The efficiency of timing models of cessation of growth of northern woody plants was defined as their ability to allow maximum utilization of the growing season without autumn frost damage. The efficiencies of three timing models were compared with the aid of computer simulations, using long-term meteorological data. A model based on night length was the most efficient, followed by one based on a joint factor of the night length and temperature sum. A model based on temperature sum alone was the least efficient. The reason for these differences in efficiency was the great year-to-year variation in the annual temperature sum accumulation. The results were compared with empirical evidence for each of the models. The relevance of the theoretical approach is discussed, as well as the application of the simulations.