Twenty-four closely spaced samples of beach sand were collected on a grid pattern and subjected to heavy mineral analysis. Each sample was sieved to obtain two grades, from which the heavies were separated in bromoform. Twelve splits of a composite sample made from quartered portions of the original samples were similarly treated. The data obtained were used to evaluate the sampling and laboratory errors of heavy mineral analysis, with the relation E 2 = (e 1 ) 2 + (e 2 ) 2 , where E is the total observed error, e 1 is the sampling error, and e 2 is the laboratory error. The results showed that the average sampling error for the beach sand is about 10 per cent, but there is a particular sampling error associated with each mineral in the suite. The laboratory error of splitting, separating, and counting is of approximately the same order of magnitude as the sampling error. The individual observations show a fair agreement with the general theory of errors. An evaluation of current routines of heavy mineral analysis is made in the light of the data obtained.