A temporal database environment, particularly if equipped with a log of the execution of predefined application-domain operations, can be regarded as a repository of narratives concerning the activities observed in the mini-world of interest. The analysis of these narratives leads to the construction of a library of typical plans, which can be used by Plan-recognition / Plan-generation algorithms to help prediction, decision-making and the adoption of corrective measures with respect to ongoing activities. The formulation of concepts and methods for identifying the typical plans, and, more generally, for discovering meaningful narrative patterns, should be based on Cognitive Science and, especially, on the various contributions of Literary Theories to the interpretation of narratives. The proposed goaloriented approach emerges as a promising research line to help understanding, and hence better modelling and using, in their many dimensions, today’s computer-based information systems.