The information-processing (IP) approach to perception and cognition arose as a reaction to behaviourism. This reaction mainly concerned the nature of explanation in scientific psychology. The "standard" account of behaviour, phrased in strictly external terms, was replaced by a "realist" account, phrased in terms of internal entities and processes. An analysis of the theoretical language used in IP psychology shows an undisciplined state of affairs. A great number of languages is simultaneously in use; no level of analysis is unambiguously referred to; and basic concepts such as information and processing remain largely undefined. Nevertheless, over the past 25 years the IP approach has developed into a disciplined and sophisticated experimental science. A look at actual practice hints at the basic reason for its success. The approach is not so much concerned with absolute or intrinsic properties of the human information processor, but with what can be called its relative or differential properties. A further analysis of this feature of the IP approach in terms of the formal language of a logical system makes explicit the basis of its success. The IP approach can be regarded as developing an empirical difference calculus on an unspecified class of objects, phrased in terms of a simulated "theory-neutral" observation language, and with operators that are structurally analogous to logical operators. This reinterpretation of what the IP approach is about brings a number of advantages. It strengthens its position as an independent science, clarifies its relation with other approaches within psychology and other sciences within the cognitive science group, and makes it independent of philosophical subtleties.