12354        13245        13254
21345        21354        21435

But exception might be taken to two even of these, namely, those that appear in the third column, where 5 is found in juxtaposition with 2 in the first case, and 4 with 1 in the second.  So great a difference between two adjacent weights would be almost sure to attract the notice of the person who was being tested, and make him dissatisfied with the arrangement.  Considering all this, together with the convenience of carriage and manipulation, I prefer to use trays, each containing only three weights, the trials being made three or four times in succession.  In each trial there are three possibilities and only one success, therefore in three trials the probabilities against uniform success are as 27 to 1, and in four trials at 81 to 1.

Values of the Weights.—­After preparatory trials, I adopted 1000 grains as the value of W and 1020 as that of R, but I am now inclined to think that 1010 would have been better.  I made the weights by filling blank cartridges with shot, wool, and wads, so as to distribute the weight equally, and I closed the cartridges with a wad, turning the edges over it with the instrument well known to sportsmen.  I wrote the corresponding value of the index of R on the wad by which each of them was closed, to serve as a register number.  Thus the cartridge whose weight was WR4 was marked 4’.  The values were so selected that there should be as few varieties as possible.  There are thirty weights in all, but only ten varieties, whose Register Numbers are respectively 0, 1, 2, 3, 3-1/2, 4-1/2, 5, 6, 7, 9, 12.  The reason of this limitation of varieties was to enable the weights to be interchanged whenever there became reason to suspect that the eye had begun to recognise the appearance of any one of them, and that the judgment might be influenced by that recognition, and cease to be wholly guided by the sense of weight.

We are so accustomed to deal with concurrent impressions that it is exceedingly difficult, even with the best intention of good faith, to ignore the influence of any corroborative impression that may be present.  It is therefore right to take precautions against this possible cause of inaccuracy.  The most perfect way would be to drop the weights, each in a little bag or sheath of light material, so that the operatee could not see the weights, while the ratio between the weights would not be sensibly changed by the additional weight of the bags.  I keep little bags for this purpose, inside the box that holds the weights.

Arrangement of the Weights.—­The weights are placed in sets of threes, each set in a separate shallow tray, and the trays lie in two rows in a box.  Each tray bears the register-marks of each of the weights it contains.  It is also marked boldly with a Roman numeral showing the difference between the register-marks of the adjacent weights.  This difference indicates the grade of sensitivity that the weights in the tray are designed to test.  Thus the tray containing the weights WR0, WR3, WR6 is marked as in Fig. 1, and that which contains WR2, WR7, WR12 is marked as in Fig. 2.