Adaptive management of waterfowl harvests accounts for uncertainties a
bout population responses to harvest, with a focus on the reduction of
uncertainties pursuant to harvest objectives and other management goa
ls. Important sources of uncertainty include limited knowledge about u
nderlying biological relationships (structural uncertainty), sampling
variation in population monitoring (partial observability), and uncont
rolled variation in the setting of harvest rates (partial controllabil
ity). We used a model for adaptive harvest management to investigate t
he use of harvests in reducing structural uncertainties. The model all
ows for both compensatory and additive relationships between harvest a
nd survival, and also includes predictors incorporating the compensato
ry and additive hypotheses, along with a procedure for updating predic
tor probabilities. The model was used to (i) characterize population c
hanges through time, and predict population changes based on simulated
data; (ii) compare predicted and observed population sizes in an effo
rt to identify the appropriate predictor for the population; and (iii)
examine the effect of harvest rate, monitoring variation, and partial
controllability on the rate of reduction in structural uncertainty. R
esults indicate that harvest can be used to learn about additive and c
ompensatory relationships, whichever is operative for a population. Wi
thin limits, learning can occur even with imprecise data about populat
ion status, and with substantial imprecision in the setting of harvest
rates. However, learning rates are depressed by high levels of monito
ring and/or harvest imprecision.