By comparing large Arctic charr Salvelinus alpinus, which had shown a persistent cannibalistic response varying from zero to very high in succeeding laboratory trials, with their individual cannibalistic behaviour after release into a natural lake inhabited by small Arctic charr, it was found that all Arctic charr had the potential to become cannibalistic, irrespective of their laboratory behaviour. More specifically, Arctic charr that never fed on prey fishes when offered them in the tank experiments turned to cannibalism when released in the lake, highlighting the potential difficulties in extrapolating laboratory results to natural settings. This was also true for naive fish that had no prior experience of eating live food. Since no significant increase in the number of prey consumed during each of the succeeding laboratory trials was found, and naive fish showed a response under natural conditions similar to that of their counterparts, the training of the Arctic charr (or experience or learning) probably had no effect upon the piscivorous or cannibalistic response after stocking. Thus, the study appeared to demonstrate that most variations in cannibalism in Arctic charr was simply a function of environmental conditions, depending on the density of conspecifics v. alternative prey, and the relative size difference between predator and prey, rather than any genetic influence.