The success of civil engineering construction in cold regions often depends on the assessment of the potential for frost heave. Over the past two decades two models have been presented to the engineering community to identify the frost heave potential, namely: (1) the segregation potential (SP), and (2) the discrete ice lens (DIL) models. These models are based on ice segregation parameters which can be obtained from simple laboratory freezing tests. However, the reliability of the ice segregation parameters obtained from the laboratory freezing tests is still in question. In recognition of the need to improve the reliability of the measurement of ice segregation parameters in a frost heave test, a study was conducted to: (1) investigate several improvements in the step freezing test used to measure ice segregation parameters, (2) examine several approximations used to interpret the laboratory step freezing test results, (3) demonstrate the compatibility of the SP and DIL parameters, (4) propose a method to obtain reliable ice segregation parameters, and (5) demonstrate the applicability of the DIL model to address field frost heave problems. The results of the study indicated: (1) the accuracy of measured ice segregation parameters may be improved if step freezing tests are conducted under shallower temperature gradients and the water intake velocity is accurately and continuously monitored, (2) inappropriate approximations may produce a significant error in the measurement of ice segregation parameters, (3) SP may be defined either by the ratios of the (i) heave velocity/1.09 to the overall temperature gradient of the frozen soil, or (ii) water intake velocity to the temperature gradient of the frozen fringe, (4) the SP and DIL ice segregation parameters are compatible, and (5) the reliability of both models is improved through the demonstration of the compatibility of the ice segregation parameters. The application of the frost heave prediction models was studied based on ice segregation parameters determined with the improved laboratory procedure. A practical method to predict the influence of the groundwater level using the DIL ice segregation parameters was proposed. A probabilistic approach for the DIL frost heave model was developed. / Graduation date: 1994