irregularity which is characteristic of errors of observation. We should expect large errors in the observed densities, on account of the difficulty of obtaining the substance in a state of purity, and because the large value of the density renders it very sensitive to the effect of impurities which diminish the density,—also because the specific heat of the vapor is great, as shown by the numerator of the fraction in the second member of (13),^[1] and because the density varies very rapidly with the temperature as seen by the numbers in the third column of Table VIII.

But at the two lowest temperatures of Cahours' experiments, the differences of the observed and calculated densities (.381 and .568) are not only great, but exhibit, in connection with the adjacent numbers, a regularity which suggests a very different law from that of the