Despite recent international research on the topic, we still have a fuzzy view of local seed networks structure and organization. This is partly due to the high variability of ways of categorising, multiplying and exchanging propagules of crops in the diversity of peasant societies, and also to the heterogeneity of scientific approaches and methodology used for studying seed networks. Beyond the fact that there exist several ways of defining what 'landraces' are, there are multiple types of practices, communities (or social units) and ways of transacting seeds that have been analysed through a great diversity of interests and methodologies. Based on various field examples drawn from MIRES partners' experiences, we will question units and scales relevant for a constructive way of analysing seed systems without harming the diversity of options farmers use for getting the genetic resources they need and/or wish.