Awareness has now become a widely accepted central feature in groupware systems. In this context, providing an awareness support consists in delivering to users involved in a collaborative task, information concerning their group, other users, their past, present or future roles or activities, etc. However, one of the potential underlying risks of such a support is to produce too much information. As a matter of fact, this amount of information may reveal itself cumbersome, leading to a cognitive overload for users which slows down the global activity of the group. In this paper, we propose to address this issue by stratifying the awareness information space into different levels of details. Such stratifications give users a progressive access to information: first, they can access to the most relevant information linked to their current activity and profile, and then, they can gradually access to the rest of awareness information. The proposed architecture relies on a generic Data Model for Awareness Information (DMAI) and on a Progressive Access Model (PAM), we have previously designed in the context of Web-based Information Systems. We show, on an example, how the coupling between the DMAI and the PAM allows groupware systems designers to describe a progressive access to awareness information. The defined stratifications can also be used as filters which save a given user, involved in a given role and performing a given activity, from useless, irrelevant or forbidden information.