The major topic of this paper is how people change their voting decisions in light of campaign effects. Our challenge to previous literature on campaign effects is twofold. First, on substantial ground, past studies have largely focused on and reported activation and reinforcement effect of campaign The flip side of a coin–deactivation– has attracted scant attention. On the other hand, much of the previous research has heavily utilized a binary dependent variable in addressing the change in voting decision. Relying on dichotomous specification, scholars have been forced to analyze either turnout decision or candidate choice, “separately.” Recently, the validity of these dichotomous specifications has been challenged and a unified specification proposed as an alternative. Utilizing a unified dependent variable that combines turnout decision and candidate choice, we investigate systematic differences between the activated and deactivated voter. The deactivated voter refers to those who changed their voting decisions toward the opposite party candidate. The analytic focus is on the question of who are more likely to deactivate their party identification as a voting cue over a campaign and of what underlying mechanism involves in this dynamic process. Regarding activation process, two conceptually distinguishable stages are assumed to intermediate between campaign stimuli and changing voting decision, perception updating and behavior converting stage. Individual party identification is hypothesized not only to biases perception updating stage, but also to exert its unbalanced influence at decision converting stage. Through these processes, party identification increases its role as a voting determinant over a campaign period, leading to prevailing activation. Attitudinal ambivalence–coexistence of positive and negative attitude toward a attitude object– is hypothesized to moderate and overcome the role of party identification as the source of bias at both stages, thus leading to deactivation. Four ambivalence measures are included as the primary independent variable of interest. Those are ambivalence toward the two major parties, issue proximity ambivalence, and emotional and trait-based ambivalence toward the two major party candidates. Because the model proposed here requires a panel data set and dynamic specification, we use the 1980 NES Major Panel Studies and employ nested-logit estimation. What we found, first of all, contradicts the conventional notion of predominance of activation over deactivation, which is mainly drawn form dichotomous specification. For instance, among the Democrats the deactivated voter (107) outnumbered the activated (84). The notion that a campaign serves mainly to activate party identification and makes them electorally relevant is not supported when a unified dependent variable is employed. Results, more importantly, confirm our expectation that attitudinal ambivalence is strongly correlated with deactivation. Specifically, those with ambivalent attitude toward two major parties and its candidates are more likely to overcome the dual barriers by nullifying or dampening the role of party identification at both stages. To sum, those with ambivalent attitude toward the two major parties and its candidate are more likely to deactivate their party identification as a voting cue, ending up with either defection or deactivated abstention. [ABSTRACT FROM AUTHOR]