48 



nately ten seconds a day. Though at the end of a week 

 the better clock would be a minute more in error than the 

 second one. The point I wish to illustrate is that with an 

 accurate time-piece we can always predict what its error 

 will be, for some days in advance. While with a poor 

 time-piece we can form no idea from the determination of 

 its error on two nights, what it will be on a third one. 

 Now it is to the elimination of the sources of error in 

 clocks and watches that the attention of the artisan is 

 directed ; and the practical form which successive im- 

 provements take is in more perfectly protecting our time- 

 pieces from the effects of temperature changes, and from 

 those resulting from variations of friction in the move- 

 ment. In the clock we endeavor to guard against the 

 effects of temperature on the pendulum, by uniting two 

 metals in such a way that one expands upwards while the 

 other expands downwards, and they are so adjusted that 

 the centre of the pendulum stays very nearly in the same 

 position. Now, although the clock is the most perfect 

 time-piece we have, yet it is still liable to the theoretical 

 objection that its pendulum swings in a circular instead of 

 a cycloidal arc. You will be interested to know, that the 

 finest clocks for astronomers' uses are so sensitive to ex- 

 ternal influences that if the barometer were to change an 

 inch in height, it would cause a variation in the clock's 

 rate of about a quarter of a second per day, and I might 

 mention that in some large observatories the standard 

 clock is kept in a cellar vault to avoid changes of temper- 

 ature and in an hermetically sealed glass case, from which 

 the air has been partially exhausted. At the Harvard 

 Observatory the clock which distributes the signals to 

 Boston and alons: the lines of the railroads and conse- 

 quently to Salem as well, is placed in the cellar inside of 

 a thick walled room which has a floor of sheet lead, its 



