English monolinguals (Experiment 1) and first language (L1)-dominant, Spanish-English and Chinese-English bilinguals (Experiment 2), who differed in L1 orthographic depth (shallow: Spanish; deep: Chinese) and second language (L2–English) proficiency, decided whether visually presented letter strings were English words, while behavioral and EEG measures were recorded. The spelling-sound regularity and consistency of stimuli were covaried such that words had either regular/consistent (e.g., GATE) or irregular/inconsistent mappings (e.g., PINT). Irregular/inconsistent words elicited more positive P200 and less negative N400 amplitudes than regular/consistent words in monolinguals, yet only a P200 response in bilinguals. English proficiency modulated L2 reading strategies, such that bilinguals employed distinct reading unit sizes in the L2 than the L1 when L2 proficiency was low, but transferred L1 reading units to the L2 when L2 proficiency was high. ERP results suggest that high L2 proficiency may be a prerequisite to the cross-linguistic transfer of reading strategies.