Abstract
Information theory as originally developed by Hartley, Shannon, and Wiener for the two-variable “Sender-Receiver” case, and further extended by McGill and Ashby, is believed to be a specific instance of a more general and abstract theory. This paper discusses possible foundations of such a theory starting from a “micro-approach” based on the primitive notions of variety and constraint also due to Ashby. Some slight inconsistencies in earlier definitions are cleared up, whereas some ambiguities in the “macro” quantities are pointed out. The applied theory, called constraint analysis, is very well suited for the analysis of constraints in multivariate systems (structure modelling or reconstructability analysis).