Rendaku, or sequential voicing, is a morphophonemic process in Japanese in which the voiceless word-initial consonant of the second element (=E2) of a compound word becomes voiced (e.g., ori + kami →origami, ‘folding’ + ‘paper’ → ‘paper folding’, /k/ becomes /g/). In adult grammar, rendaku is subject to two conditions: It applies if and only if (a) E2 is a Yamato word (native vocabulary) in the lexicon and (b) it contains no voiced consonant (e.g., b, d, & g). Recent psycholinguistic studies have revealed that Japanese-speaking preschoolers do not follow adult’s grammar; they develop their original prosodically-based rendaku processing strategy (preschooler-specific rendaku strategy). Their strategies qualitatively change in the early middle childhood to be adult-like rendaku, creating a discontinuity in children’s word-processing strategies. This study investigated factors responsible for this developmental discontinuity. We conducted an experiment using cross- modal linguistic stimuli (prosody & orthography) to see whether children’s orthographic knowledge affects their rendaku strategy or not. Our results showed that orthographic cues affected literate children’s rendaku processing. They were aware the correspondence between types of orthography and word categories in Japanese.

Hirose & Mazuka (2015 & 2017) demonstrate that Japanese speaking adults and first graders both show anticipatorycompound processing, using the language-specific compound accent rule (=CAR). That is, six- to seven-year-old childrencan exploit compound prosody to disambiguate the structure and meaning of a given compound. However, we do notknow exactly when and how children start exploiting the CAR to properly comprehend compounds. Thus, we investi-gated Japanese-speaking childrens acquisition of the CAR and their development of compound processing. We conductedlongitudinal experiments using compound comprehension tasks on 65 Japanese-speaking children aging from two- to four-years. We found that childrens compound processing strategies changed after their acquiring the CAR. Before acquiringit, children could not identify the compound head; instead they showed a language-general parsing preference for theleft-most part of a compound. Our results suggest that childrens acquisition of the language-specific CAR enables theircompound processing.