Japanese/Korean Linguistics

Japanese/Korean Linguistics Volume 32

Volume 32

The tonal correspondences between Japanese and Ryukyuan show a systematic split in which a single Middle Japanese tonal class corresponds to more than one Proto-Ryukyuan class. This pattern is crucial for reconstructing the Proto-Japonic tonal system, because it has generally been taken to imply tonal contrasts not preserved in Japanese. In contrast, the present account attributes the split to a conditioned rightward accent shift in pre-Proto-Ryukyuan. The shift merged Middle Japanese Classes 2.4–2.5 with Class 2.3 but was blocked by two factors: a word-final high vowel and the retention of an original word-internal morpheme boundary. Evidence for the latter comes from rare non-final e₁ and o₁ in Old Japanese, apparent violations of Arisaka’s Law, unexpected heavy initial syllables in modern Japanese varieties, and etymological reconstruction. The analysis accounts for most correspondences without reconstructing additional tonal classes in Proto-Japonic and highlights the role of morphological restructuring in tonal change.

This study reexamines the place of articulation of Korean coronal affricates and its relationship to post-affricate yod-dropping. Drawing on historical phonology, loanword adaptation, acoustic production, and interlanguage perception, it argues that two restrictions coexist: a gradient, phonetically grounded constraint sensitive to perceptual contrast, and a generalized lexical ban on post-coronal /j/. Production data from Seoul, Toronto Heritage, Dandong, and Qingdao Korean show that northern Chinese Korean varieties have more anterior affricates, whereas Seoul and Toronto varieties generally exhibit more posterior realizations, with age- and gender-related variation. In a Japanese-to-Korean mapping task, speakers with anterior affricates more often represented posterior affricates using contrastive ⟨j⟩, while speakers with posterior affricates showed weaker differentiation. Loanword evidence likewise reveals greater tolerance of post-affricate ⟨j⟩ in varieties with anterior articulation. These findings support a dialect-sensitive, synchronically active connection between affricate place, perceptual distinctiveness, yod-dropping, and cross-linguistic adaptation, illuminating broader mechanisms of phonological change over time. 

This paper provides a unified semantic and pragmatic analysis of the sentence mood in Korean in terms of the various updates to the common ground in structured DRT. All clauses have the uniform discourse effect of steering the common ground to a preferred subset of the possible worlds with different levels of authority. In addition, honorifics were given a novel analysis in terms of the speaker authority for her preferences and conversational negotiations. The proposed theory offered a new perspective to utterance force and social meaning as well as a concrete tool to analyze them.

This paper outlines a brief history of the evolving relationship between linguistics and computer science, ranging from the famous 1956 MIT symposium on information theory to contemporary debates over the relevance of language models for the study of the human language faculty. As a case in point, we consider the phenomenon of negative concord in children’s acquisition of negation in both English and (to a lesser extent) Korean.  Drawing on insights from both cognitive and computational approaches, we exemplify how an interactive line of inquiry may contribute to advances in the field of linguistics.

Japanese relative clauses exhibit apparent island violations. Previous analyses have attributed this pattern either to base-generated resumptive pronouns in the extraction site or to extraction of the head noun from a broad subject position. This paper presents new evidence from reconstruction effects challenging the broad subject analysis. Three diagnostics—anaphor binding, variable binding, and Condition C—show that reconstruction into the variable site inside the island is systematically unavailable, whereas reconstruction into an intermediate position below the highest subject but above the island is obligatory. I argue that these facts follow if island-insensitive relative clauses are derived by mixed chains, in which a relative operator is base-generated in the highest specifier of vP, binds a resumptive pronoun inside the island, and subsequently undergoes movement. The proposal provides new evidence that Japanese relativization obligatorily involves movement in the upper portion of the A′-dependency chain despite its apparent island insensitivity.

