308 



SCIENCE 



[N. S. Vol. LI. No. 1317 



in the three columns headed 0--E (Observed- 

 Einstein value that), relatively, the observed 

 right-ascension deflections depart more mark- 

 edly from the computed ones than do the ob- 

 served declinations-deflections. The observed 

 total deflections in every ease, except for star 

 11, exceed the Einstein values. 



nomical Expedition, at the lie of Principe, 

 west coast of Africa, where the weather condi- 

 tions were unfortunately not as favorable as 

 at Sobral, showed only a few stars and the 

 scale could not be directly determined as it 

 was not possible to remain at Principe the re- 

 quired time. Instead, plates of another region 



Comparison of Deflections of Light Bays Observed iy the British Astronomical Expedition at Soiral, 



Brazil, May £9, 1919, with Values Computed according to the Einstein Theory 



{Instruments: 4-meh lens of 19-foot focus aad S-incli coelost at. Observer: A. C. D. Crommelin) 



18. From the observational results in Table 

 I., the resulting value of the deflection, a„ at 

 the sun's limb, as published by Dr. Crommelin, 

 is l".98,i^ thus agreeing with the Einstein pre- 

 dicted value, 1".74, within 14 per cent. The 

 result from the astrographic plates taken by 

 the other British observer at Sobral, Mr. 0. 

 Davidson, using the astrographic object glass 

 of the Greenwich Observatory in conjunction 

 with a 16-inch ccelostat, was not so satisfactory, 

 the star-images being diffuse on account of a 

 probable change in figure of the coelostat mir- 

 ror; the discordance between the mean results 

 from the individual plates was said to be rather 

 large, but from the whole series an outward 

 deflection reduced to tlae limb, of 0".93, or 

 0".99, according to the method of treatment, 

 was found, with a probable error of about 

 0".3.i« 



19. The plates taken by Dr. A. S. Eddington 

 and Mr. Cottingham, the second British Astro- 



if See Nature, November 13, 1919, p. 281. The 

 probable error as given by Dr. Crommelin is 0".12, 

 whereas Dr. H. Spencer Jones, of the Greenwich 

 Observatory, in his summary {Science Progress, 

 January, 1920, p. 372) gives 0".06. 



of the sky taken at the same altitude were used 

 and compared with plates of the same region 

 and of the eclipse-field obtained previously at 

 Oxford. The determination of scale was there- 

 fore somewhat weak, though the uniformity of 

 temperature at Principe was in its favor. The 

 final result of the discussion of the plates gave 

 an outward deflection of 1".61 vnth a probable 

 error of 0".3.i^ 



20. Except then for the unsatisfactory So- 

 hral astrographic plates, the general conclusion 

 to he drawn is that deflections of light were oh- 

 served hy the British astronomers that agree 

 better with the Einstein law of gravitation 

 (Cause h) than with the N ewton-Maxwell law 

 (Cause a). This is well shown by Fig. 2, con- 

 structed by the Department of Terrestrial 

 Magnetism, giving a graphical representation 

 of the law of variation with distance followed 

 by the observed deflections for each star, as 

 well as by the computed ones on the basis of 

 causes a and 6. It is seen at once that, except- 

 ing the most distant star (56 Tauri), each star 

 shows a deflection agreeing better with the 

 Einstein value than with the Newton-MaxweU 



17 See reference to Dr. Jones's article in previous 

 footnote. 



