Theories of visual word recognition differ considerably in their assumptions about the nature of sublexical processing. In traditional dual-route models of reading (e.g., Coltheart, 1978; Coltheart, Curtis, Atkins, & Haller, 1993), sublexical orthographic units like own in real words (e.g., down, shown) or nonwords (e.g., pown, thown) are used mainly as inputs to grapheme-phoneme conversion rules whose end products are phonemes, although they are also involved in cascaded activation from letters to word representations. In most connectionist models of word reading (e.g., Plaut, McClelland, Seidenberg, & Patterson, 1996; Seidenberg & McClelland, 1989), the sublexical orthographic units are both the input to distributed networks computing phonological outputs and the input to orthography-to-semantics networks computing semantic outputs. However, these semantic outputs are primarily for the whole letter strings, not for the sublexical orthographic units, even when these units happen to be words themselves (e.g., own in down or boy in boycott). There is no obvious difference between the processing of these units and the processing of other units that do not happen to be words (e.g., int in hint or pint or mur in murder). In models of visual word recognition that stress the predominant role of phonology in access to meaning (e.g., Frost, in press; Lukatela & Turvey, 1994; van Orden, Pen-nington, & Stone,

1990), sublexical processing of orthographic units is strictly a phonological process, which activates the phonological representation of the whole word before the meaning of the word is accessed. Therefore, despite their considerable differences concerning whether the sublexical processing in reading alphabetic words is rule-based and how computation from orthography to phonology is conducted, most localist and connectionist theories of visual word recognition share the assumption that sublexical processing in reading alphabetic scripts is a process that derives the phonology and meaning of whole words from sublexical orthographic units. It is either primarily or strictly a phonological event, with no significant semantic activation associated with these sublexical units.