Two experiments involving picture classifications investigated priming behavior in the context of "wh"-movement at clause edge and in indirect object position, respectively. In Experiment 1, intermediate L1-Chinese L2-English learners produced slower classification times (inhibitions) at clause edge, apparently induced by the computation of intermediate traces, whereas faster classification times (facilitations) were detected in native speakers (NSs). In Experiment 2, advanced L1-English L2-French learners produced facilitations apparently induced by the computation of traces in indirect object position. Learners' priming pattern differed from the priming pattern produced by NSs, however. We argue that the patterns found suggest structural computations interacting with distinct activation modes in the maintenance of discourse referents, plausibly resulting from the automaticity of lexical access, or lack thereof. (Contains 7 figures and 2 tables.)