Burden and Klofstad (2005, Political Psychology) compared political party identification (PID) using »feel that you are« in the item stem with »think of yourself as« in the item stem and found significant differences in the proportions of people identifying themselves as Republicans. Neely (2007, Political Psychology) partially replicated this effect. We sought to explore this effect in more detail in a web-based survey experiment by developing a number of different item stem variants (including »think« versus »feel« along with »identify« and »attached to«). One set of respondents were randomly assigned to a branching option (with a follow-up item to determine strength of identification), while another set of respondents were assigned a single response question (with 7 graded response categories). We compared categorical identification and dimensional measurement (with the categorical then strength branching format converted into a graded scale) with a number of political opinion measures, including presidential evaluation, political ideology, along with approval of the president‘s military, economic, and foreign policy decisions, and attitudes toward a number of social and political issues (including abortion, immigration, etc.). We found substantial differences in PID resulting from nature of response format (branching versus single response) rather than the nature of the item stem presented – a single response scale took less time to complete and led to more extreme responses than a branching format (raising identification as Democrat and Republican by as much as 5%). In addition, single response PID showed a higher correlation with political opinion items. We discuss the theoretical and applied implications for using political party ID for analysis and weighting and how differences in mode of interview could influence results.