Abstract
Theories of visual sampling have been founded on the assumption that human monitors use an internal representation of the environment, which changes over time in accordance with the person's expectations. Experimental evidence supporting this view is equivocal since it is based on sampling performance averaged over long periods of a test session. Two experiments are reported that use the instrumental observing response as an index of sampling. Subjects were set to discover critical events presented to three information sources, and these events were preceded by varying amounts of extra information, which affected event predictability. A procedure for data analysis was developed to evaluate the relation between sampling behaviour and the moment-to-moment changes in the states of the sources. Results supported the view that sampling was dependent on a constantly updating internal representation of the environment.