Two experiments examined how listeners weight binaural information of individual components in a multicomponent complex when they are instructed to use this information in various ways. On each experimental trial of the first experiment, a two-component complex consisting of a 553- and 753-Hz pure tone was presented to listeners over headphones. Each component had an interaural difference of time (IDT) that was randomly chosen from a single distribution. Listeners were to indicate the apparent position of the intracranial image relative to the midline on the basis of either the IDT of the 753-Hz component, the sum of the IDTs of the two components, or the difference between the two IDTs. Observer weights for the component IDTs were derived by computing the point-biserial correlations between the IDTs of the components and the listeners' left-right judgments. Three of the four listeners were found to adjust their weighting of the binaural information appropriately for each listening task, while the fourth listener consistently gave almost no weight to the 553-Hz component regardless of the listening condition. In a second experiment, listeners were instructed to attend to the randomly selected IDT of one of three components (553, 753, or 953 Hz) and to indicate whether the intracranial image produced by that component was to the left or right of midline. Only one of six listeners gave greatest weight to the target component regardless of its frequency. The other five listeners gave significant weight to all three components in all listening conditions. In all cases, individual differences in percent correct performance could be attributed largely to individual differences in weights with little variation in internal noise.