We present a semiformal model of anonymous communication with several participants performing several anonymous actions on several messages, e.g. in digital pseudosignatures. The goal is to design a model having enough expressive power to model simple as well as very complex anonymous communication patterns.Our model concentrates on anonymity of a sender, a receiver, and on the relationship anonymity. However, the model is easy to adopt to other types of anonymity. A special anonymous channel formalism is introduced and extensively explored in this paper. The formalism builds on the top of so-called estimation procedure which takes knowledge of the adversary and processes it to find anonymous participants. Some other, already published, models of anonymity, e.g. the model of Hughes and Shmatikov or of Halpern and O'Neill, are compatible with our model – they could be used as building blocks together with (or instead of) our estimation procedure. Therefore, the tools developed for those models can be easily adapted to be used with our model.We use protocol runs and an observational equivalence on the runs which is induced by adversary's knowledge. This is a well developed area of the theoretical computer science and many tools developed therein can be adapted to work with our model. Our model is also open to many additional features, e.g. the possibility to include probability distributions on anonymity sets.