The thermal conductivity $$(\lambda )$$ of three unsaturated standard quartz sands (Ottawa C-109 and C-190, and Toyoura) was measured by a transient thermal-conductivity probe, at room temperature of approximately $$25\,^{\circ }\text{ C }$$ and at loose and tight compactions. The measurements were carried out at different degrees of saturation $$(S_\mathrm{r})$$ from dryness to full saturation. In general, a sharp $$\lambda $$ increase was observed at low $$S_\mathrm{r}$$ , followed by a moderate rise until full saturation. However, experiments on loosely compacted C-190 samples revealed $$\lambda $$ deviation from a general trend ( $$\lambda $$ vs $$S_\mathrm{r})$$ caused by water percolation. Alternatively, successful experiments were carried out on loosely packed unsaturated C-190 samples using 1 % agar gel. For loosely compacted C-109 and Toyoura, $$\lambda $$ data obtained from 1 % agar gel closely agreed with $$\lambda $$ data for water as a saturation medium. The measured data were used to verify a model by de Vries for unsaturated soils. The model largely underestimates experimental data at $$S_\mathrm{r}<0.5$$ and produces an overall root-mean-square error of about $$0.2\, \text{ W }~{\cdot }~\text{ m }^{-1}~{\cdot }~\text{ K }^{-1}$$ . Measured $$\lambda $$ data agreed with data by a steady-state technique (a guarded hot-plate apparatus) at dryness and full saturation and exceeded the steady-state data in the unsaturated region. However, TCP data can be considered more reliable due to a lower temperature increase during $$\lambda $$ measurements and a shorter testing time. Consequently, in the case of unsaturated soils, evaporation and migration of water and steam can be avoided.