The distribution of dispensing errors occurring in hospital pharmacy was analyzed statistically. The sample distribution of the daily error rate (ie, rate of the number of dispensing errors detected daily to the number of prescriptions filled daily) could not be fully analyzed, because two peaks of the distribution were observed in the histogram. The sample distribution of the risk rate of dispensing errors (ie, rate of the recorded number of daily dispensing errors to the expected number, D) was stable as compared with the sample distribution of the daily error rate. The distribution of the error risk rate was slightly biased to the left (skewness √β1>0) and flatter than a normal distribution (kurtosis β2<3). In spite of the above discrepancy, the distribution of D conformed to a normal distribution in the test of goodness of fit.