A prototype method is described which automatically analyzes exponential decay curves by using nonlinear regression analysis and deviation pattern recognition. Lifetimes and preexponential factors are computed for the proper model. The program successfully classified simulated data consisting of one, two, or three exponentials when noise followed a Poisson distribution. However, for real luminescence decay data, systematic instrumental noise of higher frequency tended to obscure low frequency deviations from incorrect models in smoothed deviation plots. This factor complicated classification of multi-exponential data, although the program easily distinguished between single and multiple exponential behavior.