A nonlinear model compared mean and variance of reaction times in a field study of farm workers exposed to organophosphate pesticides. The model had two components, one to account for learning and one to account for fatigue. Subjects were 83 men who were identified with symptoms of organophosphate poisoning and cholinesterase inhibition and 90 unexposed controls. The subjects were assessed by clinical neurological examination, evaluation of nerve conduction velocities, measurement of vibrotactile thresholds, and computerized tests of balance, mood, finger tapping speed, visual attention and simple reaction times. For the learning component there was a significant education level times exposure interaction. The age times exposure interaction was significant for the fatigue rate. For the standard error of the estimate there were significant age times exposure and education level times exposure interactions. There was an effect of education level on mean reaction time that reflected an effect on the initial performance level. The definite poisoning cases showed an increase in the fatigue rate and in performance variability as age increased, and their initial reaction times were higher if their education level was lower. The authors conclude that the model permits more sensitive detection of untoward chemical effects on reaction time than is possible with other methods.