The western Marxist tradition identifies the active engagement of human beings with their environment and with each other as a central ontological category. This physical, verbal and cognitive engagement is embodied through skill: the practical availability of what are often prediscursive modes of action, generated in collective learning processes such as conflict or alliance, materially sedimented in experience, practices, language, networks and so on, and thus continually subject to transformation or loss, but also constantly available as a resource for creative action. Movements, from above or below, are then different possible "proto-hegemonic" attempts at developing this potential from different starting-points and mobilising it around shared social projects and against others. Strategies of research into movement contexts parallel these possible organising modes: given the diversity of participants' orientations and of external interventions, there is necessarily a politics of research characterised by collusion with some participants' knowledge interests and conflict with others. The paper draws on Gramsci's conceptualisation of class consciousness to argue for a critical realism that extends the logic implicit in participants' skilled activity to a more comprehensive standpoint, using the researchers' own standpoint and knowledge interests critically as a part of this dialogue. The use of metaphor, illustration and other "hegemonising" strategies are geared to developing this two-way communication between different knowledge interests, which remains precarious unless it is developed into the coordination of shared activity. Such a politics of knowledge makes sense only given particular starting-points. A concrete example is given in the case of my own research, which moved from a participant's developing choice of priorities to a traditional intellectual's attempt to relate the milieu to externally-determined projects. The class and other relations involved in this process are examined critically, with a view to bringing out the ability of participants to "locate" the researcher and fit my activity in turn into their own perspectives and projects. The cognitive implications of this analysis enable a more complex understanding of such research activity and point to important political and ethical issues around the potential value and limitations of research for participants and researchers alike.