The percolation of water through snow is shown to be detectable with heat flux transducers (HFTs) at the base of, or embedded within, the snow. Theory indicates that the arrival of water at a HFT will result in a rapid drop in heat flow to negative values, followed by an exponential rise. Laboratory and field tests verify that the theory correctly predicts the form of the signal, and consideration of conditions in a snow cover suggest that it is unlikely the signal can be generated in any other way. Large (15¿15 cm) HFTs were installed at the snow/ground interface at three locations in Alaska where rain on snow is common. During three winters, heat flux signals indicative of water percolating to the base of the snow were detected 23 times, often simultaneously at two or more sites. All signals were coincident with rain-on-snow events, though not all rain-on-snow events (there were a total of 27) were associated with distinctive heat flux signals. Signals were observed only when rainfall rates exceeded 1.5 cm d-1, and the character of the signal differed if the snow/ground interface was at or lower than 0 ¿C. Observed delays between the start of rain and the arrival of water at the base of the snow ranged between 2 and 40 hours. A geometric analysis was used to calculate the spacing of percolation fingers or columns necessary to produce the observed number of times that water percolated onto the HFTs. The number of times there were abrupt changes in the temperature of thermistors installed in the snow adjacent to the HFTs was used in a similar manner. Results from both analyses suggest percolation columns with a mean spacing of about 25 cm, which is consistent with observations of these features in snow trenches. ¿ American Geophysical Union 1993