Advection, diffusion, total mortality and starvation mortality were measured for 2 cohorts of Pacific herring larvae Clupea harengus pallas1 in Bamfield Inlet, British Columbia, in order to determine the roles of offshore dispersal, population density and starvation in the dynamics of populations. The most strilung differences between the 2 cohorts were in initial density, total mortality, and diffusion: the initial density of the cohort that hatched in 1981 (Cohort 2) was 78 times lower than the density of the cohort that hatched in 1982 (Cohort 6); total mortality was 8 times lower in Cohort 2, 0.02 d-l, than in Cohort 6, 0.16 d-l; and Fickian diffusion was 3 to 6 times lower in Cohort 2, 0.08 km2 d-l, than in Cohort 6, 0.31 to 0.48 km2 d-' Starving larvae, identified with a morphometric index, were found in the 1 to 26 d age group of both cohorts and starvation mortality followed the same trajectory in both cohorts: an increase in magnitude from 0 d-' at hatch to a maximum of 1.0 d 1 at ages 5 to 7 d and then a decline to 0 d-' at age 20 d. That the average magnitude of starvation mortality did not vary with total mortality, as did initial density and the rates of diffusion, suggests that starvation was only one of several agents of mortality and that density-dependent predation and offshore dispersal were equally important factors.