Over the last 20 years, there has been much research aimed at improving environmental risk assessment of transgenic crops. Despite large amounts of data, decisions to allow or prohibit the release of transgenic crops remain confused and controversial. We argue that part of the reason for confusion is the lack of clear definitions of components of the environment that should be protected, and, as a consequence, there is no way to judge the relevance of data collected under the auspices of 'environmental risk assessment'. Although this criticism applies to most aspects of environmental risk assessment of transgenic crops, it is most pertinent to effects that might result from an increase in plant fitness, often referred to as increased weediness. Environmental risk assessment of weediness is regarded as complicated: an increase in the fitness of a transgenic plant compared with non-transgenic counterparts will be the result of an interaction between the altered plant phenotype and an enormous number of environmental variables. This has led to the idea that risk assessment of weediness needs to "understand" these interactions, with the implication that exhaustive data are required. Here we argue that environmental risk assessment of the weediness of transgenic plants need not be complicated. Analysis of the conditions that must be met for increased weediness to occur suggests a series of studies that starts with simple tests in the laboratory under "worst case" assumptions, and becomes increasingly complex and realistic should the simpler studies not indicate negligible risk with sufficient certainty. We illustrate how the approach might work for assessing the risks of increased weediness using the example of possible introgression of a gene for Turnip mosaic virus (TuMV) resistance from oilseed rape to certain wild Brassica species.