In collaboration with Ian Nimmo-Smith and William Marslen-Wilson, we are looking at ways in which we can model root and pattern effects in Arabic. Specifically we are using the frequency measures from Aralex to model the process of mapping an orthographic form in Arabic onto its internal morphemic representation. We are developing this modelling effort within a Bayesian framework asking whether prior knowledge about a root can influence posterior knowledge about the word pattern or the surface form as a whole, and vice versa.