Japanese Pitch-Accent Notation: How to Read 0, 1, 2, 3 and the Overline Diagrams

Japanese pitch-accent notation is the set of symbols that learner tools use to show where a word's high-to-low pitch fall lands. These symbols include accent numbers, overline diagrams, NHK drop marks, and IPA downsteps.¹² If you have already met the H/L model from the overview but still get tripped up when Jisho prints 2, OJAD draws a step diagram, and your Anki deck shows ⟨ꜜ⟩ for the same word, this page translates between those conventions.

Overview

Why one word gets three notations

The same Japanese word can appear in several incompatible visual conventions. This happened because three reference traditions evolved independently: the broadcast-Japanese tradition (NHK), the academic dictionary tradition (Kindaichi / Shin-meikai), and the linguistic transcription tradition (IPA).¹³⁴

Each tradition optimises for a different reader. The NHK dictionary serves broadcast announcers who need an at-a-glance "where does the drop fall?" marker.¹⁵ The Shin-meikai dictionary serves general consultation with a bar-style pitch diagram drawn on the kana itself.³ The IPA convention serves linguists writing about Japanese in journals that also describe African tone languages.⁴

A learner using OJAD, an Anki deck with NHK numbers, and a second-hand Kindaichi dictionary may see one standard example word, 山 (yama), printed as a number, as a bar-and-drop diagram, and with a ⟨ꜜ⟩ wedge. Tools like jcinfo.net's accent converter exist because translating between conventions is a real friction point.⁶

山やまが高たかい。²
"The mountain is tall."

The noun 山 in that sentence is printed as 2 in OJAD,⁷ as やま＼ in the post-2016 NHK convention,¹⁵ as a bar over や and ま with a fall mark at the right edge in Shin-meikai,³ and as /jamaꜜ/ in IPA transcription.⁴² All four encode the same phonological fact: the H→L drop falls on the particle that follows the second mora.

Four notations, one phonology

The notations look different because their authors choose different things to make visually prominent, not because they disagree about the pitch. Once you can read all four, the choice between them becomes a matter of which tool you are using, not which one is "correct".¹⁴²

How to use this page as a reference

The page is a translation key, not a tutorial on the H/L model. It assumes you can already parse a string like "LHH–L" as "low-high-high on the noun, low on the particle".² If that does not sound familiar, the overview article in this cluster handles the model first.

The pages on individual pitch patterns (heiban, atamadaka, nakadaka, odaka) handle drilling. This page handles symbol-to-symbol conversion. Skim the four notation sections to identify what is in front of you. Then use the conversion table at the bottom to translate any one form into the other three.

The accent-number system (0, 1, 2, 3…)

What the number marks

The accent number names the mora after which the H→L pitch drop falls. In other words, the integer is a position, not a pitch label.¹²⁸

0 is reserved for unaccented (heiban) words: there is no drop anywhere in the word, and the high contour continues onto any following particle.¹²⁸

For an n-mora word, the valid accent numbers are 0, 1, 2, … n. A word whose accent number equals its mora count is odaka: the drop falls on the particle, not inside the word itself.¹²

The accent number is the canonical lookup key in both the NHK 2016 dictionary¹ and OJAD,⁷⁹ and it is the encoding used by the Yomitan / Anki pitch-accent ecosystem.¹⁰

Worked decodings

Four standard examples cover the four patterns. Each shows the noun followed by a particle, because the particle is what separates heiban from odaka.

桜さくらが咲さく。¹
"The cherry blossoms bloom."

さくら carries accent number 0 (heiban): LHH on the noun, particle stays H, giving LHH–H.¹²

猫ねこが鳴なく。¹
"The cat meows."

ねこ carries accent number 1 (atamadaka): HL on the noun, particle stays L, giving HL–L.¹²

卵たまごを食たべる。¹
"To eat an egg."

たまご carries accent number 2 (nakadaka): LHL on the noun, particle stays L, giving LHL–L.¹²

男おとこが来くる。¹
"A man comes."

