The rapid progress in computer technology in recent years has enabled the development of increasingly complex simulators, which can handle large amounts of data. It is often assumed that this automatically leads to more accurate static and dynamic reservoir models. In reality, however, there is still much evidence that the predicted performance of a reservoir often differs vastly from the actual production behaviour. These deviations are an indication of the failure to understand the processes involved and to recognize the uncertainty inherent in the definition of important reservoir characteristics. In this paper, a classification scheme is proposed, in which uncertainty is expressed as fuzziness, incompleteness and randomness. Each of these elements is described in detail and illustrated within the context of reservoir appraisal, although the approach can be applied to the wider aspects of petroleum geoscience. It is believed that adopting this classification scheme will enable the geoscientist to build a more extensive picture of uncertainty in reservoir appraisal. It will also be invaluable as a tool with which to inform management of the existing uncertainty, using a consistent language, thus providing guidance in the decision-making process.