The experiments of Israelachvili and Tabor are reanalysed to elucidate some of the quantitative discrepancies between theory and experiment. It is shown that, although the u.v. data for mica are incomplete, the postulation of extra u.v. absorption, or the use of interpolation techniques, in the representation of ε(iξ) is unable to explain completely the experimental data at all distances. However, a reconciliation of the theoretical and experimental results ensues, if it is assumed that the radius of curvature of the mica cylinders was underestimated by about 30 % in these experiments.