Heterogeneous aquifers are nearly impossible to characterize by conventional drilling and sampling alone, and aquifer heterogeneity introduces scale effects that cannot be identified using borehole data. Surface geophysical measurements and volumetric tomography can provide the large-scale sampling of aquifer properties needed to address the scale effect in heterogeneous aquifers. This study considers the statistical and uniqueness of inversion issues inherent in the integration of surface geophysics or tomographs with well logs and aquifer test data to generate a scale-dependent model of aquifer hydraulics. The nonuniqueness of geophysical inversion is divided into a global component related to model ambiguity, and a spatial component related to the scale of investigation. Formal statistical analysis of the degree of confidence associated with a given correlation between model predictions and data indicates that statistical significance of the interpretation is closely related to the degrees of freedom in the model. Data integration often increases the degrees of freedom in the interpretation model over those used to invert surface geophysical soundings, influencing the level of confidence in the interpretation. These considerations demonstrate that a realistically limited number of boreholes will usually provide a marginal statistical sample of aquifer-averaged properties, where it is difficult to relate a limited sample of local aquifer texture to volume-averaged aquifer properties. The dependence of statistical significance on both sampling statistics and model structure is illustrated by a synthetic resistivity sounding example, and by a series of field studies in various ground water environments. The results from these studies show that refinement of interpretation models as more data is included in the analysis often biases statistical tests. The most effective data integration results when 1) boreholes are drilled in arrays designed so that scale-dependent correlations indicated by large-scale geophysical surveys can be verified; and 2) measurements are made in an iterative program of multiple phases of drilling, data collection, and analysis so that interpretation techniques can be upgraded to remain consistent with an evolving interpretation model.