activations (ortho-graphic vs. semantic). All experiments used German (L1) homonyms translated into English (L2) to form prime-target pairs (pine-jaw for Kiefer) to test whether the L1 caused interference in an all-L2 experiment. Both RTs and ERPs were measured on targets. Experiment 1 revealed reversed priming in the N200 component and RTs for low-proficiency learners, but only RT interference for high-proficiency participants. Experiment 2 showed that once the words were processed in sentence context, the low-proficiency participants still showed reversed N200 and RT priming, whereas the high-proficiency group showed no effects. Experiment 3 tested native English speakers with the words in sentence context and showed a null result comparable to the high-proficiency group. Based on these results, we argue that cognitive control relating to translational activation is modulated by (1) L2 proficiency, as the early interference in the N200 was observed only for low-proficiency learners, and (2) sentence context, as it helps high-proficiency learners control L1 activation. As reversed priming was observed in the N200 and not the N400 component, we argue that (3) the locus of the L1 activations was orthographic. Implications in terms of bilingual word recognition and the functional role of the N200 ERP component are discussed.