This paper seeks to provide an answer to why there is so much disagreement on data instantiating the ACC-marked external possession construction (EPC) in Korean; when ACC-marked EP sentences in Korean are licensed; and why EP-encoding multiple-accusative sentences and non-EP-encoding multiple-accusative sentences have the same surface form even though the latter do not have a grammatical internal possession counterpart. By conducting a carefully-constructed grammaticality judgment task, I argue that speakers have variable acceptability judgments for the EPC because of its pragmatically marked nature; a uniform analysis of the construction is not possible but it can be classified into three subtypes depending on the semantics of the V and the Part-to-Whole Spread possibility in the sense of Dowty (2001); and in all grammatical cases, the possessor NP functions as what I call a Topic-Focus, and this explains why both EP-encoding and non-EP-encoding multiple-accusative sentences have the same surface form.

The Korean plural marker tul is often described as optional, but its distribution is sensitive to definiteness and humanness. Accounts differ on how these factors interact: some argue that definiteness makes tul obligatory regardless of humanness (Kang 1994; Kim 2005), while others report that non-human definites with plural reference do not always require tul (Park 2020; Choi, Yoon, & Dayal 2026). To investigate these factors, we conducted an elicited production task and a sentence pair judgment task with native Korean speakers. In production, speakers used tul more frequently with [+Human] than [-Human] nouns in definite plural contexts. In judgment, tul-marked nouns were preferred over unmarked nouns for plural referents, particularly in definite contexts, with this preference further modulated by humanness. These findings suggest that definiteness and humanness jointly influence tul-marking in a gradient rather than categorical manner, supporting an analysis of tul as a plural modifier (Kim & Melchin 2018).

This paper investigates Coordinated Wh-Questions (CWQs) in Japanese, in which wh-phrases with distinct grammatical roles or syntactic categories are conjoined. Since such elements cannot be base-generated in a single position, CWQs raise the question of what is coordinated. Boskovic (2024) argues that a wh-indeterminate system is a prerequisite for wh-coordination, predicting that Japanese should allow it; existing accounts, however, derive Japanese CWQs from coordination of larger constituents such as CPs or vPs (Ishii 2014, Kasai 2016). I argue instead that Japanese CWQs involve genuine wh-coordination in a monoclausal structure. I further show that Japanese sluicing deletes the lowest complementizer Fin while retaining the Force head ka, indicating that Merchant's (2001) Sluicing-COMP generalization targets the lowest complementizer. Coordinated Sluicing accordingly divides into two subtypes: monoclausal wh-coordination when the first conjunct lacks ka, and biclausal multiple sluicing when it contains ka.

Minamida (2026) shows that Osaka Japanese exhibits robust wh-island effects based on two experiments: a traditional auditory comprehension experiment with methodological refinements (Experiment 1) and a novel acceptability judgment experiment using interrogative sentence endings (Experiment 2). This study replicates that study for Tokyo Japanese to examine whether wh-island effects arise in this variety and to test the applicability of the methodology across varieties. Experiment 1 does not provide clear evidence for either the presence or absence of wh-island effects in Tokyo Japanese. In contrast, Experiment 2 provides clear evidence for wh-island effects. Given that auditory stimuli require more careful construction than written stimuli because they must reflect the prosodic properties of the language under investigation, the results of Experiment 2 support the conclusion that Tokyo Japanese exhibits wh-island sensitivity. This paper also discusses possible future refinements to the auditory stimuli for Experiment 1.

This paper argues that Korean suppletive subject honorification is licensed by adjacency between a verbal root and AgrS[+hon] within a single complex head. The proposal addresses an apparent counterexample from auxiliary verb constructions, where suppletive honorification appears on the lexical verb while the regular honorific suffix -(u)si surfaces after the auxiliary. Building on Generalized Reduplication (Arregi and Pietraszko 2021a,b), I propose that AgrS is base-generated adjacent to the lexical root, thereby licensing suppletion, and is subsequently displaced postsyntactically to satisfy a Korean morphotactic constraint requiring AgrS to follow all roots in the complex head. This analysis derives the auxiliary pattern without abandoning an adjacency-based locality condition. It also preserves the contrast with passive and causative constructions, in which intervening verbal morphology blocks suppletion. The resulting account unifies the distribution of regular and suppletive subject honorification and supports a strictly local theory of contextual allomorphy.

