The variability of the pore-interaction factor, rho, for a microscopic model proposed for predicting unsaturated hydraulic conductivity, K(theta), from soil water-retention, psi(theta), data was examined in relation to soil particle-size data and indices, bulk density, organic C, parameters for the van Genuchten psi(theta) function, an index (W) of total energy of drainage and soil-series, toposequential, and geographic groupings. The rho exhibited no trend relationship to any of the soil properties tested. However, a change in the distribution of rho was observed in relation to the geometric-mean particle diameter (G(d)) and other soil textural variables. For G(d) < 0.08 mm, variability was much larger than for soils with G(d) greater-than-or-equal-to 0.08 mm. The latter consisted entirely of soils in the sand and loamy sand textural groups. A similar distributional relationship also occurred for rho vs. W. Examination of rho on three sites of the Hecla soil series and on related toposequential soils indicated that classification of rho on the basis of soil series or of soil-association groupings is a feasible strategy for rho parameter estimation on some soils. The exponential factor eta for a macroscopic model was also investigated. The eta was found to be related to W as an exponential function over the full data range. However, eta vs. W was nearly identical to a linear function for W data extending from 0 to 400 cm-gamma (where gamma is the density of water). The eta was also strongly related to G(d) as a power function, and to other textural variables as exponential functions. The increasing slope of the power function for G(d) < 0.08 mm indicated a large potential error of prediction for K(theta) on fine-textured soils.