15 Historical Phonology

Tomoaki Takayama
15 Historical phonology
1 Introduction
Historical background is significant and useful for direct observation of language
structure. This chapter provides information on historical issues helpful to under-
standing the synchronic aspects of the phonetics and phonology of modern Tokyo
Japanese.
Due to a lack of established cognate relations with other languages or linguistic
families, reconstruction of prehistoric stages of Japanese is quite often difficult.
Thus, historical studies of the language are inevitably different from the studies of
other languages, such as the Indo-European languages, which use the comparative
method as a central tool. We have to depend largely on internal resources in the his-
torical studies of Japanese. Fortunately, the historical stages of Japanese are attested
back to about the eighth century, with a large amount of writing materials and dia-
lectal information.
In order to shed light on what historical studies reveal about the structure of
modern Japanese, this chapter addresses several historical issues and gives a review
of the results and the points of controversy from studies over the last few decades.
Note that it does not present a comprehensive review of the field nor a comprehen-
sive outline from the earliest period to the present (see Frellesvig 2010 as well as the
chapters in the History Volume of this handbook series for full discussion of the his-
torical phonology of Japanese). It briefly mentions technical treatments of writing
system and philological problems only to the extent these are helpful to understand-
ing historical phonology.
Moreover, the focus of this chapter is on segmental aspects of phonology, not on
suprasegmental or prosodic features such as tone (i.e., accent). Due to lack of suffi-
cient historical data, the tonal system of modern (Tokyo) Japanese cannot be easily
traced back to earlier stages, even though the tonal system of the Kyoto dialect can
be attested in the twelfth century in written materials (see Kubozono 2012 and
Uwano 2012 for the variety of pitch accent systems in the dialects). Intonation also
complicates the situation. The tonal history of the compounds in modern Japanese is
not yet well understood. For these reasons, prosodic aspects are not dealt with in
this chapter, even though they are undoubtedly an integral part of the phonetic and
phonological studies of Japanese.
This chapter is organized as follows. Section 2 discusses various issues relating
to voiced obstruents, with main focus on the velar nasal [ŋ] and prenasalization
in the consonantal system. Section 3 deals with the historical backgrounds of the
affricates, [ts], [tʃ ], [dz], and [dʒ], as well as the asymmetry these sounds exhibit in
modern Japanese. In section 4, we discuss some problems concerning reconstruction
Brought to you by | Universität Osnabrück

Authenticated
Download Date | 2/28/16 8:18 PM
622 Tomoaki Takayama
to point out that the phoneme /s/ of modern Japanese was realized as affricates in
old Japanese. Section 5 briefly presents a history of the phonemes /h/ and /p/, based
on recent research in this area. The final section (section 6) summarizes the main
points of the chapter, and mentions residual issues for future studies.
In this chapter, the IPA symbol [u] is used to refer to the various phonetic real-
izations of the vowel /u/ of Japanese, and therefore does not exclude the unrounded
realization that is represented by [ɯ] in regular usage. This broader representation
is quite often adopted in the literature of historical phonology, since the accurate
phonetic values cannot always be defined when discussing historical issues. The
symbol [ɯ] is used in this chapter only when it is necessary to specify the un-
rounded realization.
2 Historical issues about voiced obstruents

When we discuss the Japanese consonant system and its history, we must define
terminology. The English term “voiced obstruents” commonly used to refer to the
phonemes such as /b/, /d/, /g/, and /z/ of modern Japanese is itself problematic,
especially with respect to their phonetic realizations. In Japanese linguistics, what
we call voiced obstruent is generally referred to as a daku-on consonant (see Kubo-
zono’s introduction to this volume). While the Japanese term is not well known
except to specialists, this label is convenient for identifying a specific category in
the consonant system without suggesting any kind of phonetic realization such
as prenasalization. In modern Tokyo Japanese, daku-on consonants are largely
pronounced as voiced obstruents, but this is not necessarily true of other dialects.
Similar attention is required when comparing different historical stages of a dialect
spoken in a specific area. For example, in the history of daku-on in the Kyoto dialect,
it is necessary to refer to the category without phonetic specifications, as discussed
in the following sections. In fact, one of the important issues about the consonants
of this dialect is the de-prenasalization in the daku-on series: most members of this
series pronounced as prenasalized consonants changed to the plain voiced obstruents.
Even in modern Japanese, which is our main concern in this chapter, we face prob-
lems with phonetic realizations. The standard pronunciation of /g/ in non-initial
position is the velar nasal [ŋ] (see section 2.1 for details). In these cases, the term
“voiced obstruent” is not completely phonetically accurate, and should be avoided.
However, since it is hard to find any other suitable English term, we will continue to
use it, in a categorical sense, with no phonetic specification.1 When clarification
is required during discussion, we will add a modification, for example, such as
‘modern voiced obstruents’.
1 Frellesvig (2010: 34–36) refers to the distinction between sei-on and daku-on in Old Japanese as
tenuis (tense) versus media (lax), avoiding terms such as voiceless and voiced. Whether the tenuis
were allophonically voiced in intervocalic position is a controversial question (see note 10 and Hayata
1977).

Authenticated
Historical phonology 623
2.1 Variations of /g/

Modern Tokyo Japanese has the velar nasal [ŋ] among the variations of the phoneme
/g/, as mentioned above. From the prescriptive standpoint, this nasal sound has been
regarded as one of the significant elements that characterize the standard pronuncia-
tion of Japanese for a long time. The instructions to carefully keep on pronouncing the
nasal for the non-initial /g/ have been repeatedly given and emphasized especially in
broadcasting, singing, and classes in schools, although nowadays such prescriptive
manifestations are not observed quite as often as before (see Vance 1987: 108–132;
Vance 2008: 214–222; and Shibatani 1990: 171–173 for details). In this section, we
will focus on the synchronic and diachronic aspects of the allophonic nasal sound.2
To begin with, we show the variations of /g/ in modern Tokyo Japanese. The
initial allophone is exemplified in (1). For non-initial positions, there are two cases:
the intervocalic position as shown in (2) and (3), and the postnasal position as
shown in (4).
(1) gomi [ɡomi] ‘trash’
(2) tamago [tamaŋo] ‘egg’
(3) kagi [kaŋi] ‘key, lock’
(4) ringo [riŋŋo] ‘apple’
The oral stop [ɡ] occurs in the initial position, whereas the nasal [ŋ] appears in non-
initial positions. The oral and the nasal occur complementarily according to the
position. As opposed to /g/, other voiced obstruent phonemes shown in Table 1
have no nasal variants. That is, /b/ is not phonetically realized as [m], and the
phonemes /d/ and /z/ are not realized as [n]. Among the four voiced obstruents in
Table 1, /g/ is unique in having the nasal realization.
Table 1: Voiced obstruents and nasals in modern Japanese
labial dental velar
voiced obstruents b d g
z
nasals (nasal onsets) m n
(other phonemes omitted)
2 There is a well-known controversy in the literature about whether the difference between [ɡ]
and [ŋ] should be treated as allophonic or phonemic. In this chapter, we will not elaborate on this
problem nor on the morphophonological issues involved (see Vance 1987: 108–132; Komatsu 1981:
137–148; among others for surveys and comments).

Authenticated
The difference in the behavior of /g/ and other voiced obstruents can be easily ex-
plained by the absence of the nasal /ŋ/ in the onset. In the labial and dental places,
the nasal onset /m/ and /n/ exist, respectively. If each phoneme shares the same
place of articulation with its nasal counterpart, it does not allow for any nasal variant.
Since there is no gap in any other place than the velar, there is no room for /b/, /d/,
and /z/ to be realized as nasal variants. On the other hand, /g/ permits the variant
[ŋ] to fill the accidental gap in the nasal place: that is, the gap allows nasal realiza-
tion of the phoneme /g/.
However, the allophonic range between the oral sound and the nasal is not
straightforwardly understandable in terms of phonetic contextual effects. We cannot
find any reason why nasality should be introduced to the realization of the non-
initial /g/ (Hattori [1957] 1960: 338–341; Vance 1987: 111–112). Of course, in the case
of the postnasal position such as ringo in (4), the nasal variant is likely to occur
due to assimilation to the preceding nasal consonant. In the intervocalic environ-
ment, by contrast, a variant need not be a nasal sound. It is hard to understand
why the nasal allophone [ŋ] occurs synchronically. Instead, a plausible allophone is
the fricative [ɣ] because the intervocalic position often causes spirantization. Actually,
we quite often observe this fricative rather than the nasal [ŋ] in the present pronun-
ciation of Tokyo Japanese. For example, tamago in (2) is realized as [tamaɣo], and
kagi in (3) is as [kaɣi] (Vance 1987: 111–112; Kindaichi 1942; among others). In order
to understand the intervocalic nasal realization in modern Tokyo Japanese, it is
necessary to look at the diachronic background.
2.2 Diachronic background of velar nasal

Due to scarcity of resources, it is not feasible to trace the history of the Tokyo dialect
back more than several hundred years. As for the Kyoto dialect, it is readily known
that the modern voiced obstruents were realized as prenasalized consonants in the
beginning of the seventeenth century, as will be discussed in section 2.3. A question
arises about whether Kyoto is a special case or if similar situations are observed
generally in Japanese.
When we look at the realization of /g/ in the consonant systems of the dialects,
we quite often encounter nasal variants. While a number of dialects have no nasal
but oral [ɡ] or [ɣ], sounds such as [ŋ] and [ŋɡ] are widely observed throughout Japan.
Thus, the prenasalization in the Kyoto dialect of the seventeenth century is not
idiosyncratic. Inoue (1971) discusses issues of the phoneme /g/ in various dialects,
analyzing the nationwide geographical distribution and providing a diachronic per-
spective. Synthesizing the geographical information and the historical facts demon-
strated by the document resources, Inoue (1971) concludes that the great majority of
dialects can be interpreted by two events and their relative chronological order: (i)
the emergence of the variant [ŋ], and (ii) the de-prenasalization, i.e., the vanishing

Authenticated
of the nasal element of prenasalized consonants. If [ŋ] emerged in advance, that

variant is expected to remain even if the de-prenasalization were completed in the
following stage. By contrast, if the de-prenasalization took place in advance, i.e., if
the prenasalized [ŋɡ] changed to the oral [ɡ] or [ɣ]; the nasal [ŋ] cannot emerge in
any following stage, except for the case of contact with another dialect. To sum up,
the allophonic [ŋ] may be derived from the prenasalized realization of /g/, in a way
that it fills the velar nasal gap in the consonant system, as we saw in Table 1 (see
Kamei 1956 for historical discussions on the variation of /g/). Consequently, the
nasal [ŋ] in Tokyo Japanese can be regarded as a relic of the phonetic quality in the
earlier stage of the consonant system.
2.3 Attestation of prenasalized consonants

