Fishery processes of selectivity and recruitment to legal size vary with fish length, and are mediated by fish growth. However most fishery models are age-based. To model length-dependent change within each cohort, fish numbers must vary dynamically with length as well as age in the model population array. The fishery model formalism described here achieves this by a partition of the continuous length-at-age distribution. This method is computationally efficient and cleanly differentiates legal from sub-legal fish. Fish numbers within each cohort are partitioned into length bins, called ‘slices’. A ‘slice’ is defined and calculated as the fish in each cohort length-at-age distribution that have grown into legal size since the start of the previous time step. When growth is estimated from catch length and age samples separate from the stock assessment, biases result from the implicit assumption that catch samples are representative of the population, and from ignoring length-dependent change within cohorts. These biases are avoided by integrating recruitment, growth and selectivity estimation into a stock assessment likelihood that represents changing population numbers by both age and length. Size-dependence also permits a natural extension of fishery models to trophic interactions with the surrounding ecosystem.