INTRODUCTION
The ultimate goal in rock mechanics is to make quantitative predictions of the response of rock to changes. Measured and calculated values of changes in displacements or temperatures in the rock for six different field experiments are compared using linear regression. When values of the mean compressive stresses are large compared with the values of the deviatorial stresses, linear elasticity, using values for Young's modulus from laboratory tests, provides a good predictive model. However, if the values of the mean compressive stresses are small compared with those of the deviatorial stresses the behavior and response of the rock, though systematic, is not linear. Linear heat conduction provides good predictions of changes in rock temperatures but thermal displacements are not predicted well by linear thermoelasticity.
The ultimate goal in rock mechanics is to develop an understanding of the behavior and properties of rock sufficient to enable quantitative predictions to be made concerning the response of rock masses to changes wrought by man or nature. The practical experiences and observations of civil and mining engineers and of geologists and geophysicists are of invaluable help in interpreting field observations and laboratory test results but are themselves insufficiently complete to arrive at a quantitative predictive model. Laboratory tests on relatively small samples of rock, made over wide ranges of stresses, temperatures and pore fluid pressures, have provided a wealth of qualitative information and quantitative data about the behavior and properties of rocks (Criggs and Handin, 1960; and Carter et al, 1981). In principle, the use of this information and these data in models based on sound principles should provide the predictive capabilities about rock masses that are required. Field experiments, in which some of the responses of rock masses to known changes are measured, provide data against which the validity of predictions using models can be checked. Unfortunately, it is seldom, if ever, practicable to make sufficient measurements in the field to define completely the response of a rock mass. Nevertheless, if the behavior and properties of rock are understood sufficiently well, measurements that are made of the response of rock masses to changes should correlate well with predictions based on a sound model. If the correlation between measurement and prediction is poor: the model is inapplicable or incomplete; the behavior and properties of the rock' masses are not understood sufficiently well; the changes have not been defined properly, or the measurements are not good. The disparities between measurement and prediction seldom provide sufficient information to resolve these uncertainties, because of the incomplete nature of field measurements. Field experiments are difficult, expensive and time consuming. The number of good field experiments is, therefore, small. Two quantities that can be measured with confidence in the field are changes in relative displacement between points and changes in temperature at specific points in a rock mass In this paper, measured and calculated changes in relative displacements and temperatures are compared, using linear regression, for six different field experiments done over the past two decades.