Milton Lodge and his colleagues at Stony Brook have argued that voters process campaign information on-line, summarizing their affect toward candidates as campaign information is encountered. Consequently, recall of campaign information and vote choice are believed by Lodge to be a weak predictor of actual vote decision, which is determined almost solely by the on-line tally. The claims made by the on-line model have not been tested in a dynamic election context, however, in which two or more candidates compete for the vote. This study uses a new experimental methodology that more accurately depicts the realities of a campaign environment to assess the relative importance of memory and the on-line tally in predicting both the direction and accuracy of the vote choice. Findings do not support the pure Stony Brook on-line model, as they show that in all cases voter memory plays an important role in decision making and suggest that a mixed decision-making model is more appropriate.