Comparing scores in both setups show some lines that match their scores and other lines that have opposite scores. This filters the ones that are context dependent to the ones that show a context independent reinforcement valence.