This paper develops a semantic analysis of Japanese expressive small clauses (ESCs)—phrases comparable to English You fool!—within Portner et al.’s (2019) participant-structure framework. We analyze ESCs and the nominal particle -me as conventional-implicature operators that update the hierarchical ranking of individuals. An ESC assigns its referent the lowest rank, whereas -me requires only that at least one individual outrank its referent. This entailment relation explains why -me contributes no distinct status-related content within an ESC but remains meaningful in clausal environments. The analysis also unifies the derogatory third-person and humble first-person uses of -me. We further analyze the honorific suffix -sama as an upward operator that assigns its referent the highest rank. This captures both deferential uses such as Tanaka-sama and self-aggrandizing uses such as ore-sama, while allowing multiple individuals to share the highest rank.

This paper proposes an analysis of Japanese contrastive wa as a part-time outlook marker. Like genuine outlook markers such as nanka and kurai, contrastive wa requires a salient contrasting proposition, exhibits perspective shift under attitude predicates, and can signal conflicts between subjective stances. Unlike those particles, however, wa is not restricted to subjective judgments and can occur in factual environments. To capture this combination of similarities and differences, we formulate an analysis in a model of discourse that integrates Farkas and Bruce’s Table model with Coppock’s outlook-based semantics. We analyze contrastive wa as a discourse-level, focus-sensitive downtoner that replaces a stronger pending proposal on the Table with a weaker prejacent. When the prejacent is a discretionary proposition, focus-induced contrast yields stance-related effects resembling those of genuine outlook markers. The analysis derives the interpretation of embedded wa relative to the attitude holder’s perspective and explains its stance sensitivity in modal judgments.

This paper investigates conventionalized evaluative meaning (i.e., encoding speaker (un)desirability) in conditional expressions through a case study of Japanese conditional expression tewa. Although evaluative meaning in conditionals has received some attention in descriptive work, it has not been accounted for in the formal literature. This paper fills that gap by providing one of the first formal accounts of the evaluative inference associated with tewa. I present novel evidence that, besides undesirability, tewa also conveys a frustrative meaning related to unexpectedness. Moreover, I show that tewa-conditionals, like expressives, are sensitive to the distinction between descriptive and practical contexts; yet differ from expressives in being truth-evaluable. Given these two generalizations, I propose that both the undesirability and frustrative uses of tewa are subject to what I term a weak practicality requirement, and only differ in what Kratzerian conversational background (i.e., priority or stereotypical) is involved. 

This paper investigates adjective truncation in Japanese and identifies three classes of adjectives that systematically resist truncation: adjectives ending in /ii/, adjectives with monomoraic roots, and unfamiliar adjectives. While previous work has attributed the resistance of certain adjectives to semantic factors, I argue that the observed patterns are primarily governed by phonological structure and lexical familiarity. Building on the analysis of internal expressive sentences, I propose that adjective truncation is motivated by the resolution of vowel hiatus and can be accounted for within Optimality Theory using independently motivated constraints. Specifically, truncation is driven by the interaction of ONSET, output-to-output faithfulness constraints, prosodic constraints such as FT-BIN and GRWD=PRWD, and lexically indexed faithfulness constraints. A pilot acceptability study provides preliminary support for the proposed generalizations. The analysis demonstrates that Japanese adjective truncation follows from the interaction of general phonological principles, prosodic well-formedness, and lexical familiarity, without requiring language-specific mechanisms.

