We examined evidence for nonlinear dynamics in fishery-independent survey data for an assemblage of 26 fish species on Georges Bank spanning the period 1963 to 2008. We used nonlinear time series analysis to determine (1) the presence of nonlinear dynamics in fish populations on Georges Bank, (2) the minimum number of dimensions required to effectively describe system dynamics, (3) the strength of patterns of co-predictability among all possible pairs of fish species, and (4) identification of groups of species characterized by similar dynamics. Here, nonlinear behavior refers to non-equilibrium dynamics, including chaos. The population trajectories of all 26 species exhibited strong density-dependent feedback as indicated by a Partial Rate Correlation Function analysis. Significant evidence of complex dynamical behavior was found for approximately 1 in 5 species. Low dimensionality for many of the individual series was identified, suggesting that for a given level of predictability, this system can be represented by a relatively small number of critically important ecological variables. Further, we found high levels of co-predictability among pairwise combinations of individual species. We identified 4 major species groups sharing similar dynamic features on the basis of patterns of co-predictability, and explored potential mechanisms for interpreting the groupings in terms of trophic interactions and life history characteristics.