Support for human-to-human interactions over a network is still insufficient, particularly for professional virtual communities (PVC). Among other limitations, adaptation and learning-by-experience capabilities of humans are not taken into account in existing models for collaboration processes in PVC. This paper presents a model for adaptive human collaboration. A key element of this model is the use of negotiation for adaptation of social protocols modelling processes. A second contribution is the proposition of various adaptation propagation strategies as means for continuous management of the PVC inheritance.human interactions over a network is still insufficient and more research has to be done to provide both theoretical and practical knowledge to this field.Among various reasons for the weak support for human-to-human interactions, one may distinguish the following three reasons: first, many social elements are involved in the interaction among humans. An example of such a social element may be the role played by humans during their interactions. Social elements are usually difficult to model, e.g. integrating non-verbal communication to collaboration models. Therefore, their integration to a model of interaction between humans is not easy. A second reason is the adaptation capabilities of humans which are not only far more advanced than adaptation capabilities of software entities, but also not taken into account in existing models for collaboration processes. A third reason is the learning-by-experience capabilities of humans, i.e. the capabilities to extract know-how and knowledge from previous experience and reuse it in similar situations.Human-to-human interactions between people sharing a common practice have been studied for many years. Wenger coined the term of Community of Practice (CoP) to refer to "a set of interacting people engaged in a common practice. Practice refers to the work people do, but also to the ideas behind itthe shared understandings and the activities.