Lesson 9 · — finishing the chart

The last twenty-one, and the kanji still showing through

Twenty-one characters finish the second script — and then the marks, the small kana and the long-vowel bar finish katakana itself. Several of the new shapes barely changed from the kanji they were cut out of — is the whole of , is tilted — and that, rather than their simplicity, is what makes them easy to get wrong.

≈32 minutes · stroke data from KanjiVG · every word checked against Jisho.org · the error data is a writing test on 430 Japanese schoolchildren

Lesson 9 of the course

Twenty-five plus twenty-one is forty-six, and forty-six is the whole of katakana — the same arithmetic that finished the hiragana in lesson 2. Every sound below is one you have been saying since the first week. What is new is only which marks carry it.

1.What is left of the grid

Four consonant rows, one row of two, and one character that belongs to no row at all. The holes are in exactly the places they were in hiragana, and for exactly the same reasons.

rowaiueo
hは ハひ ヒふ フへ ヘほ ホ
mま マみ ミむ ムめ メも モ
yや ヤ—ゆ ユ—よ ヨ
rら ラり リる ルれ レろ ロ
wわ ワ———を ヲ
nん ン — a row of one, and a full beat of its own

Two things on that table are not quite what they look like. The w row's last cell, , is a character modern Japanese does not write at all — section 6 is about why it is still on the chart. And in the h row is not a printing accident: it really is the same shape as its hiragana.

2.The kanji still showing through

Lesson 8 set out the operation that made these shapes: hiragana is a whole kanji written in flowing cursive, katakana is one piece of a kanji kept straight. In this half of the chart that piece is often almost the entire character.

KatakanaCut fromWhat was kept
ハ八all of it — the two strokes are the whole kanji
ミ三all three bars, slanted
ホ保the inside it
ロ呂the top half
ワ和the on its right
Tap any character in section 3 and its card names the kanji it was cut from, and whether its hiragana came from the same one.

Trick 1 · The visible ancestor is a trap here, not a help

Because so little was cut away, several of these characters are still nearly their source kanji — and that is precisely where Japanese children go wrong. The writing study behind section 4 marks a character down whenever it could be mistaken for its lookalike kanji, and names four of them explicitly: /, /, / and /.1 Two of those are the character and its own ancestor. Knowing where a shape came from tells you why it looks like that; it does not tell you which of the two you are looking at.

Fourteen are siblings. Five are strangers. Two are arguments.

Same question as day 8, asked of the remaining pairs: did the katakana and its hiragana come from the same kanji? Where they did, the family likeness is inherited and can be leaned on. Where they did not, any resemblance is an accident of two independent simplifications.

PairKatakana fromHiragana from
ハ は八波
ミ み三美
ム む牟武
ル る流留
ヲ を乎遠
The five unrelated pairs. Fourteen more — — each share one ancestor with their hiragana. Across the whole script that is 29 siblings and 13 strangers.

The two arguments are and , and they are unsettled in different ways. For the sources split between and ; since is certainly , whether that pair are siblings is itself the dispute — the same position was in on day 8. is unsettled far more deeply: proposals include , and , and the strongest may not be a kanji at all but the old annotation mark for a nasal beat.3 Its hiragana is not in doubt — is written cursively.

Trick 2 · Katakana are still not the shorter script

Day 8's twenty-five took 62 strokes against hiragana's 57. These twenty-one come out at 46 against 47 — a dead heat. Add both halves and the whole of katakana is 108 strokes against hiragana's 104. The angular script is not the quicker one; its advantage is that every stroke is a straight line or a simple hook, so it is easier to get right.

3.Write them, hear them, use them

Tap any character below. It draws itself in the correct stroke order — numbered, one stroke at a time — says itself aloud, names the hiragana that shares its sound and the kanji it was cut from, and gives three real words that use it, each glossed in English and in Arabic.

ハ ヒ フ ヘ ホ マ ミ ム メ モ ヤ ユ ヨ ラ リ ル レ ロ ワ ヲ ン

The interactive chart needs JavaScript. The 21 characters are listed above in row order, and the katakana stroke chart has all forty-six drawn out.

Four cards carry a warning rather than a word list

, and almost never begin a word, so their cards show words that contain them instead — the same arrangement needed on day 8. never begins one at all. And has no word list because there are no words; see section 6.

4.The ones that actually get written wrong

Which katakana cause trouble is a question with a measured answer rather than an opinion. A 2019 study tested all 45 katakana in handwriting on 430 Japanese children in school years 1–3 and sorted every error into five kinds.1 The findings do not match the usual beginners' list, and the mismatch is the useful part.

