May 31, 1907] 



SCIENCE 



869 



improvement made in one hundred and fifty 

 years; but incidentally it brings out an ap- 

 parently strong contrast in the performance 

 of the clocks in use at Greenwich Observatory 

 and the Naval Observatory, Washington. 

 Clocks can novr be made to give, under ordi- 

 nary conditions, very fair and steady rates; 

 but if greater excellence is required, than 

 beyond the best possible mechanical construc- 

 tion, means more or less independent of the 

 clock, as such, have to be devised to obtain 

 it. For instance the Greenwich clock is kept 

 in a room in which the temperature does not 

 vary more than 5° F., and it has an auto- 

 matic device to correct for barometric error. 

 Thus its conditions are practically the same 

 as those of the Washington, although at- 

 tained by different means. The Washington 

 clock is kept in an air-tight case in a room 

 whose temperature varies say 5° C. A com- 

 parison then of the actual performances of 

 these two clocks is of great interest. Pro- 

 fessor Eichelberger's figures on page 451 fail 

 to do this, and for these reasons: 



1. The Greenwich Clock rates are spread 

 over a period of one year, while those of the 

 Washington clock only extend over three 

 selected months. 



2. The quantity ^.015 given in the table in 

 the second column on page 451 is not com- 

 parable with ^.051 given for the Greenwich 

 clock. 



^.051 is the mean deviation of the observed 

 rate. 



^.015 is the mean deviation of the observed 

 rate from one calculated from formulae, and 

 hence quite distinct from ^051. 



Professor Eiehelberger, while admitting 

 that his comparison is not valid, appears not 

 to realize that it is altogether vitiated by 

 dealing with periods of widely different devia- 

 tion, and the fact that ^.051 and ^015 repre- 

 sented two distinct phenomena seems to have 

 escaped his notice. 



It is, however, from the material he gives 

 a simple matter to make such a comparison: 

 the rates are taken from the table on page 451, 

 and the column ' mean deviation ' exhibits the 

 difference of the separate rates from the mean 

 rate + ^016. 



For the Greenwich clock the corresponding 

 period of 1904 is first given, and in addition 

 the same period of 1905. 



It will be seen that the difference is really 

 in favor of the Greenwich clock. 



