The densities of each a g e class of adult (>l yr old) mussels Mytilus edulis were measured in autumn and spring over 8 yr on each of 12 intertidal mussel beds in the Exe estuary, southwest England. The possibility that density-dependent changes in adult numbers in each of 9 yearly age classes occurred over the summer and winter periods was explored directly on individual beds and indirectly by combining data from all beds. There was direct evidence of density-dependent changes in both seasons in some age classes on some beds. The indirect tests suggested, however, that density-dependence was much more widespread over beds, age classes and seasons than had been suggested by the direct tests. The critical density above which density-dependent losses occurred, and to which the density-dependence tended to return numbers irrespective of the abundance of the mussels at the start of summer or winter, varied considerably among beds in several age classes. These critical densities were correlated most closely with the time for which a bed was exposed over low tide but were also influenced by the distance of a bed from the estuary mouth and the softness of its substrate. Between them, these 3 variables defined the environmental gradient of intertidal mussels in the Exe estuary. The residual, density-independent variations in the changes in numbers of several age classes of adult mussels about the density-dependent relationships were correlated with weather variables (temperature, gales and rainfall) in several cases and in both seasons. Surprisingly, there was evidence of widespread density-dependent immigration of adult mussels, especially in summer and in the younger age classes. While the passive movement by adult mussels in storms is well established, it is not known whether active movements occur in these older mussels. Although much of the density-dependence was due to mortality, immigration was therefore often a n important component of the densitydependent relationships detected on individual beds. However, immigration on some beds countered losses on others. Perhaps because of this, density-dependence was often masked, and thus not detected, at the estuary scale in all but the youngest mussels whose estuary-wide abundance was dominated by 2 large mussel beds, where their numbers tended to CO-vary. Thus, strong densitydependence at one scale may not be detected at a higher scale.