The N400 and P600 ERP components are widely used in research on language comprehension, but their functional bases are still widely debated. I this talk I will present evidence from computational modeling and ERPs to suggest that N400 amplitudes reflect an implicit prediction error at the level of meaning and a learning signal driving implicit memory formation. On the other hand, the P600 seems to reflect a more explicit and attention dependent prediction error. I will present studies investigating the P600-P3-LC/NE hypothesis, namely the idea that the P600 might be a variant of the domain general P3 component and that both components might reflect phasic release of norepinephrine (NE) in the locus coeruleus (LC) in the brain stem. In line with this idea, we recently observed that P600 amplitudes (but not N400 amplitudes) entail enhanced explicit memory formation. Based on these findings, I will suggest that the N400 and P600 reflect learning signals from complementary learning systems in the neocortex and the hippocampus driving implicit and explicit memory formation.