The heat balance method was used to estimate transpirational sap flow through 60- to 75-mm-diameter stems (trunks) of 3-year-old peach [Prunus persica (L.) Batsch. cv. O'Henry] trees under field conditions. On rare occasions, heat balance estimates agreed well with independent lysimetric measurements, but on most occasions, heat balance estimates of sap flow were unrealistic in both direction and magnitude. In some cases, the errors in sap flow approached two orders of magnitude and were always the result of a calculation involving division by a very small and sometimes negative temperature differential between the stem surface temperature above and below the gauge heater. The occurrence of negative temperature differentials under positive transpiration conditions may be inconsistent with a fundamental assumption in the heat balance model, namely that temperature differentials are solely a consequence of the dissipation of energy supplied to the gauge heater. In the absence of heating power applied to the gauge, temperature differentials exceeding - 1C were correlated with the rate of change in stem temperature, indicating that ambient conditions themselves can impose a bias in gauge signals and, hence, influence gauge accuracy. Our results suggest that the effect of ambient conditions on gauge signals should be critically evaluated before considering heat balance estimates of sap flow as reliable under any given conditions.