Airborne volcanic ash can be observed and quantified from hyperspectral infrared (IR) sounders. The retrieval process of physical quantities, such as particle radius and mass, depends critically on the assumed spectrally dependent complex refractive indices (CRIs) that are used. Traditionally, the Pollack et al. (1973) dataset was used almost exclusively. These indices are, however, based on measurements of rock slabs, and in recent years, two datasets have become available from laboratory measurements of ash in suspension, the Reed et al. ( 2018) and Deguine et al. ( 2020) datasets. Here, we compare the three datasets and quantify the extent to which each of them can be used to simulate satellite observed spectra of real volcanic ash plumes. We find that whereas the Pollack et al. indices perform worst throughout, the performance of the other two is comparable for andesitic and basaltic ash plumes. However, all three datasets have difficulty in reproducing the extinction minimum around 1250 cm −1 . The Reed et al. indices in addition yield inconsistent results in the 800-850 cm −1 range. The Deguine et al. dataset is the only one, which can be used to reproduce the very large spectral signatures often observed with rhyolitic ash across the entire thermal IR window 750-1250 cm −1 . In terms of retrieved quantities, the largest differences are seen for the radius, with the Deguine et al. dataset resulting in the smallest retrieved particle sizes.