Typically, the behavior of adaptive systems is specified by a set of rules that are hidden somewhere in the system’s implementation. These rules deal with instances of the domain model. Our purpose is to specify the adaptive response of the system at a higher level (to be applied and reused for different domains or adaptive applications) in an explicit form, called adaptation language. For this purpose we have chosen learning styles (LS) as an implementation field. We defined an XML-based adaptation language LAG-XLS for the AHA! system. In this paper we focus on the empirical evaluation of LAG-XLS.