The prenasalization in the Kyoto dialect is attested in historical resources. The most
significant document is a handbook of the Japanese language for foreign learners,
Arte da lingoa de Iapam, written by João Rodriguez, and published in Nagasaki by
the Society of Jesus in 1604–1608. According to his instruction on the Romanization
of Japanese (Rodriguez 1604–1608: 177–178 and Doi 1955), the vowel letters such as
i, e, a, o, and u followed by the letters d and g should be pronounced with the half
size of the til(tilde) ‘˜’, which is considered to suggest prenasalization (Hashimoto
1932). His instruction also says that the learners must not pronounce them with the
distinct ‘til’ (Rodriguez 1604–1608: 172). For example, the spelling toga ‘offense, sin’
should not be pronounced as tonga equivalent to tõga including the complete tilde
that indicates the coda nasal. Since the relationship between the orthography and
the sounds is complicated, we summarize the correspondences in (5).3
(5) a. The letter d is used in de, da, do, and dzu which represent the Japanese
sounds [de], [da], [do], and [dzu], respectively. Note that the absence
of [du] and [di] is due to the affrication (see section 3.1). For [ʤi], it is
represented by the spelling gi in (5b).
b. The letter g is used in gui, gue, ga, go, gu, and gi which represent the
Japanese sounds [ɡi], [ɡe], [ɡa], [ɡo], [ɡu], and [ʤi], respectively.
According to (5), the prenasalized realization occurs in the stops and the affri-
cates such as [d], [ɡ], [dz], and [ʤ]. As for the letter b that corresponds to the pho-
neme /b/, Rodriguez remarks that the half size of the tilde is observed in some cases,
but it is not so common compared to the extent of the letter d and g. His comment
suggests that the prenasalization in the labial was weaker than in the dental and
the velar. Furthermore, in the revised concise version Arte breve da lingoa de Japoa
3 The romanization of Japanese in works by the Society of Jesus is based on the orthography of
Portuguese.

Authenticated
published in 1620, Rodriguez mentions that the half size of the tilde was occa-
sionally observed before the letters j and z which correspond to the phoneme /z/
(Rodriguez 1620: 12). By contrast, there are some domestic resources suggesting that
/z/ lacked prenasalized realization (see section 3.3).4
Rodriguez also refers to the fact that the letter g in the Bizen dialect lacked the
half size of the tilde and was pronounced secamente ‘with a dried sound’ (Rodriguez
1604–1608: 171). The sound is considered to be the non-nasalized realization [ɡ] or
[ɣ] of /g/. Bizen is a part of what is now Okayama Prefecture, situated in the outer
area adjacent to Kinai, the central region where the capital city Kyoto was centrally
located. His comment on the dialect reveals an interesting sociolinguistic aspect
concerning the velar sound. It presumably suggests that the habitants of the capital
were sensitive to the rural accent and that the lack of prenasalization made a harsh
auditory impression on them.
A subsequent stage of the Kyoto dialect is demonstrated by domestic documents
written about one hundred years later after Rodriguez (1604–1608). The instructions
in these documents advise the readers to keep the prenasalization in /di/ and /du/
(see section 3.3). It reveals that de-prenasalization had already taken place. More-
over, in those texts, the authors applied the terms originally referring to the coda
nasal /N/ to the nasal element of prenasalization. A representative of these docu-
ments is Ikeisai Kōgo Kikigaki, which is assumed to date back to the beginning of
the eighteenth century. It instructs that the word midu (/du/ was realized as [dzu],
see sections 3.1 and 3.3 for the phonetic realization), which is the earlier form
of mizu ‘water’ in Modern Japanese (ModJ), should be pronounced with a shorter
coda nasal /N/ inserted before /du/. This instruction in the manuscript Ikeisai Kōgo
Kikigaki resorted to the Kana script ‘ん’, which normally represents the coda nasal,
in order to refer to the nasal element that should be inserted: “midu ‘water’ should
be pronounced in a similar way to ‘みんづ’ <mi-N-du>”.5 It is natural that the elder
generation who originally maintained the prenasalization perceived it not as two
successive sounds but as an inseparable sound [nd]. Stated conversely, the younger
generation who had not acquired the prenasalization perceived it as the sequence of
a coda nasal and a plain voiced obstruent instead of an inseparable consonant. The
notation of Ikeisai suggests the recognition by younger generations (T. Takayama
1998).
2.4 Velar nasal in the past

Although the status of the velar nasal as standard has not yet been lost, its prescrip-
tive restraint is not imposed nowadays as strictly as before.6 Based on his investiga-
4 Yamane-Tanaka (2005) discusses significant aspects of the de-prenasalization in the framework of

Optimality Theory.
5 The purpose of this instruction is to preserve the traditional recitation style of waka poems.
6 On the basis of his observations, Vance (1987: 111) pointed out that the difference between the
prestige norm of [ɡ] and the official status [ŋ] can explain the preference of native speakers.

Authenticated
tion, Kindaichi (1942) pointed out that the younger generation was going to lose the
velar nasal, and predicted that the sound would vanish in the future, no matter how
much the non-nasal [ɡ] was corrected in language education.7 In fact, a half century
later, the population maintaining the velar nasal allophone decreased by half in
the past several decades (Hibiya 1988, 1995, 2002; Inoue 1983, 1998: 162–167; among
others).
Transition from the velar nasal to the oral stop may be inevitable in the con-
sonantal system of Japanese, but, at the same time, we cannot discount the fact
that the nasal variant has lasted a long time. This fact must be discussed from a
morphophonological (see note 2; Komatsu 1981: 145–148; Ito 1997) or sociolinguistic
perspective.
As for the sociolinguistic viewpoint, fortunately, we have a historical resource
that tells us about an older situation in Tokyo (Edo) Japanese. The comic tale Ukiyo
buro, which was published in the beginning of the nineteenth century, depicts vivid
conversations among ordinary people enjoying a public bath. In one of the scenes,
an attendant of the bathhouse speaks in his rural dialect, contrastively different
from the urban speech. The author Shikitei Sanba not only adopted the rustic vocab-
ulary but utilized some speech sounds in order to emphasize the rural character. For
that purpose, he marked the kana scripts representing the syllables with the velar
/g/ with a special diacritic, i.e., a small circle instead of the usual two dots dakuten
that indicates /g/. It is assumed that the special circle represents the non-nasal [ɡ] or
[ɣ] that might cause a harsh sensation or an unsophisticated impression to the urban
native speakers.8 This fact should be taken into consideration when we discuss the
background of the status of the standard nasal [ŋ] in modern Japanese.
2.5 Prehistory of voiced obstruents

The works by Joan Rodriguez provide direct evidence for prenasalization in the
beginning of the seventeenth century, as mentioned in section 2.3. However, it is
difficult to find documents directly indicating the phonetic value in earlier stages.
Nevertheless, on the basis of indirect resources and lack of contradicting evidence,
it is generally assumed that voiced obstruents were largely realized with prenasali-
zation even in Old Japanese (M. Takayama 1992, 2012: Ch. 3).
In prehistoric stages, we can do nothing but depend on hypothetical approaches,
since there is no concrete evidence available to the reconstruction of the phonetic
values. As far as the phonotactics is concerned, there is a noticeable fact that they
occur in the native lexicon under the two distributional restrictions shown in (6).
7 Shibatani (1990: 171) remarks on the significance of Kindaichi’s (1942) contribution to socio-
linguistic studies.
8 Sakanashi (1975) claims that the special diacritic does not represent the oral sounds of /g/.

Authenticated
(6) a. Voiced obstruents do not occur in word-initial position.

b. There is a maximum of one voiced obstruent per morpheme.
These phonotactic restrictions in modern Japanese are well known in the literature,
and they operated even in the eighth century, the earliest period to which we
can date back by documental resources (see Ito and Mester 1986 for a theoretical
discussion; Kamei 1970a; Morita 1977; Yamaguchi 1988 for historical discussions).
Therefore, these distributional facts should be taken into consideration even in the
hypothetical approaches. A plausible scenario is that such phonotactic restrictions
come from some diachronic processes. In addition, the genesis of these distribu-
tions is considered to relate to the history of the voiced obstruents, i.e., prenasalized
consonants.
Another important fact concerning these restrictions is that they are closely
related to the mechanism of sequential voicing known as rendaku (see Vance, this
volume, for full discussion). In other words, the phonotactic restrictions in (6) not
only govern each simplex word in the native lexicon but also play significant roles
at the morphophonological level. Actually, rendaku formation quite often took place
even in the eighth century. We briefly show the relationship between rendaku forma-
tion and phonotactic restrictions, with modern compounds in (7) and (8). The same
mechanism is true of the attested earliest stages.
(7) a. ude b. tamesi c. ude-damesi

arm try.GER arm-try.GER
‘arm’ ‘trial’ ‘trial of skills’
(8) a. ude b. kurabe c. ude-kurabe

arm compete.GER arm-compete.GER
‘arm’ ‘competition’ ‘competition of skills’
According to Komatsu (1981: 101–107), rendaku formation is established on the basis

of the restriction in (6a). Namely, the non-initiality provides the following two merits
in rendaku formation. First, the simplex tamesi in (7), which is subject to the restric-
tion in (6a), changes to the form damesi when placed in non-initial position in
the compound in (7c). The alternation to the voiced obstruent /d/ shows that the
morpheme {tamesi} in the compound in (7c) is not in word initial position any
longer. Note that the scope of (6a) is the word, not the morpheme, and therefore,
(6a) is valid not only for simplex but also for compound words. Second, there is no
minimal pair between a voiceless obstruent and its voiced counterpart in the initial
position due to the restriction in (6a). Therefore, this gap makes easy recovery to an
original simplex from the form initially voiced by rendaku.

Authenticated
The phonotactic restriction in (6b) blocks rendaku voicing. This process is well
known as Lyman’s Law or Motoori-Lyman’s Law: for details, see Vance (this volume);
Vance (1987: Ch. 10); Yamaguchi (1988); Ito and Mester (2003); van de Weijer, Nanjo,
and Nishihara (2005); among others. If a morpheme has a voiced obstruent in medial
position, rendaku voicing does not occur in the morpheme, as shown in (8). The
simplex kurabe in (8b) has one voiced obstruent, and therefore cannot undergo
rendaku, even if this form follows the lexical element ude in (8a), as in the com-
pound ude-kurabe in (8c). By contrast, since the simplex tamesi has no voiced ob-
struent, as shown in (7), it can and actually does undergo rendaku voicing, as in
the compound ude-damesi in (7c).
In sum, the distributional properties of voiced obstruents have a close relation-
ship with the morphophonological aspect, which should date back to the prehistoric
period. Hizume (2003) proposes a scenario in which the Japanese consonantal sys-
tem had only one series of obstruents at some earlier prehistoric stage that later
bifurcated. According to this view, the prenasalized consonants arose in the initial
positions of the second lexemes in compounds in order to denote concatenation
and demarcation in compounds; and, thus, the first stage of rendaku emerged.9 On
the other hand, the plain obstruents were phonetically voiced in word-medial posi-
tion, and their voiced realization was different from prenasalization from a func-
tional viewpoint, while the voiced realization later became weakened in the dialect
of Kinai, the central area (cf. Hayata 1977).10 In this way, when we discuss the history
of the voiced obstruents in Japanese, the focal point is how we deal with rendaku
and the distributional properties in (6).
3 Affrication and merger

3.1 Affrication
When we look at the dental stops in modern Japanese, [t] and [d] do not occur
before the high vowels /i/ and /u/ in native and Sino-Japanese (henceforth SJ)
words; in other words, there are no syllables such as [ti], [di], [tu], and [du]. In the
positions preceding these two vowels, affricates occur instead. These distributional
gaps in the dental stops result from a historical change that took place in approxi-
mately the sixteenth century. Concretely, before the front vowel /i/ and the palatal
9 Hizume (2003) does not completely attribute the prenasalized consonant occurrences in the native
lexicon to the genesis of rendaku. He proposes the historical stratification of the prenasalized con-
sonants (i.e., voiced obstruents) in the native lexicon. According to this proposal, there are prenasalized
consonants that emerged after the bifurcation, in addition to their predecessors.
10 Hayata (1977) argues that the voiceless obstruents (in a categorical sense) of the Kinai dialect
were phonetically voiced in intervocalic positions even in the Heian period, probably around the
eleventh century (see also M. Takayama 1992, 2012: Ch. 3; Frellesvig 2010: 34–36).

