Semantic radicals in Japanese two-character word recognition

Abstract

Extending previous studies on sub-lexical character constituent activation in Japanese and Chinese, the present regression study investigates whether recognition of two-character words in Japanese involves activation of semantic radicals and whether the semantic radicals’ contribution is orthographic or semantic in nature. A mixed-effects model complemented with a random forest analysis provided support for the importance of the semantic radical, as witnessed by two orthogonal sets of semantic radical properties: the radicals’ semantic transparency and usefulness, as well as the semantic radicals’ combinability and token frequency. The two frequency effects were facilitatory for the head of the compound, but inhibitory for the modifier, possibly due to incompatibility of the semantic class marked by the modifier's radical with that of the compound as a whole. Thus, semantic radicals emerge as not just orthographic components but as fully fledged purely orthographic morphemes.