The argument that the personality structures obtained from retrospective ratings reflect semantic similarity structures has been as provocative as a red cape in the bull ring. High congruence between those two kinds of structures seems well established. What is less clear is how and why those structures differ from that for immediate judgments of the same behaviors. The time interval between perceiving the pertinent behavior and making a judgment about it is involved in such differences. Another is the level of abstraction involved in the variables usually used. More fundamental than this ongoing controversy, however, is the fact that there is no one personality structure: personality structures change with the variables studied and even with the instruments used to measure any one set of variables. To get dependable, replicable findings, the observer's idiosyncratic contributions must be separated from his judgments. This can be done by concentrating on simple, short actions. Alternatively, how people make attributive judgments can be studied as well as the evaluative judgments that people must make about others, in carrying out the activities of everyday living. (Author)