We investigate the effectiveness of political campaign advertisements. From findings in communications, political science, and psychology, we know that the relation between voters and campaign strategists is dynamic and evolves until voters’ views on a candidate crystallize. After that point, political campaign advertisements are ineffective. To capture this “dynamic” we develop an adaptive learning model that relates voters’ impression formation to expectations about candidate behavior, one form of which (rational expectations), renders political advertising ineffective. We treat rational expectations as a limiting result that supports the concept of crystallization. Our model assumes that voters misspecify their forecasts about a particular candidate’s attributes and campaign strategy. Over time voters can reach a rational expectations equilibrium about a candidate’s qualities and discount political advertising. We illustrate the learning dynamics using simulations. As one application of this approach we focus on the influence campaign message (strategy) volatility has on crystallization (i.e., reaching the rational expectations equilibrium). Our simulation results show that campaign message volatility has an important effect on crystallization. One implication is that crystallization is a fragile, special case result that can be altered by informational shocks during the campaign.