A data base was constructed of the % N and plant d. wts (W) in t ha-1 of C3 and C4 crops that had been grown with sufficient nitrogen to permit maximum growth rate. The % N of all crops declined sharply with increase in W but this decline differed between C3 and C4 crops. When W was greater than I t ha-1, 86% of the variance in In % N was removed by the model % N = aW-b with b = -0·5 for all crops, and a = 5·7% for C3 crops and 4·1 % for C4 crops. The same model gave a good description of data on C3 and C4 crops entirely independent of that used for developing the model. According to this relationship the fractional decline in % N with increase in plant mass was the same for both types of crops, but C4 crops contained about 72% of the nitrogen in C3 crops at equivalent d. wts. As approx. 32% more dry matter was produced per unit of intercepted radiation for C4 and C3 crops, the N uptake (or weight of plant protein produced) per unit of intercepted radiation was approximately the same for both types of crops.A small improvement in the degree of fit to % N = aW-b was obtained by allowing both a and b to vary with the crop. Values of b obtained in this way for tall fescue, lucerne and winter wheat, but not for potato and sorghum, were consistent with Hardwick's 'skin core' hypothesis (Annals of Botany, 1989, 60, 439-46). The entire data set was, however, consistent with Caloin and Yu's model (Annals of Botany, 1984, 54, 69-76) in which there is a conceptual N pool for photosynthesis and another N pool for the other processes. © 1990 Annals of Botany Company.