The rotation time of dissolved anthracene is used as a measure of the viscosity on a nanometer-sized object, and this "nanoviscosity" is compared to the ordinary macroscopic viscosity of polymers. Measurements in poly(dimethylsiloxane) (PDMS) as a function of the chain length extend from a small oligomer (M = 162 g/mol) to the entangled polymer (M = 139 000 g/mol). The results are compared to similar ones in poly(isobutylene) (PIB) [J. Phys. Chem. B 2002,106, 7385]. Despite many similarities in the static structures of PDMS and PIB, qualitative differences are found in the behavior of the nanoviscosity. The nanoviscosity in the infinite polymer limit is 15 000 times lower in PDMS than in PIB. The difference arises in the oligomeric region, where the nano- and macroviscosities diverge at a much shorter chain length in PDMS than in PIB. Because the barriers between torsional conformations are much lower in PDMS than in PIB, these findings are consistent with the hypothesis that the divergence of macro- and nanoviscosities is dependent on a dynamic correlation length involving torsional relaxation. This divergence represents a change from the Stokes-Einstein-Debye behavior characteristic of small-molecule solvents to the Rouse behavior characteristic of polymers. Literature data on n-alkanes are reexamined in this context. The alkanes are an intermediate case between PDMS and PIB, both in terms of torsional barrier height and in terms of the chain length where macro- and nanoviscosities diverge.