This paper applies the theory of Communicating Sequential Processes (CSP) to the modelling and analysis of a non-repudiation protocol. Non-repudiation protocols differ from authentication and key-exchange protocols in that the participants require protection from each other, rather than from an external hostile agent. This means that the kinds of properties that are required of such a protocol, and the way it needs to be modelled to enable analysis, are different to the standard approaches taken to the more widely studied class of protocols and properties. A non-repudiation protocol proposed by Zhou and Gollmann is analysed within this framework, and this highlights some novel considerations that are required for this kind of protocol.