Abstract For a reliable assessment of the impacts of climate change on future yields of arable crops the interpretation and extrapolation of the results from crop experiments has been a major problem. The reason is the large variation in observed crop responses to CO 2 enrichment and temperature change. This study aims at improved understanding of the causes of this variability. A large number of data sets from wheat experiments under CO 2 enrichment or temperature change were analysed, and time courses of crop growth processes have been simulated with a crop growth model. These model analyses explained the variability in observed yield responses on the basis of crop characteristics and climatic conditions. Simulation of wheat experiments with the new version of light interception and utilisation (LINTUL) for climate change studies (LINTULCC) growth model showed that the correspondence between observed and simulated yield responses to CO 2 enrichment or temperature rise was often poor. This was partly caused by the modelling approach in LINTULCC and partly by the experimental variability which cannot be reproduced by LINTULCC. This comparison of experimental with simulated results showed that their correspondence might be improved in two opposite ways. The morphological development of the crop and the limiting effects of available sinks (i.e. grains) on crop growth may be described in more detail. However, in many situations (i.e. without severe water or nutrient stress) crop variables in the experiments were more constant than simulated with LINTULCC. In such a situation, a simplified model approach may be applied and may yield better results.