The prevention of medication administration errors (MAEs) represents a central focus of hospitals' quality improvement and risk management initiatives. Because the identification and reporting of MAEs is a nonautomated and voluntary process, it is essential to understand the extent to which errors may not be reported. This study reports the results of 2 multihospital surveys in which over 1300 staff nurses in each survey estimated the extent to which various types of nonintravenous (non-i.v.) and intravenous (i.v.)-related MAEs are actually being reported on their nursing units. Overall, respondents estimated that about 60% of MAEs are actually being reported. Considerable differences in estimated rates of MAE reporting were found between staff and supervisors working on the same patient care units. A simulation based on actual and perceived rates of MAE reporting is presented to estimate the range of errors not being reported. Implications regarding the reliability, validity, and completeness of MAEs actually being reported are discussed.