Authenticated
glide /j/, the stops [t] and [d] changed to [tʃ ] and [dʒ], respectively. In addition,
before the back vowel /u/, [t] and [d] changed to [ts] and [dz], respectively. As
a result, in modern Japanese, /ti/ is phonetically realized as [tʃi], and /tu/ as [tsu]
(~[tsɯ]). Kim (2001) and Lin (2011) mention these Japanese affricates in their
cross-linguistic discussions, but they do not consider the backgrounds particular to
Japanese. We discuss some details of the affrication with focus on the historical and
structural contexts in which this sound change took place.
The affrication in question is noteworthy with respect to the environments under
which it took place. If we treat this change uniquely, we face difficulty in terms of
the phonetic motivation. Generally speaking, the frication before the front vowel
(and the glide /j/) is likely triggered by palatalization due to assimilation to the
following vowel /i/. However, this explanation cannot apply to the affrication before
/u/. Accordingly, we would need to deal separately with these two affrication paths
in phonetic discussions.
As for the unrounded realization of the vowel /u/ (see Kubozono’s introduction
to this volume), we notice that /u/ shows a notable tendency to centralization when
following non-palatal sibilants such as [s], [z], and [ts] in modern Japanese. (Note
that the difference between the voiced fricative [z] and the affricate [dz] is not dis-
tinctive in modern Japanese: see section 3.2 for historical background). /u/ in those
contexts is often described as [ɯ̈ ] in IPA, such as [sɯ̈ ], [zɯ̈ ] and [tsɯ̈ ]. If the central-
ization can also trigger frication, this may explain the affrication before /u/.
At the same time, however, it is necessary to figure out why each phonetic con-
dition simultaneously caused different types of affrication in the history of a single
language. If the two events did not take place accidentally in the same period, we
must consider other aspects in order to understand the events as one and the same
process. Although the affrications of the dental stops ([t] and [d]) had different pho-
netic motivations, it is not realistic to deny the uniformity of a historical event.
In modern Japanese, the difference between the two affricates [tʃ ] and [ts] is not
distinctive, at least in the native vocabulary, since their choice always depends on
the following vowel: /i/ or /u/ (see Kubozono Ch. 8, this volume, for exceptions in
loanwords and Pintér, this volume, for the introduction of new sound sequences in
modern Japanese). As opposed to [ts], the palatalized realization [tʃ ] is due to regres-
sive assimilation to the front /i/ and the glide /j/ (Hattori [1955] 1960: 288, [1956]
1960: 321–322).11 The difference between [tʃ ] and [ts] is regarded as phonemically
redundant. However, the practical role that the consonantal difference fulfills in
11 Hattori (1955, 1956) describes the affricate [tsu] as /cu/, and [tʃi], as /ci/; namely, he regards both
sounds as derived from the phoneme /c/ on the grounds that the latter is palatalized due to the
assimilation of the following front vowel /i/ and the glide /j/. On the other hand, he argues that
since the difference between the stop [t] and those affricates cannot be straightforwardly explained
by the phonetic environmental conditions, the two phonemes /t/ and /c/ are required in the descrip-
tion of the modern consonant system.

Authenticated
distinguishing between the syllables /ti/ and /tu/ should not be underestimated. The
qualitative difference between these consonants provides an effective phonetic cue
to the recognition of the resultant syllable. In addition, the high vowels are quite
often dropped, due to high vowel devoicing (see Fujimoto, this volume, for details).
Vowel deletion (or devoicing) frequently occurs in /ti/ and /tu/, too. For example,
tikai ‘oath’ is usually realized as [tʃkai] or [tʃi ̥kai], and tuta ‘ivy’, as [tsta] or [tsɯ̥ ta].
In such devoiced realizations, the fricative part, such as [ ʃ ] and [s], is an indispens-
able element for recognizing which syllable is intended, /ti/ or /tu/, since there is no
vowel that can carry out the distinctive function. These fricative parts, rather than
the vowels, play the essential discriminating roles. Even in words where the vowel
/i/ or /u/ is not completely dropped, the qualities of the fricative parts provide
an important cue to the distinction between /ti/ and /tu/ (see section 3.2 for voiced
obstruents). According to the phonemic interpretation (Hattori 1955, 1956), the con-
sonantal difference between the palatal sound [tʃ ] and the non-palatal [ts] is re-
dundant, since it is automatically determined by the vowel that follows it. However,
the contribution of fricative parts to the distinction should be regarded as vital.
A similar situation is observed in the differences between the palatal and dental
fricatives, such as between [ ʃ ] and [s] as well as between [ʒ] and [z]. The two sounds
of each set are allophones in the phonemic treatment, since the palatals, [ ʃ ] and [ʒ]
always accompany the front vowel (and the glide /j/) and the non-palatal [s] and [z]
do not occur before the front vowel. Nevertheless, the difference between palatal and
non-palatal consonants is significant, notably when the vowels are dropped or de-
voiced. For example, sita ‘tongue’ is usually realized as [ ʃta] (or [ ʃi ̥ta]); and sukiyaki,
as [skijaki] (or [sɯ̥ kijaki]).
The function that the fricative parts carry out in modern Japanese is noteworthy
also in the discussion on historical affrication. Of course, it is difficult to demon-
strate vowel devoicing or weakening by historical resources, but we refer to the com-
ments on the pronunciations of Japanese in Ars Grammaticae Iaponicae Linguae
of Didaco Collado, published in 1632 (see Ōtsuka 1957). It says that the vowels i
and u in word final position is hardly audible for beginners; for example, the word
“gozàru” (gozaru ‘stay.HON’) sounds like “gozàr”, the word “fitòtçu” (ModJ. hitotu
‘one’) like “fitòtç”, and the words “àxi no fàra” (ModJ. asi no hara ‘field of reeds’)
like “àx no fàra”. The latter two illustrations, in which the final /u/ or /i/ is deleted,
are particularly interesting for our discussion here. Although second language re-
sources need to be dealt with carefully, they may indicate devoicing or weakening
of vowels in Japanese in the seventeenth century.
The emergence of the fricativisations in /ti/ and /tu/ (as well as /di/ and /du/)
should be investigated taking into consideration phonetic cues, as mentioned above.
According to T. Takayama (2006, 2009), this change is recognized as a historical
trend toward activating the potential contrast in quality between palatal and non-
palatal consonants. Without considering such phonetic differences, we cannot treat
the twofold phenomenon that consists of the two processes, [ti]>[tʃi] and [tu]>[tsu]

Authenticated
(as well as [di]>[dʒi] and [du]>[dzu] in the voiced counterparts) as a unique historical
event. The status or role played by the phonetic difference between palatal and non-
palatal consonants should be further discussed from both synchronic and diachronic
viewpoints in the future.
The affrication in question involves a chronological problem, too. As mentioned
above, the two processes of affrication took place simultaneously, or more precisely,
at almost the same time. Looking into the details, the affrication in /ti/ and /di/
is presumed to have slightly preceded the affrication in /tu/ and /du/. Historically,
following these two affrications, two pair mergers took place in the voiced obstruents.
Namely, the two oppositions, /di/:/zi/ and /du/:/zu/, disappeared at the next stage
(see section 3.2). There is evidence to support the claim that the merger of /di/ and
/zi/ took place a little earlier than that of /du/ and /zu/. We find three types of
systems among the various dialects with regard to these two oppositions: (i) no
oppositions are maintained, as seen in modern Tokyo Japanese; (ii) both oppositions
are maintained, as observed in the Kōchi dialect; and (iii) only /du/:/zu/ is main-
tained, as in the Ōita dialect. In contrast, we do not find a fourth type where only
/di/:/zi/ is maintained (Itoi 1962; Kuno et al. 1995; Kuno 2006; Sugimura 2001; among
others). In addition, in the dialects which still preserve the opposition(s), it is often
observed that the affrication in /tu/ and /du/ is not complete, compared to that in
/ti/ and /di/. If it is true that there was a time lag between the two affrications, we
suggest that the affrication triggered by palatalization promoted the other affrication
(T. Takayama 2009). Such a time lag is remarkable even considering general tendencies
about affrication. Further cross-linguistic investigations into similar cases are needed.
As for the chronological problem, another question arises: why did this change
take place around the sixteenth century? This question has a close relation to the
issues regarding the phonetic development of /s/. It is generally assumed in the
literature (see section 4) that the phoneme /s/, and probably /z/, were realized as
affricates, at least in part, in earlier stages. The transition in /s/ and /z/ from the
affricate to the fricative must have taken place at some chronological point before
/t/ and /d/ were affricated before /i/ and /u/. Ogura (1998) discussed the relevant
chronological issues both from the structural and diachronic viewpoints.
3.2 Merger
There is an asymmetry in the sibilants between the voiceless and voiced series.
While the difference between the voiceless fricative and the voiceless affricate is con-
trastive, as illustrated by the minimal pairs in (9) and (10), the voiced obstruents
lack the contrast between the fricative and the affricate. This difference is summarized
in Table 2: Parenthesized palatal sounds such as [ ʃ ], [tʃ ], [ʒ], and [dʒ] occur before
the front vowel /i/ or the glide /j/.

