The deltaO-18 value of streamflow at Imnavait Creek, Alaska, shifted dramatically from -30.3 parts per thousand on 14 May, the first day of streamflow in 1990, to -22.5 parts per thousand on 22 May, at the end of the snowmelt. These results suggested a large degree of mixing between snow (deltaO-18 value = -27.6 parts per thousand +/- 1.6 SD) and underlying soil moisture (estimated deltaO-18 value = -20.6 parts per thousand). Nevertheless, independent hydrological measurements of snow redistribution by wind, snow ablation, snow and soil moisture content, and snowmelt runoff indicate there cannot be significant mixing of meltwater with underlying ice-rich soils. An alternative explanation is that isotopic fractionation during the phase change from solid to liquid dominates the isotopic variation in streamflow during snowmelt and prevents a straightforward application of O-18 as a conservative hydrological tracer. By contrast, under dry antecedent conditions in late summer, O-18 appeared to be a suitable tracer following rain contributions to streamflow. Streamflow increased as a result of rainfall, but stream isotopic composition did not change until at least two hours after streamflow increased, implicating a wave, or piston-like mechanism for forcing ''old'' water into the stream channel. Analyses of the stable hydrogen and oxygen isotope composition of various hydrological components within the watershed indicate the importance of evaporation as a dominant factor in the hydrological cycle; soil moisture, streamflow, permafrost, and the snowpack all showed varying degrees of isotopic alteration as a result of evaporation. Our analyses indicate that caution would be advised for any application of stable isotopes to hydrological studies in arctic watersheds. Proportions of snowmelt mixing with underlying soil water may be subject to overestimation because isotopic fractionation as snow melts can be similar in direction and magnitude to the isotopic mixing of snowmelt and soil waters.