Multivariate techniques allow data sets to be summarised, but are not deslgned to causally model ecological phenomena. The ambiguity of the interpretation results from the ambiguity of the scaling space. The phytosociological method proposed by Grandjouan (1982; These d'etat, Universite Louis Pasteur, Strasbourg) is a general interpretive framework allowing separate consideration of 3 distinct reference spaces. a geographical space to visualise the spatio-temporal distribution of the descriptors, an ecological space showing the effects of the environmental variables on the biological descriptors, and a bio-ethological space showing the statistically meaningful species cornmunities. It is thus possible to define the position of each biological category and each observation along the scale of variation of each environmental variable (hydrological position, HP) or each geographic coordinate (geographical posltlon. GP), as well as its density relatlve to these scales (indicative value, IV). Using continuous zooplankton data collected along a transect across the Llgunan Sea front, pnncipal component analysis and correspondence analysis were compared to this technique. The phytosociological method produced the most meaningful ecological results about the structure of thls frontal ecosystem.