This paper describes a category of information systems called ‘Soft information systems’. Basically these lie between conventional information systems in which data are well defined, and management support, and decision support systems. A soft system is a system which operates on soft data. Such data cover categories of information such as conjectures, predictions, feelings and opinions and are typically found in systems where there is some element of anticipation of the future. Such systems include career counselling systems, where clients have rather ‘fuzzy’ views of what they might wish to do, through to economic based systems used to predict levels of employment, inflation levels and interest rates etc. In between are management-type systems in which it is normally assumed that hard (i.e. factual) data are used, for instance in budgeting systems. The paper describes some previous work and the experimental systems which tested and explored the ideas involved as well as attempting to refine some of the techniques that can be used to deal with soft data.