ABSTRACT Theoretical and empirical models of congressional voting assume that legislators vote with the sole purpose to move policy closer to their ideologically ideal point. While NOMINATE correctly classifies most votes cast by members of Congress, a significant number of votes are misclassified and coded as spatial error. The literature on congressional voting assumes this error to be random and idiosyncratic across members. We argue that spatial errors are systematic across members, with spatial error being more likely on roll-call votes tackling salient policy issues and among members representing constituencies with greater ideological divergence between the median voter and the member's primary election constituency. Using Aldrich-McKelvey scaling to place legislators and constituencies in the same ideological space, we find support for our theory. We attribute this finding to the electoral uncertainty faced by members of both the House and Senate representing constituencies with greater ideological divergence between these key electoral principals.