This paper presents a comprehensive laboratory study that examines the effects of porosity, water content, density and grain size distribution on the thermal conductivity of soils which were sampled from 16 synoptic stations of Korea. The experimental results clearly demonstrate that porosity and water content are important parameters which strongly affect the thermal conductivity of soils. Soils with lower porosities and higher water contents have higher thermal conductivities. On the contrary, increase of the matrix density slightly increases the thermal conductivity, and grain size distribution hardly affects the thermal conductivity. Dry soils with the same porosity tend to have more scattered values of thermal conductivity than wet soils. Based on the experimental results, a multiple linear regression model and a nonlinear regression model, having two regression variables of porosity and water content, were presented to predict thermal conductivity. Both models show a high accuracy of prediction with $R^2$ values of 0.74 and 0.82, respectively. Thus, it is expected that the suggested empirical models can be used for predicting thermal conductivity of soils by measuring porosity and water content.