Environmental risk assessment of genetically modified organisms requires determination of their fitness and invasiveness relative to conspecifics and other ecosystem members. Cultured growth hormone transgenic coho salmon (Oncorhynchus kisutch) have enhanced feeding capacity and growth, which can result in large enhancements in body size (>7-fold) relative to nontransgenic salmon, but in nature, the ability to compete for available food is a key factor determining survival fitness and invasiveness of a genotype. When transgenic and nontransgenic salmon were cohabitated and competed for different levels of food, transgenic salmon consistently outgrew nontransgenic fish and could affect the growth of nontransgenic cohorts except when food availability was high. When food abundance was low, dominant individuals emerged, invariably transgenic, that directed strong agonistic and cannibalistic behavior to cohorts and dominated the acquisition of limited food resources. When food availability was low, all groups containing transgenic salmon experienced population crashes or complete extinctions, whereas groups containing only nontransgenic salmon had good (72.0 +/- 4.3% SE) survival, and their population biomass continued to increase. Thus, effects of growth hormone transgenic salmon on experimental populations were primarily mediated by an interaction between food availability and population structure. These data, while indicative of forces which may act on natural populations, also underscore the importance of genotype by environment interactions in influencing risk assessment data for genetically modified organisms and suggest that, for species such as salmon which are derived from large complex ecosystems, considerable caution is warranted in applying data from individual studies.