An experiment was performed to obtain judgments of 324 college students about a list of 18 causal explanations for good or poor performance on an examination. These stimulus causes were judged with respect to a description of two hypothetical situations in which a student either did well or did poorly on the examination. Half the sample judged the similarity or difference of the 153 possible pairings of these 18 causes for success or failure. The other half of the sample was asked to rate the stimulus causes on 14 bipolar scales. Multidimensional scaling analysis of the similarities data in the success and failure conditions indicated that the subjects in each condition distinguished the causes on two dimensions: internal versus external and intentional versus unintentional. Analysis of the bipolar scale ratings provided statistical support for the interpretation of these dimensions. The results are discussed in relation to the a priori causal distinctions made in current attribution models of success-failure. (Author/CTM)