190 CARNEGIE INSTITUTION OF WASHINGTON. 



The signs of the first two quantities N — D were erroneously inter- 

 changed in the last report. 



The above corrections were applied, leaving residuals consisting 

 of accidental errors and effects of barometer and thermometer upon 

 the clock-rate. Consequently an attempt was made to test the 

 dependence of the rate upon the mass of the air. Unfortunately the 

 ■Riefler clock employed in San Luis was not sealed, nor were any 

 records of temperature kept in the clock-room, which was subject 

 to considerable changes in temperature. As the changes in the 

 clock-room and in the observing-room would necessarily differ, the 

 attempt to directly measure the dependence of the clock-rate upon 

 the mass of the air proved a failure. 



It therefore became necessary to devise some other means for the 

 elimination of such a term. It is natural to suppose that the diurnal 

 variation of temperature, acting upon a more or less constant air- 

 pressure, would in the mean result introduce a diurnal variation in 

 the clock-rate dependent upon the mean time. Under such an 

 assumption we would also expect that in the long run the corrections 

 due to such a factor would be approximately equal and of opposite 

 sign 12 hours apart, and consequently the methods employed in the 

 preliminary determination of clock correction should eliminate this 

 term. That is, by the method employed we should expect to derive 

 clock corrections free from the effects dependent upon right ascen- 

 sion (Aa„) and free from the diurnal term, but containing the term 

 night minus day N — D, the constant error dependent upon the 

 illumination of the sky. For the first approximation the latter term 

 was neglected. 



Appljdng the resulting clock corrections, the residual error would 

 include the terms Aa„, N — D, and the diurnal correction. Solving 

 f or Aa„ and N — D, there would remain accidental errors and the diurnal 

 term. As the effect N — D is generally considered to be a constant 

 term, but as in reality the change in the illumination of the sky is grad- 

 ual, there should be a small residual term dependent upon this effect. 



Consequently, after applying the corrections already evaluated, 

 the residuals were collected according to mean solar time, hourly 

 means taken, and curves drawn. It was considered advisable to draw 

 off the results according to seasons, both for the purpose of testing 

 any variations in the diurnal rate due to seasonal changes and to 

 demonstrate that the residuals thus arranged bear no resemblance to 

 the Aa„ term, which was supposed to have been completely taken out. 

 In the following table the diurnal term as determined by seasons 

 bears no apparent trace of a residual Aa^ term. The variation is 

 similar in each case and strictly according to mean time. There 

 also seems to be a certain degree of similarity between the spring 

 and fall terms, and between the summer and winter terms. Some 



