ABSTRACT
Models of word recognition are attempts to characterise some of the mental processes that allow a reader to identify, comprehend and pronounce written words. They try to decompose the act of word recognition into its component parts and describe the workings of those parts. More recent models of word recognition have been expressed in the form of computer programs which aim to mimic aspects of human word recognition. The visual analysis system must encode letter identities and positions before the reader can determine whether a word being fixated is familiar or unfamiliar and, if familiar, which word it is. The procedure for connecting letters to sounds must do more than associate individual letter identities with individual phonemes. If the sublexical procedure is active when a word with a regular spelling is being read, then it should generate a pronunciation which agrees with that produced by the lexical procedure.