ErrorWorst charactersWhat it means for you
Wrote a different katakana ソ→ン, ル↔レ, リ→ソ, マ→ユ, ヨ→ヲ, ワ→ウ Shape confusion proper — the pairs worth drilling against each other.
Mirror writing ヨ (18), ミ (10) Not knowing which side a stroke starts or attaches on.
Malformed shape モ (27), ヒ (21), ミ (11) Getting a stop, hook or sweep wrong at the end of a stroke.
Wrote the hiragana instead モ, ヤ, リ These three genuinely resemble their twins — see below.
Left it blank ハ, ワ (21 each), ロ, ユ, ヨ Could not recall the shape at all. A different problem entirely.
Counts are children out of 430, across all three school years. Only characters from this lesson are listed; , , and from day 8 top several of these columns too.

Trick 3 · and are the same puzzle as and , with one mark instead of two

Day 8 promised this rule would settle when arrived. It does, unchanged — read the direction the pen travels, not the finished picture:

The giveaway on a finished letter is where the strokes begin: 's start down the left-hand side, 's start along the top. This substitution — writing for — was the single commonest one in the study, and it was still happening in year three.

Trick 4 · For , and , check which side the horizontals attach to

is the most mirror-written character in the whole script — 18 of 430 children drew it backwards. Its three horizontals all meet the upright on the right, making a comb whose teeth point left. is second, for a related reason: all three of its strokes sweep right to left, and reversing that habit reverses the character. Neither is a memory problem — write them slowly once, in the right direction, and watch the start of each stroke.

Trick 5 · Five of these fail on a blank page, not on a lookalike

, , , and lead the "left it blank" column: the children were not confusing them with anything, they simply could not produce them. That needs a different exercise from the drill below. Recognition — seeing and knowing it says wa — will not fix it. Write them from the sound, on paper, with the chart face down.

One pair worth demoting, on the same evidence that demoted / on day 8: against is in every beginner's list and appears nowhere in this study's error data. What Japanese children actually write instead of is . Both are worth a glance; only one is attested.

5.Read them cold

A character appears, you name its sound — four options, by tapping or by pressing 1–4. There is deliberately no play button on the question, because hearing it is the answer. Miss one and it stops, names the hiragana that shares the sound, shows how the strokes go, and lists the words from the cards above that contain it.

The wrong answers come from the table in section 4 before anywhere else — against , against , against , against . Only then does it fall back to the same row, and only then to anywhere. Wrong answers that are obviously wrong measure nothing.

The drill needs JavaScript. All twenty-one characters, their sounds and their words are in section 3 above.

6., the character that retired

is on the chart, has a stroke order, and partners — which means, as lesson 2 established, that it is pronounced simply o. It is also a character you can read a great deal of modern Japanese without ever meeting. Three independent signs of that:

  1. The object particle is always written , in hiragana, even in a sentence otherwise thick with katakana. That was 's only job — so it took the whole job with it.
  2. Foreign words with a wo sound are now spelled , not .4
  3. The dictionary has no entry for a single word containing it, and the writing study in section 4 tested 45 characters, not 46 — it left out on the grounds that children are not taught to write it.1

So learn it to read — it turns up in documents written before the 1946 reforms, when ordinary Japanese was set in kanji and katakana rather than kanji and hiragana — and do not go looking for a use for it. This is the same kind of honest gap as day 5's row and day 7's row: the absence is the fact, not a hole in the chart.

7., the mark that holds a vowel

Day 8 ended by naming the one thing its words needed and its chart did not contain. Here it is. — the chōonpu — is not a letter and not a dash. It repeats whatever vowel came before it for one more beat.

WrittenReadsBeats
コヒ ko · hitwo
コーヒー ko · o · hi · ifour
Same two consonants, twice the length. The mark is doing the same work does — taking a full mora of its own.

This is why so many of the words on the cards above carry it: , , , . Leave it out and the word is not merely mispronounced, it is misspelled.

Trick 6 · One mark, whichever vowel — and that is the difference from Arabic

Arabic writes a long vowel by writing a letter, and which letter depends on the vowel: ا lengthens a in بَاب, و lengthens u, ي lengthens i. Katakana uses one neutral mark for all five — simply inherits the vowel to its left. So you never choose which lengthener to write; you only decide whether the beat is there. Hiragana, confusingly, does it the Arabic way instead, by writing the vowel out: , not .

One practical note: in vertically-set Japanese the mark turns with the text and is written as a vertical stroke, not a horizontal one. It is the only character on the chart that changes direction with the writing.

8.The marks, in the other script

Twenty-five more characters, and not one new shape among them. Lesson 5's rule carries over whole: two small strokes voice a row, one small circle turns h into p. Everything you learned about is true of , which is why this section is short.

