In the coming decade, low-frequency radio arrays will begin to probe the epoch of reionization via the redshifted 21-cm hydrogen line. Successful interpretation of these observations will require effective statistical techniques for analysing the data. Due to the difficulty of these measurements, it is important to develop techniques beyond the standard power-spectrum analysis in order to offer independent confirmation of the reionization history, probe different aspects of the topology of reionization and have different systematic errors. In order to assess the promise of probability distribution functions (PDFs) as statistical analysis tools in 21-cm cosmology, we first measure the 21-cm brightness temperature (one-point) PDFs in six different reionization simulations. We then parametrize their most distinct features by fitting them to a simple model. Using the same simulations, we also present the first measurements of difference PDFs in simulations of reionization. We find that while these statistics probe the properties of the ionizing sources, they are relatively independent of small-scale, subgrid astrophysics. We discuss the additional information that the difference PDF can provide on top of the power spectrum and the one-point PDF.