When in use the platinum wire is adjusted so that the end in the tube indicates the temperature at which the room is to be kept, and the small gas jets are lighted. Suppose the temperature of the room to be below that required, the mercury column will not be in contact with the platinum wire, and the circuit will be open. In this condition the valve will be open, gas flowing and burners burning, which will continue until the temperature rises and closes the circuit by contact between the mercury and platinum; this attracts the armature and closes the valve. This continues until the temperature falls enough to break the circuit, when the same action is repeated.

This is found to work in a very satisfactory manner during the six colder months of the year; the temperature of the room is kept within a range of two degrees, and not unfrequently for forty-eight hours within one degree. By covering the circulating coils with large sheets of close wrapping paper, folded in the middle and dropped over the pipes, leaving the bottom open, the range of temperature is decreased about one-half.

The table on which the chronometers are placed is circular, and stands in the middle of the room, with the control and automatic thermometers in its centre. Each chronometer has a separate compartment, large enough to receive the chronometer and leave a space of an inch or more between its case and the walls of the compartment, each compartment being fitted with a separate lid. Holes are bored in the bottoms of the compartments to allow a free circulation of air, and the lids are left open except at the time of comparing, when they are all closed except that of the chronometer under comparison. The object of this is to permit the comparisons by ear to be made more easily.

A hygrometer is used for testing the moisture in the air, and is left in the room only long enough to determine the daily percentage.

All of the chronometers are compared daily between 11 and 11.40 A. M. with the mean time standard clock, and the errors and rates are worked up every seventh day, called term day; from these mean rates all calculations are made. Comparisons are made to the nearest quarter of a second. The temperature is closely observed each day at comparison and recorded for the previous twenty-four hours, by a chronometric thermometer (a chronometer not compensated for temperature), and by self-registering maximum and minimum ther

mometers.

All chronometers received, either new, or after having been cleaned

and repaired, are placed on trial for six months before their purchase, if new, or before their issue, if old ones. The trials should commence near the middle of summer or the middle of winter in each year, so that the natural temperatures of both the extremes can be used as well as artificial ones.

All chronometers on being placed on trial have the tops of their boxes removed for convenience in comparing, and in order to make them more sensitive to the surrounding temperature. They are examined to see that they fit properly in their gimbals, work perfectly free and without jar, and hang with their faces horizontal.

Some time during the cooler months of their trial they are placed in the temperature room for about fifty days, during which time. they are given two tests at three different temperatures, one set going from a lower to a higher, and the other from a higher to a lower temperature, always beginning with one extreme and ending with the same. By this means the effect of time on the rate is eliminated.

Any three temperatures between 45° and 90° Fahr. may be used, as between these points with the ordinary chronometer the changes of rate, owing to temperature alone, are proportional to the squares of the differences of temperature from the temperature of compensation, or fastest running. Fifty-five, seventy, and eighty-five degrees are good temperatures to use, as between these extremes are included all the temperatures through which chronometers will pass in ordinary navigation. These temperatures need not be equidistant, as is the case in Hartnup's method, but may be taken as most convenient.

Suppose we begin this test with 55° as the lowest temperature. The chronometers are placed in the room with the temperature at 52° or 53° for a day or two, when it is raised to 55°, and kept so for a term of seven days; it is then raised slowly to 70° and allowed to remain a day or two at that point before beginning the term at that temperature. After seven comparing days at 70°, the temperature is raised slowly to 85°, and allowed to stand a day or two, and then a seven day term is noted as before. The temperature is then raised to about 90°, allowed to stand a day or two, then lowered to 85°, and the same tests made again, only in reverse order.

Great care must be exercised, especially in lowering the temperature, to keep the hygrometric state of the air in the room about the same, and in no case to allow it to approach saturation. At each change of temperature a day or two is allowed the chronometers in