in 



L. A. Batter — Gravity Determinations at Sea. 17 



ment is shown to be fallacious. A source of error also not 

 considered is that due to possible imperfections of the vapor 

 tension tables. 



3. Insufficient evidence has been given to prove that, in the 

 reduction of the observations, it is best to omit those made on 

 board vessels at anchor. A method of adjustment which 

 must assume practically what is to be proved, and which 

 necessitates the rejection of data secured under supposedly the 

 best conditions, weakening thereby the connecting link between 

 the ocean results and the shore pendulum stations, can hardly 

 be regarded as the best possible one. Instead some logical 

 method of observation and of adjustment must be striven for, 

 which will take advantage to the fullest possible extent of the 

 shore and harbor results. 



4. The problem of obtaining sufficiently reliable ocean grav- 

 ity results still awaits solution. 



Method to be Tried on the "Carnegie.'''' 

 The method it is proposed to try on the " Carnegie," begin- 

 ning, if possible, at Cape Town in about April of 1911, is 

 practically the same as that employed in the magnetic work. 

 At all ports visited there will be both shore and harbor 

 observations, especially at those places where g has been 

 observed with pendulums and where accordingly the anomaly 

 Ag is known, thus permitting a logical determination of purely 

 instrumental constants. Our provisional equation of condition 

 for such stations will be of the following form, j3 having the 

 same significance as in equation (1) above : 



^"ill = * + «(*-*«,) + » (■»-*„)■ (10) 



k—JCt + 7c\ = constant part (Jc t ) of the relation between the 

 zero of the thermometer and the zero of the barometer plus 

 the constant part Jc\ of the errors of the vapor tension tables. 

 It is hoped also by zero point determinations of the thermom- 

 eters and by comparisons of barometers with port stand- 

 ards wherever there are such, to determine Tc x independently 

 of k\ and thus gradually get some idea of the various errors. 



a (t — t ) is to represent the change in instrumental constants 

 with elapsed time from some mean epoch, t a ; it may later be 

 found necessary to introduce a quadratic term, a' {t—t^f, but it 

 is believed that, with proper care of instruments and with 

 sufficiently frequent zero determinations of the thermometers, 

 this term may be avoided. 



b(jB—fi ) is to take account of the variations not included 

 in the time term, but dependent upon barometric height or upon 



Am. Jour. Sci.— Fourth Seeies, Vol. XXXI, No. 181. — January, 1911. 



2 



