ABSTRACTThe purpose of this paper is to highlight some important ways in which standard information‐processing models of language, and specifically of reading, are deficient. It is proposed that the standard models do not provide an adequate framework for dealing with a number of central issues in reading. Their deficiencies are considered under four headings, (1) theoretical rigour, (2) the interaction of different levels of information during reading, (3) the parallel nature of much of human information processing, and (4) reading comprehension. The source of the problems is considered to be the view of information as an inert encoding and of information processing as a series of these encodings. A distinction is made between models which emphasize a flow of information and models which emphasize a flow of control. Production systems are proposed as examples of the latter type of model. Their use is seen as a qualitatively different approach to language processing which provides a solution to many of the problems inherent in the standard model