584 



SCIENCE 



[N. S. Vol. LI. No. 1328 



2".43 and l".ll, thus exhibiting a range of 75 

 per cent.) 



(b) The observed non-radial component, 

 which according to the Einstein law (1) 

 should be zero, varies from — 0".l7 to 

 + 0".16 ; it amounts at times to one tenth of 

 the Einstein radial deflection at the sun's 

 limb and is from 3 to 7 times the probable 

 error. 



(c) The value of the deflection at the sun's 

 limb as deduced from stars Nos. 3, 4, 6 and 

 10, near the sun's axis, is 2".02, and from 

 stars Nos. 2, 5 and 11, near the sun's equator, 

 1".88; the two values differ 0".14 or 8 per 

 cent. (The observed deflection therefore is a 

 function not simply of distance alone, as re- 

 quired by the Einstein law, but also appar- 

 ently of the position angle.) 



6. After various trials the following pre- 

 liminary formulae were found to represent the 

 observed quantities with good approximation :^ 



„, = 1!IZ + 5!?? sin^ {A - 239°) , (2) 



P P 



dp = 0''0323p sin 2 (A - 233°). (3) 



The close agreement in the independently- 

 derived phase angles, 239° and 233°, led to 

 the impression that some common cause pro- 

 duced the superposed radial effect, represented 

 by the second term in (2), and the non-radial 

 effects represented by (3). Now the position 

 angle of the zenith for Sobral at mid-totality 

 of the eclipse, projected on the plate, is 241°. 6, 

 which value could be substituted with fair 

 approximation in place of the phase angles 

 for (2) and (3). Thus the second term of 

 (2) and the single term of (3) were found to 



■1 The sum of the residuals squared on the basis of 

 formula (1) was 0.093, whereas on the basis of 

 (2) the sum was reduced to 0.037. "Were the non- 

 radial effeots regarded solely as errors of observa- 

 tion, then the sum of the squares amounts to 0.106; 

 however, the sum of the squares of the residuals 

 resulting by applying formula (3) is but 0.016. 

 Other formulas were also established giving a still 

 closer representation of the observed quantities 

 than do (2) and (3), however, they did not admit 

 of physical interpretation as readily as those given. 

 This matter will be discussed more fully in the 

 complete paper. 



be related in some manner to the local zenith. 

 The effect of terrestrial atmospheric refrac- 

 tion on the sun and the stars is to shift them 

 apparently all towards the zenith, those 

 farthest from the zenith being shifted most. 

 The question accordingly arises whether the 

 superposed effects with which we are con- 

 cerned may not have resulted from incom- 

 plete elimination of differential refraction 

 effects in the earth's atmosphere. It may be 

 observed also that by the introduction of our 

 second term in (2), the value of the deflection 

 at the sun's limb was reduced from 1".98 to 

 1".77, which agrees closely with the Einstein 

 value. 



7. With the effective aid of my colleague, 

 Mr. W. J. Peters, in charge of the reduction 

 of the atmospheric-refraction cybservations 

 made aboard the Carnegie, the iwssible out- 

 standing effects resulting from incomplete 

 elimination of differential refraction effects 

 in the earth's atmosphere have been investi- 

 gated. The differential terrestrial refraction 

 effects between the sun and each of the 7 

 stars were rigorously computed by two differ- 

 ent methods for the time of exposure of the 

 eclipse plates and the prevalent meteorological 

 conditions. Lacking complete details regard- 

 ing the precise times of exposures of the 

 check-plates obtained before sunrise between 

 July 12 and 18, our computed differential- 

 refraction effects for the check-plates are for 

 the present only tentative ones. The exami- 

 nation as far as it can be made at present 

 indicates that outstanding effects in the 

 differences between the differential terres- 

 trial refraction effects for the eclipse-plates 

 and the check-plates, may largely, if not com- 

 pletely, account for the non-radial effects in 

 the observed light deflections, as also decrease 

 the value (1".98) of the radial deflection at 

 the sun's limb. This is a matter that can 

 be more definitely determined when the 

 original data and complete details regarding 

 the reductions of the measures are available. 

 The present indications are that precise allow- 

 ance for differential terrestrial refraction 

 effects may bring Crommelin's results into 

 closer accord with the Einstein law of gravi- 