おとこ carries accent number 3 (odaka): LHH on the noun, drop on the particle, giving LHH–L.¹²

Here is the general rule, applied to a 3-mora word with accent n: mora 1 is L unless n == 1; morae 2 through n are H; morae n + 1 to the end are L. If n == 0, the contour stays H from mora 2 through the end of the word and through any following particle.²⁸

On a 4-mora word with accent 2, morae 1–4 surface as L–H–L–L, with the drop between mora 2 and mora 3.² Accent 0 on the same 4-mora word gives L–H–H–H, particle H. Accent 4 gives L–H–H–H, particle L: odaka, exposed only by the particle.²

The off-by-one trap

Accent number 2 does not mean "mora 2 is high and the rest are low". It means the drop falls after mora 2: mora 2 is the last H, mora 3 is the first L.²⁸

The same misreading causes errors at every position. Accent 3 means "drop after mora 3", not "mora 3 is high". Initial lowering adds to the trap: mora 1 is L unless the accent is 1, so a 4-mora word with accent 2 is L–H–L–L, not H–H–L–L.²⁸

Accent number is a drop position, not a stress mark

An English-trained ear instinctively reads "accent 2" as "stress on the second syllable". In Japanese, the integer is the index of the last H mora: the H→L drop lands on the boundary between mora n and mora n + 1.²⁸ Every common decoding error a learner makes is some variant of this mistake.

Where you meet this notation

OJAD prints the accent number alongside its visual contour for every word and verb form.⁷⁹ Major online learner tools, including Jisho, the Yomichan / Yomitan ecosystem, and the Migaku pitch-accent add-on for Anki, index pitch accent by integer.¹⁰ The NHK 2016 dictionary continues to list the accent number even after switching its primary visual to ＼ / ￣.¹⁵

In practice, the accent-number system is the closest thing the pitch-accent ecosystem has to a universal interchange format. If you only learn to read one notation, this is the one.

Overline-and-drop diagrams

How an overline encodes pitch

An overline drawn over a mora (or over a run of morae) signals "this mora is H". Absence of a bar signals "this mora is L".³²

The convention is mora-aligned: the bar starts above the first H mora and ends above the last H mora. Long vowels, the moraic ん, and the first half of geminates each get their own bar segment.²

Pre-2016 NHK dictionary editions used a continuous overline drawn across the H portion of the word. The Japan Foundation's review of the 2016 revision describes the old convention as tango no naka de oto ga takaku hatsuon sareru bubun ni jōsen wo hiku to iu hōshiki: "a method of drawing an overline on the portion of the word pronounced with higher pitch".⁵

The drop mark on the overline

The downstep is marked at the right edge of the overline. In older NHK and learner-textbook conventions, the mark is a small vertical fall stroke or a hooked end on the overline.⁵²

In Shin-meikai's bar-and-graph layout, the drop is signalled by where the bar terminates: the bar ends after the last H mora, and a small fall hook drops to the L baseline.³

In OJAD's web display, the drop is drawn explicitly as a downward step on the per-mora contour line, with the accent number printed beside the word.⁷⁹

The heiban overline

A heiban (accent 0) word receives a bar that starts above mora 2 and runs to the end of the word without a terminal fall mark; in older conventions, an additional bar segment may continue over any following particle to signal "the particle stays high".³²

The visual contrast with an odaka word is the absence of the fall mark and the continuation of the bar onto the particle.²

Reading worked examples back into accent numbers

If a bar runs from mora 2 to the end of a 3-mora word with no fall mark, and continues onto the particle, it is heiban, accent 0 (example: さくら 0).¹²

If a bar sits only over mora 1 with a fall mark at its right edge, it is atamadaka, accent 1 (example: ねこ 1).¹²

If a bar sits over morae 2 and 3 of a 4-mora word with a fall mark at its right edge, it is nakadaka, accent 3 (example: おかやま 3 in the variant traditions that mark it so).²⁸

If a bar sits over morae 2 and 3 of a 3-mora word with a fall mark on the particle, it is odaka, accent 3 (example: おとこ 3).¹²

Dictionary conventions: NHK vs. Shin-meikai vs. OJAD

A learner who consults more than one reference will eventually see the same word printed three different ways. The three reference traditions below each occupy a distinct niche: the NHK broadcast dictionary, the Kindaichi-tradition Shin-meikai dictionary, and OJAD.

