The classical Rayleigh model assuming isotopic equilibrium fails to explain the deuterium and oxygen 18 contents of polar snow. This model leads to too high temperature-isotope gradients (both for δD and δ18O), to too low δD - δ18O slopes, and consequently to an excessively large range of deuterium excess values (d = δD - 8δ18O). We present a new model that takes into account the existence of an isotopic kinetic effect at snow formation as a result of the fact that vapor deposition occurs in an environment supersaturated over ice. This kinetic effect is thoroughly discussed from a microphysical point of view and tested against experimental data and field observations. This new formulation reconciles predicted and observed values both for the temperature-isotope and δD - δD18O relationships for reasonable values of supersaturation over ice.