Atlantic salmon (Salmo salar) exhibit extreme diversity in the age of smelt migration and sexual maturation, both within and among populations. Theoretical analyses reveal the adaptive significance of such variation, but models of the underlying physiological and behavioral mechanisms are also needed. I summarize one such proximate model that suggests that smelting and maturation are only triggered if expected performance trajectories exceed threshold levels during sensitive periods. The probability of a threshold being exceeded is therefore dependent on an individual fish's ability to acquire and utilize resources efficiently, which in turn depends on a range of physiological and behavioral traits. Spatial variation in life histories is chiefly caused by differences in growth opportunity, although there is evidence of geographical variation in genetically determined expected growth trajectories. Simulations show that small changes in young fry growth rates can have pronounced effects on mean smelt age and sex ratio, by influencing the proportion of males that fail to exceed the threshold for early smelting and mature as parr (and so are less likely to survive to smelting). More sophisticated proximate models should allow powerful predictions of smelt production based on simple environmental parameters, due to their influence on growth trajectories and hence life history decisions.