MarkDoesRows
゛ tenten voices the consonant カ→ガ · サ→ザ · タ→ダ · ハ→バ
゜ maru turns h into p ハ→パ — this row only

ガ ギ グ ゲ ゴ ザ ジ ズ ゼ ゾ ダ ヂ ヅ デ ド バ ビ ブ ベ ボ パ ピ プ ペ ポ

The interactive chart needs JavaScript. The 25 characters are listed above in row order, and the katakana stroke chart has all seventy-one drawn out.

Trick 7 · The row was the hard one in hiragana. Here it is the easiest

Lesson 5 had to bend its own rule for this row: there is no common two-kana word in the language containing , because in native Japanese a p only turns up after or . In katakana the same five characters are among the most useful on the chart — , , , , .

The reversal is not a coincidence, it is the same fact twice. A word can start with p in Japanese essentially only if it was borrowed — and katakana is the script borrowed words are written in. The row hiragana could barely fill is the row katakana lives in.

Trick 8 · and are dead here too

sounds exactly like and exactly like , and modern katakana simply does not use them — the dictionary has no word containing either, so both cards below are empty on purpose. This is the same absence and had on day 5, and the same kind of absence as in section 6. Learn them to recognise; never reach for them when writing. When you hear ji or zu, it is or .

The arithmetic of a mark

A tenten always adds exactly two strokes to its base and a maru always exactly one — true of all 25 here, as it was of all 25 in hiragana. So these 25 katakana take 103 strokes against their hiragana's 104: another dead heat, and the same answer as Trick 2. Each card below splits the count for you — the strokes of the base, then the mark.

One character exists in katakana that hiragana has no version of at all: , which is wearing the tenten and gives a v sound for foreign words. It is not on the chart above because it is not one of the standard forty-six-plus-twenty-five, and in practice you will meet the same word spelled both ways — and are both current, and the dictionary tags both common. Read it; you do not need to write it.

Tell the marked from the unmarked

A second drill, over the four base rows and their marked forms together — against . This is the only thing in the section that is genuinely new to practise: the shapes are known and the rule is known, so what is left is noticing a diacritic the eye is trained to skip.

The drill needs JavaScript. All 25 marked characters and their bases are in the chart above and in section 3.

9.The small , and the holes they fill

Same rule as lesson 7, and the same arithmetic. Eleven characters end in -i and can be palatalised — plus — times three small vowels, is 33. The small kana replaces the i, it does not follow it: + is kya, one beat, not ki-ya.

キャ キュ キョ · シャ シュ ショ · チャ チュ チョ · ニャ ニュ ニョ · ヒャ ヒュ ヒョ · ミャ ミュ ミョ · リャ リュ リョ · ギャ ギュ ギョ · ジャ ジュ ジョ · ビャ ビュ ビョ · ピャ ピュ ピョ

The interactive chart needs JavaScript. All 33 combinations are listed above in row order.

Trick 9 · The two yōon charts are complements, not copies

Lesson 7's chart had six dashed cells it could not fill — — and said the first three "live in katakana loanwords". They do, and here they are:

Empty in hiraganaFilled in katakana
ひゅヒュ — ヒューズ fuse
みゅミュ — ミュージック music
びゅビュ — インタビュー interview
ぴゅピュ — コンピューター computer
ぴょピョ — ピョンヤン Pyongyang

And it runs the other way. The chart above is empty at — and hiragana fills every one: , , , , . Those are native words, and native words are not written in katakana; the cells katakana fills are loanwords, and loanwords are not written in hiragana. Neither script needs all 33. Between them they use them all — except one.

Trick 10 · is the last empty cell in the language

Of the 33, exactly one has no common word in either script: /. This is the fourth time the p row has been the awkward one — day 5 had to bend its word rule for it, day 7 dashed out all three of its cells, and section 8 above showed the full-size row reversing into the easiest row on the chart. All four are the same underlying fact: a Japanese p is either borrowed, or it follows or . Nobody has borrowed a pya.

10.The small — a beat of silence

The last small character, and the only one still outstanding in both scripts. is a small and makes no tsu sound at all. It doubles the consonant that follows it — and it takes a full beat while doing nothing audible.

WrittenReadsBeats
カプ ka · putwo
カップ ka · (hold) · puthree
The middle beat is silence with the mouth already closed on the p. Skipping it is as wrong as skipping a vowel.

That makes three things in katakana that take a beat without being a syllable: , the long-vowel from section 7, and now . Japanese counts beats, not letters — this is the same point lesson 2 made about , arriving for the third time.

Trick 11 · You already know this from شَدَّة

Arabic doubles a consonant with the shadda — مُدَرِّس (mudarris), where the mark sits on the doubled letter itself. Japanese doubles a consonant by putting a whole separate character in front of it. Same idea, opposite method — and it is exactly the split lesson 2 found with against the accusative ending: Japanese hangs a character off the word, Arabic changes the letter.

