Lookalike Hiragana: How to Tell the Most-Confused Kana Apart
Lookalike hiragana are the small set of kana whose outlines overlap so closely that beginners may read them as the same character. The fix is to train the eye to scan for one specific feature.1 This article walks through the five core within-hiragana confusion groups, plus two honorable mentions. It names the single structural feature that tells each pair apart and anchors every diagnosis in stroke order or stroke count, so you can verify it yourself.
Overview
Why these pairs trip everyone up
Hiragana descends from man'yōgana, an early system that used kanji for their sounds, written in the highly cursive sōsho (草書) style. Each modern hiragana is a simplified rendering of a whole kanji.12 Because the simplification followed the speed of brush motion rather than visual contrast across the set, several characters ended up with near-identical outlines. They differ only in one structural feature: a loop, a crossbar, a closure, or a stroke-ending direction.1
The shared-ancestor signal is real in a few cases. ね, れ, and わ all simplify cursive forms whose right-hand component resolves with a vertical drop plus a flourish on the right. Their kanji of origin (祢, 礼, 和) all carry such a right-side element.3
Most lookalike pairs are convergent simplifications rather than genealogically related shapes. In other words, the cursive ancestors did not share strokes, but the simplification process still landed on similar outlines.1 The Hiragana History article on this site covers the man'yōgana derivation in depth.
How this article is organized
Each section below takes one confusion group and follows the same four steps: name the pair or triplet, point to the one structural feature that tells them apart, give the stroke-order or stroke-count anchor, and offer one durable distinguisher. You can jump directly to the pair you are struggling with.
What this article does not cover
Hiragana ↔ katakana cross-script lookalikes (り vs リ, へ vs ヘ, か vs カ, や vs ヤ) are not in scope here. For that family, the key question is "which script is this?" rather than "which kana within hiragana?" The structural-feature approach used in this article does not transfer.3
Pronunciation, reading drills, and stroke-by-stroke writing instruction also live elsewhere. The Hiragana Chart article handles the full inventory. The Hiragana Stroke Order article walks through how each character is drawn. This page is only about telling similar shapes apart by sight.
The five core within-hiragana lookalike groups
The five groups at a glance
The five within-hiragana confusion groups widely cited in beginner pedagogy are listed below with the one-word diagnostic for each.45
| # | Group | Structural diagnostic |
|---|---|---|
| 1 | ぬ vs め | Bottom-right loop vs. no loop |
| 2 | わ vs れ vs ね | Right-side ending: open sweep vs. hook vs. closed loop |
| 3 | さ vs ち | Mirror-image bottom curve: left vs. right |
| 4 | は vs ほ | Number of crossbars on the right shape: one vs. two |
| 5 | る vs ろ | Bottom of the stroke: closed loop vs. open tail |
Two honorable-mention groups (き vs さ, い vs り) follow the core five; both are disambiguated by gross features that the eye picks up quickly with regular reading.5
ぬ vs め: the bottom-loop test
What they share
Both kana are written in 2 strokes,6 and both end with a curving right-hand stroke that crosses an initial diagonal. At small font sizes, their outlines are almost identical.
The kanji of origin are unrelated (ぬ from 奴, め from 女),3 so the outline overlap is convergent rather than genealogical. The two kana came from independently simplified cursive forms that happened to end up with the same overall shape.
The one structural difference
ぬ closes a small loop at the bottom right, where the second stroke crosses itself and tucks back inside. め has no loop. Its second stroke crosses the first and exits as a clean tail.75
The sci.lang.japan FAQ states the diagnostic as "The end of ぬ (nu) makes a loop."5 LearnTheKana frames the same feature as "The ぬ (nu) character simply curls to the right, whereas the め (me) character doesn't."7
Stroke-order anchor
Both are 2-stroke kana.6 The first stroke is the short diagonal at the upper left; the second stroke is the long curving return. The second stroke of ぬ returns through itself to form the closed loop; the second stroke of め does not return.7
If you can already write either kana correctly, you already know the difference. The visual test usually fails only when a reader has not yet practiced writing the characters by hand.
