The construction of personalised information systems is becoming more important as the quantity of information to be searched increases. The approach taken in this paper involves the implicit tracking of user behaviour over the long-term, with an emphasis on how the collected information can be processed and interpreted to build up a good user model for the required task. Parallels are drawn between such systems and the study of non-verbal communication within the physical world, before an interpretive framework based on semiotics, the study of signs, is outlined. The system task considered is that of recommendation within a hypertext environment such as the web.