Makch 22, 1907] 



SCIENCE 



451 



days, with those computed by means of the 

 above formula. I find during the year 

 1900 the mean of these difference is 0^.028 

 and the largest difference is 0^071. 



About 1867 F. Tiede installed at the 

 Berlin Observatory a weight-driven clock 

 enclosed in an air-tight case. The original 

 escapement was replaced in 1876 by a 

 gravity escapement and the clock has con- 

 tinued to give satisfaction, certainly up 

 until 1902, when it was dismounted for 

 cleaning. The only published rates that I 

 have been able to secure are those during 

 twelve weeks in 1877-8. During this 

 period the average deviation of the ob- 

 served daily rates, the average interval for 

 each rate being six days, from the mean 

 daily rate for the entire period is 0=.030. 



In connection with the publication of 

 these rates^ the statement is made that for 

 weeks at a time this mean deviation will 

 not exceed 0^02. 



DAILY BATE OP BIEFLEE SIDEBEAL CLOCK KO. 70. 



being communicated to the pendulum 

 through the suspension spring. This clock 

 was enclosed in an air-tight glass case and 

 was mounted in a vault where the tem- 

 perature was artificially controlled. The 

 definitive rates have been determined from 

 September, 1903, to May, 1904, but, un- 

 fortunately, during this entire period we 

 were unable to prevent the glass case leak- 

 ing and there was a variation of tempera- 

 ture in the vault of about 5° C. However, 

 from the first of February to the middle 

 of May the pressure was kept nearly con- 

 stant by reducing it each day by means 

 of an air pump by from 1 to 2 millimeters. 

 During this period the following results 

 were obtained: 



The formula from which the computed 

 rate is obtained. 



Daily rate = + 0".0161 — O'.OOIOS (T-Mar. 29.0) 

 — 0'.0456(t-27°.0) 



is the result of a least square solution of 

 the twenty observed rates. 



Collecting together the results given 

 above, we have: 



MEAN DEVIATION OF DAILY CLOCK BATE. 



Mean ±0.015 



In 1903 there was installed at the Naval 

 Observatory one of Riefler's clocks, No. 70, 

 with a nickel-steel pendulum, the impulse 



• A. N. Nr. 2184. 



In considering this table I hope the par- 

 ticular method by which each of the num- 

 bers given was obtained will be remem- 

 bered. The reason for using different 

 methods has already been given. If the 

 rates of the first three clocks had been 

 treated in a manner similar to that used 

 in the last three cases, the numbers given 

 might have been slightly smaller, and if 

 the period under consideration with the 

 Berlin and Washington clocks had been 



