In an effort to improve patient safety, some hospitals require that two nurses rather than one administer all medications. Such a policy has not been subjected to an economic evaluation. The authors conducted a cross-over study comparing error rates when medication was administered by two nurses as against a single nurse. During a 46 week study period, 319 errors were detected among 129,234 medications administered in a geriatric assessment and rehabilitation unit, giving an overall error rate of 2.5 per 1000 medications. The vast majority of errors detected were relatively minor and had no serious adverse consequences. The error rate per 1000 medications administered by a single nurse was 2.98 (95% CI: 2.45-3.51) and was statistically significantly higher than 2.12 (95% CI: 1.69-2.55) per 1000 for two nurses. A time and motion study indicated that when two nurses rather than one administered medication an additional 17.1 h of nursing time was required per 1000 medications administered. The authors conclude that the use of two nurses to administer medication does statistically significantly reduce the medication error rate, but the clinical advantages are dubious and in our view such a policy cannot be recommended.