( 437 ) 



The \arialioiis L L. LH, Ll, L f . Lu and L ti! ai'c cxpi'cssod in 

 socoikIs »>t' liiii(\ llio oilier (»n('s in seconds of arc. and llic dcri- 

 \'ali\'Os arc formed accord in<iiy. Xaliii'allv llie lernis in LH, L I. 

 L f' , L ((. L ('.' . L <S and L(\' now can oidv relate lo conslani eri'ors 

 in diese elenuMils. The L .T and L U of the separate (Mpiations ri^'o- 

 rously e.\|iress the \ariations of the |»aralla.\ \alnes and senddianieters 

 for the actnal distances, hnl iniuhl appi'itxiniately Ix^ considered as 

 those of .T„ and A^,, the |»aralla\-constant and \\\v seniidianietei' for 

 distance 1 '). 



We now |.iercei\"e that the inHnence of an ei'ror in the lalitnde 

 is very sliuhi and thai this is also the case with constant errors in 

 the declination of the nnxni or the stars. An e(|nino.\-ei-i'or common 

 to tlie moon and the stars is not eiitirelv eliminated, hnt its inllnence 

 is immaterial and my [)ei'sonal error in transit ol)S(M'\ations, which 

 iidlneiices H, f and /'. is eliminated in so far as the error is the 

 same foi- the moon and the stars. The only terms which \v(> have 

 lo consider fnrther ai'c therefore those dependiiiLi' on L rr and L I*. 



If we accept for tlu^ two linibs as mean error for wei.ulit J =b 5\()5 

 and d= 4^S4 respectively. a> it may l)e that indeed the observations 

 of the second limli are a little more accnral(\ then we have: 



\ L L — O.St) L .T„ - 2.00 Z. //„ = ]>.0<S ± 1^2.S 

 11 L L ^ 1.72 L .1^, - 2.0:5 L //„ = + 0-.70 ± 1^4(; 



The mean of the two resnils withont reuard to their weights is: 



Z. A 4- 0.4:} L .T„ -]- 0.02 L //„ = — 0\ill ± 0\tl7 



and the resnltin.U' lon^uitnde fonnd in this \\ay, a|)arl from the fact 

 that it is not wholly free from jiersonal error, is alfecled oidy Uy 

 the ujic(M'lainty in the constant of the parallax. If instead of 11 axskn's 

 constant we adopt that of Xkwcomü ( Asti-onoiiticul ( 'mishiiils p. 1*.).')) 

 \\ hicli is lariicr l»y 0."4. oni' re-ult hecomes: 



L Lr= ^- U^;](; ± 0^{I7 



Hence we see that the inllnence of the nncertainty in the conslani 

 of the jiarallax is probably not <>reat. 



As my linal resnlt 1 act-ept L L z= — 0\:). antl hence: 



[>()N(ilTri)K OF TFIK OHSKKVATIOM'IKK 



') If tlio coefticionfs (if A tt,, and ^ i?„ ai'e (ielcnuiiicil lijioniiislv v-.c liiid : 

 foi' jiinl, 1 4-0. si:', and +1.'.>3. for lirni) 11 - I .r.(i and — l.!>r». or piaclicaliy 

 the -same values as fonncily {added 1904 Jan). 



