Using Insulin RIA Kit (double antibody method, Dainabot Co., Ltd.), individual components of error due to pipetting of reagents, bound-free separation and counting were evaluated experimentally and theoretically. They were statistically synthesized to get the total response error relationship (RER). The resulting RER agreed well with the pooled RER obtained by the method reported previously. This good agreement indicates that individual components of error were reasonably and precisely estimated. Based on this result, contribution rates of individual components of error to the total response error were evaluated. The averaged contribution rates over the range to measure, when pipetting precision, C.V. (coefficient of variation) = 0.005, were 64% for counting error, 24% for bound-free separation error, and the error related with reagent delivery was only 12%. When C.V. = 0.01, however, which is considered normal pipetting precision, the contribution rate for the latter went up to 35%. Effect of poor precision in pipetting on final assay results is so great that much more attention than we usually think of should be paid to pipettes and pipetting techniques.