This study analyses data from over 20 000 haddock stomachs collected during the 1981 and 1991 Stomach Sampling Projects of the North Sea. Sampling was within the framework of the Multispecies Virtual Population Analysis (MSVPA) for fisheries stock assessment. In 1981 stomachs were collected to calculate input feeding parameters from main predators. During 1991 the sampling exercise was repeated to test stability of the parameters in the region. We investigate stability of haddock stomach content weight between years accounting for ontogenic, spatial and temporal variations within the scope of the survey resolution. Analysis using generalized linear and additive models is performed for weight of the stomach content including and excluding empty stomachs and also for proportion of stomachs without food. Results indicate that haddock stomach contents varied significantly between years, quarters and North Sea roundfish areas. Content weights were overall higher in 1981 than in 1991. In 1981 levels were high from spring to fall and in 1991 mostly in spring. During both years levels were lowest in the central North Sea and in winter. Most of the deviance in modelling the stomach weight is explained by the length of the predator. Stomachs sampled in 1981 were most frequently empty in spring and those sampled in 1991 in winter. In both years, proportions decreased with fish length except in winter when levels increased. Proportion of empty stomachs was highest in the central region of the North Sea. Most of the proportion variation is explained by quarterly fluctuation. Variation of content weight can be related to prey composition found in the stomachs. High stomach contents are generally associated with high contribution of fish prey to the total stomach content, mainly sandeels. We propose that sandeels were more vulnerable to predation in 1981 than in 1991 due to colder temperatures. ?? 2002 International Council for the Exploration of the Sea. Published by Elsevier Science Ltd.