Why /r/ is not a special, empty consonant in Japanese Thomas Pellard To cite this version: Thomas Pellard. Why /r/ is not a special, empty consonant in Japanese. Journal of East Asian Linguistics, Springer Verlag, 2016, 25 (4), pp.351-383. 10.1007/s10831-016-9147-4. hal-01356834v3 HAL Id: hal-01356834 https://hal.archives-ouvertes.fr/hal-01356834v3 Submitted on 5 Oct 2016 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Why /r/ is not a special, empty consonant in Japanese 2016, Journal of East Asian linguistics 25(4), doi: 10.1007/s10831-016-9147-4 Thomas Pellard cnrs (crlao) Abstract In recent work on Japanese phonology, /r/ has been argued to be a unique consonant in the Japanese phonological system, characterized by its default, unmarked and featureless nature. These peculiar features have been claimed to derive from the original epenthetic status and the late historical phonologization of /r/ in Japanese. However, a review of all the relevant evidence shows that there is actually no solid basis for that hypothesis, which proves to be not only inadequate, but also directly falsified by the data. No r-epenthesis process can be reconstructed in earlier Japanese, and /r/ cannot be analyzed as a unique, default-empty consonant. Keywords: Japanese; Phonology; Liquids; /r/; Underspecification; Markedness; Epenthesis 1 Introduction oretical and empirical grounds. In the following, I shall reexamine all the arguments brought up by the Hr con- The recent article “The phonology of Japanese /r/: A cerning the phonetic properties, the distribution pat- panchronic account” published by Laurence Labrune in tern, the frequency, the phonological and morphopho- the Journal of East Asian Linguistics (2014, 23: 1–25) ar- nological behavior of /r/, and add additional evidence gues for the unique status of the consonant /r/ in Japa- from L1 acquisition data and the typology of consonant nese. Labrune’s hypothesis of a unique status for /r/ in epenthesis. The evidence purporting to support the Hr Japanese, henceforth Hr, synthesizes and develops pre- does not actually corroborate that hypothesis, and even vious research in both Japanese traditional linguistics sometimes directly refutes it. The evidence adduced by and generative phonology. It comprises several differ- the Hr can be shown to be overstated, not uniquely char- ent claims: /r/ is claimed (a) to be an unmarked conso- acterizing /r/, amenable to alternative explanations, or nant, (b) to lack any phonological feature specification simply inaccurate or wrong. Morever, both the method- at the underlying level, and (c) to be an originally epen- ology and the argumentation suffer from important lim- thetic consonant. itations and inadequacies, in particular the absence of a Together, these three characteristics are taken to systematic comparison with other consonants and the uniquely define /r/, a special default consonant in Jap- scantiness of quantitative and statistical evidence. The anese. The explanation for that unique status is sought hypothesis that /r/ is a special empty consonant in Jap- in diachrony,1 and the Hr proposes that /r/ was not a anese has thus to be rejected, and alternative accounts phoneme in proto-Japanese but the default realization should be preferred. of an empty onset in hiatus position, which was phonol- ogized only later. 2 Phonetic properties The Hr presents many empirical claims about /r/ in Japanese and raises several important issues that need 2.1 Variability to be addressed. Besides its relevance for synchronic and diachronic studies on Japanese phonology, the Hr The Hr first presents phonetic arguments for the unique has also wider typological and theoretical implications featureless status of /r/. Japanese /r/ is said to exhibit a as well. In particular, it should be considered within wide array of contextual, stylistic, sociolinguistic, and di- the broader context of debates in theoretical phonology alectal2 realizations, and that variability is tacitly inter- about underspecification, markedness, and epenthesis. 2 Labrune (2012: 92) states that [ɮ] is the “most common realiza- Nevertheless, the Hr is open to challenge on both the- tion of /r/ in some Ryûkyûan dialects.” To the best of my knowl- edge, there is not a single Ryukyuan variety where this is true, 1 Hence the subtitle “a panchronic account.” The use of and no dialect has ever been reported to possess such a lateral “panchronic” is however idiosyncratic here and does not fricative (Uemura 1997; Karimata 1999; personal fieldnotes). The refer to the program of Panchronic Phonology by Hagège & symbol [ɮ] was sometimes used in an ad hoc manner in older de- Haudricourt (1978), which aims at formulating universal laws of scriptions to transcribe the Southern Ryukyuan fricative vowel sound change. 