ncreasing data spaces, relation-based data structures, and individual customized data is challenging the current data representations, mainly limited to a list-based approach, which does not describe the relationships between data objects. When data structures are changing towards semantic markup, the relations between data objects become valuable and can be stated as "meta-affordances", which the user can utilize in understanding relationships between data. Adaptive systems that relies on User Models have previously been based on user history or models of the complete environment. Open environments, such as the Internet, requires a different approach & this project provides a method to generically evaluate differences in user behavior that could refine existing User Models. Data from a simple visual indexation tasks that the user solves in a set of different environments, serves as a basis for detecting user behavior. The results are that detection of user behavior using simple generic calibration is possible in visual interfaces - the method could also be applied in HCI in general which will be encouraged by the authors.