The ultimate goal of early life studies of fish over the past century has been to better understand recruitment variability. As evident in the Georges Bank haddock (Melanogrammus aeglefinus) population, there is a strong relationship between recruitment success and processes occurring during the planktonic larval stage. This research sought new insights into the mechanisms controlling the recruitment process in fish populations by using biological-physical modeling methods together with laboratory and field data sets. We created the first three-dimensional model of larval haddock on Georges Bank by coupling models of hydrodynamics, lower trophic levels, a single copepod species, and larval haddock. Interactions between feeding, metabolism, growth, vertical behavior, advection, predation, and the physical environment of larval haddock were quantitatively investigated using the coupled models. Particularly, the model was used to compare survival over the larval period and the sources of mortality in 1995 and 1998, two years of disparate haddock recruitment. The results of model simulations suggest that the increased egg hatching rates and higher food availability, which reduced starvation and predation, in 1998 contributed to its larger year-class. Additionally, the inclusion of temperature-dependent predation rates produced model results that better agreed with observations of the mean hatch date of survivors. The results from this biophysical model imply that food-limitation and its related losses to starvation and predation, especially from hatch to 7 mm, may be responsible for interannual variability in recruitment and larval survival outside of the years studied.