This study investigated whether candidate images could be designated as primarily either stimulus- or perceiver-determined and if a multiple regression model could be constructed to predict candidate image ratings from pre-stimulus perceptions of the candidate's party and post-stimulus ratings of the advertisement. One hundred twenty subjects were selected in six groups of 20 each from the University of Texas--Austin, Texas Lutheran College, and Randolph Air Force Base. A set of 14 semantic differential scales which had been demonstrated in previous research to be predictive of voting behavior was modified for the purpose of quantifying party and candidate images. Multiple discriminant analysis was used to compare the pre-stimulus party image measures and the post-stimulus candidate image scales. Based on the rationale for the study, significant differences were attributed to the advertisements (stimulus-determined) and similarities were attributed to perceiver effects. In every case, message evaluation variables dominated the predictive models, indicating the existence of perceiver-determined effects and confirming the multivariate nature of the perceptions voters form about political candidates. (Author/RB)