Recent studies have demonstrated a strong empirical relationship between political discussion and political knowledge. However, as of yet there has been no clear discussion or demonstration of how political discussion translates into increased political knowledge. The present study proposes three explanations--exposure (similar to the two step flow), anticipatory elaboration (linking work on uses and gratifications and news information processing), and discussion-generated elaboration (focusing on how discussion itself can influence information processing)--for this observed empirical relationship. In order to test these three explanations, data from the 2000 ANES and a local community survey during the 1996 presidential election were employed. Findings suggest that the direct relationship between discussion and knowledge may be mediated through motivations and information processing behaviors. These findings support the anticipatory elaboration and discussion-generated elaboration explanations while questioning the exposure explanation, and link well with recent findings on the cognitive mediation model.