Authenticated
(9) a. siru ([ ʃ ]) : tiru ([tʃ ])

‘learn, know’ ‘scatter, fall’
b. husi : huti
‘node, joint’ ‘edge, brim’
(10) a. sumi ([s]) : tumi ([ts])

‘corner’ ‘guilt, sin’
b. hanasu : hanatu
‘speak, talk’ ‘shoot (an arrow or a bullet)’
Table 2: Asymmetry between voiceless and voiced obstruents in modern Japanese
fricative affricate difference between

fricative and affricate
voiceless obstruent [s] ([ʃ]) [ts] ([tʃ]) contrastive

voiced obstruent [z]([ʒ]) [dz]([dʒ]) non-contrastive
The asymmetry results from the mergers, as discussed in section 3.1. As far as the
Kyoto dialect is concerned, the contrast between the two kinds of voiced obstruents
was confused to a large extent at the end of the sixteenth century, and completely
merged in the next century. Namely, /di/ ([dʒi]) merged with /zi/ ([ʒi]), and /du/
([dzu]) merged with /zu/ ([zu]). For example, the word kuzu ‘trash’ in modern Japanese
comes from /kudu/ of the earlier stage, spelled ‘くづ ’ in kana script, and the word
kuzu ‘kudzu vine’ comes from /kuzu/, spelled ‘くず’ in kana script (the English spell-
ing of “kudzu” bears no relation to the original kana spelling).12 While these words
are homonyms in modern Japanese, they were distinguished from each other in the
earlier stage. On the other hand, the voiceless contrasts /ti/:/si/ and /tu/:/su/ are
still preserved, as shown in (9) and (10).
Considering the chronological fact that the affrication took place in the period
just prior to the merger, the merger was no doubt caused by the affrication. The affri-
cation before the high vowels made the qualitative distance between /d/ and /z/
closer, and eventually the opposition became neutralized in those positions. Note
that the opposition between /d/ and /z/ itself is maintained, since the affrication
did not occur with other vowels such as /e/, /a/ and /o/; namely, the contrasts
/de/:/ze/, /da/:/za/, and /do/:/zo/ are preserved. As to the partial merger of /d/
and /z/, there is a historical question that should be explained: why is the conse-
quence in the voiceless series different from that in the voiced series? The affrication
took place in the voiceless [t] as well as in the voiced [d] and, therefore, /t/:/s/
12 Although pre-modern kana letters were quite often spelled without the diacritical mark dakuten
indicating a voiced obstruent, we will show here the kana spelling with dakuten added.

Authenticated
should have been subject to the same phonetic condition that the qualitative dis-
tance had become closer; nevertheless, the voiceless opposition /t/:/s/ has been
entirely preserved in modern Japanese. The asymmetry pointed out above concerns
not only the system of modern Japanese but also the process of the historical
change. This is an interesting theoretical point for phonetics and phonology in
general, as well as for Japanese in particular, some relevant issues of which are
discussed here.
First, while there are fricative variations among the realizations of voiced obstru-
ent phonemes, there is no voiced phoneme that is exclusively realized as fricative in
modern Japanese. For example, the phoneme /b/ is realized not only as the stop [b]
also as the bilabial fricative [β] especially in intervocalic positions. Moreover, the
phoneme /z/ is realized not only as the fricative [z] but as the affricate [dz]. In con-
trast, some voiceless phonemes are exclusively realized as fricatives, such as /s/ and
/h/.13 When we look at the voiceless consonants in the stage around the fifteenth
century, the period just before the affrication and merger, it is probable that there
were two fricative phonemes, /s/ and /ɸ/, in the system. Among the voiced (or
prenasalized) obstruents, on the other hand, there is only the phoneme /z/ as a
candidate that would have been exclusively realized as fricative; however, since
there is no clear evidence, it is difficult to establish such reconstruction. In addition,
it is assumed in the literature that /z/ was realized mainly as prenasalized cons-
onants (probably affricates) in old Japanese (see sections 3.3 and 4). If the situation
had not changed in the fifteenth century, there would be no phoneme realized exclu-
sively as voiced fricative in the consonant inventory. Thus, the voiced obstruents
compared to the voiceless obstruents do not have fricative sounds that bear a dis-
tinctive function in the system. This fact should be taken into consideration when
discussing the asymmetry in the merger, which may come from the asymmetry in
the consonant system.
A second issue concerns the difference in the functional load between the voice-
less contrasts and the voiced contrasts, shown in (11).
(11) a. voiceless contrasts: /ti/:/si/ (including /tj/:/sj/), and /tu/:/su/

b. voiced contrasts : /di/:/zi/ (including /dj/:/zj/, and /du/:/zu/
The distribution of voiced obstruents is constrained by two morpho-phonotactic

restrictions in the native lexicon, as discussed in (6) above. These restrictions have
lasted through the known history of Japanese. Thus, the voiced obstruents occurred
less frequently than the voiceless obstruents, which suggests that the functional load
of the voiced contrasts was fairly low. In fact, there were only a few minimal pairs
for the voiced obstruents in (11b) in the lexicon of Japanese of the sixteenth century
13 Whether /h/ is a typical fricative or not is problematic, but at least, it is not definitely realized as
a stop.

Authenticated
(T. Takayama 1993).14 A contrast may be vulnerable when the functional load is
extremely low. The functional load may not always be a decisive factor in historical
events, but we must seriously consider it when discussing the merger in the series of
voiced obstruents.
3.3 Merger and prenasalization

Another topic about the merger quite often discussed in the literature concerns the
historical relation to the loss of prenasalization. Against the usual confusion caused
by merger, historical documents give instructions about the prescriptive pronuncia-
tion as well as how to correct deviations there from. Such documents often appeared
from the end of the seventeenth century to the beginning of the eighteenth century.
They instructed that the distinctions between /di/ /du/ and /zi//zu/ should be made
in the way shown in (12).15
(12) a. /di/ and /du/ should be pronounced with a shorter coda nasal
immediately before them, and with the tip of tongue touching to the
roof of the mouth.
b. /zi/ and /zu/ should be pronounced with no shorter coda nasal, and
with the tip of tongue not touching to the roof of the mouth.
The instructions in (12) reveal the earlier stage just before the distinction was lost.
The phonetic difference is described in (13).
(13) a. /di/ and /du/ were realized as prenasalized affricates.

b. /zi/ and /zu/ were realized as plain voiced fricatives.
The instructions suggest, first of all, that the prenasalization was being lost in the
seventeenth century (see section 2.3 for detailed discussion), and secondly, that
the prenasalization of /z/ vanished earlier than that of /d/ and /g/. The time lag
eventually provided a chance for the distinction between /d/ and /z/ to be carried
by the difference in prenasalization, in addition to the difference between affricates
14 There were two minimal pairs of words sharing the same pitch accent of the Kyoto dialect in the
sixteenth century. One is a pair of native words, udi ‘family, clan’ and uji ‘maggot’, and the other is a
pair of the initial positions of the SJ words, di ‘ground’ and ji ‘letter, Chinese character’. The words
mentioned in the text, kudu ‘trash’ and kuzu ‘kudzu vine’, do not have the same pitch accent.
15 Although there are some differences among the kinds of instructions, we will not give the details
and differences that are observed among documents. The instruction shown in (12) is a summarized
version (see T. Takayama 2003 for details of differences).

Authenticated
and fricatives. In other words, the difference of prenasalization reinforced the dis-
tinction between affricates ([dz][dʒ]) and fricatives ([z][ʒ]). This is summarized in
Table 3, which schematically illustrates phonetic realizations in intervocalic posi-
tions. The variants, [dʒi] and [dzu], occurred in the initial positions and after the
coda nasal /N/ in the fourth stage.
Table 3: The processes of the mergers between /di//du/ and /zi//zu/
/di//du/ /zi//zu/
1. the initial state [ndi][ndu] [ndʒi][ndzu]

2. the stage after the vanishing of the nasal portion in /z/ [ndi][ndu] [ʒi][zu]
3. the stage after the affrication of /d/ [ndʒi][ndzu] [ʒi][zu]
4. the stage after the merger resulting from the loss of [ʒi][zu] [ʒi][zu]
the nasal portion in /d/
As a result, the contrast between /di/:/zi/ and between /du/:/zu/ temporarily

resisted merger. In this way, the de-prenasalization and the merger accidentally
overlapped in the historical context. A kind of accidental synchronization between
independent diachronic events reveals dynamic characteristics of the phonological
history. See Kamei (1950), M. Takayama (2006, 2012: 147–162) and T. Takayama
(1993, 2010) for historical discussions of the processes shown in Table 3; see Steriade
(1993) and Riehl and Cohn (2011) for theoretical discussions on the relationship
between prenasalization and the affricates.
4 Issues concerning the phonetic realization

of /s/
The phonemic status of /s/ has been relatively stable with no drastic changes until
modern Japanese; that is, we do not find any merger or split through history, apart
from some speculations mentioned about the prehistoric period (section 4.1). In
modern Japanese, the phoneme /s/ is realized exclusively as fricatives [s] and [ ʃ ].
However, investigation into the phonetic value of /s/, dating back to the earliest
stage attested by documented resources, shows a different situation. It is widely
accepted that the phonetic realizations of /s/ in the eighth century can be recon-
structed as affricates, or at least, basically as affricates (Arisaka 1936; Kamei 1970b;
Mori 1991; Ogura 1998; Takeuchi 1995; Hayashi 2002).
Since the affrication of /ti/ and /tu/ took place in the sixteenth century, as dis-
cussed in section 3 above, it is assumed that the transition of the phonetic realiza-
tions of /s/, from affricates to fricatives, should have taken place in some earlier time
before the sixteenth century, although attestation of an accurate date is extremely
difficult.

Authenticated
In this section, we note a few problems that arise as consequences of old Japanese
reconstruction of /s/. Although technical discussions about historical data analysis
or philological issues are relevant to such reconstruction, they are not discussed
here (see the History Volume). The focus here is on the phonological aspects of the
reconstruction of /s/.
4.1 /s/ in the consonant system of Old Japanese

The first problem concerns how the phonemes are organized in the consonant system
of Old Japanese. Since it is generally assumed in the literature that the phonemes that
correspond to /s/ and /z/ in modern Japanese were realized as affricates, we will use
the symbols /ts/ and /dz/ to represent them. The consonant system of Old Japanese
is shown in Table 4.
Table 4: The consonant system of Old Japanese
voiceless (or non-prenasalized) p t k

ts
prenasalized b d g
dz
nasal m n
liquid r
approximant j w
We briefly comment on each member of the consonant system in this table. First,
the labial /p/ corresponds to /h/ of modern Japanese. The fact that the modern /h/
comes from the labial sound is unquestionable in the literature (section 5). Second,
/d/, /g/, and /b/ at the beginning of the seventeenth century had prenasalized real-
izations, as discussed in section 2.3. Whether Old Japanese had prenasalized /b/, /d/,
/g/, and /dz/ is not well-documented, but there is no negative evidence against such
realizations (sections 2.3 and 2.5). Therefore, these segments are generally recog-
nized as prenasalized in the earliest attested stage. For the other phonemes /m/,
/n/, /r/, /j/, and /w/, there is essentially no controversy about their phonemic status
and phonetic values, and they are thought to be essentially no different from their
modern Japanese counterparts.
A look at the obstruents in the inventory shows a remarkable characteristic:
there is no fricative but there are affricates such as /ts/ and /dz/. For the voiceless
obstruents, except for /ts/, there are the stops /p/, /t/, and /k/, as discussed at the
beginning of section 4. The situation of the voiced obstruents (or the prenasalized
obstruents) may be the same as that of the voiceless ones. Note that the prenasalized
realization may associate closely with the realizations of the stops (see Steriade 1993
and Riehl and Cohn 2011).

