Few studies have determined whether formal estimates of selection explain patterns of trait divergence among populations, yet this is one approach for evaluating whether the populations are in equilibria. If adaptive divergence is complete, directional selection should be absent and stabilizing selection should prevail. We estimated natural selection, due to bear predation, acting on the body size and shape of male salmon in three breeding populations that experience differing predation regimes. Our approach was to (1) estimate selection acting within each population on each trait based on an empirical estimate of reproductive activity, (2) test for trait divergence among populations, and (3) test whether selection coefficients were correlated with trait divergence among populations. Stabilizing selection was never significant, indicating that these populations have yet to attain equilibria. Directional selection varied among populations in a manner consistent with trait divergence, indicating ongoing population differentiation. Specifically, the rank order of the creeks in terms of patterns of selection paralleled the rank order in terms of size and shape. The shortest and least deep-bodied males had the highest reproductive activity in the creek with the most intense predation and longer and deeper-bodied males were favored in the creeks with lower predation risk.