The goal of this project is to determine how well our current models of public opinion---the on-line and memory-based---predict the course of public opinion during political campaigns. Unfortunately, the dynamic implications of these public opinion models have not been explored to the point where they can provide an answer to this question and the dynamic implications of these models have not been leveraged in the empirical evaluation or theoretical refinement of the models themselves. My approach to this task is two-pronged. I first formalize the theoretical arguments into mathematical equations to produce dynamic maps of the movement of public opinion. Consequently, I test the theoretical models by collecting data on campaign communications in eight congressional and gubernatorial campaigns and use these data as the inputs in the equations. The result is a predicted course for public opinion over the campaign, given the campaign communications that actually occurred. I then examine public opinion data to evaluate which of the two models accurately predicts the course of public opinion over the campaign. The results suggest that neither model can adequately account for the dynamics of a political campaign; therefore, I suggest a path for future research aimed at understanding the relationship between memory for campaign information and candidate evaluation.