NHK Nihongo Hatsuon Akusento Shin-Jiten (2016)

The 2016 revision replaced the older overline-on-H-portion convention with a position-only marker. The new system uses ＼ immediately after the kana of the last H mora to signal the drop. It uses ￣ at the end of the word to signal "no drop".¹⁵

The Japan Foundation review of the 2016 edition documents the change verbatim: oto no sagarime wo ＼ to iu kigō de arawasu (sagarime ga nai tango ni wa saigo ni ￣ wo tsukeru): "the pitch drop is represented by the symbol ＼; words without a drop receive ￣ at the end".⁵

The dictionary continues to print the accent number alongside the ＼ / ￣ symbol. The number remains the standard lookup key.¹ Coverage runs to approximately 75,000 headwords, with accent symbols printed in red.¹

The 2016 revision also changed several recommended primary entries. 父 (chichi, "father") shifted from チチ＼ to チ＼チ, and 熊 (kuma, "bear") expanded from only クマ＼ to additionally accepting ク＼マ.¹⁵ These changes reflect both the new symbol set and the dictionary's observation of akusento heibanka (accent flattening), a trend toward heiban patterns in contemporary Tokyo Japanese.⁵¹¹

The recommended accent can shift between editions

The recommended primary entry for a word is not a fixed phonological fact. It is a lexicographer's call about what the broadcast register currently prefers. A 2010 teaching note that lists 父 as チチ＼ (accent 2) is not wrong for its date; the 2016 NHK new edition simply lists チ＼チ (accent 1) as the primary entry now.¹⁵ If you are working from older notes, check the current NHK or OJAD entry before drilling.

Shin-meikai Nihongo Akusento Jiten (Kindaichi tradition)

Shin-meikai was edited under the supervision of Kindaichi Haruhiko. The current edition (2014, 2nd edition) is edited by Akinaga Kazue and published by Sanseidō.³

The dictionary holds approximately 76,600 entries with pitch information and is the principal academic alternative to NHK; it is cited as a reference in both Vance 2008 and Kawahara 2015.¹²²

The visual convention is a horizontal bar drawn directly over the kana for the H span. The bar ends after the last H mora, and the fall is shown either as a small hook at the end of the bar or as a step down to the L baseline. The accent number is printed but is visually secondary to the bar diagram.³

The Kindaichi tradition predates and competes with the NHK 2016 revision. It has retained the bar-style notation even as NHK moved to the ＼ symbol.⁵³

OJAD (Online Japanese Accent Dictionary, University of Tokyo)

OJAD is hosted by the Minematsu laboratory at the University of Tokyo's Graduate School of Engineering. The project's framework is described in Nakamura et al. (Interspeech 2013).⁷¹³

Its primary display is an interactive per-mora contour diagram (a step graph showing H or L at each mora and an explicit downward step at the drop position) with the accent number printed alongside the word.⁷⁹

A pitch-contour overlay is available alongside the bare accent kernel (the drop position) for learners who want the full mora-by-mora picture.⁹ This is OJAD's name for the per-mora step diagram beyond the bare accent number.

Coverage extends to verb and adjective conjugations and to multi-word phrases via the Suzuki-kun prosody tutor.⁷⁹¹³ OJAD is the standard online reference in practice, and its accent numbers track the NHK 2016 conventions.¹⁷

When two dictionaries disagree

The NHK 2016 revision's published changelog documents entries moving away from their older positions. 父 (chichi) moved from accent 2 (チチ＼, HL→L on a 2-mora word) to accent 1 (チ＼チ, drop after mora 1) as the primary recommendation. This means a 2014 Kindaichi printing will list one accent for "father" and a 2016 NHK printing will list another.¹⁵³

熊 (kuma) is documented as an explicit double-listing case in the 2016 NHK edition: the new edition accepts both クマ＼ (accent 2, odaka) and ク＼マ (accent 1, atamadaka), where prior editions accepted only the former.¹⁵

Loanwords show the most active divergence between traditions. The heibanka (flattening) trend documented in Shin Meikai across editions is uneven across publishers. In this trend, words like "trainer", "guitar", "rehearsal", "address", and -ing formations shift from accented to heiban (accent 0). As a result, a single katakana word can carry one accent in a 2014 Kindaichi printing and another in the 2016 NHK printing.⁵¹¹

The practical lesson is simple: pick one reference and stick with it across a study session. Mixing the NHK 2016 entry for one word and the 2014 Kindaichi entry for another will eventually surface contradictions that are dictionary artefacts, not pronunciation rules.