Authenticated
Regarding the voiceless obstruents, Kamei (1970b) pointed out that the recon-
structed system does not accord with a cross-linguistic implication about the affri-
cate: if there is an affricate in a consonant system, then there should be a fricative
as its counterpart. Namely, if the affricate /ts/ existed in the system, the fricative /s/
would be expected to exist, too. However, the attestations from historical resources
do not indicate that the fricative /s/ existed in Old Japanese, apart from /ts/, which
was a predecessor of modern /s/. Arisaka (1955: 489–490) speculates that the frica-
tive /s/ might have existed in a prehistoric stage, but later vanished. Possibly, the
initial /s/ vanished via /h/, and the intervocalic /s/ merged into /ts/, which would
often be weakened in intervocalic positions. Kamei (1973a) discusses a morphological
phenomenon that may support such a hypothesis, that is, the alternation between
the zero consonant and the /ts/ as observed between the simplex /ame/ ‘rain’ and
the derived form /tsame/ in compounds such as haru-same ‘spring rain’ in modern
Japanese.16
As Kamei (1970b, 1973a) points out, the gap of the dental voiceless fricative in
the consonant system of Old Japanese remains a mystery in the historical studies of
Japanese phonology. Cross-linguistic approaches may help us solve the mystery.
State-of-the-art theoretical studies may impact on the reconstruction of the affricate
in Old Japanese.
As for the details of phonetic realizations, it should be mentioned that the
proposed reconstruction does not necessarily exclude the possibility that there were
fricative allophones. In fact, Ogura (1998) argues that the fricative sounds would
have existed in intervocalic positions. The stop portions of affricates may have weak-
ened between vowels, but remained stable in initial position. Intervocalic weakening
may have triggered the further weakening of the intervocalic labial /p/ which was
realized as [ɸ], and consequently caused the merger between the intervocalic [ɸ]
and the approximant /w/, which has been attested to have taken place around the
eleventh century (see section 5 for details). Another account of phonetic realizations
has been proposed by Hayata (1977), as mentioned in notes 1 and 10. He argues that
/p/, /t/, /k/, and /ts/ were voiced in intervocalic positions, and that these voiced
allophones were phonemically distinguished from the prenasalized voiced realiza-
tions of /b/, /d/, /g/, and /dz/, respectively (see also Frellesvig 2010: 34–38). These
arguments are important for future discussions on this topic.
4.2 Phonetic value and rendaku

Let us now address how sequential voicing, i.e., rendaku, may affect the phonetic
value of /ts/. On the basis of the data of rendaku in Old Japanese, Moriyama (1962)
16 Kamei (1973a) discusses various possibilities concerning the hypothetical s, and shows the possi-
bility that the initial consonant ts of the simplex *tsame would be sporadically confused with s that
vanished afterwards.

Authenticated
pointed out that the number of words that undergo rendaku in /ts/ is drastically
smaller than the number in any other voiceless segment such as /p/, /t/, and /k/.
This discrepancy between /ts/ and the other voiceless obstruents should not be
ignored when discussing the phonetic realizations of these phonemes. However,
we face a paradoxical problem. If the disparity comes from the difference in the
phonetic realization between an affricate and a plain stop, a question arises as to
why an affricate is less prone to rendaku than a plain stop. In order to accept the
assumed reconstruction of /ts/, it is necessary to explain why only the affricate /ts/
behaves differently from the other voiceless obstruents, despite the fact that voice-
less obstruents, including /ts/, have a stop element in common. Furthermore, pre-
nasalization triggered by rendaku should be taken into consideration (note that
rendaku is not simple voicing in old Japanese). These problems remain unsolved.
The relationship between rendaku voicing and phonetic conditions in old Japanese
needs further investigation.
5 Issues on the labial voiceless stop

A quite well-known fact in the history of Japanese is that the phoneme /h/ in
modern Japanese comes from the labial sound in old Japanese. The facts involving
/p/ or /h/ were taken up quite frequently in textbooks to the extent that it might
seem that there would be no room for further consideration. Nevertheless, a number
of significant studies have been carried out in the recent decades (Komatsu 1985;
Kida 1989; Mori 1991: 97–135; Hayashi 1992; Takeuchi 1995; Ogura 1998; among
others). Their concerns include, for example, estimations of phonetic realizations
by means of historical documents, investigations into the dialectal variations of
relevant sounds, considerations about the dates of the relevant changes, the way in
which the changes took place, and factors or conditions under which the relevant
changes took place. These subjects are, of course, complicatedly related with each
other, and cannot be discussed separately. Since there is a large body of literature
about the historical issues on /p/ and /h/, it is impossible to provide a comprehen-
sive survey. We focus here on a few issues closely related to the general phonological
studies.
5.1 Changes from old /p/ to modern /h/

Let us briefly sketch the history of this phoneme, addressing important points in
chronological order. The consonant system of the earliest stage attested by historical
records, including the phoneme /p/, was shown in Table 4 above. In the history, the
first change relevant to /p/ is the spirantization in (14). It is still not clear whether

Authenticated
this change took place in every phonological context at once or in intervocalic posi-
tion before it occurred in word-initial position.
(14) The spirantization in /p/: [p] > [ɸ]
The change in (15) targeted the labial fricative [ɸ] in the intervocalic positions,
which merged with the phoneme /w/ around the eleventh century. (16a), (16b), and
(16d) illustrate the words involved in this change. The targeted lexical items were
native words as in (16a) and (16b) as well as Sino-Japanese morphemes as in (16d).
In contrast to these words, the native word in (16c) and the SJ morpheme in (16e)
illustrate the forms with the initial /p/, unaffected by the change in (15).
(15) [ɸ] > [w] / V_V
(16) a. [kapa] > [kaɸa] > [kawa] (ModJ. [kawa] ‘river’)

b. [jupu] > [juɸu] > [juu] (ModJ. [ju:] ‘evening’)
c. [pana] > [ɸana] (ModJ. [hana] ‘flower’)
d. [kapu] > [kaɸu] > [kau] (ModJ. [ko:] ‘tortoise shell, instep [甲 ]’)
e. [pa] > [ɸa] (ModJ. [ha] ‘wave [波 ]’)
As a result of the change in (15), the segmental sequence /Vpu/ merged into
/Vu/ (where V indicates a vowel), because phonotactics did not allow /wu/. This
is illustrated in (16b) and (16d). All SJ morphemes involved in (15) went along this
path, since the intervocalic [ɸ] was followed by no vowels other than /u/. The inter-
vocalic [ɸ] in SJ words corresponds to the coda p in classical Chinese. The high
vowel /u/ is epenthetic in loanwords, such as [kaɸu] in (16d) that come from kap
in Classical Chinese (see Kubozono, Ch. 8, this volume for details about epenthetic
vowels in loanwords).
The date of the merger in (15) can be collaborated by the confusion observed in
writing between pa, pi, pu, pe, po and wa, wi, u, we, wo (Tsukishima 1969, among
others).17 By contrast, the date of (14) is controversial. Since Hashimoto (1928), it
has been generally assumed that /p/ had already been spirantized in or before the
Nara period. However, Kida (1989) reexamined the evidence and pointed out the
possibility that even if the intervocalic /p/ was spirantized, the initial /p/ may still
have been realized as a stop even in the beginning of the Heian period. Hayashi
(1992) argued that the bilabial stop [p] remained at least until the ninth century,
i.e., the beginning of the Heian period, pointing out that the two changes in (14)
and (15) should have occurred in fairly quick succession, because the spirantization
17 [wi], [we], [wo], and [je] are not permitted in the native and SJ words of modern Japanese (Chapters
3 and 8).

Authenticated
in (14) probably naturally triggered the further weakening of stricture in the inter-
vocalic position, and the intervocalic [ɸ] was further loosened and voiced, resulting
in the approximant [w] (see also Frellesvig 2010: 34–38).
What is significant in relation to modern Japanese is that the change in (15)
brought a phonotactic rearrangement to [ɸ]. The result of the change confined [ɸ]
to word-initial positions. Komatsu (1985) interprets this as meaning that this phoneme
acquired a demarcative function, which is one of the significant factors behind the
change in (15). The phonotactics of this sound established in the eleventh century
was transferred over to modern Japanese. In fact, we find /h/ in medial positions in
neither the native nor SJ lexicon, except for a few native words such as ahiru ‘duck’
and ahureru ‘overflow’, and a large number of loanwords such as sohuto ‘soft’
borrowed from English.
The third change shown in (17) took place in the period around the second half
of the seventeenth century to the eighteenth. As a result of this change, the labial
feature vanished from the articulation of this phoneme.
(17) delabialization in /ɸ/(=/h/): [ɸ] > [h]
As for the phonetic realization of /h/, it is quite often pointed out even in some
textbooks of Japanese phonetics and phonology that it should be regarded as a de-
voiced vowel whose quality is the same as that of the vowel that follows it; namely,
the sounds should be described as [i ̥i], [e̥e], [ḁa], [o̥o], and [ɯ̥ ɯ] rather than current
IPA representations such as [çi], [he], [ha], [ho], and [ɸɯ]. At any rate, the phoneme
/h/ is not straightforwardly specified by any place of articulation. /h/ is an idio-
syncratic consonant that is different from other obstruents such as the labial /b/,
dental /t/, the velar /k/ in the inventory. The transition from the labial sounds to
/h/ as shown in (17) is a remarkable phenomenon in terms of the relationship
between phonotactics and sound change in a general sense. The question of how
confining /ɸ/ to initial position triggered delabialization has not been substantially
discussed in the literature. As to this problem, Kamei et al. ([1976] 2007: 72–87)
points out that it took a very long period, i.e., several hundred years, for the initial
[ɸ] to change to /h/ after the disappearance of the intervocalic [ɸ] in (15) in the
Kyoto dialect, and argued that [ɸ] may have lasted due to surrounding dialects that
still preserved the labial realization. He also suggested that the initial bilabial [ɸ] did
not change to the labio-dental [f], which is phonetically more stable than [ɸ], under
the socio-geographic condition he assumed. Further research is needed, especially
concerning how phonotactic properties relate to some sound changes. It is relevant
not only to historical studies but also to theoretical considerations.
5.2 Geminate of the labial stop

As mentioned in Kubozono’s introduction to this volume and Kawagoe (this volume),
the geminate /pp/ (quite often described as /Qp/ in traditional Japanese linguistics)

