The feasibility of using empirical state space models such as canonical variates analysis (CVA) models for the scale-up and scale-down of direct methanol fuel cells (DMFC) is explored. The development of such models to predict voltage responses of DMFCs has been previously investigated with a small-scale single cell and a three-cell stack assembly. Each system was subjected to a range of varying dynamic loads and operating conditions. A model was developed for each of the two systems for varying cell operating conditions, and these models were then validated against the other "unseen" data sets collected from the two cells. The models developed were found to give acceptable inferential estimates and one-step-ahead predictions. One outcome of that study was the possibility of using such models to aid cell scale-up and scale-down. This paper addresses the feasibility of using a dynamic state space canonical variates analysis (CVA) model of the small-scale single cell to predict the voltage responses of the scaled up three-cell stack system. In addition the feasibility of using the three-cell system model to predict the dynamic behaviour of the small system is also studied (scale-down). The results achieved are encouraging and indicate the potential of using a CVA state space representations as an aid to cell scale-up and scale-down.