ABSTRACTCategorical time series regression was applied to 55 fish stocks in the Potomac, Hudson, Narragansett, Delaware, and Connecticut estuaries for the period 1929–1975. Interannual variability in catch per unit effort (CPUE) was related to CPUE, hydrographic variables, and pollution variables, lagged back in time to represent the conditions contributing to the multiple ages comprising each fishery. Hydrographic variables included water temperature and flow in the estuary– and, for offshore spawning stocks, wind direction and magnitude–during the months of spawning and early life stage development. Pollution variables included measures of dissolved oxygen conditions in the estuaries, volume of material dredged, and sewage loading (or human population). Lagged CPUE, hydrographic variables, and pollution variables all played important roles in explaining historical variability in CPUE. Lagged CPUE was significant in 45 of 55 stocks generally accounting for 5–35% of the variability. Lagged hydrographic variables were significant in 53 of 55 stocks, explaining an additional 5–40% of the variability unaccounted for by lagged CPUE. Lagged pollution variables were significant in 35 of 55 stocks, generally accounting for an additional 5–30% of the variability not explained by lagged CPUE and hydrographic variables. Results did not exhibit expected patterns of consistency in the importance of lagged CPUE for a species across estuaries or consistency in the importance of pollution variables across estuaries. Results did exhibit the expected north‐to‐south longitudinal pattern in the importance of timing of the hydrographic variables, the months of importance being one or two months later in more northerly estuaries. Higher‐order interaction effects were important in almost all stocks that were well‐modeled by categorical time series regression. Of the 30 stocks with final regression models having R2 > 0.55, 26 stocks involved significant interaction effects, five had only significant interaction effects (no significant main effects), and 20 stocks had significant interactions involving variables not significant as main effects. The difficulties involved in analyzing long‐term trends in fish populations and partitioning variability between natural and anthropogenic sources are discussed.