The mean extinction law A(lambda)/A(V) over ultraviolet wavelengths, down to lambda = 0.12 mum, is known to be well described by a function of the optical quantity R [= A(V)/E(B - V)]. This paper investigates deviations from that mean law, with emphasis on differences among various sight tines through bright nebulosities and those through dense, dark clouds. We consider <delta(x)/delta(x1)>, the mean ratio of the deviation at wavenumber x = 1/lambda to that at x1. If <delta(x)/delta(x1)> greater-than-or-equal-to sigma(x, x1), the standard deviation of the ratio among the sight lines in the sample, the deviations and x and x1 are correlated. The deviations at x are correlated with those at the 2175 angstrom bump, x1 = 4.6 mum-1, only over the rather narrow range of wavenumbers 4 mum-1 < x < 5.0 mum-1. Deviations at large wavenumbers (the "FUV rise") are correlated over the whole range 5.25 mum-1 < x < 8 mum-1, where the upper limit is set by the lack of suitable observations. The form of the FUV rise has been determined by two separate procedures. Figure 8 gives the results. The deviations in the FUV rise for sight lines through dark clouds, normalized at 8 mum-1, differ substantially from deviations for sight lines through bright nebulae. The mixed or diffuse sight lines lie between the bright and dark nebulae. If the deviation of a particular sight line is known from observations, the correlations in Figure 8 can be used to estimate the deviations at other wavenumbers. Small silicates do not have enough extinction to explain the FUV rise for x < 6.5 mum-1. A mixture of uncoated small and large silicates can provide the FUV rise for various sight lines only if the extinction increases much more rapidly with x in the wavenumber region 7 mum-1 < x < 8 mum-1 than the optical constants by Draine & Lee suggest. The observations might be consistent with sight lines through H II regions having uncoated silicates, while silicates along other sight lines are coated enough to produce drastic extinction differences. Alternatively, polycyclic aromatic hydrocarbons might well have the steeply increasing extinction that is required by the deviations. The 2175 angstrom bump is narrower in the deviations than in extinction, meaning that the deviations tend to be confined to the. center of the bump. If small graphite particles are responsible for the bump, the deviations occur over a much narrower interval in wavenumber than standard optical constants for graphite suggest.
