Researchers in online learning are hampered in their efforts to improve the visibility and validity of their work, to promote in-depth discussion of their analyses, to compare methods, tools or even findings, by the fact that access to corpora in their original state is often impossible. In other fields, the need for corpora to be made available across different teams of researchers was expressed before we made it a research orientation in online learning. But in these domains, for example in ATNL, the artefacts in the data collections are less complex and more homogeneous (tablets, texts, books, recordings of telephone conversations etc) than in ours. Building a corpus from an online experiment involves synchronising and linking together many data fragments (log-on archives, traces of interactions, student products, tests, interviews etc) so that an analysis can be carried out post factum. This methodological challenge has recently gained scientific recognition, as is shown by the support of the Agence Nationale de la Recherche for the Mulce project (http://mulce.univ-fcomte.fr/). In order to further explore this area and to expose this methodological orientation to wider scrutiny, we offer a practical workshop (symposium). We are pleased to invite researchers from different horizons to discuss a wide range of issues such as types of data to be collected, tools and methods to be used in the processing and analysis of these data, such as might best contribute to the construction of corpora suitable for cross-disciplinary exchange and for treatment by different teams.