Participatory rural appraisal (PRA) methods in India are increasingly taken up by public sector organisations as well as by NGOs among whom they have been pioneered. While PRA methods are successfully employed in a variety of project planning contexts — and with increasing sophistication — in other situations, the practice of PRA faces constraints. This paper examines the constraints as experienced in the early stages of one project, and suggests some more general issues which these point to. In particular, it is suggested that as participatory exercises, PRAs involve 'public' social events which construct 'local knowledge' in ways which are strongly influenced by existing social relationships. The paper suggests that information for planning is shaped by relations of power and gender, and by the investigators themselves; and that certain kinds of knowledge are often excluded. The paper is not, however, to be read as a generalised critique of PRA. Social dominance and gender are not universally experienced as constraints in the practice of PRA. The paper arises from a particular moment in one project's own critical analysis of its methods. In this sense it is not a conclusion or a judgement, but an indication of the continuing need for context-specific methodological adaptation, especially as PRA is more widely employed in the public sector. Finally, the paper suggests that as a method for articulating existing local knowledge, PRA needs to be complemented by other methods of 'participation' which generate the changed awareness and new ways of knowing, which are necessary to locally-controlled innovation and change.