Volume 32, Issue 1

In Japanese, subjective emotion predicates such as kanashii ‘sad’ and tsurai ‘painful’ are exclusively used to describe the inner state of the speaker, reflecting a highly subjective construal. This study investigates systematically how emotion is expressed in formal but confidential narrative texts, in which the egocentric perspective normally conveyed by these predicates would be inappropriate. In the data examined, subjective predicates occurred very infrequently in the form of direct assertions. Instead, a range of syntactic structures is used which reflect a departure from a subjective construal; the emotion is either presented as severed in some way from the experiencer or a separate reporting self emerges, distanced from the experiencing self. These syntactic structures reflect a less egocentric, more objective construal of emotion, suitable for this narrative genre. In this way, syntax is shown to work to obscure the linguistic privilege lexically encoded in subjective emotion predicates in Japanese.

The present study investigated a hypothesis proposing the involvement of three operations in processing Japanese sequential voicing (rendaku): a lexical-specific operation, an etymology-specific operation, and a rule-based operation. Second language (L2) learners of Japanese are in the process of constructing an L2 mental lexicon of Japanese, and this lexicon is assumed to display a clear contrast between rule-based and etymology-specific occurrences of rendaku in early-stage learning and lexical-specific rendaku at later stages as a result of memory-based lexical learning. Native Chinese (N=32) and Korean (N=32) speakers learning Japanese, matched for their lexical and grammatical knowledge, avoided applying rendaku in compounds with a medial voiced obstruent in the second element, indicating that Lyman’s Law is an active principle even in L2 acquisition. Both L2 learner groups also showed sensitivity to lexical strata by distinguishing Japanese-origin words (wago) from Sino-Japanese words (kango) and alphabetic loanwords (gairaigo). Thus, as factors of rendaku processing, Lyman’s Law is considered rule-based while lexical stratum is considered etymology-specific. In contrast, both L2 learner groups showed a low occurrence of rendaku both for Lyman’s Law exceptions (i.e., X+basigo) and for infrequent or rare words (i.e., X+zyootyuu). These instances can be considered memory-based, lexical-specific rendaku, which L2 learners must acquire as individual lexical items. This study indicated that all three described operations were used by L2 Japanese speakers to process rendaku.

In Japanese loanwords, voiced geminates can be devoiced in the presence of another voiced obstruent (e.g., /doggu/ → /dokku/ ‘dog’). This devoicing pattern has been studied extensively in the recent phonological literature in terms of theoretical modeling as well as from the perspective of experimentation and corpus studies. Less well-known is the observation that /p/ may cause devoicing of geminates as well (e.g., /piramiddo/ → /piramitto/ ‘pyramid’), although to date no objective evidence has been offered to confirm this observation. The current study thus attempts to test this observation objectively by way of a corpus study and two phonological judgment experiments. The results generally support the idea that /p/ can cause devoicing of geminates in Japanese loanwords; in other words, /p/ may trigger Lyman’s Law in causing devoicing of geminates. In addition to this descriptive discovery, throughout the paper we discuss intriguing task effects in phonological experimentation, by comparing the corpus data and the results of the two judgment experiments. Although our aim is primarily descriptive, we offer some analytical possibilities for the /p/-driven devoicing of geminates at the end of the paper.

This paper examines the phonology of Sino-Japanese within the framework of Optimality Theory, incorporating elements from Itō and Mester’s (1999) coreperiphery model. A major component of Itō and Mester’s model-the unmarked status of Yamato words-was challenged by Kawahara, Nishimura, and Ono (2002), who argued that Sino-Japanese must be the unmarked option. While it is true that Sino-Japanese is the least marked stratum, automatically declaring the least-marked stratum as the default incorrectly predicts that speakers will overgeneralize alternations regardless of their productivity. Taking inspiration from Kurisu’s (2000) analysis of Sino-Japanese geminates and Mascaró’s (2007) formalization of allomorphy, I propose that where a lower stratum’s alternations are unproductive, speakers actually store all allomorphs in the underlying form, thus exempting the stratum from the implicated faithfulness constraint. This allows the unindexed faithfulness constraint to be ranked higher as needed to ensure that only productive alternations are extended to novel items.