Quantifying long-term size-selective harvest patterns is necessary for understanding the potential evolutionary effects on exploited species. The comparison of fishery selection patterns on the same species subject to different gear types, in different areas, and over multi-decadal periods can reveal the factors influencing selection. In this study we quantified and compared size-selective harvest by nine Alaskan sockeye salmon (Oncorhynchus nerka) fisheries to understand overall patterns. We calculated length-specific linear selection differentials (the difference in average length of fish before vs. after fishing), which are produced by different combinations of exploitation rates and length-selectivity values, and nonlinear standardized differentials, describing disruptive selection, across all years for each fishery. Selection differentials varied among years, but larger fish were caught in 73% of years for males and 84% of years for females, leaving smaller fish to spawn. Disruptive selection was observed on female and male fish in 84% and 92% of years, respectively. Linear selection was stronger on females than males in 77% of years examined, and disruptive selection was stronger on males in 71% of years. Selection pressure was influenced by a combination of factors under and beyond management control; analyses using mixed-effects models indicated that fisheries were less size selective in years when fish were larger than average and had lower exploitation rates. The observed harvest of larger than average sockeye salmon is consistent with the hypothesis that size-selective fishing contributes to decreasing age and length at maturation trends over time, but temporal variability in selection and strong disruptive selection suggests that the overall directional pressure is weaker than is often assumed in evolutionary models.