ABSTRACT
Cognitive psychologists are masters of hypothesis testing, but rank amateurs at discovering new hypotheses. If a group of capable cognitive psychologists were asked to devise a series of experiments relating to a topic with an established research history, then the group could quickly devise a sequence of well-designed studies. One discovery method is to gather observations of naturalistic behavior then analyze and categorize these observations to yield verifiable hypotheses. The Schank and Neisser studies just described represent rather unconstrained production tasks in which the subjects produce verbal answers to questions asked by the researcher. Production tasks in general seem appropriate for discovery research but it is often possible to constrain the types of responses and thus obtain a rich but more focused set of data than that obtained by these naturalistic studies. One type of production task that have used is an augmented clustering task.
