556 



SCIENCE 



[Vol. LVI, No. 1455 



This value is more than a second of arc smaller 

 than the one derived from our meridian circle 

 ohservations, beginning in 1893. Errors of 

 the old observations, including those of the 

 star declinations, the lack of corrections for 

 latitude variations, and the errors of the geodet- 

 ic triangulation probably account fully for this 

 difference. It appears not to have been quoted 

 as an illustration of geological change. Pos- 

 sibly it is so large as to be out of bounds. The 

 range of movements under discussion at present 

 is usually from one to five feet. There was a 

 notable earthquake in 1868, possibly, due to a 

 fault slip. It would, however, be difficult to 

 assign the proper proportions of a difference 

 of a second of arc to the errors of observation 

 and to the slipping, if the 1868 shock is to be 

 made responsible for its part. 



The meridian circle results include the epochs 

 of the earthquakes of 1895, 1903, 1906 and 

 1911. The first one of these was less intense 

 than the others. The observations at Ukiah 

 include the epochs of the last three. 



On request for our data, the results given in 

 the column Obsei-ved cpo below were originally 

 compiled in quarterly values, the observations 

 of each month having been combined into 

 means of the values of two or three months. 

 The table gives the number of successive quar- 

 terly values that make up the respective annual 

 values of cpo. In the annual ipo some of the 

 periodic errors due to declination have been 

 eliminated. There are some gaps in the se- 

 quence of observations, one from 1908 to 1912 

 due to my absence at San Luis, Argentina, for 

 the work of the Carnegie Southern Observa- 

 tory. 



The system of standard stars in use varied 

 from time to time, for special reasons of pro- 

 gram and international projects of observation, 

 and in the column of Corrected <fo the sys- 

 tematic corrections derived from the various 

 authorities have been applied as far as is pos- 

 sible. The reduction has thus been made to 

 the system of auwers, for whiofi a correction 

 of approximately +0.1" may still be required, 

 to give us the true astronomical latitude of our 

 instrument. The average residual of the last 

 column is two thirds the size of tha/t of the 

 column preceding. 



There are more stars in the catalog of 

 Newcomb, and rigorous systematic correc- 

 tions can not be derived for the extra stars. The 

 latitude results during the use of his system, in 

 the years 1901 to 1904, are nearly 0.1" higher 

 than those of the remaining years. . Some 

 additional results are included in the last 

 column. 



Treated as a consecutive series, from 1893 to 

 1921, there is no sensible continuous progres- 

 sive change of latitude from the mean, 

 37° 20' 25.6". To illustrate the effect of 

 abrupt changes at the epochs of earthquakes, 

 the yearly results have been combined in 

 groups, corresponding to each of three shocks. 

 The small differences thus exhibited are such as 

 would be expected, from the accidental errors 

 of the yearly means. The influence of the high 

 point in the series, where the declinations of 

 Newcomb were employed, is felt in the com- 

 binations. There will be an apparent progres- 

 sive rate up to that point and a drop there- 

 after which are probably fictitious. The mean 

 values of 90 preceding 1901 and following 

 1904 have excluded the effect of the systematic 

 difference due to Newcomb. 



The errors in determining the declinations of 

 stars are smaller than those involved in lati- 

 tudes, since some of the important systematic 

 errors can be eliminated. The probable error 

 of the annual 90 is evidently too large to permit 

 the detection of real changes that do not ex- 

 ceed 0.05". Differences of 0.1" would be an- 

 ticipated for half the comparisons between sep- 

 arate yearly means, according to the ";i\v of 

 distribution of errors, and three differences as 

 large as 0.2" might occur in a run of 18 dif- 

 ferences. 



The following computed latitudes for Ukiah 

 differ from the true values by a systematic 

 constant. The effect of the fourteen month 

 term is eliminated in the mean of any fourteen 

 consecutive monthly values. The effect of the 

 annual term is eliminated in twelve. By taking 

 means of fourteen consecutive values, be- 

 ginning each series with the first month of a 

 year, each mean below is subject to a correction 

 for two extra months of the annual term — a 

 small constant for all of them. No correc- 

 tions need be computed for either term in this 



