Presented at the 2015 Aquarius/SAC-D Science Team MeetingWe present a method that allows for a realistic uncertainty assessment of the Aquarius salinity retrievals. This method involves modeling the errors of all input parameters that are ingested by the salinity retrieval algorithm. These parameters include: the noise in the radiometer measurements, knowledge errors in the pointing angle, uncertainties in the auxiliary geophysical fields (SST, wind speed, wind direction), uncertainties in the model for the dielectric constant of sea water, uncertainties in the atmospheric absorption model, uncertainties in the surface roughness correction algorithm and uncertainties in the corrections for solar, lunar and galactic radiation. It also includes estimates for uncertainties due to contamination from land, sea ice and undetected RFI. The sensitivity of Aquarius salinity retrievals to each input parameter can be calculated by performing multiple salinity retrievals wherein the value of each input parameter is slightly perturbed, creating a sensitivity analysis. It is essential to distinguish between random and systematic uncertainties. We compare the results of the formal uncertainty assessment with an empirical error assessment based on triple collocations between Aquarius, ARGO and HYCOM salinities.