Nitrogen dilution curves relate a crop’s critical nitrogen concentration (%Nc) to biomass (W) according to the allometric model %Nc = a W-b. This model has a strong theoretical foundation, and parameters a and b show little variation for well-watered crops. Here we explore the robustness of this model for water stressed crops. We established experiments to examine the combined effects of water stress, phenology, partitioning of biomass, and water-soluble carbohydrates (WSC), as driven by environment and variety, on the %Nc of wheat crops. We compared models where %Nc was plotted against biomass, growth stage and thermal time. The models were similarly scattered. Residuals of the %Nc - biomass model at anthesis were positively related to biomass, stem:biomass ratio, Δ13C and water supply, and negatively related to ear:biomass ratio and concentration of WSC. These are physiologically meaningful associations explaining the scatter of biomass-based dilution curves. Residuals of the thermal time model showed less consistent associations with these variables. The biomass dilution model developed for well-watered crops overestimates nitrogen deficiency of water-stressed crops, and a biomass-based model is conceptually more justified than developmental models. This has implications for diagnostic and modeling. As theory is lagging, a greater degree of empiricism might be useful to capture environmental, chiefly water, and genotype-dependent traits in the determination of critical nitrogen for diagnostic purposes. Sensitivity analysis would help to decide if scaling nitrogen dilution curves for crop water status, and genotype-dependent parameters are needed.