The ability of viscous layer models to describe the attenuation of waves propagating in grease‐pancake ice covered ocean is investigated. In particular, the Keller's model (Keller, ; https://doi.org/10.1029/97JC02966), the two‐layer viscous model (De Carolis & Desiderio, ; https://doi.org/10.1016/S0375-9601(02)01503-7) and the close‐packing model (De Santi & Olla, ; arXiv:1512.05631) are extensively validated by using wave attenuation data collected during two different field campaigns (Weddell Sea, Antarctica, April 2000; western Arctic Ocean, autumn 2015). We use these data to inspect the performance of the three models by minimizing the differences between the measured and model wave attenuation; the retrieved ice thickness is then compared with measured data. The three models allow to fit the observation data but with important differences in the three cases. The close‐packing model shows good agreement with the data for values of the ice viscosity comparable to those of grease ice in laboratory experiments. For thin ice, the Keller's model performance is similar to that of the close‐packing model, while for thick ice much larger values of the ice viscosity are required, which reflects the different ability of the two models to take into account the effect of pancakes. The improvement of performance over the Keller's model achieved by the two‐layer viscous model is minimal, which reflects the marginal role in the dynamics of a finite eddy viscosity in the ice‐free layer. A good ice thickness retrieval can be obtained by considering the ice layer as the only source in the wave dynamics, so that the wind input can be disregarded.