The useful part for reading: if you can already hear the difference between دَرَسَ and دَرَّسَ, you can already hear the difference between and . It is not a new sound, only a new spelling of one you have.

Two rules worth having: never begins a word, and it never sits before a vowel — there has to be a consonant for it to double. It is everywhere in the vocabulary already on this page:

WordReadsMeans
クッキーku·k·kīcookie
ベッドbe·d·dobed
ペットpe·t·topet
サッカーsa·k·kāfootball
バッグba·g·gubag
キッチンki·t·chinkitchen

11.The sounds katakana had to invent

Hiragana has five small vowels too, and essentially never uses them. Katakana uses them constantly, because katakana has a job hiragana does not: spelling foreign words that contain sounds Japanese does not have. There is no fa, no ti, no je in native Japanese — so katakana builds them.

The method is the one you already know from the yōon. Take the character whose consonant is right, and swap its vowel with a small one:

WantBuild fromGivesAs in
faフ + ァファファイル file
fiフ + ィフィフィルム film
feフ + ェフェカフェ café
foフ + ォフォフォーク fork
tiテ + ィティティッシュ tissue
diデ + ィディディナー dinner
cheチ + ェチェチェック check
sheシ + ェシェシェフ chef
jeジ + ェジェジェット jet
wiウ + ィウィウィスキー whisky
weウ + ェウェウェブ web
woウ + ォウォthe replacement for
Read the middle column and the logic is visible: is the only f-ish character, so every f sound is built on it; is chi, so che is built on it and ti has to come from instead.

This closes the loop on section 6. The reason retired is sitting in the last row of that table: does the job, is built from characters you know, and follows a rule that generalises — where was a single character with a single use.

12.Katakana, finished

Every katakana word on this page is now fully readable, with nothing taken on trust:

WordWhat it needed
コンピューターン · small ュ · long ー — sections 1, 9, 7
キャンセルsmall ャ · ン — section 9
ショッピングsmall ョ · small ッ · marked グ — sections 9, 10, 8
フォークsmall ォ · long ー — sections 11, 7
ジェットコースターmarked ジ · small ェ · small ッ · long ー

What genuinely remains is outside the chart: a handful of archaic characters ( in katakana form, and , which you will not meet), and kanji — which is a different kind of learning altogether and belongs to a different day.

Both halves of the script, the marks and the small kana are on the katakana stroke chart; the 33 combinations have their own sheet at yōon — the small ゃ ゅ ょ, which now shows both scripts.

The katakana stroke chart now covers all forty-six and is built to print — keep it beside the paper while you practise. Its companion is the hiragana stroke chart.

13.Sources

  1. 森田安徳 (2019), 「小学1〜3年児童のカタカナ書字の誤りに ついて — 字形の誤りを中心に」, 神戸親和女子大学 児童教育学研究 38, pp. 301–311 — a handwriting test of 45 katakana on 430 children in school years 1–3, with every error classified and counted. Section 4's whole table comes from its 表5 to 表11, as does the marking rule quoted in Trick 1 and the fact that was excluded from the test. It is also why / is demoted here: it does not appear in the data.
  2. Tofugu, “Learn Katakana” — the clearest free walkthrough of the full set. Read this one before the next class, and use it to drill the second half against the first.
  3. Man'yōgana ancestry for all twenty-one pairs, cross-read across six sources rather than taken from one: en.wikipedia's derivation table, sljfaq.org, en.wiktionary, and the Japanese-language ja.wiktionary 字源 lines, ja.wikipedia 片仮名 and ja.wikipedia ン, which is where 's competing theories and the annotation-mark proposal are set out. They disagreed in three places and the disagreements are recorded with the data rather than resolved silently.
  4. ja.wiktionary, — the source for section 6's second point: once wrote the /wo/ of European words and 「現在ではこれに変えて「ウォ」が用いられる」.
  5. KanjiVG, © Ulrich Apel, CC BY-SA 3.0 — every stroke order and direction on this page. It is what settles against : prose can describe the difference, but the path data records which end of the stroke the pen starts from.
  6. Jisho.org, serving JMdict/EDICT — every reading and gloss, and the commonness tag that decided which words earned a place. It refused and , exactly as it refused their hiragana forms in chapter 5, and it has no entry at all containing .
  7. NHK World — Easy Japanese — free lessons with native recordings. Worth using for section 7: the extra beat a adds is something a speech synthesiser renders only approximately.
  8. Madinah Arabic and Almaany — the Arabic glosses, their harakat, and the long-vowel comparison in Trick 6.