A connectionist approach to Japanese kanji word naming

Connectionist approaches have been providing new views for understanding bo th normal and disordered reading processes (e.g., Seidenberg & McClelland, 1989; Plaut, McClelland, Seidenberg, & Patterson, 1996). The present resear ch was designed to simulate normal naming processes of Japanese Kanji words using a connectionist network. The network was trained to map the orthogra phy of two character Kanji words onto their pronunciations or phonology. Ea ch Kanji character was represented by a 16 x 16 grid pattern on the input l ayer, and the word's phonology at the output layer consisted of phonologica l codes for the two component Kanji characters. The training corpus include d 4,136 two-character Kanji-words with frequencies higher than four occurre nces per million. After 900 training epochs, the network could correctly na me 99.8% of the 4,136 words, including those with inconsistent or atypical character-sound correspondences. In terms of efficiency of word naming, the network showed frequency and consistency effects and an interaction betwee n these variables, largely comparable to these effects in the naming latenc ies of Japanese skilled readers (Fushimi, Ijuin, Patterson, & Tatsumi, 1999 ). On the other hand, when naming nonwords, the network's performance was s ubstantially worse than that of-skilled readers. Properties of the connecti onist approach for Japanese Kanji word naming are discussed.