Just as these experiments were concluded in 1882, M. Magnus Nyrén, of St. Petersburg, published an elaborate investigation of the small annular displacements of the stars due to the successive transmission of light, involving an increase of Struve's "constant of aberration" from 20·445" to 20·492". And from the new value, combined with Newcomb's light-velocity, was derived a valuable approximation to the sun's distance, concluded at 92,905,021 miles (parallax = 8·794"). Yet it is not quite certain that Nyrén's correction was an improvement. A differential method of determining the amount of aberration, struck out by M. Loewy of Paris,[792] avoids most of the objections to the absolute method previously in vogue; and the upshot of its application in 1891 was to show that Struve's constant might better be retained than altered, Loewy's of 20·447" varying from it only to an insignificant extent. Professor Hall had, moreover, deduced nearly the same value (20·454") from the Washington observations since 1862, of Alpha Lyræ (Vega); whence, in conjunction with Newcomb's rate of light transmission, he arrived at a solar parallax of 8·81".[793] Inverting the process, Sir David Gill in 1897 derived the constant from the parallax. If the earth's orbit have a mean radius, as found by him, of 92,874,000 miles, then, he calculated, the aberration of light--Newcomb's measures of its velocity being supposed exact--amounts to 20.467". This figure can need very slight correction. Professor Harkness surveyed in 1891,[794] from an eclectic point of view, the general situation as regarded the sun's parallax. Convinced that no single method deserved an exclusive preference, he reached a plausible result through the combination, on the principle of least squares--that is, by the mathematical rules of probability--of all the various quantities upon which the great datum depends. It thus summed up and harmonised the whole of the multifarious evidence bearing upon the point, and, as modified in 1894,[795] falls very satisfactorily into line with the Cape determination. We may, then, at least provisionally, accept 92,870,000 miles as the length of our measuring-rod for space. Nor do we hazard much in fixing 100,000 miles as the outside limit of its future correction. FOOTNOTES: [Footnote 748: Airy, _Month. Not._, vol. xvii., p. 210.] [Footnote 749: Mars comes into opposition once in about 780 days; but owing to the eccentricity of both orbits, his distance from the earth at those epochs varies from thirty-five to sixty-two million miles.] [Footnote 750: J. D. Cassini, _Hist. Abrégée de la Parallaxe du Soleil_, p. 122, 1772.]