Pannetier et al. respond.

Pannetier et al Respond: The statement in our original Letter that there is "no theory for a quantitative analysis of the Al'tshuler, Aronov, and Spivak (AAS) effect in an extended network" means simply a lack of an analytical expression for the magnetoresistance in order to fit the data. Such an expression was derived some time ago for the hollow-cylinder problem and has been used by different authors. Now a general theory for the magnetoresistance in an arbitrary geometry is available. For instance, in the limit kFl » 1, which is actually the relevant one for the experiments, a compact expression for AT? (//) has been obtained for the square network and the fit of data is now possible. In this respect, the Comment of Browne et al does not provide any quantitative prediction and the argument used is only a qualitative one. Actually, the experiments are performed in the so-called weaklocalization regime and then Ref. 2 provides an answer to the question raised by Browne et al: The periodicity of transport coefficients, with a period 0 = o/2, in a regular network is really the counterpart of the hollow-cylinder behavior. A similar situation on superconducting networks tells us the same phenomenology. However, the origin of the effect is very different from the Cooper pairing in superconductors and was attributed to the interference effect resulting from the enhancement of backscattering in the weak-localization regime. The recent experiment done by Webb et al on a single ring shows the presence of oscillations with both periods 0/2 and 4>0. Therefore, the question of self-aver aging in cylinders and arrays and the lack of self-averaging in single rings call for further investigations. At present there is no theory of this self-averaging and the ensembleaveraged theory used in Refs. 2, 3, and 5 deserves further studies. Note that in perturbative calculations, it is always the average conductance that is being calculated, which is not the same as the typical conductance. In general, fluctuations from sample to sample can therefore arise and this depends on the experimental setup.