Modeling water and energy movement in salt-affected soils requires knowledge of heat, salt, and water interaction. Heat flow in soil may be determined from a knowledge of the thermal conductivity and temperature gradient. The effects of water and soluble salts on the apparent thermal conductivities (lambda) of two Ca-saturated soils were examined and compared with estimates of the effects from the De Vries model. Apparent thermal conductivities were determined as the values giving the best fit of measured temperatures at the axes of soil cores immersed in a water bath with temperatures calculated from a finite difference solution to the heat flow equation. Samples of Weswood clay loam (fine-silty, mixed, thermic Fluventic Ustochrept) and Padina loamy sand (loamy, siliceous, thermic Grossarenic Paleustalf) were prepared to produce a range of soil solution contents with either CaCl2, MgCl2, NaCl, or Na2SO4. Concentrations of the solutions applied to the soil samples ranged from 0.1 mol kg-1 to solubility limits. At any given soil solution content, the apparent thermal conductivity of the soils decreased with increased salt concentration in solution. Compared with the soils wet with salt-free water, thermal conductivity decreased by as much as a 20% when the concentration of salt solution was increased by 1 mol kg-1. When solid shape factors (g(s)) determined from soil wet with salt-free water were used in the De Vries model, values of lambda as a function of salt concentration were underestimated. most or the underestimation was from the initial decrease in conductivity with salt concentrations < 2 mol kg-1, suggesting the differences between the calculated and measured values were due primarily to microstructural changes from the addition of the salts and not to change in the thermal conductivities of the soil solution and soil air. For the loamy sand, the reduction of thermal conductivity could be matched with the De Vries model by using values of g(s) calculated from the salt-affected soils. For the clay loam, mathematical limits on the value of g(s) prevented matching calculated and measured values of lambda.