ABSTRACT

The use of masked priming techniques to investigate basic questions in visual word recognition has become increasingly common in recent years (e.g., Andrews, 1996; Bodner & Masson, 1997; Bowers, Vigliocco, & Haan, 1998; Castles, C.Davis, & Letcher, 1999; C.Davis & Forster, 1994; C.J.Davis, 2001; de Moor & Brysbeart, 2000; Forster, 1998; Forster & C.Davis, 1984, 1991; Grainger & Jacobs, 1993, 1994; Hinton, Liversedge, & Underwood, 1998; Humphreys, Besner, & Quinlan, 1988; Humphreys, Evett, & Quinlan, 1990; Kinoshita, 2000; Lukatela & Turvey, 1990; Perea & Rosa, 2000; Rastle, M.Davis, Marslen-Wilson, & Tyler, 2000; Segui & Grainger, 1990; Sereno, 1991; Van Heuven, Dijkstra, Grainger, & Schriefers, 2001). This methodological trend intensifies the need for detailed theoretical accounts of masked priming. The present chapter analyzes the account of masked priming effects offered by competitive network models of visual word recognition. The chief goal is not to report simulations of specific experiments, but rather to develop a precise description of the factors underlying masked priming effects in a competitive network model. As such, most of the chapter is concerned with the development of a general formula for predicting masked priming effects in a specific competitive network model-the interactive activation (IA) model (McClelland & Rumelhart, 1981). Because the resulting expression is formulated in terms of standard psycholinguistic variables, the analysis presented here helps to bridge the divide between purely computational accounts and verbal theories of visual word recognition and priming. The analysis of masked priming effects in the model is developed in stages, as a function of the nature of the relationship between the prime and the target. Thus, the analysis commences with the relatively straightforward case of identity primes, and then proceeds to different types of orthographic form primes: partial word form primes, nonword form primes, and word form primes. The main emphasis throughout is on the development of a precise description of priming effects in the IA model, although the correspondence between these predictions and empirical findings is also discussed, where relevant. A number of novel predictions, which await experimental investigation, also are discussed. Prior to this analysis, I make some brief introductory remarks concerning competitive network models in general, and the IA model in particular, and describe the assumptions that are required to enable this model to simulate masked priming in the lexical decision task.