Using previous years’ yield patterns may be one method of breaking a field into management zones for the purpose of site-specific management. For this method to be useful there must be temporal stability of yield patterns and there must be a sound method of assessing the spatial-temporal stability of yield in a field. To this end, a method was developed to give a non-biased estimate of the within-field spatial-temporal stability of yield. The method determined the probability that the normalized yield for all years available at a given location in the field fit within the accuracy limits of the combine. Combine accuracies of ± 2.5%, 5% and 10% of the field mean yield and gridded data of 3 m, 6 m, and 9 m cell sizes, as well as crop choice were all included in the model to assess its sensitivity to changes in these factors. The resulting spatial-temporal stability maps were well correlated with visual estimations of the spatial yield patterns. The model results were highly influenced by the inputted combine accuracy, but grid size and crop choice had little affect on the proportion of the field or the spatial pattern of temporal stability in the two field sites examined. The sensitivity of the model to changes in the input value for the combine accuracy indicated that a good estimate of this value is required for the determination of the stable patterns in a field. Key words: Spatial-temporal stability, yield