OF CLOCK RATES UPON LONGITUDE WORK. 133 



The latitude from this partial summation of our fundamental 

 observations is ^o^37° 20' 25".6. This was the value derived from 

 the first effective meridian circle work here. Previous to the begin- 

 ning of that work, in 1893, the adopted latitude of the instrument, as 

 furnished by the U. S. Coast and Geodetic Survey, was a full second 

 of arc smaller. This difference could be due to the errors of declina- 

 tion of the stars employed in the earlier determination, and to the lack 

 of corrections for the periodic variation of latitude. 



As to possible sources of systematic error in our zenith distance 

 observations, there might be a sensible difference between nadir read- 

 ings during daytime and night. 



This would probably not appear as a periodic term, however. All 

 accidental errors, even those of circle readings, are larger in the day- 

 time. 



It is more difficult to concede the probability of a shift in the 

 zenith point, due to a variation in the refraction at the zenith. Espe- 

 cially is this less probable during the night hours, when the atmos- 

 pheric conditions are most stable. 



Whether our observed periodic variation is in the clock rates, or 

 in some term affecting the position of the meridian, the determination 

 of the difference of longitude between two widely separated stations 

 will show the effect if the phase is the same for both stations. 



The usual procedure for an exchange of longitude signals is tO' 

 observe the same list of stars at each station, in order to eliminate 

 errors of right ascension. 



If the stations differ 90° in longitude, the rate of one clock will 

 be carried forward six hours to the epoch of observation at the other 

 station. 



If the established daily rate of the clock be used, the accumulated 

 error would amount to 0^.03 in this interval. 



Every exchange of signals would have this error, but if signals 

 are sent each six hours during one day the errors would occur in 

 pairs, two successive plus errors being followed by two minus errors. 



The double amplitude of o^o6 would occur in two consecutive 

 exchanges, twelve hours apart. 



If two stations differ 180° in longitude, exchange of signals every 

 six hours would give differences alternately of o^o6 and zero. 



