Research into the innovation process requires models of the process which accurately describe, and can reliably be applied to, real examples of innovation in organizations. In this paper two models are compared empirically in terms of reliability and accuracy. The first (Zaltman, Duncan & Holbek, 1973) follows the traditional approach of depicting the process in a series of discrete stages. The second (Schroeder, Van de Ven, Scudder & Polley, 1989) shows the process to be more fluid, without a fixed sequence of stages. The models are applied to case histories of seven innovations observed in a longitudinal study of a hospital ward. Inter-rater reliability is higher for Schroeder's model, while overall support across raters tends to be stronger for the individual stages of Zaltman's model. However, there are problems with the sequence of stages in the latter model. Future research should aim to develop empirically-grounded models based on a wide range of different innovations.