Extended feature models enable expressing powerful constraints by involving feature attributes in cross-tree relations. However, most of the existing methods for the automated analysis of feature models are not devised to handle such models. In this paper we define a transformation to remove such cross-tree relations, which is applicable when certain restrictions hold. This transformation takes an extended feature model with cross-tree relations involving attributes and constructs a semantically equivalent model that includes only basic cross-tree constraints. Those models with the restricted form of cross-tree relations are allowed to have feature attributes with continuous domains. Thus, the proposed transformation enables automated analysis on an equivalent model without facing extra challenges introduced by the inclusion of attributes in cross-tree relations. © 2011 IEEE.