Recent theoretical efforts have provided strong evidence that the effective conductivity varies with distance between monitoring wells and a point source. In the present study, two artificial confined aquifers were subjected to steady state pumping tests. Analysis of these experimental results indicates dependence of the estimated hydraulic conductivity on distance from the pumping well and the form of the physical heterogeneity present. As the precision of the measured hydraulic heads within the experimental apparatus was limited to ~0.5 mm, the experimental results were reproduced numerically. Although some minor discrepancies were noted between the numerical and experimental results, the primary observations were consistent. In all cases the Thiem equation was used in combination with image well theory (to account for the boundary conditions) to analyze the response at 25 observation points (87 observation points used in the numerical experiments), providing a number of estimates of the hydraulic conductivity. Within each experiment it was observed that the variance of the conductivity estimate decreased with the distance between the two observation wells used in the calculation. Further, the mean conductivity varied with distance from the pumping well. In an experiment in which the medium was constructed to mimic a second-order stationary random field the mean conductivity approached, at large distance from the pumping well, the conductivity for a random, anisotropic medium under mean uniform flow. In an experiment in which the medium was constructed to mimic a structured (scaled) medium the mean conductivity decreased, at large distance from the pumping well, below the geometric mean and approached a value between the geometric and the harmonic means of the conductivity field. While the boundary conditions make it difficult to fully analyze the large-dimension behavior of these two media, it is clear from the results that the structured medium performs quite differently than the random field at large-measurement scale. ¿ 1998 American Geophysical Union