In my ERP study, where I implicated mixed design "oddball - passive perceptual training (aka classical conditioning) - oddball", I obtained strange pattern of correlation: changes in MMN were straightly correlated to changes in P3a. It means the bigger the MMN became after training, the smaller the P3a became.
Honestly, I don't know how to interpret this... Absolute majority of studies showed that after discrimination training increase in MMN associated with increase in P3a.
Will appreciate any ideas/links!