IPA-style high/low and downstep marks

The downstep symbol ꜜ

The downstep mark, ⟨ꜜ⟩ (Unicode U+A71C), is listed in the IPA chart's "Tones and Word Accents" block.⁴¹⁴

It marks the H→L transition at the exact point in the segmental string, or sequence of sounds. For Japanese, the standard convention is to place the symbol immediately after the last H mora: /jamaꜜ/ for 山, /okajamaꜜ/ for 岡山 in the convention used by Kawahara 2015 and the Wikipedia treatment.²⁸

In IPA, the downstep mark normally precedes the syllable whose tonal level it modifies. This carries over from the pre-Kiel convention for stress and accent marks.⁴ In Japanese linguistic transcriptions, the mark is placed immediately after the last H mora (that is, before the L mora it triggers), which is consistent with the "before the affected element" reading.⁴²

Unicode also defines a near-identical glyph, ⟨ꜝ⟩ (U+A71D, modifier letter raised down arrowhead), used as a typographic fallback where the U+A71C glyph is unavailable. Both forms appear in published Japanese linguistic work.¹⁵

The classic minimal-pair set for はし shows how the downstep maps onto the three accent positions.

橋はしが長ながい。²¹⁰
"The bridge is long."

In IPA: /haɕiꜜ ga nagai/. 橋 is odaka (accent 2 for a 2-mora word), and the downstep symbol lands at the right edge of はし.²¹⁰

箸はしを使つかう。²¹⁰
"To use chopsticks."

In IPA: /haꜜɕi o tsɯkaɯ/. 箸 is atamadaka (accent 1), and the downstep falls between mora 1 and mora 2.²¹⁰

Optional upstep ꜛ and per-mora H/L diacritics

The upstep mark, ⟨ꜛ⟩ (Unicode U+A71B), is the IPA's mirror symbol to the downstep and is rarely used in Japanese learner materials. It appears in transcriptions of tone-language pitch resets and in linguistic discussions of phrase-initial pitch rise.⁴

Per-mora level-tone diacritics (high acute ⟨◌́⟩, low grave ⟨◌̀⟩, mid macron ⟨◌̄⟩) are listed in the same IPA block.⁴¹⁴ An academic transcription may render 山 as /ja̋ma/ or /jámà/, with the diacritics placed over each mora's nucleus, instead of /jamaꜜ/. Both encode the same H→L drop.⁴²

These diacritics are most commonly seen in phonology papers, in the IPA pronunciation columns of Wiktionary entries, and in transcribed examples in Vance 2008 and Kawahara 2015.¹²²

Where you meet this notation

Wikipedia phonology articles for Japanese use ⟨ꜜ⟩ in transcribed examples for atamadaka, nakadaka, and odaka words.¹⁵⁸ Linguistic papers, including Kawahara 2015 and the Pierrehumbert / Beckman analyses, use the same convention.² Forvo and some Yomitan / Yomichan flashcard pipelines render IPA with ⟨ꜜ⟩ in their transcription field. Anki pitch-accent add-ons typically show both the accent number and the IPA-style transcription.¹⁰

Recognising ⟨ꜜ⟩ is enough for almost any learner workflow. Producing IPA from scratch is a skill needed only by writers of phonology papers.

Conversion table: same word, four notations

Table header and rows

The table below shows one standard example per pattern. It uses the same set as the cluster's overview article (さくら, ねこ, たまご, おとこ, with はし as the minimal-pair anchor), so this table is a strict superset of the overview's three-mora schematic.

Word	Pattern	Accent number	Overline-and-drop (legacy)	NHK 2016	IPA-style
桜さくら ("sakura")	heiban	0	bar over くら, continues onto the following particle	さくら￣	/sakɯɾa/ (no ꜜ)
猫ねこ ("neko")	atamadaka	1	bar over ね with fall mark; こ is L	ね＼こ	/neꜜko/
卵たまご ("tamago")	nakadaka	2	bar over ま with fall mark; た is L, ご is L	たま＼ご	/tamaꜜɡo/
男おとこ ("otoko")	odaka	3	bar over とこ; fall mark on the particle, not on the word	おとこ＼	/otokoꜜ/ (drop on prt)
橋はし ("bridge")	odaka	2	bar over し; fall on the particle	はし＼	/haɕiꜜ/
端はし ("edge")	heiban	0	bar over し, continues onto the particle	はし￣	/haɕi/
箸はし ("chopsticks")	atamadaka	1	bar over は with fall mark; し is L	は＼し	/haꜜɕi/

Accent numbers and ＼/￣ marking follow the NHK 2016 conventions;¹⁵ overline interpretation follows the Shin-meikai tradition;³ IPA transcription follows the conventions of Kawahara 2015 and the IPA Handbook.⁴²

Reading the table both ways

Given any one cell, you can derive the other three. The accent number is the most compact encoding. The IPA form gives the underlying phonology. The overline-and-drop and ＼/￣ forms are visual layouts of the same fact.

