INTRODUCTION AND OBJECTIVES: Many dispensing errors occur in the hospital, and these can endanger patients. The purpose of this study was to assess the rate of dispensing errors by a unit dose drug dispensing system, to categorize the most frequent types of errors, and to evaluate their potential clinical significance. METHODS: A prospective study using a direct observation method to detect medication-dispensing errors was used. From March 2007 to April 2007, "errors detected by pharmacists" and "errors detected by nurses" were recorded under six categories: unauthorized drug, incorrect form of drug, improper dose, omission, incorrect time, and deteriorated drug errors. The potential clinical significance of the "errors detected by nurses" was evaluated. RESULTS: Among the 734 filled medication cassettes, 179 errors were detected corresponding to a total of 7249 correctly fulfilled and omitted unit doses. An overall error rate of 2.5% was found. Errors detected by pharmacists and nurses represented 155 (86.6%) and 24 (13.4%) of the 179 errors, respectively. The most frequent types of errors were improper dose (n = 57, 31.8%) and omission (n = 54, 30.2%). Nearly 45% of the 24 errors detected by nurses had the potential to cause a significant (n = 7, 29.2%) or serious (n = 4, 16.6%) adverse drug event. CONCLUSIONS: Even if none of the errors reached the patients in this study, a 2.5% error rate indicates the need for improving the unit dose drug-dispensing system. Furthermore, it is almost certain that this study failed to detect some medication errors, further arguing for strategies to prevent their recurrence.