Time domain reflectometry (TDR) offers a unique opportunity to measure liquid water content &thgr;L in frozen soil, since the permittivity of ice is much lower than that of water. However, calibrations of TDR derived from drying unfrozen soil, where water is replaced by air, may not apply to a freezing soil, where water is replaced by ice, since the permittivity of ice is greater than that of air. We designed a gas dilatometer to calibrate TDR for &thgr;L in frozen soil. A soil sample is hermetically sealed in the gas dilatometer; subsequent soil freezing reduces total air space and hence increases pressure inside the gas dilatometer, since ice is less dense than water. The amounts of soil water frozen were computed from measured pressure change as temperature was incrementally decreased. TDR calibrations for two samples of the same soil at different total water contents had identical slopes but different intercepts, supporting our hypothesis that there exists no unique calibration of TDR for &thgr;L for frozen soil, but rather a family of calibration curves, each curve corresponding to a different total water content. ¿ American Geophysical Union 1995