Background, aim, and scope We strive to predict consequences of genetically modified plants (GMPs) being cultivated openly in the environment, as human and animal health, biodiversity, agricultural practise and farmers' economy could be affected. Therefore, it is unfortunate that the risk assessment of GMPs is burdened by uncertainty. One of the reasons for the uncertainty is that the GMPs are interacting with the ecosystems at the release site thereby creating variability. This variability, e. g. in gene flow, makes consequence analysis difficult. The review illustrates the great uncertainty of results from gene-flow analysis. Main features Many independent experiments were performed on the individual processes in gene flow. The results comprise information both from laboratory, growth chambers and field trials, and they were generated using molecular or phenotypic markers and analysis of fitness parameters. Monitoring of the extent of spontaneous introgression in natural populations was also performed. Modelling was used as an additional tool to identify key parameters in gene flow. Results The GM plant may affect the environment directly or indirectly by dispersal of the transgene. Magnitude of the transgene dispersal will depend on the GM crop, the agricultural practise and the environment of the release site. From case-to-case these three factors provide a variability that is reflected in widely different likelihoods of transgene dispersal and fitness of introgressed plants. In the present review, this is illustrated through a bunch of examples mostly from our own research on oilseed rape, Brassica napus. In the Brassica cases, the variability affected all five main steps in the process of gene dispersal. The modelling performed suggests that in Brassica, differences in fitness among plant genome classes could be a dominant factor in the establishment and survival of introgressed populations. Discussion Up to now, experimental analyses have mainly focused on studying the many individual processes of gene flow. This can be criticised, as these experiments are normally carried out in widely different environments and with different genotypes, and thus providing bits and pieces difficult to assemble. Only few gene-flow studies have been performed in natural populations and over several plant generations, though this could give a more coherent and holistic view. Conclusion The variability inherent in the processes of gene flow in Brassica is apparent and remedies are wished for. One possibility is to expose the study species to additional experiments and monitoring, but this is costly and will likely not cover all possible scenarios. Another remedy is modelling gene flow. Modelling is a valuable tool in identifying key factors in the gene-flow process for which more knowledge is needed, and identifying parameters and processes which are relatively insensitive to change and therefore require less attention in future collections of data. But the interdependence between models and experimental data is extensive, as models depend on experimental data for their development or testing. Recommendations More and more transgenic varieties are being grown worldwide harbouring genes that might potentially affect the environment ( e. g. drought tolerance, salt tolerance, disease tolerance, pharmaceutical genes). This calls for a thorough risk assessment. However, in Brassica, the limited and uncertain knowledge on gene flow is an obstacle to this. Modelling of gene flow should be optimised, and modelling outputs verified in targeted field studies and at the landscape level. Last but not least, it is important to remember that transgene flow in itself is not necessarily a thread, but it is the consequences of gene flow that may jeopardise the ecosystems and the agricultural production. This emphasises the importance of consequence analysis of genetically modified plants.