tion. The ore in question was a gold-silver ore in which the gold, more especially, occurred in a notoriously "spotted" manner. The assays given are combined gold and silver results. The sample for the twenty-four hours was taken by hand from the battery feeder, a portion being caught every hour in a specially shaped scoop as it fell off the edge of the disc. The total weight was about 200 lb. and maximum size of particles from one to one and a half inches. In the course of an investigation into a continued and serious shortage of bullion as compared with the amount expected from a consideration of head and tail assays, a special series of tests was made on the method of quartering down the daily ore head sample.

Each day's sample for a number of days in succession was either mixed and quartered direct or the whole was crushed to pass 14-in. mesh before mixing and quartering. Duplicate samples were made at the first coning-up and quartered down and assayed separately, forming checks on the method of quartering. The results are summarized below.

The above indicates a possible error of about $14 a ton when the original sample is not crushed fine before quartering, while in the other case the difference shown is only $2.00.

It may be asked, in regard to "A," why the error shown should be supposed to account for a uniformly high result in the long run, instead of neutralizing itself over a long period of time, and it should be noted in reply that the source of the error is not likely to operate on the side of defect, or only to a slight extent, for this reason: since it is known that the ore is spotted in value and that a small stone may assay several hundreds of dollars right alongside of quite low-grade rock, let us suppose that a certain battery sample is fairly uniform but contains one stone of "high-grade" value; if, in coning and quartering the sample, that stone happens to be in the rejected quarter, the sample taken for assay will be representative, whereas if it happens to fall in the quarter saved the assay sample will be thrown up to the high limit of error shown. Thus, it is not so much a question between excess and defect (neutralizing one another) as it is between excess and mean, yielding an average in excess, over a period.

It is obvious that the greater the number of pieces into which that stone of "high-grade" is broken before the sample is quartered, the better the chance of its being distributed equally between the portion rejected and the portion saved, but on the other hand assuming the sample to have been broken sufficiently fine for safety at this stage, in continuing to quarter down this same sample the original danger appears again and increases in proportion as the sample gets smaller so that it will soon become necessary to subdivide the particles still further in order to obtain an even distribution of the high-grade ore throughout the mass.

A. Harvey (Mining and Scientific Press, Jan. 30, 1904) gives the following table as a guide to the margin of safety in size of ore pieces.

These methods of sampling the heads are necessarily expensive and as it is not always easy to point to any material gain in monthly profits resulting from their adoption, and as moreover, their use does not invariably result in the establishment of an accountant's balance between the bullion debited and credited to the mill, many companies have asked themselves whether they were not wasting time and money in trying to ascertain the true value of their ore heads and have finally decided to be content with a nominal value for the ore, obtained by adding the amount of bullion actually recovered to the aggregate amount of gold going to the dump in residues and dividing the sum of these by the tonnage for the month. The degree of satisfaction to be derived from this solution of the difficulty is a matter of individual opinion, and while admitting that it is not possible to ascertain head and tail assay values and tonnage with sufficient accuracy to furnish a check on small losses through carelessness or petty pilfering, yet the present writer confesses to a preference for making the best attempt possible to ascertain the real figures in each case, so as to act as a check against serious losses and afford an indication of any radical defects in the operation of the plant. Sampling of Residues. This is comparatively easy as the pulp at this stage is finely comminuted and usually homogeneous, so that only ordinary care and common sense should be needed to obtain a reliable sample.

In the case of sand the charge may be bored from top to bottom in a number of places after having leached dry preparatory to discharging, or if the residue be discharged into trucks each load

may be bored as it passes to the dump. In boring a tank charge care should be taken when using the ordinary split tube sampler not to take too much at a time, since if the rod be worked down from top to bottom without being withdrawn and emptied there is danger that the lower strata may escape sampling. It is better to drive down about 12 inches, then withdraw and empty the pipe, and then after carefully inserting in the hole without breaking down the sides, sample the next 12 inches and so on to the bottom. If on account of the method of filling or any other reason it seems probable that the charge is not homogeneous it is better to divide the area of the tank into concentric rings containing an equal tonnage and bore in each ring an equal number of holes so that the quantity of sand taken from each ring will be representative of the tonnage contained therein.

For slime residues the use of the automatic sampler is probably the most satisfactory method where the pulp is discharged in a homogeneous fluid condition.

The sample of filter cake is not so easy to obtain. One method is to cut a section of the cake from a leaf of each charge or in the case of the rotary filters to take a grab sample from the scraper periodically. It sometimes happens, however, that a sample dipped out of the discharge launder as the watery cake is flowing to the dam does not give an assay check on the sample cut from the cake before discharging, the latter being usually lower. The reason for this is probably that, assuming the sample taken from the launder to be a correct average, there may be with almost any type of filter a loss of a small amount of valuable solution which passes out with the cake and in that case the sample taken from the leaf does not represent the whole of the lost metal going to the dams.

Sampling Slime Pulp at Different Stages of the Treatment.-If this is to be of any value at all it is absolutely necessary that the action of the cyanide on the ore should be stopped immediately the sample is taken. This may be ensured in two ways: (1) A sample taken every hour, for instance, at a certain point in the treatment may be at once placed on a laboratory vacuum filter

and filtered and water washed, the several cakes composing the twenty-four hour sample being dried, mixed, and quartered down for assay as one sample; (2) or a method devised by H. A. White1 may be used. The sample as soon as taken is placed in an enamelled bucket and enough saturated solution of potassium permanganate added to maintain a permanent pink coloration in the solution during the whole period of sampling. This destroys the free cyanide and stops further action on the ore. At the end of the period of sampling, usually twenty-four hours, the contents of the bucket are well mixed, washed by decantation or filtration and sent for assay. The reaction between the permanganate and the cyanide takes place without precipitation of the gold but whether this is true of the silver content of the solution the writer has never determined as he has been accustomed to use the immediate filtration and washing method.

Discrepancy in Assay of Slime Residues. In the case of gold ores there is a possible source of error in the assays which will tend to show a higher extraction than has actually taken place, and therefore an apparent shortage of gold in the clean-up. W. A. Caldecott has shown that when a sample of slime containing potassium auro-cyanide solution as moisture is dried for assay it is almost impossible in the subsequent fusion to collect in the lead button all the gold present in this form probably on account of its being deposited in such minute particles as to approach the molecular state, there being therefore not enough globules of reduced lead to come in contact with every molecule of gold. It was found that the finer the litharge in the flux was ground the higher was the resulting assay in gold, but that under the most favorable conditions the gold content found was lower than that obtained by washing the pulp thoroughly and assaying the ore and washings separately. Whether the above reason be the correct one or whether the loss of gold be due, as is maintained by some assayers, to a deposition of the gold bearing solution on the walls

Journal Chem., Met. and Min. Soc. of S. A., Sept., 1911.

2 Proceedings of Chemical, Metall. & Mining Society of S. A., Vol. II, page 372.