Abstract Parallel distributed processing (PDP, also known as artificial neural network) was introduced as an alternative for modeling regular noncatastrophic individual tree mortality. A two hidden-layered PDP system was created using back-propagation as the learning procedure and the sigmoid function as the transfer function. An empirical data set was used to test the performance of the system. Using the performance of a logistic regression as a benchmark, the new system had a better fit to the data than that of the logistic regression. It was also found that, though the system was not instructed to fit the data with logistic curves, the response surface of the model closely followed a logistic response surface. This finding suggests that, based on the goodness-of-fit measure employed, the best function to model individual tree mortality may indeed be the logistic function. A PDP system can be regarded as a procedure which attacks the problems of parameter estimation and model selection simultaneously. Topics regarding the potential use of a PDP system as an alternative to modeling individual tree mortality were also discussed. For. Sci. 37(3):871-885.