Fish scales from seven species of demersal fish in an archival collection were analyzed for stable isotopic compositions of carbon and nitrogen to study long-term changes in trophic structure of the Georges Bank food web. Nitrogen isotopic compositions are often used to infer trophic level. In the case of haddock, Melanogrammus aeglefinus (Linnaeus), there was a trend towards feeding at 2/3 of one trophic level (2.45% in ¹⁵N) lower in 1987 than in 1929. Values of °¹³C, frequently employed to identify sources of organic carbon to consumers, declined by 1.5 from 1929 to 1960, and then increased again toward the present, suggesting changes in the food web at the level of the primary producers. Superimposed on long-term isotopic trends were short-term variations (1 to 10 yr).To identify potential causes for these isotopic trends, canonical correlation analysis was performed between isotopic data and a suite of environmental and population factors including sea surface temperature, the Greenland Regional Pressure Anomaly (GRPA), the North Atlantic Oscillation (NAO), and the following haddock stock parameters; stock size, fishing mortality, recruitment, and weight-at-age-2 (a measure of growth rate). Isotopic variation was significantly correlated with a combination of environmental and population variables: GRPA, NAO, weight-at-age-2, stock size, and fishing mortality. On the basis of published gut content analyses, the seven fish species were predicted to vary in trophic level (TL) from American plaice, Hippoglossoides platessoides (Fabricius), (TL 2.9) to summer flounder, Paralichthys dentatus (Linnaeus), (TL 4.5), whereas measured °¹⁵N values suggested smaller differences in the trophic levels of these species (less than one TL). Four species showed good agreement between gut-predicted and measured ¹⁵N values, while three species did not. Inclusion of information on ontogenetic dietary shifts in our predictions improved the agreement in some cases but not in other. Differences between stable isotope analysis and gut content analysis in terms of what they measure, i.e., integrated assimilated diet vs short-term ingested diet, respectively, may account for some of the differences in results. Based on our analyses and previous studies, feeding habits of these fish may undergo considerable year-to-year and geographic variation, some of which may have been missed in gut content analyses. To the extent that these fish are representative members of the food web, trophic variation in these fish may indicate more general changes in the food web.