Japonic varieties exhibit considerable diversity in their pitch accents, defined here as the tones associated with lexically specified accent positions. However, this diversity has rarely been analyzed within a cross-linguistically comparable framework. This study recasts Japonic pitch-accent systems in Autosegmental-Metrical terms. It shows that the four alignment-defined bitonal pitch accents—L+H*, L*+H, H*+L, and H+L*—correspond closely to the four accent-kernel types developed independently within Japanese linguistics, and that Japonic also attests monotonal pitch accents, H* and L*. Since modern Japonic varieties generally permit only one pitch-accent type within a given system, these types do not contrast paradigmatically and need not be underlyingly specified. Instead, cross-variety differences are analyzed as differences in tone-assignment rules, parameterized by the identity of the starred tone and the presence or absence of default leading and trailing tones. This approach provides a unified typology of Japonic pitch accent and facilitates comparison with general phonological theory.

This study examined whether the selectivity of Korean L-Tensification (LT) extends to nonce words. LT applies only to a subset of eligible words, and this selective application has conventionally been attributed to Sino-Korean (SK) etymology. As an alternative, this study tested whether phonotactic patterns condition LT application. Forty-three speakers and a transformer model trained only on phonemic input-output pairs were tested on the same 144 nonce words, constructed to reflect SK, non-SK, or neutral phonotactics. LT application was quantified using acoustic measures for speakers and confidence scores for the transformer model. Both speakers and the model applied LT selectively. For speakers, non-SK phonotactics suppressed LT in plosive targets, whereas SK phonotactics promoted LT in affricate targets under specific structural conditions. However, the transformer diverged from the speakers, with both phonotactic manipulations lowering LT confidence. These results support a phonotactic contribution to LT selectivity, while the speaker-model divergence motivates future research.

The current study focuses on the second language (L2) acquisition of bi-clausal wh-questions in Japanese by English-speaking learners. While the scope of bi-clausal wh-questions is determined by the position of the wh-phrase in English, in Japanese it is determined by the type of morphosyntactic marker (Complementizer -to vs. Question-marking particle -ka). Considering these crosslinguistic differences between the two languages, we examined how L2 learners acquire the wh-scope-marking strategy in Japanese, adapting the design of previous studies (Lieberman et al. 2006; Umeda 2008). Our results indicate that mastering the interpretive difference between the two morphosyntactic markers is challenging for many English-speaking learners, although a subset of learners was successful. As we did not observe any L1 transfer effects, we speculate that the challenges of mastering the property may come from other factors such as complexity of the structure and lack of exposure to bi-clausal wh-questions in the classroom context.

The Japanese language, throughout its attested history, has a good number of transitive-intransitive verb ‘pairs’. In Old Japanese, the e-endings participate in transitivity alternations, but they apparently exhibit two opposite functions in different verb forms, namely transitivizing and detransitivizing. This paper provides a novel Distributed Morphology-based account of the Old Japanese facts, arguing that the e-endings have an asymmetrical dual status at this stage of the language. While some e-endings involve a v head -y-, which may be directly suffixed to the root, other e-endings manifest a separate suffix -e- which is located higher in the verbal domain. The v head -y- is an intransitivizer which invariably derives intransitive verbs, whereas the higher head -e- is in fact an inner-aspectual element underspecified regarding transitivity, whose real function is to license change of state.

This paper proposes that nominalization structures in early Japanese and Korean display syntagmatic parallels in three important inflections and/or suffixes: a base adnominal inflection *-r (and its agentively-marked variant *-or), a resultative infinitival and substantivizing inflection *-a, and a nominal suffix *ku expressing circumstances that I label ‘absolute’. I show that two of these nominalizations in Proto-Japanese, the fossilized nominal [ADN+*a] and the ku-nominal [ADN+*a+*ku], match part-for-part the construction of complex sequential nominalizations in Proto-Korean, and proposes Proto-Japano-Koreanic reconstructions for these patterns and their constituent elements. In so doing, this paper refutes the strong claim by Vovin that Japanese and Korean share “no common paradigmatic morphology,” and concludes with providing a framework for understanding how morphological mismatches between the two languages might shed light on grammatical changes in their proto-languages.

