ABSTRACT
Language is structured at multiple levels. Sentences are composed of words, words are composed of meaningful constituents called morphemes, and morphemes are composed of letters or phonemes. In the domain of single word recognition, investigations into both phonological and morphological aspects of constituent structure have had to address similar issues, and the phonological and morphological models that have been proposed are analogous in many respects. In an alphabetic writing system, orthographic units correspond to phonological units, yet often there are ambiguities. There are one to many mappings between written units (graphemes) and phonology so that there can be many phonemic interpretations for a single grapheme. Compare for example the pronunciation of OW in ROW, COW, GOWN, OWN, and FOWL. Complementarily, there are many mappings from phonology back to form. For example note the many ways to write the sound that is common to EIGHT, GREAT, MATE, STRAIGHT and WAIT. With a few notable exceptions (e.g., Stone, Vanhoy, & Van Orden, 1997), models within the phonological domain of word recognition typically focus on the relation of written to spoken form rather than from spoken to written form. A common assumption is that there is an analytic processing option for words that are compositional and a second more holistic process for words that are irregular (e.g., Coltheart, Curtis, Atkins, & Haller, 1993; Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001). The presence of a single letter whose mapping to a spoken form is not consistent with the most common mapping classifies a word as “irregular.” In the dual route framework, neither the consequences of multiple atypical grapheme-phoneme mappings within a word nor complexities in the mapping from pronunciation back to written form are of interest.
