Japanese speakers can infer specific sub-lexicons using phonotactic cues

Abstract Phonotactic restrictions do not usually hold uniformly across the entire lexicon of a particular language, and thus the lexicon of a natural language is often assumed to consist of different sub-lexicons. A question that arises is how specific these sub-lexicons can be. A classic, conservative approach is to posit only broad distinctions, such as the distinction between native words versus borrowed words. An alternative approach is to posit more specific miniature lexicons, such as a set of morphemes that show a particular morpheme-specific pattern or a set of loanwords from a particular language. With this general theoretical issue in mind, this paper first points out that there are phonotactic restrictions that can cue a very specific class of words in Japanese; for example, geminate /ɾɾ/ occurs almost exclusively in loanwords from Italian. Building upon these novel observations, the current experiments tested whether Japanese speakers can infer specific sets of words such as “snack names”, “Italian restaurant names”, and “German names” based on particular phonotactic cues. The results of the two experiments support the idea that the lexicon of a natural language can consist of very specific sub-lexicons, at least going beyond the often posited native versus loanword distinction.