A technique is described for examining the mode of operation of radiation detectors and estimating the losses and gains caused by random summing effects in spectra acquired at high counting rates. The technique is applied to estimating the errors caused by random summing when measuring the amounts of carbon and nitrogen in the human body by the prompt gamma ray neutron activation analysis system installed in Leeds, UK. Our examination shows that the counting rate would have to increase to more than twice the value used at the method's calibration, before an error of greater than 10% is caused in the estimation of carbon by random summing effects, whilst for nitrogen a change of only 18% in the counting rate will cause a 10% error in its estimation.