A spatial model for legislative roll call analysis is presented, using a nonlinear logit model to analyze political choice data. The model assumes probabilistic voting based on a spatial utility function, allowing for the estimation of spatial coordinates and utility parameters from observed choices. Ordinary Guttman scaling is a special case of this model. The NOMINATE program is used for one-dimensional analysis of two alternative choices with no nonvoting, and its robustness is evaluated using U.S. Senate roll call data from 1979-81. Monte Carlo studies are also presented, and substantive applications are briefly illustrated. The model assumes that individuals occupy positions in a spatial space, and their choices are based on proximity to their ideal points. The model accounts for errors in voting behavior and allows for the recovery of both individual and choice coordinates, as well as utility function parameters. The model is tested against various scenarios, including perfect voting, unanimous roll calls, and random roll calls. It is shown that the model can handle these cases and provides reliable estimates of spatial coordinates and utility parameters. The model is implemented in the NOMINATE program, which performs unidimensional nominal unfolding. The program is tested on U.S. Senate roll call data and is shown to be robust to various data conditions. The model is also tested against alternative methods of generating starting values and is shown to produce reliable estimates of spatial coordinates and utility parameters. The model is compared to other methods, such as Guttman scaling, and is shown to provide more accurate estimates of spatial coordinates and utility parameters. The model is also shown to handle cases where the true space is multidimensional, and it is demonstrated that the model can recover the true dimensionality of the space. The model is also shown to handle cases where there is strategic voting, and it is demonstrated that the model can recover the true spatial coordinates and utility parameters even in the presence of strategic voting. The model is shown to be robust to various data conditions, including the inclusion of nonscalable roll calls and the exclusion of near perfect senators. The model is also shown to be robust to changes in the utility function and to alternative iteration sequences for parameter estimation. The model is shown to produce reliable estimates of spatial coordinates and utility parameters, even in the presence of errors in voting behavior. The model is also shown to be robust to the inclusion of roll calls with small minorities and to the assumption of a common utility function for all senators. The model is shown to be robust to the inclusion of nonscalable roll calls and to the exclusion of near perfect senators. The model is shown to be robust to the inclusion of roll calls with small minorities and to the assumption of a common utility function for all senators. The model is shown to be robust to the inclusion of nonscalable roll calls and to the exclusion of near perfect senators. The model is shown to be robust to the inclusion of roll calls with small minorities and to the assumption of a common utility function forA spatial model for legislative roll call analysis is presented, using a nonlinear logit model to analyze political choice data. The model assumes probabilistic voting based on a spatial utility function, allowing for the estimation of spatial coordinates and utility parameters from observed choices. Ordinary Guttman scaling is a special case of this model. The NOMINATE program is used for one-dimensional analysis of two alternative choices with no nonvoting, and its robustness is evaluated using U.S. Senate roll call data from 1979-81. Monte Carlo studies are also presented, and substantive applications are briefly illustrated. The model assumes that individuals occupy positions in a spatial space, and their choices are based on proximity to their ideal points. The model accounts for errors in voting behavior and allows for the recovery of both individual and choice coordinates, as well as utility function parameters. The model is tested against various scenarios, including perfect voting, unanimous roll calls, and random roll calls. It is shown that the model can handle these cases and provides reliable estimates of spatial coordinates and utility parameters. The model is implemented in the NOMINATE program, which performs unidimensional nominal unfolding. The program is tested on U.S. Senate roll call data and is shown to be robust to various data conditions. The model is also tested against alternative methods of generating starting values and is shown to produce reliable estimates of spatial coordinates and utility parameters. The model is compared to other methods, such as Guttman scaling, and is shown to provide more accurate estimates of spatial coordinates and utility parameters. The model is also shown to handle cases where the true space is multidimensional, and it is demonstrated that the model can recover the true dimensionality of the space. The model is also shown to handle cases where there is strategic voting, and it is demonstrated that the model can recover the true spatial coordinates and utility parameters even in the presence of strategic voting. The model is shown to be robust to various data conditions, including the inclusion of nonscalable roll calls and the exclusion of near perfect senators. The model is also shown to be robust to changes in the utility function and to alternative iteration sequences for parameter estimation. The model is shown to produce reliable estimates of spatial coordinates and utility parameters, even in the presence of errors in voting behavior. The model is also shown to be robust to the inclusion of roll calls with small minorities and to the assumption of a common utility function for all senators. The model is shown to be robust to the inclusion of nonscalable roll calls and to the exclusion of near perfect senators. The model is shown to be robust to the inclusion of roll calls with small minorities and to the assumption of a common utility function for all senators. The model is shown to be robust to the inclusion of nonscalable roll calls and to the exclusion of near perfect senators. The model is shown to be robust to the inclusion of roll calls with small minorities and to the assumption of a common utility function for