This report documents the theory, structure, and implementation of a performance processor (written in FORTRAN IV) that can accept performance demonstration data representing various levels of operator's skill and, under user control, analyze data to provide candidate performance measures and validation test results. The processor accepts two types of information: (1) sample performance data on magnetic tape, and (2) user information reflecting knowledge about features of the performance that are considered to be important to measurement. The sample performance data input is smoothed by the processor in order to remove or reduce noise factors in accordance with information provided by the user. Criterion performance functions are, optionally, provided by the user or are computed by the processor using skilled performers' data. The processor then develops a discrete representation of the continuous performance data based on observed deviations from the criterion functions. This discrete representation, in turn, is used to model each performance using state-space techniques. The processor operates on the state-space model to compute vectors which form generators of various conceivable measure spaces. Candidate performance measures are then generated by operating on the vectors with multiple regression algorithms. Empirical validation tests of several types are applied to the candidate measures for assessment of their validity-likelihood. (Author/DEP)