Authenticated
occurs in modern Japanese. In morphological alternations, this geminate occurs as

the counterpart of the singleton /h/ as exemplified in (18).
(18) hai ‘cup’

ip-pai (< iti+hai) ‘one cup’
The geminate /pp/ occurs quite often in SJ words and occasionally in the native
lexicon (hai in (18) is a SJ morpheme; see Nasu (this volume) and Ito and Mester
(Ch. 7, this volume) for morphophonological aspects of SJ morphemes). Where does
the geminate come from, or how did the geminate emerge? In the Japanese Portu-
guese dictionary Vocabulario da Lingoa de Iapam published in Nagasaki in 1603,
the geminate is spelled by the roman letters pp, similar to common Romanization
of modern Japanese. Therefore, the geminate can be dated back at least to the begin-
ning of the seventeenth century, but earlier stages are difficult to verify due to the
lack of a distinctive mark in the domestic writing system.
However, we find a few words that may indicate that the geminate /pp/ or the
phoneme /p/ with longer duration existed even in an earlier stage, as shown in (19).
(19) a. ModJ. appare ‘admirable’

cf. aware ‘pathos’
b. ModJ. moppara ‘exclusively’
The word appare is completely different from the word aware in modern Japanese,
but the two words are doublets, etymologically derived from a common word.
The former originates from the emphasized form of the latter, i.e., apare, which was
probably pronounced in old Japanese with a longer closure of lips. The form apare
(or, in the later stage, the spirantized aɸare) became aware via the merger between
intervocalic [ɸ] and [w](=/w/) in (15). While the spirantization and the following
merger occurred, the bilabial closure has been preserved in appare up to modern
Japanese. The merger did not involve appare, compared to the other member of the
doublet aware, and therefore the form with a longer or emphasized /p/ may date
back to the stage before the merger that took place around the eleventh century
(see section 5.1).
The longer /p/ may have resisted the spirantization due to the solid closure,
which may explain why the labial stop in appare remains. A similar condition may
have been operative in the native word moppara in (19) (Kamei et al. [1976] 2007: 82–
86; T. Takayama 2002).
5.3 Mimetic p
As often mentioned, the Japanese language has a great number of mimetic expres-
sions, and their role in the lexicon is quite important. The properties of their forms

Authenticated
are also extremely significant from phonological or morphophonological points of

view (see Kubozono’s introduction to this volume). It is well known that the phono-
tactics of mimetic forms are different from the phonotactics of plain native words, as
discussed in general phonological studies (Ito and Mester 1999, 2008; Nasu 1999; see
Nasu, this volume, for details on mimetic phonology). Especially, the initial singleton
/p/ in mimetic forms is remarkable (Nasu 1999), and occurs in a large number of
mimetic words as shown in (20).
(20) a. piyo piyo ‘cheep of chick’

b. pikari(to) ‘a flash of light’
Where did the stop /p/ of mimetics come from, if /p/ of Old Japanese changed to
the non-labial /h/? The Vocabulario da Lingoa de Iapam (1603) lists nine mimetic
words (no other words than mimetic) as the entries beginning with /p/. Even in the
seventeenth century, both /p/ and /ɸ/ occurred in the consonant system, similar to
modern Japanese. The difference between /ɸ/ and /p/ is attested from romanized
Japanese as written by Portuguese missionaries (/ɸ/ is represented by the letter f,
and /p/ by the letter p). Directly demonstrating the situation before the seventeenth
century is difficult because there was no distinction made in the domestic writing
system. However, on the basis of indirect resources, Kamei (1959, 1960) argued that
the labial stop probably remained in the mimetic expressions even after the spiran-
tization of /p/. He suggests that the labial voiceless stop was preserved through
the history of Japanese since the quality of a sound itself is crucial to mimetic
expressions or sound symbolism.18 Komatsu (1981: 249–283) refers to a kind of
morpho-semantic effect quite often observed between voiceless obstruents and voiced
(prenasalized) sounds, which is characteristic of Japanese mimetic expressions, as
illustrated in (21).
(21) a. pota pota

‘with (something like liquid) dripping lightly’
b. bota bota
‘with (something like liquid or others) dripping heavily’
Komatsu (1981) argues that since such semantic difference is significant in

Japanese sound symbolism, the phonetic parallelism between [p] and [b] should
have been preserved in mimetic words, in spite of the spirantization in the plain
18 Kamei (1970b) suggests another historical change involving sound symbolism in Japanese. That
is the transition from the affricate /ts/ to a fricative (=/s/ in ModJ), as discussed in section 4.1. He
pointed out that the change did not necessarily target mimetic expressions for a reason similar to
that seen with the labial stop.

Authenticated
native words, and as a result, that the phoneme /p/ in Old Japanese bifurcated into
/ɸ/ and /p/ after the spirantization. Such a split in the history of Japanese provides
an important and valuable resource for the discussion of the relationship between
sound symbolism and sound changes, both from a general point of view as well as
for cross-linguistic investigation of sound symbolism.
6 Conclusion
This chapter first discussed historical issues regarding voiced obstruents in section
2, focusing on the velar nasal [ŋ] and prenasalization in the consonantal system.
After introductory remarks on terminology, we dealt with the velar nasal variant [ŋ]
of /g/ (section 2.1), which has been given the status of standard in modern Japanese,
and looked at the prenasalization that forms the historical background of the velar
nasal (section 2.2). Issues concerning the attestation of the prenasalization were also
discussed (section 2.3). We returned to the topic of the velar nasal in section 2.4 to
discuss some problems in its history. In section 2.5, we dealt with the prehistory of
the obstruents in Japanese with focus on the phonotactics of the voiced obstruents
and rendaku. Section 3 dealt with the historical background of the affricates, [ts],
[tʃ ], [dz], and [dʒ], in modern Japanese. Specifically, we looked at two successive
sound changes: the affrication of [tu], [ti], [du], and [di], and the mergers between
/du/([dzu]) and /zu/([zu]) and between /di/([dʒi]) and /zi/([ʒi]). Furthermore, we
looked at the diachronic overlap between the mergers and the loss of prenasaliza-
tion in the voiced obstruents. Section 4 dealt with issues regarding the phonetic
value of the phoneme /s/ at the stages before the affrication of [t] and [d], which
were discussed in the previous section. In addition, we also pointed out a dis-
sonance between the reconstructed phonetic value of /s/ and the scarcity of rendaku,
i.e., the sequential voicing of /s/ (=/ts/ in Table 4). Reviewing the recent research and
the noticeable arguments therein, section 5 presented a history of the phonemes, /h/
and /p/, which resulted from the bifurcation of the labial /p/ in Old Japanese. The
historical issues surrounding the geminate /pp/ of modern Japanese as well as those
concerning the phoneme /p/ in mimetics were also discussed.
In sum, we have dealt with the main topics of the historical phonology of
Japanese in this chapter, especially the consonantal issues that are helpful in under-
standing the synchronic aspects of modern Japanese. However, there are many
important issues that were not discussed in this chapter.
Vowel coalescence is a significant historical event shaping the modern Japanese
vowel system and phonotactics. As we saw in section 5.1, the vowel combination
/Vu/ occurred in many words, resulting from the merger of intervocalic [ɸ] into /w/
in (15). In addition, there were a large number of SJ morphemes that were not
involved in (15) but which originally had /Vu/ arrangements such as kiu (九) ‘nine’,
seu (少) ‘young’, tau (<taũ 唐 ) ‘Tang dynasty’, and you (<yoũ 用 ) ‘use’. Furthermore,

Authenticated
the vowel combination /Vu/ also emerged from other sources (see T. Takayama 1992
for details). Regardless of the source, these combinations have all been replaced by
long vowels in modern Japanese. The situation results from the vowel coalescence
which is assumed to have ended at the beginning of the seventeenth century. How-
ever, modern Japanese has the vowel combination /Vi/. Details on vowel combina-
tions and coalescence are discussed by Kubozono (Ch. 5, this volume) and T.
Takayama (1992).
Naturally, the historical background of SJ words is significant for various aspects
of the history of Japanese. Phonological problems of SJ words are discussed by Ito
and Mester (Ch. 7, this volume) in their relation to modern Japanese.
Finally, there are morphophonological phenomena not addressed here, including
so-called “onbin” (Frellesvig 1995). Moreover, the various questions and controversial
points concerning the reconstruction of the phonological system of the eighth cen-
tury were not discussed in this chapter. These problems are dealt with in the History
Volume of the same handbook series.
Acknowledgements
I would like to express my special thanks to Donna Erikson for improving the
English expressions. I am greatly indebted to Gábor Pintér, Akio Nasu, and Haruo
Kubozono for many valuable comments and corrections on an earlier version of this
chapter. Any remaining inadequacies and errors are of course my own responsibility.
References
Arisaka, Hideyo. 1936. Jōdai ni okeru sagyō no tōon [/s/ of Old Japanese]. Kokugo to Kokubungaku
13(1). Reprinted in Hideyo Arisaka 1957, 145–159.
Arisaka, Hideyo. 1955. Jōdai on’inkō [A study of sounds of Old Japanese]. Tokyo: Sanseido.
Arisaka, Hideyo. 1957. Kokugo on’inshi no kenkyū: Zōho shinpan [Studies of phonological history of
Japanese: revised version]. Tokyo: Sanseido.
Collado, Didaco. 1632. Ars grammaticae iaponicae lingae. Rome. Reproduced in Takanobu Ōtsuka
1957.
Doi, Tadao (trans.). 1955. Nihon daibunten [A Japanese translation of Rodriguez 1604–1608]. Tokyo:
Sanseido.
Fujimoto, Masako. this volume. Chapter 4: Vowel devoicing. In Haruo Kubozono (ed.), The handbook
of Japanese phonetics and phonology. Berlin: De Gruyter Mouton.
Fukushima, Kunimichi (ed.). 1989. Rodorigesu nihon shōbunten (A reproduction of Rodriguez. 1620,
in the collection of the school of Oriental and African studies of London University). Tokyo:
Kasama Shoin.
Frellesvig, Bjarke. 1995. A case study in diachronic phonology: The Japanese Onbin sound changes.
Aarhus: Aarhus University Press.
Frellesvig, Bjarke. 2010. A history of the Japanese language. New York: Cambridge University Press.

