A very interesting EEG/ERP study was carried out by Morgan-Short et al. (2010), in which participants learned an artificial language either explicitly (formally) or implicitly (immersion), and the differences in the processing was observed and compared to native patterns. This study is unique in its use of an artificial language, and as a result can therefore control for amount/type of L2 exposure, which is a common obstacle in SLA research. I'm attaching a link to the abstract and article. Hope that helps!