SYNOPSIS AND INTRODUCTION: Much of the past research in accounting ethics has focused on whether accountants conform to prescribed codes of professional ethics. Other research has been normative in nature, recommending what constitutes appropriate ethical conduct or focusing on the accountant's responsibility in society. This study selects a different approach by testing a multivariate measure of how accountants make ethical judgments. Data were gathered with the assistance of the institute of Certified Management Accountants and the Institute of Management Accountants (formerly the National Association of Accountants). Accountants were asked to respond on bipolar scales to realistic scenarios involving ethical decisions. Several tests for construct validity produced supportive results for the hypothesized three-dimensional measure, with the dimensions being moral equity, relativism, and contractualism. First, we developed a questionnaire with four scenarios concerning ethical issues. Each scenario ended in a particular action taken by an individual. Responses to that action were recorded on eight bipolar scales, representing the three dimensions above. The questionnaire was mailed to 500 randomly selected certified management accountants resulting in a 62.8 percent response rate. Second, the results from a factor analysis and a traditional reliability coefficient test suggest that a high degree of internal consistency exists for each dimension of the measure. The appropriate factor loadings ranged from a low of 0.68 to a high of 0.92, while the reliability coefficients varied from 0.75 to 0.92. Next, the content validity of the three-dimensional measure was checked by comparing it with a global ethical/unethical measure. Again, the results support the hypothesis that the multivariate measure captures the appropriate domain of content. Adjusted R2-values ranged from 0.59 to 0.76 when the global measure was regressed against the multivariate measure. Finally, a sense of predictive validity was obtained by comparing the multivariate measure with a behavioral intention measure for the respondent. Adjusted R2-values ranged from 0.45 to 0.76 for the four scenarios tested, indicating that the three-dimensional measure "explains" a respectable portion of the variance in the behavioral intention of the individual. The multidimensional measure developed in this study may be a guide for future research into how accountants make ethical judgments. Such knowledge can be used in turn to develop useful codes of conduct, create ethical organizational cultures, and direct ethical training for and by accountants.