Deep dyslexic (DD) readers produce semantic errors during word naming
and are impaired at nonword naming. Previous models of DD have explain
ed this co-occurrence of deficits by postulating damage to both lexica
l and nonlexical pathways in the reading system. Buchanan, Hildebrandt
and MacKinnon (1994) offered an alternative explanation that resulted
in the prediction that words with several semantic and phonological n
eighbors would be read with less success by DDs than would words with
few neighbors. This paper briefly describes a test of the semantic nei
ghborhood hypothesis using HAL, a computational model of semantic spac
e developed by Lund and Burgess (in press).