A study was made of the problem of differential thermal analysis utilizing statistical experimental design. A factorial type of design was chosen to yield maximum efficiency of the data obtained. “Lakeland”, Florida Kaolin of commercial whiteware grade was selected because of its relative purity. Due to the large number of samples to be analyzed, it was found necessary to choose a commercial brand rather than a very pure kaolinite which was found impossible to obtain in sufficient quantity. A statistical 2x3x3 factorial design was used and the calculations incident thereto made in detail. Data, which to the eye appeared erratic, was found to be acceptable within the probability limits of error of the experiment. The effects of two different furnaces, three rates of heating, and three densities of the samples were analyzed to determine the area under the endothermic reaction peaks as measured by Cr-Al differential thermocouples. It was found that densities and rates both yielded highly significant values and that the interactions of the two showed significant results when examined by a distribution of “F”¹. These effects were analyzed for the regressions therein and were found to be significant in the linear trends only. From this indication it would seem evident that the optimum density and rate of heating had not been reached in the limits of this experiment. Replications and furnaces did not yield significant results when tested by “F” distribution. From this it was deduced that the experiment could be conducted with sufficient precision under the present procedure and conditions and reproducibility obtained irrespective of the furnace used. Examination of the data will indicate that the level of the two furnaces is different, but the trends within are in the same areas. A certain amount of erratic data were obtained due to a lag in heating rates and this difficulty will be discussed. / Master of Science