One durable distinguisher
Tofugu's beginner mnemonic for ぬ is "some noodles … 100% smooth and bendable, like noodles," and for め it is "a beautiful eye … because of the makeup on it."8 LearnTheKana's variant frames the same image as a noodle held by chopsticks: ぬ has the "extra long" noodle (mnemonic "no" for "noodle"), and め has dropped the noodle on the ground (mnemonic "me" for "mess").7
The structural test is binary, no matter which mnemonic you prefer: look at the tail. Loop = ぬ; clean exit = め.
ぬいぐるみがほしい。9
"I want a stuffed toy."
めがねをかけます。9
"I wear glasses."
A clean noun-level minimal pair on a single mora, a rhythm unit in Japanese, is hard to find. め appears most often as a verb-ending morpheme rather than as a standalone word. The useful classroom contrast is therefore the loop test inside ぬいぐるみ vs めがね, as above.10
わ vs れ vs ね: the right-side ending
What they share
All three kana are written in 2 strokes.6 In all three, the first stroke is a short vertical drop on the left. The second stroke begins at the top, runs right, drops down, and finishes on the right side. The difference lives in the final few millimeters of that second stroke.5
The cursive ancestry partly explains the convergence: わ derives from 和, れ from 礼, and ね from 祢.3 All three source kanji carry a right-side component (口, 礼's right radical, 尔). That is why the cursive simplifications converged on the same "left vertical + right element" outline.
The three endings, in order of openness
The three kana sit on a single openness axis. わ ends in a clean curve that sweeps out to the right, with no hook and no loop. It is the most "open" ending. れ ends in a curve that hooks back up like a fishhook: open, but with a flick. ね ends in a fully closed loop, the most "closed" ending.
The sci.lang.japan FAQ states the ね case as "The end of ね (ne) makes a loop."5
The whole diagnostic is the progression open → hook → loop.
Stroke-order anchor
All three are 2-stroke kana.6 The first stroke is identical across the three. Only the second stroke's final motion differs.
Tofugu explicitly flags the lookalike cluster on each character's mnemonic page. ぬ "looks like some noodles. There are several other kana that are similar to this one (れ, め, ね, わ)." わ "looks especially similar to ね." れ is "similar to め, わ, ぬ, and ね. What makes this one different is the curve at the back." ね is distinguished from "(ぬ, れ, め, わ) … Because it has a loop at the end for the tail."8
One durable distinguisher
The "open / hook / loop" left-to-right reading order maps loosely to the kana names' Latin onsets w → r → n. わ is wide open, れ has a small hook, and ね closes into a loop, matching Tofugu's "Nelly the cat … the loop is for ne" image.8
The binary test for any single character: trace the second stroke's tail. No flourish = わ; small upward flick = れ; closed circle = ね.
私は学生です。9
"I am a student."
毎日練習します。9
"I practice every day."
猫が好きです。9
"I like cats."
The triplet does not produce a clean three-way minimal set on a single mora, because わ, れ, and ね have different lexical roles. A useful drill is to hold three single words side by side: わたし / れんしゅう / ねこ. Reading them in immediate sequence forces the eye to land on the end of the second stroke each time.
さ vs ち: the mirror-image second stroke
What they share
Both kana have a top horizontal stroke crossed by a vertical-ish stroke, plus a curving bottom stroke. Their overall outlines are mirror images of each other along the vertical axis.45
Their stroke counts differ: さ is 3 strokes,6 and ち is 2 strokes.6 In modern print fonts, the top crossbar plus vertical of さ can connect into what looks like a single stroke, but the canonical count remains 3.
As with ぬ/め, the outline overlap is convergent rather than genealogical. さ derives from 左 (also attested 散), and ち from 知 (also attested 千).3
The one structural difference
さ curves toward the left at the bottom; ち curves toward the right.4 The bottom stroke is a true mirror image along the vertical axis.
A secondary cue appears in many handwritten styles and several print styles: in さ, the bottom curve is often detached from the upper element. In ち, the bottom curve is continuous with the vertical stroke. This holds in most teaching fonts, but not all. The mirror direction is the more reliable cue.
