The difficulty may be met by concentrating the whole of the coarse gold in a small fraction of the ore, by sifting and making a separate assay of this fraction. A portion of the ore, of about 1000 grams, is ground to a very fine powder and passed through an 80 sieve, re-grinding when necessary, until only 20 or 30 grams is left of the coarser powder. This is mixed with fluxes and carried through as a separate assay. The sifted portion is _thoroughly mixed_, and a portion of it, say 30 or 50 grams, taken for assay. The weights of the two portions must be known, and care must be taken that nothing is lost in the powdering. The method of calculating the mean result from the two assays is shown on page 109. In this way of working there is no advantage in continuing the grinding until the coarser fraction is reduced to a gram or so--rather the contrary; and rubbing on until all the gold is sent through the sieve is to be distinctly avoided. The student must bear in mind that what he is aiming at is the exclusion of all coarse gold from the portion of ore of which he is going to take only a fraction. The question of the smaller sampling of gold ores has been dwelt on at considerable length, as befits its importance, in order that the student may be impressed with a sense of its true meaning. Sampling is not a mystery, nor does the art lie in any subtle manner of division. It is, of course, absolutely necessary that the stuff to be sampled shall be well mixed, and the fractions taken, so that each part of the little heap shall contribute its share to the sample. Moreover, it must be remembered that tossing about is a poor sort of mixing, and that everything tending to separate the large from the small, the light from the heavy, or the soft from the hard (as happens in sifting), must be avoided, or, if unavoidable, must be remedied by subsequent mixing. With a well-taken sample, we may rely on a great majority of our results falling within normal limits of error; but nothing can be more certain than that, in a moderately large experience we shall get, now and again, deviations much more considerable. These erratic assays can only be met by the method of working duplicates, which call attention to the fault by discordant results. Such faulty assays should be repeated in duplicate, so that we may rest the decision on three out of four determinations.