Models that rely on the assumption of simple scalability have dominated multiattribute choice and preference modeling for almost 20 years, despite the fact that simple scalability is violated in many instances. Several attempts have been made to construct choice models that do not involve this assumption; however, their application to individual-level analysis has been limited by data requirements and computational problems. Furthermore, few attempts have been made to build similar models of preference structures. The authors present a method for analyzing individual-level hierarchical preference structures that is estimated by a nested logit formulation. The method is very tractable with a large number of brands; it requires only paired comparison data and it provides significantly better fits on calibration data and predictions on holdout data than a Luce model for most respondents in an illustrative application.