This essay proposes a new definition of scientific "misconduct," which is broader than the definition recently adopted by the U.S. government. According to the proposed definition, misconduct is a serious and intentional violation of accepted scientific practices, commonsense ethical norms, or research regulations in proposing, designing, conducting, reviewing, or reporting research. Punishable misconduct includes fabrication of data or experiments, falsification of data, plagiarism, or interference with a misconduct investigation. Misconduct does not include honest errors, differences of opinion, or ethically questionable research practices.