We assess the ability of multivariate statistical analyses applied to event hydrographs parameters, to characterize a catchment hydrological behaviour. Motivation for such an approach lies in the fact that streamflow records have yet to be exploited to their full potential towards hydrological interpretation and can be used to infer a catchment state of connectivity from a qualitative standpoint. We have therefore processed 96 event hydrographs from a small headwater temperate humid forested catchment using principal component analysis, variation partitioning and classification tree analysis. These techniques prove to be promising in discriminating contrasted types of hydrologic responses (e.g. low- vs high-magnitude events, slow vs quick timing events), identifying the main hydro-meteorological variables that control these responses and determining thresholds values of the hydro-meteorological variables leading to a switch between catchment response types. Copyright © 2010 John Wiley & Sons, Ltd.