Current interest in multicatchment hydrologic studies challenges the use of geochemical mixing models across scale, where changes in stream water chemistry from catchment to catchment may indicate (1) changes in the proportional contributions of end-members, (2) changes in the geochemical signatures of end-members in space, or (3) changes in the geochemical signatures of end-members in time. In this study we examine stream water chemistry from a series of eight nested catchments in a 1.47 km2 temperate forest watershed in southern Quebec for evidence of contributing end-members. We use eigenvector and residual analysis (Hooper, 2003) of the multivariate stream water chemistry records to estimate the dimensionality of the mixing space for each individual catchment, indicating the number of contributing end-members. Using the mixing space of the largest, highest-order catchment (1.47 km2), we evaluate its ability to predict stream water chemistry in the seven upstream catchments, representing progressively smaller areas. We observe significant spatial variation in ionic mixing ratios within the 147 ha watershed. Only spatial testing across catchments allowed us to identify appropriate conservative tracers most compatible with the application of a single mixing model across scale. On the seasonal timescale, groundwater geochemistry changes significantly due to the recharge from spring snowmelt, indicating a mixture of two groundwater end-members of varying age. On the timescale of storm events, shallow perched water and throughfall provide geochemical signatures consistent with physical mixing while unsaturated zone soil water sampled from local pockets of glacial till does not. Our results suggest cautious application of end-member mixing analysis (EMMA) for multicatchment studies.