There is a large and growing body of literature on the political polarization of socio-politicalattitudes. This work, however, often treats parties as homogeneous groups, missing underlying within-partyvariation. This paper argues that social scientists should measure attitudinal trends at three levels: 1) theaggregate, 2) parties, and 3) within parties. Using several waves of ANES data between 1992 and 2016, Iapply this analytic strategy to the study of immigration attitudes—one of the most salient issues currentlyoccupying the U.S. political arena. The results suggest a number of important findings. First, immigrationattitudes have been fairly stable at the aggregate level, and have, if anything, become more pro-immigrationover time. Second, while immigration attitudes have become slightly more favorable over the past twentyyears, these trends in the aggregate mask drastic party polarization since 2008, with Republicans becomingsharply more anti-immigration and Democrats becoming more pro-immigration. Third, while the partieshave clearly grown apart, the attitudes within the parties have not trended uniformly: while Democrats havebecome more pro-immigration over the last several years, they have also grown more variable on this issue.The growing divide among Democrats can be explained by emergent ideological and partisan cleavages, withliberals and “strong” Democrats largely driving the party’s pro-immigration trends. In contrast, Republicanshave become more anti-immigration and consistently so; while there were previously significant differencesamong Republicans along regional, religious, and racial lines, these differences have collapsed, and the resultis a largely uniform Republican party on immigration. I conclude with a discussion of how studying withinpartyvariation may provide valuable insight into how parties establish policy platforms and selling pointsin campaigns; further, clear within-party divides may serve as signs of future party realignment along thelines of salient, dividing issues.