Meltwater mixing in small arctic lakes at Saqvaqjuac (63°68′N, 90°40′W) was studied in 1980 and 1981 to evaluate the applicability of theoretical lake water renewal times to the modeling of ice-covered lakes. Two 370-GBq tritium additions were made to 7.09-ha P&N Lake. One was mixed with the unfrozen water at the time of maximum lake-ice thickness (May 1980) and the other was mixed with the lake immediately after freezing (October 1980). Dye experiments were also performed at four lakes to define the spatial and temporal distribution of the inflow and icemelt layers. Results from the tritiated water and dye addition experiments, as well as conductance and temperature profiles, showed that during ice-on, the cold low-density meltwater floated in a thin layer 0–100 cm beneath the ice, extended over the entire subice-surface area, and left the lake without mixing with the heavier subice water. These results imply that (1) lake models incorporating a lake flushing rate term need to be reevaluated to accommodate the lack of meltwater mixing beneath spring ice and (2) more attention should be given to the early spring meltwater chemistry and its distribution within the upper lake strata.