Two experiments examined the role of pronunciation rules and of lexical information in pronouncing letter strings. In Experiment I, subjects pronounced pseudowords varying in the strength of the rules needed to pronounce them, as well as in the availability of a lexical model. In Experiment 2, the stimuli were words varying in rule strength and in usage frequency. The pronunciation times from both experiments displayed an interaction between rules and lexical information: When the rules necessary were strong, the relative availability of lexical information was less important than when the rules were weak. The results were discussed with respect to both traditional dual-process models of pronunciation and models proposing the use oflexical analogies.A salient feature of skilled reading is the rapid generation of pronunciations, both of known words and of novel but pronounceable nonwords, or pseudowords. Currently, a debate exists concerning the appropriate underlying model for these pronunciation processes. A major piece of the controversy concerns the unit of analysis in pronunciation. A traditional view (e.g., Coltheart, 1978; Coltheart, Davelaar, Jonasson, & Besner, 1977;Forster & Chambers, 1973) has been that readers possess pronunciation knowledge of two types: lexical representations that contain pronunciations for known words, and an independent set of pronunciation rules (typically presumed to be grapheme-phoneme correspondences) used to pronounce unfamiliar words and pseudowords. The current debate concerns the existence of grapheme-based pronunciation rules, with a variety of researchers suggesting that lexical information may be sufficient to support the pronunciation of both known and novel strings.
Are Pronunciation Rules Necessary?The fact that readers easily generate pronunciations for completely novel words has often been taken as prima facie evidence for the existence and use of pronunciation rules. Typically, readers' reliance on such rules has been incorporated into a dual-process model (Baron & Strawson, 1976;Forster & Chambers, 1973) in which readers look up lexical items and apply pronunciation rules in parallel, with the former generally faster for known items, but the latter required for novel or infrequent words. Recently, however, researchers have begun to question this view of pronunciation.This research was submitted to the University of Texas at Austin in partial fulfillment of the requirements for a doctorate in psychology. Thanks go to Tom Carr, two anonymous reviewers, and the editor for constructive comments on earlier versions of the paper. One particularly damaging piece of evidence involves the contribution of lexical information to the pronunciation of novel words or pseudowords. The proposal has been that a novel string might be pronounced by analogy to a known word that is visually similar (Baron, 1977;Brooks, 1977;Glushko, 1979;Marcel, 1980, Rosson, 1983. So, for example, Kay and Marcel (1981) found that the pronunciation chosen for an ambiguous pseudoword (a nonword that ...