A method of estimating accuracy of test scoring

When errors of test scoring obey a Poisson frequency law (theoretical considerations suggest that they do), the method described may be used for finding the upper fiducial limits of scoring errors per paper. A criterion is suggested for establishing tolerance limits on scoring errors, and a method is given (1) for finding the probability of being wrong in the statement that the tolerance limit is being met for a given size sample or (2) for finding the size of sample that will make this probability not greater than some fixed value.