AbstractThis paper examines the way in which reality is abstracted in the preparation and design of a game or simulation model. Two modes and four dimensions of abstraction are identified. Of particular importance is the construal dimension, which depicts the subjective content of a model as perceived by a participant. It is argued that anything more than trivial abstraction in this dimension alters the meaning of models to such an extent that human players will not be able properly to understand them. This limits the types of games with human players that can validly be played for research or education to those with relatively specific rather than general results. The argument is developed in terms of a spectrum of result generality for game and simulation models.