An improved Jones sample splitter is described which is faster to operate than the conventional type, yet requires no special skill or training. Samples obtained with this instrument are free from the occasional erratic samples obtained with ordinary Jones sample splitters. Working drawings are given for this improved sample splitter, which can be built with hand tools. The sample splitter was tested with small gravel, granules, and sand in which the percentage of each constituent was accurately known. Three series of 24 samples each were taken and sieved; in each series one detail of technique was varied. The components separated by sieving were returned to the original mixture in each test to keep the conditions of the experiment nearly constant. Summarized instruction are given for the best technique of operation. The second part of the paper deals with methods used to measure the performance of the improved sample splitter. The concept of a state of control is discussed. The calculations required to detect lack of constancy in the chance cause system, biased percentages, or the presence of a predominating cause of variation are shown. The principle that variability in comparable series of samples tends to diminish as the state of control is approached was found to apply in the present study. These methods of statistical analysis are very general and may be applied to field sampling studies or wherever quantitative measurements are made on a series of samples suspected of coming from an essentially homogeneous unit or constant chance cause system.