Icing measurements were carried out in natural winter clouds with an instrumented wind tunnel set up at the summit of Puy de Dame (1 500 m a.s.1.). The microphysical data (liquid water content, drop-let spectra) were obtained by using the particle measuring system PMS ASSP 100. The ice density was measured on a rotating cylinder and the ice deposit of the cross-section was photographed on a fixed cylinder. The density measurements ranged from 300 to 900 kg m-3 during the experiment and are in agreement with Macklin's results (1962). The profile of the ice deposit is compared to the profile predicted by the model of Lozowski and others (1979), which considers a fixed density. ~Ie propose to improve this model by taking into account the variation of ice density with the angle of impact on the cylinder. This calculation is based on Macklin's results and on the determination of the 1 oca I impact speed by us i ng the result of La ngmu i r and Blodgett (1960). The variation of ice density with the angle depends on various parameters: pressure, temperature, air speed, liquid water content, and, especially, the droplet distribution. The improvement is not sufficient to explain some observed profiles; this may be attributed to the fact that the model is not time-dependent.