The free head on the spring was turned until the reading showed it to be a trifle too strong. It was then secured, and afterwards brought to the exact strength required by running it rapidly in a lathe and rubbing its surface over its entire length with fine emery cloth. This reduced the strength of each coil equally. This was a delicate operation, requiring great care to reduce the strength enough and not too much. A great many springs had to be made, several being generally required, often a full set of ten, with each indicator. This testing apparatus was convenient and reliable, and the workmen became very expert in its use. The spring when in use was always exposed to steam of atmospheric pressure. At this temperature of 212° we found by careful experiment that all the springs were weakened equally, namely, one pound in forty pounds. So the springs were made to show, when cold, 39 pounds instead of 40 pounds, and in this ratio for all strengths. This system of manufacture and testing was examined in operation by every engineer who ordered an indicator, the shop on St. Martin’s Lane being very convenient. They generally required that the indicator should be tested by the mercurial column. The Elliotts, being large makers of barometers, had plenty of pure mercury, so this requirement was readily complied with, and the springs were invariably found to be absolutely correct. We never used the mercurial column in manufacturing, but were glad to apply it for the satisfaction of customers. I employed the following test for friction. The indicator when finished was set on a firm bracket in the shop. The spring was pressed down as far as it could be, and then allowed to return to its position of rest very slowly, the motion at the end becoming almost insensible. Then a fine line was drawn with a sharp-pointed brass wire on metallic paper placed on the drum. The spring was then pulled up as far as possible and allowed to return to its position of rest in the same careful manner. The point must then absolutely retrace this line. No indicator was allowed to go out without satisfying this test. The workmanship was so excellent that they always did so as a matter of course. Mr. Henry R. Worthington once told me, long after, that on the test of an installation of his pump in Philadelphia, after he had indicated it at both steam and water ends, the examining board asked him to permit them to make a test with their own indicator, which they did the next day. They brought another indicator, of Elliott’s make like his own, but the number showed it to have been made some years later. “Would you believe it,” said he, “the diagrams were every one of them absolutely identical with my own!” I replied that the system of manufacture was such that this could not have been otherwise. [Illustration: Plan of Spring-testing Instrument.]