ABSTRACT
Until recently, a large majority of the research on the processing of visual words has ignored whether morphemes are involved in the recognition or the understanding of words. This research either has been conducted with morphemically simple words or, if not, has largely ignored whether the words are morphemically complex or not. Among other things, this seems to have shaped views of how words are processed; processing is often discussed in terms of lexical entries, and identification and/or access to a word’s meaning are often seen as occurring when the excitation of a single lexical entry either rises above some threshold or is sufficiently greater than the excitation of any competitor. That is, identifying words is seen as a memory-retrieval process that is quite distinct from the obviously productive processes involved in computing syntax and putting together the meanings of words to understand the meaning of discourse. Parallel distributed models have a somewhat more complex view of how identification of words and access to their meaning occur, but these networks usually have only letters, letter clusters, and words as the meaningful units which enter into the processing of visual words, and if morphemes are used, they are likely to have no special status beyond syllables or other sequences of letters that may co-occur frequently in words.
