Rendaku is not blocked by two nasal consonants: A reply to Kim (2022)

Abstract

Rendaku is a morphophonological process in Japanese, in which the first consonant of the second element of compounds becomes voiced (e.g. /ao-sora/ → [ao-zora] ‘blue sky’). In a recent paper, Kim (2022) argues that the presence of two nasals—but not one nasal— in the second element reduces the probability of Rendaku application (e.g. /umi-kamome/ → [umi-kamome], *[umi-gamome] ‘seagull’). This claim has an important consequence for grammatical theories, as it would constitute a case in which the influences of two segments add up to block a phonological process, a pattern that is referred to as “super-additive counting cumulativity.” We have reexamined the empirical basis of this claim, and have found that there are only four relevant lexical items. Moreover, two of these items actually undergo Rendaku almost all the time, and there are thus only two items whose Rendaku is clearly blocked. The empirical basis of the theoretical claim based on Rendaku thus seems to be weak at best. With this, we move on to the main contribution of this paper, which is a new nonce-word experiment that aimed to resolve the conflicting results from the previous studies on this topic. The results of the current experiment show that the applicability of Rendaku is not reduced by the presence of two nasal consonants.

Lyman’s Law can count only up to two

“…One possibility is that this difference could be an unintentional consequence of participants not wanting to have too many positive Rendaku responses overall (recall that there was no condition which involved a clear violation of Lyman's Law-those E2s with a voiced obstruent-in this experiment). Kawahara and Kumagai (2023) report an experiment which more directly compared forms with two voiced obstruents which would clearly Lyman's Law and those that contain two nasals.…”

“…13 See also Kawahara and Kumagai (2023) for a follow-up study with a larger number of items and participants, which showed the opposite trend, in which forms with two nasals showed slightly higher, rather than lower, Rendaku response rates than the control condition, although that trend was not credible. This result offers an additional reason to believe that the effects of two nasal consonants are suspicious.…”

“…The Buy Response function, however, allows us to include only up to 50 questions. Kawahara and Kumagai (2023) report a similar experiment on the effects of two nasals which used all the segments that can potentially undergo Rendaku.…”

“…See Kim (2022) for a renewed analysis of the two nasal effect using the Rendaku database (Irwin, Miyashita, Russel, & Tanaka, 2020). See also Kawahara and Kumagai (2023) for a detailed reexamination of this reanalysis presented in Kim (2022).…”