No abstract

No abstract

This work was supported by the Natural Sciences and Engineering Research Council of Canada Grant AO998 to D.B. and was submitted to the University of Waterloo in partial fulfillment of the requirements of a master's degree for M.A.R. We are particularly grateful to Marco Zorzi for providing the consistency analysis reported here and for helpful discussion. We are also grateful to Peter Kwantes for providing the LEX model simulation data reported here and to Ken Forster, Chris Kello, Ken Paap, and an anonymous reviewer for their helpful comments and comprehensive reviews. Low-frequency irregular words are named more slowly and are more error prone than low-frequency regular words (the regularity effect). Rastle and Coltheart (1999) reported that this irregularity cost is modulated by the serial position of the irregular grapheme-phoneme correspondence, such that words with early irregularitiesexhibit a larger cost than words with late ones. They argued that these data implicate rule-based serial processing, and they also reported a successful simulation with a model that has a rule-based serial component-the DRC model of reading aloud (Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001). However, Zorzi (2000) also simulated these data with a model that operates solely in parallel. Furthermore, Kwantes and Mewhort (1999) simulated these data with a serial processing model that has no rules for converting orthography to phonology. The human data reported by Rastle and Coltheart therefore neither require a serial processing account, nor successfully discriminate among a number of computational models of reading aloud. New data are presented wherein an interaction between the effects of regularity and serial position of irregularity is again reported for human readers. The DRC model simulated this interaction; no other implemented computational model does so. The present results are thus consistent with rule-based serial processing in reading aloud.

A number of highly successful computational models of basic processes in reading implement multiple levels of representation (modules) that get activated when a letter string is presented (e.g., Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001;Grainger & Jacobs, 1996;McClelland & Rumelhart, 1981). A central feature of these models is that activation across different modules is cascaded. In systems that operate by cascaded processing, there are no thresholds within modules. As soon as there is even a small amount of activation in an early module this flows on to later modules. The present work demonstrates a previously unappreciated consequence when serial processing follows cascaded parallel processing. More specifically, a simple manipulation of processing rate in the model produces a qualitatively different pattern from that produced by university-level readers. This effect is illustrated in the context of nonlexical processing by Coltheart and colleagues' computational model. The Dual Route Cascaded ModelThe dual route cascaded model (hereafter DRC) is very successful at accounting for a wide variety of findings in naming and lexical decision tasks. Indeed, its authors list 18 phenomena from the naming task that the model simulates successfully. Coltheart et al. (2001) comment that the set of phenomena that the DRC model can simulate is much larger than the set that any other current computational model of reading aloud can simulate; and, to the best of our knowledge, there is no effect seen in reading aloud that any of these other models can simulate but that DRC cannot. (p. 251) Coltheart et al. thus consider DRC to be the most successful computational model of reading aloud.DRC consists of three routes; the lexical semantic route, the lexical nonsemantic route, and the nonlexical GPC (grapheme-phoneme conversion) route (see Figure 1). A semantic system is not yet implemented in the model. The knowledge bases in the two remaining routes differ; the lexical system is based on word-specific knowledge, whereas the nonlexical system is based on a set of sublexical spelling-sound correspondence rules. Briefly, the correct pronunciationfor all words known to the model can be produced by the lexical route's operating in isolation, and the correct pronunciation of a nonword can be produced by the nonlexical route's operating in isolation. The lexical route's operating in isolation can not produce the correct pronunciationof a nonword. The nonlexicalroute's operating in isolation produces the correct pronunciation of all words that obey the spelling-sound correspondence rules, but regularizes the pronunciation of words that are exceptions to these rules (e.g., PINT is pronounced /pInt/). When the intact model is operating, both the lexical and the nonlexical routines affect the time to construct a pronunciation for both words and nonwords. A more detailed discussion of the architecture and how the full model operates is provided by Coltheart et al. (2001). For our present purposes, only the operation of the nonlex...

Nine experiments show that in the context of Stroop dilution the extent to which flanking distractors are processed depends on the nature of the material at fixation. A Stroop effect is eliminated if a word or a nonword is colored and appears at fixation and the color word appears as a flanker. A Stroop effect is observed when the color carrier at fixation is from a different domain than the color word distractor (e.g., Arabic digits). It is argued that when the material at fixation is in the same domain as the color word distractor, the distractor is not processed. Taken together, these results implicate a role for material-specific, limited-capacity processing in the context of this variant of the Stroop paradigm.