Authenticated
Hashimoto, Shinkichi. 1928. Hagyō shiin no hensen ni tsuite [A history of the voiceless labial
consonant]. In Sanki Ichikawa (ed.), Okakura sensei kinen ronbunshū [Festschrift of professor
Okakura]. Reprinted in Shinkichi Hashimoto 1950, 30–45.
Hashimoto, Shinkichi. 1932. Kokugo ni okeru biboin [Nasal vowels in Japanese]. Hōgen 2(1).
Reprinted in Shinkichi Hashimoto 1950, 1–9.
Hashimoto, Shinkichi. 1950. Kokugo on’in no kenkyū [Studies on Japanese phonology]. Tokyo:
Iwanami.
Hattori, Shiro. 1955. On’inron (1) [Phonology (1)]. Kokugogaku 22. Reprinted in Shiro Hattori 1960,
279–301.
302–322.
323–352.
Hattori, Shiro. 1960. Gengogaku no hōhō [Methodologies in linguistics]. Tokyo: Iwanami.
Hayashi, Chikafumi. 1992. Hagyō tenko on wa naze Heian jidai ni okotta ka: Nihongo on’inshi no
shiten to kijutsu [Why did the intervocalic ɸ merge into w in the Heian period?]. Kokugo to
Kokubungaku 69(11). 110–119.
Hayashi, Chikafumi. 2002. Kyū-seiki nihongo no shiin onka: Nihongo on’inshi ni okeru bunkengaku-
teki kōsatsu no imi to hōhō [Phonetic values of the consonants of Japanese in the 9th century].
Kokugo to Kokubungaku 78(4). 1–14.
Hayata, Teruhiro. 1977. Seisei akusentoron [Generative analysis of pitch accent]. In Susumu Ōno and
Takesi Sibata (eds.), Iwanami kōza nihongo 5: On’in [Iwanami Japanese series 5: Phonology],
323–360. Tokyo: Iwanami.
Hibiya, Junko. 1988. Bariēshon riron [Variation theory]. Gengo Kenkyu 93. 155–171.
Hibiya, Junko. 1995. The velar nasal in Tokyo Japanese: A case of diffusion from above. Language
Variation and Change 7(2). 139–152. Reprinted in Natsuko Tsujimura 2005.
Hibiya, Junko. 2002. Gengo hen’i no chiriteki sai [Regional differentiation of Linguistic Variation].
Onsei Kenkyū [Journal of the Phonetic Society of Japan] 6(3). 60–68.
Hino, Hiroshi. 1993. Nihon shōbunten [A Japanese translation and reproduction of Rodriguez. 1620,
in the collection of the Ajuda library, Lisbon]. Tokyo: Shinjinbutsu Ōraisha.
Hizume, Shūji. 2003. Seidaku bunka to sokuon, hatsuon [The establishment of sei-daku opposition
and moraic obstruents/nasals]. Kokugogaku 54(2). 95–108.
Inoue, Fumio. 1971. Gagyōshiin no rekishi [The distribution and history of the g-sound in Japanese].
Kokugogaku 86. 20–41.
Inoue, Fumio (ed.). 1983. Shinhōgen to kotoba no midare ni kansuru shakai gengogakuteki kenkyū:
Tokyo, Shutoken, Yamagata, Hokkaido [A sociolinguistic study of “new dialect” and “language
deterioration”: Tokyo, Metropolitan area, Yamagata, Hokkaido]. (KAKENHI report).
Inoue, Fumio. 1998. Nihongo uocchingu [Watching Japanese]. Tokyo: Iwanami.
Ito, Junko and R. Armin Mester. 1986. The phonology of voicing in Japanese: Theoretical consequences
for morphological accessibility. Linguistic Inquiry 17(1). 49–73.
Ito, Junko and R. Armin Mester. 1997. Correspondence and compositionality: The ga-gyō variation in
Japanese phonology. In Iggy M. Roca (ed.), Derivations and constraints in phonology, 419–462.
New York: Oxford University Press.
Ito, Junko and R. Armin Mester. 1999. The structure of the phonological lexicon. In Natsuko Tsuji-
mura (ed.), The handbook of Japanese linguistics, 62–100. Malden, MA, and Oxford: Blackwell.
Ito, Junko and R. Armin Mester. 2003. Japanese morphophonemics: Markedness and word structure.
Cambridge, MA: MIT Press.

Authenticated
Ito, Junko and R. Armin Mester. 2008. Lexical classes in phonology. In Shigeru Miyagawa and
Mamoru Saito (eds.), The Oxford handbook of Japanese linguistics, 84–106. Oxford: Oxford
University Press.
Ito, Junko and R. Armin Mester. this volume. Chapter 7: Sino-Japanese phonology. In Haruo Kubo-
zono (ed.), The handbook of Japanese phonetics and phonology. Berlin: De Gruyter Mouton.
Itoi, Kanichi. 1962. Ōita-ken no yotsugana ben ryakuhō: Kusu-machi Kimuta hōgen ni tsuite [A report
on the contrasts of zi and di, zu and du in the Kusu machi Kimuta dialect of Ōita prefecture].
Kokubungakukō 27. 48–52.
Kamei, Takashi. 1950. Kenshukuryōkoshū o chūshin ni mita yotsugana [On the contrasts of zi and di,
zu and du in Kensyukuryōkosyū]. Kokugogaku 4. Reprinted in Takashi Kamei 1984, 291–312.
Kamei, Takashi. 1956. Gagyō no kana [On kana letters for the syllables with the g-sound]. Kokugo to
Kokubungaku 33(9). Reprinted in Takashi Kamei 1984, 1–25.
Kamei, Takashi. 1959. Shunnōten. Kokugogaku 39. Reprinted in Takashi Kamei 1984, 141–153.
Kamei, Takashi. 1960. Kodaijin no warai goe [Laughter of ancient people]. Nihon Koten Bungaku
Taikei Geppō 37. Reprinted in Takashi Kamei 1984, 371–375.
Kamei, Takashi. 1970a. Kana wa naze dakuon sen’yō no jitai o motanakatta ka o megutte kataru [On
the reasons why the kana system does not have letters for daku-on]. Jinbun Kagaku Kenkyū 12.
Reprinted in Takashi Kamei 1986, 63–150.
Kamei, Takashi. 1970b. Suzume shiu shiu [Sparrows chirp ‘siu siu’]. Seikei Kokubun 3. Reprinted in
Takashi Kamei 1984, 447–464.
Kamei, Takashi. 1973a. Bunken izen no jidai no nihongo [Prehistoric Japanese]. Reprinted in Takashi
Kamei 1973b, 265–399.
Kamei, Takashi. 1973b. Nihongo keitōron no michi (Kamei Takashi Ronbunshū 2) [Perspectives on
the cognation between Japanese and other languages (Selected works of Kamei Takashi vol.
2)], 265–399. Tokyo: Yoshikawa Kōbunkan.
Kamei, Takashi. 1984. Nihongo no sugata to kokoro 1 (Kamei Takashi Ronbunshū 3) [Meaning and
form in Japanese 1 (Selected works of Kamei Takashi vol. 3)]. Tokyo: Yoshikawa Kōbunkan.
Kamei, Takashi. 1986. Gengo bunka kusagusa (Kamei Takashi Ronbunshū 5) [Considerations on Lan-
guage and Culture (Selected works of Kamei Takashi vol. 5)]. Tokyo: Yoshikawa Kōbunkan.
Kamei, Takashi, Tokihiko Ōtō and Toshio Yamada (eds.). 2007 [1976]. Nihongo no rekishi 5: Kindaigo
no nagare [History of the Japanese Language vol. 5: Trends in Modern Japanese], 2nd edition.
Tokyo: Heibonsha.
Kawagoe, Itsue. this volume. Chapter 2: The phonology of sokuon, or geminate obstruents. In Haruo
Kubozono (ed.), The handbook of Japanese phonetics and phonology. Berlin: De Gruyter Mouton.
Kida, Akiyoshi. 1989. P’on zokukō [Further consideration on the p sound]. In Okumura Mitsuo kyōju
taikan kinen ronbunshū kankōkai (eds.), Okumura Mitsuo kyōju taikan kinen ronsō [Festschrift
in celebration of the retirement of Professor Mitsuo Okumura], 415–429. Tokyo: Ōfūsha.
Kim, Hyunsoon. 2001. A phonetically based account of phonological stop assibilation. Phonology 18
(1). 81–108.
Kindaichi, Haruhiko. 1942. Gagyō bionron [On the Japanese velar nasal]. Reprinted in Haruhiko
Kindaichi 1967, 168–197.
Kindaichi, Haruhiko. 1967. Nihongo on’in no kenkyū [Studies in Japanese Phonology]. Tokyo: Tōkyōdō
Shuppan. Reprinted in Haruhiko Kindaichi 2005.
Kindaichi, Haruhiko. 2005. Kindaichi Haruhiko chosakushū 6 [Selected works of Haruhiko Kindaichi
vol. 6]. Tokyo: Tamagawa University Press.
Komatsu, Hideo. 1981. Nihongo no sekai 7: Nihongo no on’in [The world of Japanese 7: Japanese
phonology]. Tokyo: Chūōkōronsha.
Komatsu, Hideo. 1985. Bogo no rekishi o toraeru shiten [A viewpoint to deal with the history of
native language]. In Shirō Hayashi (ed.), Ōyō gengogaku kōza 1: Nihongo no kyōiku [Series in
applied linguistics 1: Japanese teaching], 132–147. Tokyo: Meiji Shoin.

Authenticated
Kubozono, Haruo. 2012. Varieties of pitch accent systems in Japanese. Lingua 122(13). 1395–1414.
Kubozono, Haruo. this volume. Introduction to Japanese phonetics and phonology. In Haruo Kubo-
Kubozono, Haruo. this volume. Chapter 5: Diphthongs and vowel coalescence. In Haruo Kubozono
(ed.), The handbook of Japanese phonetics and phonology. Berlin: De Gruyter Mouton.
Kubozono, Haruo. this volume. Chapter 8: Loanword phonology. In Haruo Kubozono (ed.), The hand-
book of Japanese phonetics and phonology. Berlin: De Gruyter Mouton.
Kuno, Mariko, Makoto Kuno, Makio Ōno and Takao Sugimura. 1995. Yotsugana tairitsu no shōshutsu
katei: Kōchi-ken Nakamura-shi, Aki-shi no baai [The mergers of zi and di, zu and du in the
Nakamura and Aki dialects of Kōchi Prefecture]. Kokugogaku 180. 84–96.
Kuno, Makoto. 2006. Kōchi hōgen no zenbion [Pre-nasalization of voiced plosives in Kochi dialect].
Onsei Kenkyū [Journal of the Phonetic Society of Japan] 10(1). 60–69.
Lin, Yen-Hwei. 2011. Affricates. In Marc van Oostendorp, Colin J. Ewen, Elizabeth Hume and Keren
Rice (eds.), The Blackwell companion to phonology 1, 367–390. Malden, MA and Oxford: Wiley-
Blackwell.
Mori, Hiromichi. 1991. Kodai no on’in to Nihon shoki no seiritsu [Sounds of Old Japanese and pro-
duction of Nihon shoki]. Tokyo: Taishukan.
Morita, Takeshi. 1977. Nippo jisho ni mieru go’on renketsu jō no ichi keikō [A combinatorial tendency
of sounds in Vocabulario da Lingoa de Iapam]. Kokugogaku 108. 20–32.
Moriyama, Takashi. 1962. Rendaku: Jōdai ni okeru [Rendaku in Old Japanese]. Gobun Kenkyū 14. Re-
printed in Takashi Moriyama 1976, 279–297.
Moriyama, Takashi. 1976. Jōdai kokugo on’in no kenkyū [Studies on sounds in Old Japanese]. Tokyo:
Ōfūsha.
Nasu, Akio. 1999. Onomatope ni okeru yūseika to [p] no yūhyōsei [Voicing in onomatopoeia and the
markedness of [p]]. Onsei Kenkyū [Journal of the Phonetic Society of Japan] 3(3). 52–66.
Nasu, Akio. this volume. Chapter 6: The phonological lexicon and mimetic phonology. In Haruo
Kubozono (ed.), The handbook of Japanese phonetics and phonology. Berlin: De Gruyter Mouton.
Ogura, Hajime. 1998. Sagyō shiin no rekishi [Ancient Japanese sibilants]. Kokugogaku 195.
Reprinted in Hajime Ogura 2011, 137–160.
Ogura, Hajime. 2011. Nihongo on’inshi ronkō [Research on the history of Japanese phonology].
Osaka: Izumi Shoin.
Ōtsuka, Takanobu. 1957. Koryādo nihon bunten [A Japanese translation and a reproduction of Ars
grammaticae iaponicae linguae]. Tokyo: Kazama Shobō.
Pintér, Gábor. this volume. Chapter 3: The emergence of new consonant contrasts. In Haruo Kubo-
Riehl, Anastasia K. and Abrigail C. Cohn. 2011. Partially nasal segments. In Marc van Oostendorp,
Colin J. Ewen, Elizabeth Hume and Keren Rice (eds.), The Blackwell companion to phonology 1,
550–576. Malden, MA and Oxford: Wiley-Blackwell.
Rodriguez, João. 1604–1608. Arte da lingoa de Iapam. Nagasaki. Japanese annotation by Tadao Doi,
1976. Nihon bunten. Tokyo: Benseisha.
Rodriguez, João. 1620. Arte breve da lingoa de Japoa. Macao. Reproduced by Kunimichi Fukushima
1989 and reproduced with Japanese translation by Hiroshi Hino 1993.
Sakanashi, Ryūzō. 1975. Sanba no shiroki nigori ni tsuite [On the diacritical circle of Sanba].
Okayama Daigaku Hōbungakubu Gakujutsu Kiyō 36. 142–127. Reprinted in Ryūzō Sakanashi
2004, 253–280.
Sakanashi, Ryūzō. 2004. Kinsei no goi hyōki [Aspects of spellings in the Edo period]. Tokyo: Musa-
shino Shoin.
Shibatani, Masayoshi. 1990. The languages of Japan. Cambridge: Cambridge University Press.

