Radical lookup is how you find a kanji you can see and cannot type. The traditional method uses the indexing component plus a stroke count; the modern methods are faster and more forgiving, and knowing both means you are never stuck in front of a character.
The traditional method
Identify the indexing radical, the single component under which the character is filed. Count its strokes, find that radical’s section in the dictionary, then locate your character by the number of strokes outside the radical.
Two things make this harder than it sounds for a learner. Knowing which of several components is the official index requires familiarity rather than logic, since the choice was made historically. And counting strokes correctly requires the corner-turning rule, where a single movement that changes direction counts as one stroke.
It is still worth understanding, because dictionaries, indexes and character databases are organised this way, and because it works with no device at all. The Unicode Han database records the same radical assignments in machine-readable form, which is where the digital tools get their data.
| Step | What you do | Where it goes wrong |
|---|---|---|
| Identify the radical | Pick the indexing component | Several components look plausible |
| Count its strokes | Standard counting rules | Turning strokes counted twice |
| Find the section | Radical index in the dictionary | Variant forms listed separately |
| Count remaining strokes | Everything outside the radical | Same counting trap |
| Scan the entries | Find your character | Similar characters adjacent |
Multi-radical search, which is better
Digital dictionaries let you select several components you can see, and narrow the results with each one. This removes the hardest part of the traditional method, which is guessing which component is the official index.
The practical technique: pick the most distinctive component first, then add a second only if the list is still long. Distinctive beats common, so choosing 麻 narrows faster than choosing 口.
It also tolerates being wrong. If a component you selected is not actually present, you can remove it and try another, whereas a wrong radical guess in a paper dictionary sends you to the wrong section entirely.
For most learners this is the default lookup method, and the traditional one is a fallback for paper.
Handwriting input, which is better still when it works
Draw the character on a touchscreen and the system offers matches. When you can approximate the shape, this is the fastest method available.
It has a useful side effect: it tests whether you can produce the character, which is a different question from whether you recognise it. Learners often discover they cannot draw something they can read, which is exactly the gap handwriting practice closes.
The caveats are worth knowing. Recognition is tolerant of shape and often sensitive to stroke count and order, so a character drawn with an invented sequence may not be found even when it looks correct. And a character you cannot approximate at all cannot be looked up this way, which is when the component methods take over.
| Method | Best when | Weakness |
|---|---|---|
| Radical plus stroke count | Paper dictionary, no device | Slow, needs correct counting |
| Multi-radical search | You can see the components | Needs a component-selection tool |
| Handwriting input | You can approximate the shape | Order-sensitive, needs a screen |
| Optical lookup from a photo | Printed text in front of you | Fails on handwriting and odd fonts |
| Asking someone | A name, or a genuinely rare form | Requires someone to ask |
Choosing a method by situation
You are reading printed text on a screen: copy and paste, or use an optical lookup. No component knowledge required.
You are reading paper and have a phone: handwriting input first, multi-radical search if the shape defeats you.
You are reading paper with no device: the traditional method, which is the reason it is still taught.
The character is in a name: expect the lookup to give you the character and not the reading, because name readings often do not follow from the characters. That is normal, and native speakers ask too.
A worked lookup
Walking one character through shows how the methods compare in practice.
Say you meet 燃 in printed text and cannot type it. By the traditional method you would identify 火 as the indexing component, count its four strokes, find the fire section, then count the remaining twelve strokes to locate the character. That works and takes a couple of minutes with practice.
By multi-radical search you select 火 and, if the list is still long, add another visible part. The character appears in seconds, and it does not matter whether the part you chose is the official index.
By handwriting input you draw it, approximately, and the system offers it among the first few candidates provided your stroke count is roughly right.
By optical lookup, if the text is on a screen or you can photograph it clearly, you get the character with no analysis at all.
The general-use status of whatever you find is worth checking against the jōyō kanji table, since that tells you whether it is a character you will meet again or a one-off in a name or a specialist term.
What lookup practice teaches you
Repeated lookups are a slow but genuine way to learn components, because identifying a radical forces you to decompose the character deliberately.
Repeated lookups also build a feel for which components are load-bearing, since the ones you keep selecting in multi-radical search are the ones doing the most identifying work across the characters you actually read.
They also expose the difference between recognition and production. Handwriting input fails when your mental image is vague, and that failure is informative: it tells you which characters you know only well enough to recognise.
