We have measured the phase decoherence rate, τ deviates from theory with a flatter T -dependence. We speculate that this latter behavior is due to incomplete screening of the s=2 Fe impurities by the conduction electrons.PACS numbers: 72.15. Qm, 73.20.Fz The Kondo problem is a paradigm many-body problem in condensed matter physics. In recent years, Kondo physics has been observed in semiconducting quantum dots [1] and carbon nanotubes, while Kondo lattices play a major role in some strongly-correlated materials [2]. The original problem that motivated Kondo's 1964 paper [3] was the increase in resistivity at low temperature of metals containing magnetic impurities. Although Kondo solved this puzzle, his perturbation theory diverged in the limit of zero temperature. Wilson [4] showed that, at temperatures far below the Kondo temperature T K , the magnetic impurity forms a spin-singlet with the surrounding conduction electrons and behaves as a nonmagnetic scatterer with cross section given by the unitarity limit.The brief history given above might leave the impression that the the behavior of dilute magnetic impurities in metals is completely understood. That is, however, not the case. An important aspect of the Kondo problem concerns the distinction between elastic and inelastic scattering. This distinction is extremely important in the context of quantum transport and mesoscopic physics, where it is known since 1979 that elastic scattering from static disorder preserves quantum phase coherence, whereas inelastic scattering destroys it. The magnetic impurity contribution to the conduction electron phase coherence rate, τ −1 φ , was first measured explicitly by two groups in 1987 [5,6], and has received renewed attention recently [7,8] in the context of the debate over zero-temperature decoherence in disordered metals [9,10]. Until very recently [11,12], however, there was no theoretical expression for the temperature dependence of the inelastic scattering rate valid for T not too far below T K , and very little reliable data in that temperature range [13].The most reliable estimates of τ −1 φ come from analysis of low-field magnetoresistance data in the context of weak-localization theory. In disordered metals without magnetic impurities, τ −1 φ is dominated by electronphonon scattering at temperatures above about 1 K, and by electron-electron scattering at lower T . In the presence of magnetic impurities, τ −1 φ contains the additional contribution γ m , which peaks at T ≈ T K . In order to observe this peak in γ m while keeping the magnetic impurity concentration low enough to avoid interactions between impurities, one must choose a system with T K below about 10 K; otherwise τ −1 φ is dominated by electronphonon scattering. In order to acquire data far below T K , however, it is important to keep T K as high as possible. The optimal range for T K is a few Kelvins, which is achieved with Fe impurities in Ag [14].We fabricated Ag wires of dimensions L = 780 µm, w = 0.1 − 0.2 µm and t = 47 nm on oxidized Si...