Various methods have been at different times proposed for employing the test of reduction. In the ruder periods of analytic chemistry we find Hahnemann recommending a retort as the fittest instrument, and stating ten grains as the least quantity he could detect.[507] Afterwards Dr. Black substituted a small glass tube, coated with clay and heated in a choffer; and in this way he could discover a single grain.[508] In a paper published in the Edinburgh Medical and Surgical Journal, I showed how to detect a sixteenth of a grain; and afterwards even so minute a quantity as a hundreth part of a grain.[509] The process is performed in a glass tube; which, when the quantity of the oxide is very small, should not exceed an eighth of an inch in diameter, and may be conveniently used of the form first recommended by Berzelius, and represented in Fig. 3.—The best material for reducing the oxide is recently ignited charcoal, if the quantity of suspected substance be very small. For when any of the ordinary alkaline fluxes is used, more than half of the arsenic is retained, probably in the form of an arseniuret of the alkaline metalloid. But when the quantity of matter for analysis is considerable, charcoal is inconvenient, as it is apt to be projected up the tube on the application of heat; and an alkaline flux is on that account preferable. For this purpose soda-flux,—made by grinding crystals of carbonate of soda with an eighth of their weight of charcoal, and then heating the mixture gradually to redness, so as to drive off all water,—is better than the more familiar black flux, which contains carbonate of potash; because the latter attracts much moisture when kept for some time.—If the quantity operated on is large it should be mixed with the flux before being introduced into the tube; if it is small, it may be dropped into the tube and covered with charcoal. The materials are to be introduced along a little triangular gutter of stiff paper, if the tube is large; but with a small tube it is preferable to use the little glass funnel represented in Fig. 2, to which a wire is previously fitted, for pushing the matter down when it adheres. The material should not be closely impacted. Heat is best applied with the spirit-lamp, first to the upper part of the material, with a small flame, and then to the bottom of the tube, the flame being previously enlarged. A little water, disengaged in the first instance, should be removed with a roll of filtering paper, before a sufficient heat is applied to sublime the metal. As soon as the dark crust begins to form, the tube should be held steady in the same part of the flame. With these precautions a well defined crust will be procured with facility.