And they build tolerance for not knowing. The habit of looking a character up rather than skipping it is what turns unfamiliar text from a wall into something navigable, which matters more for reading progress than any individual character does.
Building a lookup habit that helps
How you handle an unknown character changes how much you learn from meeting it.
Guess before looking. Identify the components, guess the category from the semantic part, guess a reading from the phonetic part if there is one, then check. The guess costs five seconds and it is what makes the lookup memorable rather than transactional.
Note the word rather than the character. You met it inside something, and the word is what you will meet again.
Keep a short running list of characters you have looked up more than once. Repeated lookups of the same character mean it is common in what you read and has not stuck, which makes it an obvious candidate for deliberate practice.
And resist looking up everything. Reading with too many interruptions stops being reading, and a character you can skip without losing the sentence can wait for its second or third appearance.
The fastest way to stop needing lookups
Learn components deliberately, since a character made of parts you can name is one you can search for immediately by any method.
Learn characters inside words, because context usually tells you what a character means even before you look it up, and it narrows the candidates when the reading is ambiguous.
And practise production, since the ability to draw a character approximately is what makes handwriting input work. A brain-wave study found words learned by handwriting were encoded more strongly than the same words typed, and the same production practice that builds writing also builds the mental image lookup depends on.
Consistency in how you record what you looked up also matters more than the method itself, since a lookup that leaves no trace teaches you the character once and then lets it go.
Reading widely does the rest. Most lookups a learner performs are of characters that are genuinely common, and after enough encounters the lookup stops being necessary.
{{appName}} builds the production side directly: prompt first, character hidden, stroke order checked on your attempt, and misses returned sooner. It is free in early access. For the structural half, kanji radicals explained and stroke count cover what lookup depends on.
Looking up a character you cannot see clearly
Two awkward cases come up often enough to plan for.
Handwritten text. Optical lookup usually fails on handwriting, multi-radical search struggles when components are joined or abbreviated, and the practical route is context first: work out what the word probably is from the sentence, then check that guess in a dictionary by reading rather than by shape.
Stylised or decorative forms. Shop signs, logos and older printing use forms that deviate from the standard, sometimes substantially. The same approach applies: guess from context, confirm by reading, and treat the shape as a variant rather than a character you have never met.
In both cases the fallback is asking someone, which is what native readers do and which is worth remembering when a single character is holding up a whole sentence.
Why lookup speed matters more than it sounds
Reading progress is limited less by vocabulary than by friction, and lookup is most of the friction.
A learner who takes two minutes to find each unknown character reads a page in an hour and stops. A learner who takes ten seconds reads the same page in ten minutes and continues, and the second learner meets far more characters per week as a result.
That compounding is the practical argument for getting comfortable with the fast methods early rather than treating lookup as a chore to be endured. Time invested in knowing how to search well returns itself within a few reading sessions.
It also changes what you are willing to read. Material slightly above your level becomes approachable when unknown characters cost seconds, and reading slightly above your level is where most vocabulary growth happens.
Frequently asked questions
How do you look up a kanji by radical?
Identify the indexing component, count its strokes, find that radical’s section in the dictionary, then locate the character by the number of strokes outside the radical. Digital multi-radical search is easier, since you select several components you can see rather than guessing which one is the official index, and it tolerates a wrong guess.
What is the fastest way to look up an unknown kanji?
Handwriting input, when you can approximate the shape, because drawing it takes seconds and needs no component analysis. Optical lookup from a photo or screen is faster still for printed text. Multi-radical search is the reliable fallback when the shape defeats you, and the traditional radical method is for paper with no device.
Why can I not find a character with handwriting input?
Usually because recognition is sensitive to stroke count and order rather than only to shape, so a character drawn with an invented sequence may not match even when it looks right. The other common cause is a vague mental image producing an approximate drawing, which is itself a useful signal that the character is recognised rather than known.
Do I still need to learn radicals with modern tools?
For lookup specifically, less than before. For learning, more than ever, because components are what make characters cheap to learn and produce, and multi-radical search works best when you can name what you are seeing. The tools removed the tedium of the traditional index without removing the value of knowing the parts.
What is the best app for learning the components lookup depends on?
{{appName}} focuses on production, which is where component knowledge becomes usable: prompt first, character hidden until you have written it, stroke order checked, and missed characters returned sooner. Being able to draw a character approximately is also what makes handwriting lookup work. It is free in early access.