Western Old Japanese is notable for the formal diversity of its adjectival modification constructions. While past literature has identified the extent of this diversity, there is a paucity of research uncovering its underlying mechanisms. This study aims to address this research gap using a corpus-driven methodology with data from the Oxford-NINJAL Corpus of Old Japanese. It finds that the formally diverse adjectival modification constructions of Western Old Japanese can be placed into two groups based on how they exhibit parallels in their structural (i.e., morphosyntactic) and functional (i.e., semanticopragmatic) properties. This finding has potential implications for comparative research concerning the Japonic relatives of Western Old Japanese.

This paper investigates resultative constructions with intransitive verbs in Korean. It presents novel empirical evidence that unergative and unaccusative resultatives differ in the distribution of result predicates, as shown by their divergent behavior with respect to fronting, right dislocation, and omission. This paper argues that both unergative and unaccusative resultatives involve a small clause as the result clause, yet are best analyzed as having distinct internal structures. The asymmetry is attributed to the presence or absence of a null subject in the small clause: unergative resultatives contain an overt subject, whereas unaccusative resultatives contain a null subject. This study refines the typology of Korean small clauses by providing novel empirical evidence from intransitive resultatives and, on the theoretical side, lends support to cyclic Spell-out and the linearization of small clauses.

This paper argues that ellipsis is licensed morphologically as well as syntactically. It focuses on native Japanese verbs related to the main verb su ‘do’ (√DO), including its potential and honorific suppletive forms. Unlike other native Japanese verbs, these verbs permit an adjunct-inclusive interpretation under ellipsis. To account for this asymmetry, I propose that ellipsis is licensed if and only if the licensing head is not adjacent to a phonologically overt Root, assuming that √DO in Japanese lacks a phonological exponent. This condition restricts VP ellipsis involving native Japanese verbs to those derived from √DO. It also correctly licenses particle-attached VP ellipsis and Sino-Japanese VP ellipsis and explains why English permits vP ellipsis rather than VP ellipsis. Furthermore, the analysis supports a PF-deletion approach to ellipsis over a no-insertion analysis.

This study investigates how Korean what-so (mwue-l kulehkey ‘what-ACC so’) constructions are disambiguated through prosodic and morphosyntactic cues. Previous studies have shown that these constructions permit three non-canonical interpretations (why-question, wh-exclamative, rhetorical question), in addition to the canonical what-question interpretation. Based on an analysis of naturally occurring speech, we demonstrate that cues to these interpretations are found in both morphosyntax and prosody. Morphosyntactically, topic-marked subjects are incompatible with rhetorical question interpretations, while the presence of an additional accusative-marked object favors wh-exclamatives and why-question readings. Prosodically, wh-question interpretations (what- and why-questions) are associated with rising Intonational Phrase (IP) boundary tones, whereas non-wh-question interpretations (wh-exclamatives and rhetorical questions) exhibit falling IP boundary tones. Overall, these findings demonstrate how Korean employs morphosyntactic marking and prosodic boundary tones to resolve interpretive ambiguity in what-so constructions.

Relative clause attachment can be shaped by prosodic and morphosyntactic cues, but how these sources of information interact remains unclear. Korean provides an informative test case because constituent length can influence prosodic grouping, while honorific agreement can grammatically favor attachment to a particular noun phrase. This study examined their interaction by manipulating constituent length and honorific marking in Korean relative clause attachment. The results revealed a strong preference for high attachment and a robust effect of honorific agreement. A length-based prosodic effect emerged only when honorific marking was absent, indicating that length influences attachment when grammatical guidance is weak but does not override morphosyntactic licensing. These findings suggest that prosodic length functions as a secondary, context-dependent cue in sentence comprehension.

This paper aims to deepen our understanding of constructions with a quote followed by an accusative object in Japanese. We propose to argue that the structural configuration of such constructions is analogous to argument doubling in the sense of Saito (2017, 2024). In constructions with argument doubling, two accusative objects are construed as arguments of a single predicate and assumed to receive the same theta-role. It will be shown that syntactic and semantic properties exhibited by constructions with a quote and an accusative object are naturally explained by identifying them as argument doubling. We also explore the implications of our analysis, suggesting that the unavailability of their English counterparts stems from the fundamental differences between Japanese and English—that is, the presence or absence of phi-feature agreement and weak heads comparable to overt case-markers.