Stroke-order anchor
さ is 3 strokes: the top horizontal, the vertical-ish stroke crossing it, and the bottom left-curving stroke.6 ち is 2 strokes: the top horizontal, then a single continuous stroke that drops down and curves to the right.6
Stroke count is the diagnostic when you can see both kana clearly: 3 = さ; 2 = ち. The mirror-direction test handles small fonts or fast handwriting, where the stroke count may be ambiguous.
One durable distinguisher
Tofugu frames さ as "two hands stirring a bowl of salsa," with the hands pulling the bottom curve leftward toward the body. It frames ち as "that forced smile you have to make every time you're in a group photo" when told to say "cheese," with the smile pulling to the right.8
The structural test does not depend on the mnemonic: the tail of the bottom curve points left for さ and right for ち.
桜が咲く。9
"The cherry blossoms bloom."
父は先生です。9
"My father is a teacher."
No clean single-mora noun pair exists for さ vs ち. In practice, beginners meet the contrast in the verb stem さく ("to bloom") and either the doubled-character word ちち ("father") above, or the given name さち (e.g. 幸 Sachi, "happiness"). The name さち puts both kana in one word, so the eye must flip from left-curving to right-curving in adjacent positions.10
は vs ほ: count the crossbars
What they share
Both kana belong to the h-row (は ひ ふ へ ほ). Both have a vertical stroke on the left, a small horizontal stroke crossing it, and a closed shape on the right. Their left halves are structurally identical.11
Cursive ancestry: は derives from 波 (also attested 八), and ほ from 保.3 Both source kanji carry a right-side component with horizontal cross-strokes. The modern hiragana right-side rectangle and its crossbars are simplifications of those features.
The one structural difference
The right-side closed shape is crossed by one horizontal bar in は (ha) and by two horizontal bars in ほ (ho).11 Counting the crossbars on the right side is the whole diagnostic.
Tofugu encodes this image as は = "uppercase H plus lowercase a → 'Ha!'" and ほ = "a chimney on the left, Santa Claus on the right uttering 'ho ho ho.'"8
Stroke-order anchor
は is 3 strokes: the left vertical, the top horizontal, and the right closed shape drawn as one continuous stroke.6 ほ is 4 strokes: as in は plus a second horizontal bar across the right shape.6
At writing speed, stroke count is the diagnostic: 3 strokes = は, 4 strokes = ほ.
One durable distinguisher
The test that works at any size or angle is "count the rungs on the ladder": は has one rung on the right, and ほ has two. The image works in print or handwriting, and it does not require the Tofugu chimney-and-Santa image.8
花が咲きました。9
"The flowers bloomed."
本を読みます。9
"I read a book."
A clean lexical minimal pair on a single mora: はし ("bridge" 橋, "chopsticks" 箸, "edge" 端; disambiguated by pitch and context) vs ほし ("star" 星).10 The following vowel is the same, and only the h-row consonant kana changes. Reading はし and ほし side by side forces the crossbar-count test on a real-word contrast. Another useful pair is はな ("flower" 花 or "nose" 鼻) vs ほね ("bone" 骨),10 which adds the わ/れ/ね test on the second character.
る vs ろ: closed loop vs. open tail
What they share
Both kana are written in a single stroke.6 The stroke starts at the top left, sweeps across, comes down, and curves at the bottom. Their outlines are nearly identical until the very last millimeter.
Cursive ancestry: る derives from 留 (also attested 流), and ろ from 呂.3 Both source kanji have rounded or stacked lower elements. The modern hiragana's curling bottom is the cursive simplification of those forms.
The one structural difference
る closes the bottom into a small loop. ろ ends with the curve open, with no loop.4 The loop on る is small enough to disappear in low-resolution screen fonts or fast handwriting. That is exactly why the pair is confusing.
Stroke-order anchor
Both are 1-stroke kana.6 The only differing motion is at the very end: る tucks the pen back to close the loop, while ろ lifts the pen.
