Recent theoretical developments in cognitive psychology imply both a need and a possibility for methodological development. In particular, the theory of problem solving proposed by Allen Newell and Herbert A. Simon (1972) provides the rationale for a new empirical method for the processing of think-aloud protocols--trace analysis. A detailed example is presented in which trace analysis is applied to human performance on a spatial reasoning task. The study, a portion of a larger study, involved 12 subjects who were required to solve a variety of spatial arrangement problems under different conditions. The protocol produced by a single subject--a 30-year-old female psychology student--while solving the block problems was selected for analysis on the basis of completeness and interest. The relations between trace analysis, on the one hand, and the psychometric ideas of measurement and standardization, on the other, are discussed. A non-psychometric approach to standardized testing, referred to as "theory referenced test construction," is proposed. The main premise of the approach is that test items should be validated against computer-implemented information processing models of the relevant cognitive functions. Eight figures and a 64-item list of references are provided. A computer simulation program, a bibliography of Knowledge and Understanding in Human Learning reports, and a report distribution list are appended. (TJH)