We fitted hierarchical Bayesian models to coded wire tag data and studied the variation in the probability that male coho salmon Oncorhynchus kisutch released at 72 locations in 41 basins and 8 regions mature early. We used data from at least 10 cohorts at each location; therefore, our analysis was both spatially and temporally more extensive than those conducted previously. We accounted for variation in the log-odds of early maturation with (1) mean smolt weight, (2) covariates (developed specifically for this paper) that indexed marine environmental conditions during the first few months after out-migration, and (3) hierarchical means that identified each cohort's release location, basin, and region. There was posterior support for location-specific effects of smolt size on the probability of early maturation across the geographic range of this species, but not for location-specific marine environmental effects. Nevertheless, substantial residual variation remained from our final analysis; we argue tliat this variation can be attributed to environmental effects that were not detected here. The spatial variation in the probability of early maturation was dominated by region-level differences in freshwater rearing conditions, as evidenced by positive relationships between mean smolt weight and the expected log-odds of early maturation at all levels in the geographic hierarchy. We identified more than one such relationship, and the differences between the center and the edges of the geographic range are considered here. Ultimately, since we defined regions on the basis of existing evolutionarily significant units for coho salmon, our results further support the use of these units for management and conservation purposes.