Worked walkthrough on the おとこ row: the accent number 3 says "drop after mora 3". The word has three morae (お, と, こ), so the drop falls on the particle. The legacy overline runs from mora 2 to the end of the word, with the fall mark sitting on the particle. The NHK 2016 form writes おとこ＼. The ＼ symbol falls after the last H mora of the word, which is exactly where the drop lands. The IPA form writes /otokoꜜ/.¹²

To reverse the walkthrough, start with the IPA /tamaꜜɡo/. The downstep sits between mora 2 (ま) and mora 3 (ご), which is accent 2. The legacy overline is a bar over ま with a fall mark at its right edge. The NHK 2016 form is たま＼ご. Pick any starting cell, and the other three follow mechanically.

Good to know

Accent numbers are not stress numbers

The accent number is a pitch-drop position, not a stress mark. An English speaker reading "accent 2" may instinctively interpret it as "stressed on mora 2". The actual reading is "high through mora 2, drop after mora 2".²⁸ The same misreading is the source of every common off-by-one error in pitch-accent decoding by L2 learners.²

Pre-1985 dictionaries used different conventions

Older printings of pitch-accent dictionaries (and academic Japanese-linguistics textbooks from the early Shōwa period onward) used the 甲乙丙 (kō-otsu-hei) tradition and a variety of bar-graph styles. You may still encounter these in second-hand copies of older editions. The conventions are intelligible but should not be adopted for current work.¹²²

Heiban looks ambiguous in citation form

A noun pronounced in isolation cannot be classified as heiban vs. odaka from its own contour alone. Both surface as L–H...H. The contrast appears only on a following particle.²⁸

This is a notation problem as much as it is a phonology problem. A notation that omits the particle cannot disambiguate, and the same surface contour has two underlying analyses. When a reference shows the noun without a particle, the accent number is the only thing that distinguishes 端 (0, heiban) from 橋 (2, odaka).²⁸

Particles get their own pitch cell

Some notations, such as OJAD's step diagram and the Shin-meikai bar style, draw the particle as a separate cell with its own H/L label.⁷³ Others, such as the bare accent number and the NHK ＼/￣ form, omit the particle entirely and require the reader to infer it.¹⁵

This is a structural observation about the notations, not a phonological claim. When a reference shows only the noun, the reader must use the accent number to decide whether the following particle is H (heiban) or L (every other pattern).¹²

A reading order for new learners

The cluster's overview article covers the H/L model and the four patterns. This page is a second-pass reference for translating between the conventions you encounter in study tools.²

Below N4, the recommendation is to read only the overview article and defer notation translation until you are actively using two or more reference tools that disagree on the same word.

See also

• Japanese Compound-Word Pitch Accent: How Two Words Combine into One Accent Pattern
• Regional Pitch Accent in Japanese: Kansai (Keihan), Tohoku, and the Accentless Dialects
• The Japanese Vowel Inventory: Five Vowels, Done Right
• The Japanese Consonant Inventory: Phonemes, Allophones, and the Kana Chart

References

Footnotes

1. NHK放送文化研究所 (ed.). 『NHK日本語発音アクセント新辞典』. NHK出版, 2016. ISBN 978-4-14-011345-5. Publisher product page: https://www.monokakido.jp/en/dictionaries/nhkaccent2/index.html ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶ ↩¹⁷ ↩¹⁸ ↩¹⁹ ↩²⁰ ↩²¹ ↩²² ↩²³ ↩²⁴ ↩²⁵ ↩²⁶ ↩²⁷ ↩²⁸ ↩²⁹ ↩³⁰ ↩³¹ ↩³² ↩³³

