[en] If we are to believe our predictions of the thermomechanical behavior of the material surrounding a nuclear waste repository in granite, we must test the computational methods used in making the predictions. If thermal loadings appropriate to a real repository are used, thermally induced displacements and strains are quite small, and available geotechnical instrumentation is only marginally able to measure these effects to the accuracy desired to make thorough tests of the predictions. We outline a three-step program to address these issues. (1) Conduct experiments in which the thermal loading is large compared to that induced by a real repository. This will permit us to make accurate measurements with available instrumentation. (2) Simultaneously, develop improved instrumentation that will enable us to make accurate measurements of motions induced by thermal loadings appropriate to a real repository. (3) Finally, conduct a second set of experiments, with the improved instrumentation and thermal loading similar to that of a real repository in granite. If we can predict the effects of this thermal loading to a few percent over distances of tens of meters for time periods of a few years, and demonstrate that these predictions are correct, we can have reasonable confidence that, using the same methods, we can predict the behavior over thousands of meters for hundreds of years to an order of magnitude. That accuracy should be satisfactory for those distances and times