Unlike single response formats that use a single dimension to measure attitudes, the branching technique of attitude measurement separates the rating task into two different sequential tasks for measuring bipolar attitudes: direction of attitude (positive versus negative) and intensity of attitude (strength). Some research (Malhotra, Krosnick, and Thomas, 2009, POQ) has indicated that superior outcomes are obtained with branching that has more articulation (3 categories rather than 2 categories) on the endpoints rather than in the middle. However, no comparisons were made with single response formats and, further, the nature of the midpoint may have affected the results. We report here 2 studies comparing a number of branching alternatives, including a consideration of alternative middle responses, with a series of single response measures to determine relative efficacy of the scales. In Experiment 1, we had over 6,000 respondents who indicated their liking and approval for President George W. Bush in the U.S. as well as the extent to which they thought that spending for the military should be increased or decreased using 1 of 5 response formats (2 branching with 2 or 3 endpoints and 3 single response – 5, 7, 9 categories). In Experiment 2, we extended the study internationally in 6 countries (US, UK, France, Germany, Spain, Italy) and used a different set of topics – attitudes and behaviors toward a number of different actions (e.g. drinking coffee, exercise, etc.). We also included numeric box entry and Visual Analog Scales as 2 additional alternatives to the single response and branching response formats in Experiment 2. Overall, we replicated Malhotra, Krsonick, and Thomas in Experiment 1, but also found that single response format took less time and was generally similar in criterion-related validity. We found that some transformations of the criteria were associated with some superiority of the branching tasks.