1 preted as deriving from its lack of phonological specifi- usually considered to target marked segments rather cation. Notwithstanding its alleged variability, it is com- than unmarked ones (Trubetzkoy 1939: 236–237, Hock- monly agreed that /r/ in Japanese is consistently voiced, ett 1955: 166–167, Flack 2007), so that in this case the apical, alveolar or post-alveolar, with a short or weak clo- data would suggest, to the contrary, that /r/ is marked sure (Akamatsu 1997: 105–116, Saitō 2003: 11, Vance 2008: in Japanese. That contradiction needs to be clarified one 89). The existence of such invariant features for /r/ that, way or the other by the Hr. cover laryngeal setting, manner, stricture, and place, di- In any case, the two pieces of evidence presented rectly contradicts the emptiness argued for by the Hr. by the Hr, word-initial prohibition and co-occurrence Another Japanese consonant that should be consid- restrictions, do not affect only /r/, but voiced obstru- ered together with /r/ is /h/. It is important to note ents as well. For the Hr, the existence of phonotactic that the consonant /h/ (< *p) has actually more con- constraints targeting /r/ demonstrates its unique un- textual allophones than /r/,3 since it can be realized as marked status, while the fact that the same constraints [h], [ɸ], [ç], or even [x], [], or [ɦ], depending on the apply to other consonants (the voiced obstruents) is context (Saitō 2003, Labrune 2012: 69). Some abstract deemed irrelevant, because those consonants are con- analyses even include [p] as an allophone of the same sidered to be “heavily marked” (Labrune 2014: 15, fn. 11). phoneme, especially in geminates (McCawley 1968: 77– This reasoning is a logical fallacy, both circular and self- 78, Labrune 2012: 70). If we follow the logic of the Hr that contradicting. For the sake of discussion, I will neverthe- “a wide range of phonetic realizations is suggestive of less examine below the evidence adduced by the Hr in the unmarked status” (Labrune 2014: 3) of a consonant, more detail. then /h/ should be considered as even less marked than /r/, which runs counter to the Hr. 3.1 Word-initial prohibition 2.2 Other characteristics In Old Japanese4 (OJ) “Yamato” native words,5 /r/ is banned from word-initial position, while the “empty on- The description of /r/ as “the shortest of all Japanese con- set” or “zero consonant” /’/ (Labrune 2014: 5–6)6 is re- sonants” (Labrune 2014: 3) is a truism: The basic real- stricted to that position, with no word-internal vowel se- ization of /r/ is a tap, i.e., a “momentary” (Catford 2001: quence allowed (#’V.CV, †V.’V, V.rV, †#rV).7 Though the 67) sound defined by a “short closure” and a “brief con- Hr does not mention it, this is not an absolute rule since tact between the articulators” (Ladefoged & Maddieson there are several exceptions. Most Modern Standard Jap- 1996: 230–231). Whether /r/ is the shortest consonant of anese (MSJ) exceptions result from the historical loss Japanese also depends on one’s stance about the zero of an intervocalic consonant, but we already find in OJ consonant /’/ posited by the Hr (see below): If /’/ has no forms with word-internal vowel sequences such as kai substance and is empty, one wonders what its duration ‘oar,’ kui ‘regret,’ ko2i ‘lie down (infinitive),’ oi ‘age (infini- is, and whether /r/ can be said to be shorter than zero tive),’ uuru ‘to plant (adnominal)’ (Vovin 2009: 420–422, (/’/). Frellesvig 2010: 39).8 Though /r/ is said to fail to exhibit “any significant in- Even if we set aside such counter-examples, contrary fluence on the neighboring segments” (Labrune 2014: 3), to the statement by the Hr (Labrune 2014: 5–6), the ab- no actual data is presented and no comparison is made 4 I will follow common practice (Martin 1987; Vovin 2005a, 2009; with other consonants. That claim is thus unsubstanti- Frellesvig 2010) and use “Old Japanese” instead of Labrune’s ated. (2014) “Ancient Japanese,” and I will restrict its usage to refer to the language of the 8th century only. 3 Distribution pattern 5 Sino-Japanese loanwords from the 8th century (Frellesvig 2010: 284–286, not the 4th or 5th contra Labrune 2014: 2, 21) onwards and contemporary loanwords do show initial /r/. The distribution pattern of /r/ in Japanese is notewor- 6 The Hr refers to similar analyses by other scholars, but actually thy for its well-known restrictions. Though the Hr in- Hattori (1961) defines /’/ as a voiced laryngeal and not as a zero, terprets these constraints as evidence for the unique Wenck (1966: 15) as a smooth onglide, and McCawley (1968: 88– unmarked status of /r/, distributional restrictions are 93) uses /ʔ/, which he treats as a glide.