Authenticated
Steriade, Donca. 1993. Closure, release, and nasal contours. In Marie K. Huffman and Rena A.
Krakow (eds.), Nasals, nasalization, and the velum (Phonetics and phonology 5), 401–470.
San Diego: Academic Press.
Sugimura, Takao. 2001. Kyūshū hōgen no yotsugana [On the contrasts of zi and di, zu and du in the
phonemic systems of the Kyushu dialects]. Onsei Kenkyū [Journal of the Phonetic Society of
Japan] 5(3). 10–18.
Takayama, Michiaki. 1992. Seidaku shōkō [A consideration on sei-daku]. In Ikudō Tajima and Kazuya
Niwa (eds.), Nihongo ronkyū 2: Koten nihongo to jisho [Studies of Japanese vol. 2: Classical
Japanese and dictionaries], 17–56. Osaka: Izumi Shoin.
Takayama, Michiaki. 2006. Yotsugana to zenbion [Prenasalization and merger]. In Chikushi kokugo-
gaku danwakai (ed.), Chikushi gogaku ronsō II: History of Japanese and dialects, 158–174.
Tokyo: Kazama Shobō.
Takayama, Michiaki. 2012. Nihongo on’inshi no kenkyū [A study of phonological history of
Japanese]. Tokyo: Hituzi Syobo.
Takayama, Tomoaki. 1992. Nihongo ni okeru rensetsu boin no chōboinka: Sono rekishiteki imi to
hassei no onseiteki jōken [The changes of Vu hiatuses into long vowels in the history of
Japanese language]. Gengo Kenkyu 101. 14–34.
Takayama, Tomoaki. 1993. Hasatsuon to masatsuon no gōryū to dakushiin no henka: Iwayuru
yotsugana gōryū no rekishiteki ichizuke [Transition and merger in the voiced obstruents of
late middle Japanese]. Kokugo Kokubun 62(4). 18–30.
Takayama, Tomoaki. 1998. 17 seiki matsu no zenbion no jittai ni tsuite: Ikeisai kōgo kikigaki and
Waji shōran shō no saikenshō [On the prenasalization in the end of the seventeenth century:
Re-examinations of Ikeisai kōgo kikigaki and Waji shōran shō]. Kagawa Daigaku Kokubun Kenkyū
23. 1–8.
Takayama, Tomoaki. 2002. Nihon kango no shiteki on’inronteki kadai [Some issues in historical
phonology of Sino-Japanese]. Onsei Kenkyū [Journal of the Phonetic Society of Japan] 6(1).
44–52.
Takayama, Tomoaki. 2003. Kenshukuryōkoshū kara ukagaeru zenbion’yōso no ichi kyokumen
[An aspect of the de-prenasalization seen in Kenshukuryōkoshū]. Kagawa Daigaku Kokubun
Kenkyū 28. 94–101.
Takayama, Tomoaki. 2006. Hasatsuonka to bointaikei [Affrication and vowel system], In Jōo Hakutarō
hakushi kanreki kinen ronbunshū henshū iinkai (ed.), Jikken onseigaku to ippan gengogaku: Jōo
Hakutarō hakushi kanreki kinen ronbunshū [Experimental phonetics and general linguistics:
Festschrift in celebration of the 60th anniversary of Dr. Jōo Hakutarō], 218–228. Tokyo: Tōkyōdō
Shuppan.
Takayama, Tomoaki. 2009. Tagyō dagyō hasatsuonka no on’inronteki tokushitsu [Contrastive aspects
of the affrications in late middle Japanese]. Kanazawa Daigaku Kokugo Kokubun 34. 203–215.
Takayama, Tomoaki. 2010. Yotsugana konran to zenbishiin suitaika to no kanshō: Kobetsu gengoshi
no shiten no jūyōsei [Accidental synchronization between de-prenasalization and merger]. In
Tsukimoto, Masayuki, Toshihiro Fujii and Shūji Hizume (eds.), Kotengo kenkyū no shōten,
851–871. Tokyo: Musashino Shoin.
Takeuchi, Nobuo. 1995. Yōshū oyobi chōan ni okeru Ennin no shittan gakushū: Tokuni Hōgetsu
‘shiddan jibo’ o megutte [Ennin’s learning of siddham in Yang-zhou and Chang-an and his
note on the pronunciation of siddha-mātrikā letters]. Tokyo Daigaku Kyōyōgakubu Kiyō: Hikaku
Bunka Kenkyū 33. 1–146.
Tsujimura, Natsuko (ed.). 2005. Japanese linguistics 3. Pragmatics, sociolinguistics, and language
contact, 1–14. New York: Routledge.
Tsukishima, Hiroshi. 1969. Heian jidaigo shinron [A novel approach to Japanese of the Heian
period]. Tokyo: Tokyo University Press.

Authenticated
Uwano, Zendo. 2012. Three types of accent kernels in Japanese. Lingua 122(13). 1415–1440.
Vance, Timothy J. 1987. An Introduction to Japanese phonology. Albany, NY: State University of New
York Press.
Vance, Timothy J. 2008. The sounds of Japanese. Cambridge: Cambridge University Press.
Vance, Timothy J. this volume. Chapter 10: Rendaku. In Haruo Kubozono (ed.), The handbook of
Japanese phonetics and phonology. Berlin: De Gruyter Mouton.
van de Weijer, Jeroen M., Kensuke Nanjo and Tetsuo Nishihara (eds.). 2005. Voicing in Japanese.
Berlin and New York: Mouton de Gruyter.
Yamaguchi, Yoshinori. 1988. Kodaigo no fukugōgo ni kansuru ichi kōsatsu [On compounds in Old
Japanese]. Nihongogaku 7(5). 4–12.
Yamane-Tanaka, Noriko. 2005. The implicational distribution of prenasalized stops in Japanese. In
Jeroen M. van de Weijer, Kensuke Nanjo and Tetsuo Nishihara (eds.), Voicing in Japanese,
123–156. Berlin and New York: Mouton de Gruyter.

Authenticated

15 Historical Phonology

Uploaded by

Copyright:

Available Formats

15 Historical Phonology

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

15 Historical Phonology

Uploaded by

Copyright:

Available Formats

Tomoaki Takayama

Brought to you by | Universität Osnabrück

2 Historical issues about voiced obstruents

Brought to you by | Universität Osnabrück

2.1 Variations of /g/

(1) gomi [ɡomi] ‘trash’

(2) tamago [tamaŋo] ‘egg’

(3) kagi [kaŋi] ‘key, lock’

(4) ringo [riŋŋo] ‘apple’

Table 1: Voiced obstruents and nasals in modern Japanese

labial dental velar

(other phonemes omitted)

Brought to you by | Universität Osnabrück

2.2 Diachronic background of velar nasal

Brought to you by | Universität Osnabrück

of the nasal element of prenasalized consonants. If [ŋ] emerged in advance, that

2.3 Attestation of prenasalized consonants

Brought to you by | Universität Osnabrück

2.4 Velar nasal in the past

4 Yamane-Tanaka (2005) discusses signiﬁcant aspects of the de-prenasalization in the framework of

Brought to you by | Universität Osnabrück

2.5 Prehistory of voiced obstruents

Brought to you by | Universität Osnabrück

(6) a. Voiced obstruents do not occur in word-initial position.

(7) a. ude b. tamesi c. ude-damesi

(8) a. ude b. kurabe c. ude-kurabe

According to Komatsu (1981: 101–107), rendaku formation is established on the basis

Brought to you by | Universität Osnabrück

3 Aﬀrication and merger

Brought to you by | Universität Osnabrück

Brought to you by | Universität Osnabrück

Brought to you by | Universität Osnabrück

Brought to you by | Universität Osnabrück

(9) a. siru ([ ʃ ]) : tiru ([tʃ ])

(10) a. sumi ([s]) : tumi ([ts])

Table 2: Asymmetry between voiceless and voiced obstruents in modern Japanese

fricative aﬀricate diﬀerence between

voiceless obstruent [s] ([ʃ]) [ts] ([tʃ]) contrastive

Brought to you by | Universität Osnabrück

(11) a. voiceless contrasts: /ti/:/si/ (including /tj/:/sj/), and /tu/:/su/

The distribution of voiced obstruents is constrained by two morpho-phonotactic

Brought to you by | Universität Osnabrück

3.3 Merger and prenasalization

(13) a. /di/ and /du/ were realized as prenasalized aﬀricates.

Brought to you by | Universität Osnabrück

Table 3: The processes of the mergers between /di//du/ and /zi//zu/

1. the initial state [ndi][ndu] [ndʒi][ndzu]

As a result, the contrast between /di/:/zi/ and between /du/:/zu/ temporarily

4 Issues concerning the phonetic realization

Brought to you by | Universität Osnabrück

4.1 /s/ in the consonant system of Old Japanese

Table 4: The consonant system of Old Japanese

voiceless (or non-prenasalized) p t k

Brought to you by | Universität Osnabrück

4.2 Phonetic value and rendaku

Brought to you by | Universität Osnabrück

5 Issues on the labial voiceless stop

5.1 Changes from old /p/ to modern /h/

Brought to you by | Universität Osnabrück