Aim. This paper is a report of the development and validation of an instrument to measure nurses' knowledge of high-alert medications and to analyse known administration errors. Background. Insufficient knowledge is a factor in nurses' drug administration errors. Most errors do not harm patients, but incorrect administration of high-alert medications can result in serious consequences. Sufficient knowledge about high-alert medications is vital. Method. A cross-sectional study was conducted in 2006 in Taiwan using a questionnaire developed from literature review and expert input, and validated by subject experts and two pilot studies. Section 1 of the questionnaire (20 true-false questions) evaluated nurses' knowledge of high-alert medications and section 2 was designed to analyse known administration errors. Snowball sampling and descriptive statistics were used. Findings. A total of 305 nurses participated, giving a 79·2% response rate (305/385). The correct answer rate for section 1 was 56·5%, and nurses' working experience contributed to scores. Only 3·6% of nurses considered themselves to have sufficient knowledge about high-alert medications, 84·6% hoped to gain more training, and the leading obstacle reported was insufficient knowledge (75·4%). A total of 184 known administration errors were identified, including wrong drug (33·7%) and wrong dose (32·6%); 4·9% (nine cases; 9/184) resulted in serious consequences. Conclusion. The questionnaire was valid and reliable. Evidence-based results strongly suggest that nurses have insufficient knowledge about high-alert medications and could benefit from additional education, particularly associated with intravenous bolus administration of high-alert medications. Further research to validate the instrument is needed.