This study analyzed the harvesting behavior of perennial cash-crop growers in Sri Lanka. These growers face two alternatives with respect to harvesting; premature and mature harvesting. The objectives of the study were: to determine the reasons for premature harvesting, to explain this behavior with socio-economic factors as explanatory variables, and to describe the behavior with decision theory. The first objective was achieved by surveying a sample of 240 farmers. Fear of theft and immediate money needs were the reasons why most farmers harvested their crop at a premature stage. A logit probability model was used to explain this behavior. Education of the farmer, ratio of lowland to total land operated by the farmer, and the total family income were significantly related to harvesting behavior. Expected utility theory, expected profit maximization, and a lexicographic safety-first model were used to predict farmer behavior. The expected utility approach used the exponential utility function, the quadratic utility function, and the cubic utility function. The lexicographic safety-first model minimized the probability of regret as the first objective and maximized the expected income as the second objective, in that order. The expected utility model with the exponential utility function made the largest number of correct predictions followed by the expected profit maximizing model. The conclusions of this study, while providing more evidence of the poor predictive ability of the expected profit maximizing model, further supports the usefulness of expected utility theory in describing and predicting farmer behavior. A majority of the studies on farmer behavior have concentrated on resource allocation. This study has demonstrated that even harvesting behavior can be explained by expected utility theory. There was no consistency in the way the utility functions ranked the two harvesting alternatives; for certain farmers the ranking of one function reversed the ranking of other functions. This study has, thus, demonstrated the influence of utility functional forms on the ranking of prospects. The results were sensitive to changes in the discount rate and the results of the safety-first model were sensitive to changes in the expected income. / Ph. D.