June 11, 1920] 



SCIENCE 



583 



The deflection by this law should every- 

 where be the same on a circle concentric with 

 the sun, i. e., condition (1) the deflection 

 should vary alone with the inverse distance, 

 not also, for example, with heliographic lati- 

 tude; firrthermore, the deflection should be 

 strictly radial, i. e., condition (2) the deflec- 

 tion should coincide in direction with a line 

 drawn to the center of the sun. Plotting 

 Crommelin's actually observed deflections, for 

 each of the 7 stars, in magnitude and direc- 

 tion, as was done in Fig. 1 by the Department 

 of Terrestrial Magnetism, a careful examina- 

 tion shows that there are systematic depar- 

 tures from both conditions (1) and (2) which 

 apparently can not be explained wholly by 

 errors of observation.- In addition we have 

 the fact that the resulting value of a,,. No. 3 

 in Table I., is 1".98 instead of 1".74, or about 



axis; the north end of this axis for Sobral at 

 mid-totality was about 16°. 8 east of the north 

 end of the sun's axis of rotation. The two 

 columns giving the probable errors, as deduced 

 by us from the individual data derived from 

 Crommelin's 7 plates, show that the average 

 probable error for both the radial and non- 

 radial components is about + 0".04:. The 

 angular departure, (3, it will be observed, 

 varies from — 28° to + 37° ; a plus value 

 means an angular departure in the positive 

 direction of the angle A, i. e., in the direction 

 N., E., S., W. The sign of a^ corresponds 

 with that of p. How many of the 7 plates 

 gave a plus or a minus fi is shown in the last 

 two columns. It will be seen that for stars 

 6 and 10, the minus sign greatly predominates, 

 and for stars 2 and 11, the plus sign greatly;- 

 predominates. 



Madial and Non-Badial Component s of Observed Light Deflections at Sohral, Brazil, May 29, and Angu- 

 lar Departures for Badiality 

 Based on results from 7 photographic plates obtained by Dr. A. C. D. Crommelin with a 4-inch lens 

 of 19-foot focus and using an 8-inch coelostat. 



14 per cent, larger than the theoretical value. 

 What was the chief cause of the superposed 

 effects ? 



4. Table 2 contains our resolved components 

 of the observed light deflections, namely, the 

 strictly radial component, an and ap, the com- 

 ponent i>erpendicular to the radius, represent- 

 ing the non-radial effects or angular depar- 

 tures, /?, from radial ity, exhibited in Fig. 1. 

 A is the positon angle of the star counted 

 continuously in the direction IST, E., S., W., 

 from the north end of the declination or geo- 



2 Dr. Silberstein has also directed attention to 

 the existence of the non-radial effects. Monthly 

 Notices, E. A. S., Vol. 80, pp. 111-112. 



5. Table II. shows the following facts : 

 (a) The observed radial component is 

 greater than the Einstein theoretical value 

 for the first five stars (ISTos. 3, 2, 4, 5 and 6) 

 and less for the two most distant stars (ISTos. 

 10 and 11). (The observed radial deflections 

 for the two stars, Nos. 6^ and 11, which depart 

 most from the Einstein values, correspond, 

 respectively, to deflections at the sun's limb of 



3 Curiously, Eddington 's observed deflection for 

 star 6, according to data kindly supplied recently, 

 also departs most markedly from the Einstein law; 

 in his case, however, the deflection reduced to the 

 sun 'a limb is about 55 per cent, too low for that 

 star. 



