The analysis of the evolution of crop yield reveals a change in grain composition: increases in yield have led to a decrease in the protein to starch or oil ratios. This negative relationship reflects the higher increase of C assimilation compared with N assimilation. For wheat (Triticum aestivum L.), flour protein concentration is the main quality criterion, especially for bread making. Therefore, a critical question for the future is how to manage the relationship between yield and N concentration, where the objective is to increase both the level and stability of yield and N concentration. To answer these questions, we need a better understanding of the mechanisms involved in the variations of yield and of its composition. First of all, we need to analyse genetic variability in different environments that allow the identification of genetic sources of variation that can be used for breeding or in more reductionist approaches. In this paper, we used data from controlled environment and field experiments, at canopy and plant level, and at different sink:source ratios to analyse the genetic and environmental relationship between grain productivity and composition. These experiments confirmed the strong negative relationship between grain yield and N concentration. Post-anthesis temperature and water deficit had significant effects on grain yield and protein concentration, but they did not modify the negative relationship between these variables. However, pre-anthesis water deficit decreased the sink:source ratio resulting in a lower intercept, while the slope was unchanged. Nitrogen deficiency also modified the intercept of the negative relationship, but more importantly it decreased the slope three to four-fold. Thus under limiting N conditions, grain N concentration is more sensitive to yield variation than under non-limiting N conditions. Genetic variation of single grain dry mass and of the sink:source ratios had similar effects to the environmental variation. Three major conclusions can be drawn from these results: (1) the negative relationship between grain yield and protein concentration is primarily determined at the stem level; (2) the grain itself is more limiting for starch synthesis than for protein synthesis and (3) overall any increase in yield is followed by an increase of N utilisation and use efficiency.