The observatories of Greenwich and Cambridge despatched expeditions to test this prediction at the eclipse of 1919. It was found that the stars which appeared near to the sun at the instant of eclipse showed an appreciable displacement, as com pared with their normal positions, of the type required by Ein stein's theory. Exact measurement confirmed that the displace ment varied approximately as the inverse distance from the sun, and that the displacement at the limb was sensibly equal to Ein stein's predicted value of 1.745". The Cambridge observers, hampered by cloudy weather, obtained for this quantity the value 1.61" but with a probable error of 0.30". The Greenwich observ ers obtained a value of 1.98" with a probable error of 0•12", but Prof. H. N. Russell subsequently pointed out that their photo graphs contained intrinsic evidence of a horizontal and vertical scale difference of the order of one part in 12,000 almost cer tainly due to a distortion of the coelostat mirror under the sun's rays, and if the measures are corrected for this the result is brought much closer to the theoretical prediction. Three years later an expedition, sent out by Lick observatory to observe the 1922 eclipse, was favoured by good weather and obtained for the displacement at the sun's limb, a value of 1-72" with a prob able error of 0-I I". None of the expeditions had of course meas ured the deflections of stars actually at the sun's limb ; most of the stars were several diameters away from the limb, the observed deflections being corrected so as to bring them to the limb. The deflections of stars at all distances were found to agree well with the predictions of Einstein's theory.
cise, the Fraunhofer lines in the solar spectrum ought to show a displacement to the red ; this displacement ought to be homolo gous, and should be of amount o-oo8 A units at the cyanogen band X 3883 at which observations have been chiefly made. Early attempts to test this prediction led to strangely discordant results. All observers found some effect of the kind predicted, but its amount was generally substantially less than the predicted amount, varying from almost nil (St. John, 1917) to nearly the full amount to be expected (Evershed, 1918 and 1923, Grebe and Bachem, 1919). The latest measurements of the displacement are those of Evershed (1927). Measuring the mean displacement shown by the lines of 29 prominences, Evershed found that the H and K lines of calcium showed the same mean displacement of 0.009 A, as compared with 0.0408 A required by theory.
The want of success of the earlier observations is explained by the circumstance that the relativity shift of o-oo8 A is liable to be masked by larger shifts of uncertain amount arising from other causes. Larger shifts would be expected in stars which were of larger mass than the sun or of smaller radii. Many known stars have masses far larger than the sun, but they also have larger radii, and so ought to show only about the same spectral shift as the sun. On the other hand the class of stars known as "white dwarfs" have masses comparable with that of the sun, but far smaller radii. For instance, calculation shows that the companion to Sirius, a star whose radius is only about of that of our sun, ought to show a shift of as much as 0.30 A. On putting the matter to the test at Mount Wilson observatory, W. S. Adams found an actual shift of 0-32 A. It is hardly possible any longer to doubt that the spectral shift predicted by Einstein really exists.