One durable distinguisher
Tofugu frames the pair as る = "a crazier route. There is a loop at the end" vs ろ = "a plain old road" with no loop.8 The road-route pairing is mnemonically efficient because both words start with R, and the more-decorated character (route, with a loop) maps to る.
The test does not depend on the mnemonic: look at the bottom of the stroke. Closed circle = る; open tail = ろ.
家に帰る。9
"I'm going home."
廊下を走る。9
"I run in the hallway."
The second sentence is a small stress test for the diagnostic, because る and ろ appear in the same line.
A clean lexical minimal pair: るす (留守, "absent from home," JLPT N4) vs ろうか (廊下, "hallway," JLPT N5).10 The contrast at the first mora forces the loop test in word-initial position, where the reader's eye does the most work. A smaller second pair is もる (盛る, "to pile up / serve in a bowl"; or 漏る, "to leak") vs もろい (脆い, "fragile; tender-hearted"),10 which puts the loop test at the second mora.
Honorable mentions: two pairs that occasionally cause trouble
き vs さ: count the strokes
Both kana have a horizontal element at the top crossed by a vertical-ish element, so their outlines resemble each other.5 き has two horizontal crossbars at the top. In modern print fonts, they are sometimes drawn as one connected stroke, but conceptually there are two. さ has one.5 The sci.lang.japan FAQ states it directly: "き (ki) has two lines at the top but さ (sa) only one."5
Stroke counts: き is 4 strokes (or 3 in connected handwriting styles), and さ is 3 strokes.6 Tofugu's mnemonic anchors the diagnostic visually: "Notice how much it [き] resembles a key,"8 with the two horizontal crossbars as the key's teeth.
い vs り: which side is taller, and which way they lean
Both kana are 2-stroke characters made of two short vertical-ish elements with a gap between them.6
い (i) has two strokes of roughly the same height. The left curves down-right, and the right is a small upright tick.8 り (ri) has a shorter left stroke and a noticeably longer right stroke that hooks downward and to the right. It often extends past the baseline of the left stroke.5 The right stroke of り is the diagnostic.
Tofugu pairs the mnemonics neatly: い is "a couple of eels hanging out … upright … trying to mimic the letter 'i'," and り is "the reeds are swaying in the wind."8 The image captures the asymmetry as wind bending one reed more than the other.
Why these are honorable mentions, not core
Both pairs are distinguished by large, easy-to-see features: stroke count for き/さ, and stroke length plus hook for い/り. The reader's eye usually picks these up quickly with a few weeks of reading exposure. No consistent pedagogy source treats them as part of the same "core five" that includes ぬ/め and わ/れ/ね.45
The core five rely on features that stay subtle even with practice and benefit from explicit naming: a small loop, a single crossbar, or a mirror direction. The honorable-mention pairs do not.
Good to know
Train the silhouette, then the diagnostic feature
The brain's first-pass kana recognition is based on outline. For lookalike pairs, the outlines overlap, so the second pass has to land on the diagnostic feature within a fraction of a second. The fastest way to build that second pass is to read the pair side by side inside real words, not on flashcards: ぬいぐるみ vs めがね, わたし vs れんしゅう vs ねこ. Word context primes the brain for the correct kana before the outline match even completes.
Small-font screens hide the diagnostic before the eye does
The bottom-right loop on ぬ and the closed bottom on る are both smaller than the kana's overall outline. If you read hiragana on a tiny screen and conclude "the closure isn't there," the issue is usually font rendering, not kana identification. Increase the font size or zoom before deciding which kana you are looking at.
Stroke order is not optional for lookalikes
Two of the five core groups (ぬ/め and わ/れ/ね) are distinguished by where the second stroke ends.6 A learner who has never written the kana will hit a reading-speed plateau on those pairs early, because the eye does not know where to land.
