Data obtained from routine analytical radioimmunoassays were processed using five curve-fitting procedures, viz. ‘Amersham’, single binding site, four parameter logistic, a linear logit-log and a polynomial logit-log. The polynomial logit-log procedure gave the best fit, but this was probably due to the inherent flexibility of this curve-fitting process since the analytical precision achieved with it was no better than what was obtained with most of the other procedures. A limited study failed to show that statistical weighting of data before curve fitting had any practical advantage.