The scale of variation in population densities and other demographic variables is an important consideration in the design and interpretation of experiments and sampling programs. Here, we studied spatial and temporal variation in populations of Corophium volutator, an intertidal amphipod that is the most abundant macro-invertebrate on mudflats in the upper Bay of Fundy, Canada. Variables that were quantified included density (peaked in mid-summer at 10,000s per m²), proportion juvenile (an index of population structure), sex ratio (female-biased throughout the year), proportion of females that were ovigerous (reproduction occurs between May and August), and fecundity (number of embryos per female). We studied populations at 4 different sites, i.e., separate mudflats, for 1–2 years, and estimated variance components at 3 spatial scales: metres (samples), 100s of metres (transects) and many kilometres (mudflats). Population density exhibited low variation between years (< 10% of the random variation), but showed high variation at our smallest and largest spatial scales (45% between samples, 2% between transects, and 40% between mudflats), i.e., the distribution of C. volutator within a mudflat was aggregated at the scale of metres, but not at that of 100s of metres. Thus, relatively few transects, but many samples per transect, are required for good representation of density on a mudflat. Fecundity was similar between mudflats, but proportion juvenile, sex ratio, and proportion of ovigerous females were strongly affected by site or the interaction between month and site (> 60% of the random variation in each case). The high variation observed between mudflats for most demographic variables demonstrates that several control sites are necessary for measuring natural variation, a critical consideration in studies of environmental impacts.