We introduce here a new experimental setup that provides temporally aligned linguistic and behavioral data together with physiological activity time-series recorded during social interactions. It brings the experimental approach closer to ecological social interaction. Such endeavour requires the aggregation of linguistic, physiological and neuro-cognitive information. Compared to measurement of activity grounded on existing linguistic material our setting presents some additional challenges as we are dealing with conversations. In addition to present the rationale, setup and preliminary analyses, we discuss (i) the challenges caused by the spontaneous and interactional nature of the activity recorded ; (ii) the problem of balancing experimental setup between the technical needs and the desire to keep some level of naturalness in the task ; and (iii) the difficulties in relating in a temporal way linguistic events with physiological signals that have their own biological dynamics.