This paper re-examines the syntactic properties of Japanese non-coreferential sentence-final Noun-Copula constructions (Mermaid Constructions, MMCs). I first revisit previous analyses of Japanese MMCs by reconsidering the structural position of the nominative subject. Furthermore, I argue for a non-unified analysis of Japanese MMCs by presenting asymmetries regarding the presence/absence of syntactic island effects (examined by long-distance scrambling and PP topicalization) and the (im)possibility of gap-filler dependencies. To account for these asymmetries, I argue that Japanese MMCs involve two syntactically distinct types: one with the MMC Noun (i.e., the ‘Noun’ slot in non-coreferential sentence-final Noun-Copula constructions) functioning as an N head and the other with the MMC Noun functioning as a C head. The proposed non-unified analysis also accounts for the asymmetry concerning the (im)possibility of Nominative/Genitive Conversion in Japanese MMCs.

This paper investigates the sentence-final -ki in Korean, focusing on its distribution and interpretation in the root context. Although -ki has been standardly treated as a nominalizer, its distribution in both embedded and root clauses calls for a deeper analysis beyond its role as a nominalizer. In embedded contexts, it is licensed only under verbs of priority meanings, and in root contexts, it has directive readings. Based on this observation, I argue that when -ki appears in clausal environments, it functions not merely as a nominalizer but also as a mood marker, encoding comparison-based modal meaning

This paper investigates Japanese-speaking children’s comprehension of the negative-sensitive items hitotu-mo ‘(not) even one’ and nani-mo ‘(not) anything,’ which are treated as negative concord items (Watanabe 2004). Previous studies have shown that children interpret nani-mo correctly from an early age (Sano et al. 2009), but they may have difficulty with interpreting the additive particle mo (Kurokami et al. 2021). We therefore examined whether children interpret the minimizer hitotu-mo as adults do, or misinterpret it as the numeral hitotu ‘one’ by ignoring mo. We conducted the Truth Value Judgment Task with 14 Japanese-speaking children (4;4–6;9). The results show that four- to six-year-old children comprehend both hitotu-mo and nani-mo in negative sentences in an adult-like manner.

This study proposes a framework for comparing the historical development of loanword prosody, focusing on loanwords from Middle Chinese in three Sinosphere languages: Vietnamese, Korean, and Japanese. By examining the influence of the source language, its retention in contemporary forms, and the direction of historical change in irregular correspondences, this study reveals divergent prosodic trajectories. Although all three languages originally reflected Middle Chinese tonal categories, their historical developments differ significantly. Sino-Vietnamese words exhibit high retention of Middle Chinese tones with sporadic exceptions. Sino-Korean words maintain substantial source influence while reshaping irregular patterns through analogical change toward a dominant accentual pattern. In contrast, Sino-Japanese words show a structural departure from Middle Chinese tonal categories, developing a novel accentuation system based on mora count in the Modern period. Finally, this study explores typological and sociolinguistic factors, such as the degree of Chinese cultural influence, to account for these divergences.

An investigation of 106 Gwangju Korean Sign Language (KSL) toponyms by Ji (2022) describes most names as bimorphemic compounds indexed to the source Korean names, thus loan translations. Roughly a quarter of the KSL names simply represent the initial source morpheme semantics; however, four names represent only the second source morpheme semantics, apparently flouting a constraint for monomorphemic outputs that requires indexing the initial source morpheme semantics. At JK31, two hypotheses were proposed to explain preference for the second morpheme in the KSL toponym: 1) the greater semantic salience of the second source morpheme makes it favored over the initial source morpheme; and 2) an easily assimilated KSL Hangul character sign initialization indexing the initial source morpheme satisfies a general constraint supporting indexing any feature of the initial source morpheme, thus allowing the second morpheme to appear alone while preventing ambiguity. This work examines these two hypotheses.