The Wikimedia Stroke Order Project notes that since 1977 there is no single mandated stroke-order standard in Japan, and that "stroke orders … should follow commonsensical orders which are widely accepted in society."12 In practice, every modern Japanese textbook teaches the same hiragana stroke orders. The stroke counts reported in this article (2 strokes for ぬ, め, わ, れ, ね, ち, い, り; 3 for さ and は; 4 for ほ and き; 1 for る and ろ) match the consensus used in Japan Times' Genki, Tofugu, and the hiragana stroke order app.86 The Hiragana Stroke Order article on this site walks through every kana's stroke sequence in full.
Mnemonics work, but they are scaffolding
The noodle-and-eye image for ぬ/め, the open-hook-loop progression for わ/れ/ね, and the rung-counting trick for は/ほ are all useful at the absolute-beginner stage. The widest-reaching mnemonic set is Tofugu's "Learn Hiragana" guide.8 LearnTheKana publishes per-pair mnemonic pages for the same set with slightly different imagery.711
The mnemonics are teaching aids, not part of the language. Their job is to be discarded once recognition is automatic. If the mnemonic is still doing the work at the six-month mark, the bottleneck is reading volume rather than the mnemonic itself.
Font matters, especially at small sizes
The る vs ろ closure and the ぬ vs め loop can vanish in some low-resolution screen fonts. Japanese-targeted fonts (Hiragino Kaku Gothic, Yu Gothic, Noto Sans JP) render the diagnostic features more crisply than Latin-first fallback fonts that fake Japanese glyphs. This is a widely observed problem rather than a citable principle. If a study text shows two kana as visually identical, suspect the typeface before suspecting your eyes.
Mirror-image pairs are a separate problem
さ vs ち is the only true mirror-image pair in the core five. The others are loop-vs-no-loop (ぬ/め, る/ろ, the ね end of わ/れ/ね), crossbar-count (は/ほ), or stroke-ending differences (within わ/れ/ね). The mirror direction in さ vs ち is not random, but it is not preserved from the source kanji either. 左 ("left," for さ) and 知 ("know," for ち) have unrelated left-right structure in their full kanji forms, so the mirror-image overlap is convergent simplification.3
Teaching reports suggest that learners with a history of left-right reversal in their native script, such as d/b confusion in alphabetic literacy, find さ vs ち disproportionately difficult. The durable fix is writing the pair in the same word repeatedly, for example the name さち, until the directional asymmetry becomes automatic.
Hiragana ↔ katakana lookalikes are a different article
The cross-script confusions (り vs リ, へ vs ヘ, か vs カ, や vs ヤ, せ vs サ in some fonts) are catalogued in Wiktionary's "Appendix: Easily confused Japanese kana" reference page.3 For that family, the diagnostic is "which script is this?" rather than "which kana within hiragana?" It depends on word context, surrounding script, and font weight rather than on a single structural feature. The structural-feature approach used in this article does not transfer.
The confusion is normal and short-lived
Misreading several of these pairs in the first few weeks is the common starting point, not a sign of a problem. There is no published acquisition study that quantifies the timeline, so this is practitioner consensus across Tofugu, LearnTheKana, and the sci.lang.japan FAQ rather than a peer-reviewed finding. The shared observation is that recognition speed in connected text, not for any single kana shown alone, is the real bottleneck after the first month.
Drilling lookalike kana on flashcards past the first month
After the first month, the bottleneck is recognition speed inside real sentences. Continuing to drill lookalike kana on isolated flashcards for longer than that period tends to plateau, because flashcards do not train the eye to spot the diagnostic feature inside surrounding script. Once the binary tests above feel automatic on flashcards, the standard next step is more reading: a graded reader, a children's book, or NHK Easy News.
See also
- Dakuten and Handakuten: The Voicing Marks on Hiragana
- Yōon: Contracted Sounds in Hiragana (きゃ, しゅ, ちょ and the Full 33)
- Long Vowels in Hiragana: How to Read and Write ああ, いい, うう, ええ/えい, and おう/おお
- The Small つ (Sokuon): How to Read and Pronounce the Geminate Consonant
- The Three Hiragana Spelling Exceptions: は, へ, and を as Particles
- Hiragana, Katakana, or Kanji First? A Beginner's Script Order