2. Kawahara, Shigeto. "The Phonology of Japanese Accent." Chapter 11 in Haruo Kubozono (ed.), Handbook of Japanese Phonetics and Phonology, De Gruyter Mouton, 2015. https://user.keio.ac.jp/~kawahara/pdf/HandbookAccentPublished.pdf ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶ ↩¹⁷ ↩¹⁸ ↩¹⁹ ↩²⁰ ↩²¹ ↩²² ↩²³ ↩²⁴ ↩²⁵ ↩²⁶ ↩²⁷ ↩²⁸ ↩²⁹ ↩³⁰ ↩³¹ ↩³² ↩³³ ↩³⁴ ↩³⁵ ↩³⁶ ↩³⁷ ↩³⁸ ↩³⁹ ↩⁴⁰ ↩⁴¹ ↩⁴² ↩⁴³ ↩⁴⁴ ↩⁴⁵ ↩⁴⁶

3. 金田一春彦 (Kindaichi, Haruhiko) (supervisor); 秋永一枝 (Akinaga, Kazue) (ed.). 『新明解日本語アクセント辞典』 (Shin Meikai Nihongo Akusento Jiten), 第2版. 三省堂 (Sanseidō), 2014. ISBN 978-4-385-13672-1. Library catalogue record: https://catalog.princeton.edu/catalog/6971989 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹²

4. International Phonetic Association. Handbook of the International Phonetic Association: A Guide to the Use of the International Phonetic Alphabet. Cambridge University Press, 1999. Tones and Word Accents section. Reference chart: https://www.internationalphoneticassociation.org/content/ipa-tones-and-word-accents ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹

5. 国際交流基金. 「『NHK 日本語発音アクセント新辞典』」. 日本語教育通信 本ばこ. https://www.jpf.go.jp/j/project/japanese/teach/tsushin/bookshelf/201610.html ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶

6. jcinfo.net. "Japanese Pitch Accent Notation Tool." https://www.jcinfo.net/en/tools/ja-accent (limitation: third-party converter; used here only as evidence that learners encounter the same word printed in three notation styles online). ↩

7. Online Japanese Accent Dictionary (OJAD). Minematsu Lab, Graduate School of Engineering, The University of Tokyo. https://www.gavo.t.u-tokyo.ac.jp/ojad/eng/pages/home ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹

8. "Japanese pitch accent." Wikipedia. https://en.wikipedia.org/wiki/Japanese_pitch_accent (tertiary cross-check; primary references therein are Vance 2008 ¹², Kawahara 2015 ², Shin Meikai Nihongo Akusento Jiten ³, and NHK Nihongo Hatsuon Akusento Jiten ¹). ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹²

9. Online Japanese Accent Dictionary (OJAD). "Intro to 4 Features of OJAD." https://www.gavo.t.u-tokyo.ac.jp/ojad/eng/pages/usage ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

10. tatsumoto. "Japanese Pitch Accent Primer." https://tatsumoto.neocities.org/blog/japanese-pitch-accents (limitation: independent blog; cited only for the 橋 / 端 / 箸 accent-number triple where it agrees with the NHK ¹ and Shin-meikai ³ dictionary entries). ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

11. 白岩 広行 (Shiraiwa, Hiroyuki). 「新明解国語辞典からみたカタカナ語アクセントの変遷（三上）」 (research paper on katakana-word accent change as documented in the Shin Meikai kokugo jiten across editions). Hiroshima University. https://home.hiroshima-u.ac.jp/jshira/papers/ ↩ ↩²

12. Vance, Timothy J. The Sounds of Japanese. Cambridge University Press, 2008. Pitch-accent chapter (the work surveys vowels, consonants, syllable structure, accent, and intonation of standard Tokyo Japanese). ↩ ↩² ↩³ ↩⁴

13. Nakamura, Nobuaki; Hirano, Hiroya; Minematsu, Nobuaki; et al. "Development of a Web Framework for Teaching and Learning Japanese Prosody: OJAD (Online Japanese Accent Dictionary)." Interspeech 2013. https://www.isca-archive.org/interspeech_2013/nakamura13_interspeech.html ↩ ↩²

14. International Phonetic Association. IPA chart, revised to 2005 (Kiel 2015 reproduction). PDF: https://www.internationalphoneticassociation.org/sites/default/files/IPA_Kiel_2015.pdf ↩ ↩²

15. "Downstep." Wikipedia. https://en.wikipedia.org/wiki/Downstep (tertiary cross-check on the Unicode codepoint U+A71C and on the typographic-fallback glyph U+A71D ⟨ꜝ⟩; primary reference for the IPA convention is the IPA Handbook ⁴). ↩ ↩²