270 
MESSRS. THOMAS MARTIN LOWRY AND PERCY CORLETT AUSTIN 
solution volumes (Bousfield and Lowry, ‘ Phil. Trans.,’ A, 1905, vol. 204, p. 283) 
the errors in the measurement of specific-rotatory power increase with dilution, so that 
the course of the curves becomes uncertain as the proportion of water increases towards 
100 per cent. It is, however, quite evident that the curve of errors, fig. 3 b, for the 
linear formula intersects the zero-line at three points between e = 0-45 and e—\. 
Any attempt to flatten this curve, by the addition of a third term containing e 2 , is 
obviously doomed to failure, since the curvature from the straight line is in opposite 
directions at the two ends, and cannot be covered by a section of a parabola. The type of 
formula used by Winther is therefore only of value for interpolation over a narrow range, 
and is even worse than a simple linear formula for extrapolation. This is true not only 
for dilute solutions, where the errors immediately become larger, because the curvature 
