Wa or ga: the difference is what the listener already knows

は and が differ in what your listener already knows. は marks something the two of you are already holding in mind and puts the new information after it; が marks the thing itself as new and puts the weight before it. The particle is not reporting the grammar of your sentence, it is reporting the state of the conversation.

That one difference resolves most of the cases a rule sheet leaves ambiguous, including the sentences where both particles are correct and only the meaning moves.

The particle marks where the new information sits

The quickest way to feel it is to answer a question.

Someone asks 誰が来ましたか — who came? The unknown is the person, so the answer puts the weight there: 田中さんが来ました. Answer it with 田中さんは来ました and you have said something slightly off — you have treated 田中さん as old news and offered came as the new part, which is not what was asked.

Now flip the question. 田中さんは何をしましたか — what did Tanaka do? Tanaka is established, so he takes は, and the new information lands after it: 田中さんは帰りました.

That is the whole mechanism in miniature. が throws the spotlight backwards onto what precedes it. は clears what precedes it out of the way and throws the spotlight forwards.

Which explains a rule that otherwise has to be memorised: question words never take は. 誰が, 何が, どれが, いつが — always が. A question word is by definition the thing your listener does not know, and は exists to mark what they do. 誰は is not a subtle error, it is a contradiction.

は is not a subject marker, which is why the rule you were taught keeps failing

Most textbooks introduce は and が side by side, which quietly implies they are competing for the same job. They are not on the same layer of the sentence at all.

が marks the grammatical subject. は marks the topic — the thing the sentence is about — and the topic is frequently not the subject:

• 寿司は食べません — 寿司 is the object. は has displaced を.
• 東京には行った — は is sitting on top of に, not replacing it.
• 今日は寒い — a time expression, and no subject in sight.

は overwrites が and を when it lands on them, and stacks on top of に, で, から, と. That behaviour makes no sense for a case marker and perfect sense for something operating one level up: first name the frame, then say something inside it.

So the pair you are actually choosing between is not subject A or subject B. It is put this in the frame or put this in the spotlight.

A folk tale performs the handover in two sentences

Japanese stories open the same way so consistently that the opening is a grammar lesson:

昔々、ある所におじいさんとおばあさんが住んでいました。おじいさんは山へ芝刈りに、おばあさんは川へ洗濯に行きました。

First sentence: が. The old man and old woman have just been introduced, and the listener knows nothing about them yet. Second sentence: は, twice. They now exist between you, so they become the frame and the new information is where each of them went.

English does the same work with articles. A man came into the shop. The man asked for directions. Nobody taught you to switch from a to the; you did it because the second mention is no longer new. は and が carry that same distinction, and they carry it further, because Japanese has no articles to do the job.

Once you have seen the handover you start noticing it everywhere: the first line of a news item takes が, the follow-up sentences take は.

Some sentences take が away from you as a choice

A large share of the agonising over this pair is spent on sentences where the grammar has already decided. Three cases are worth knowing cold, because they remove the question entirely.

Question words and their answers, as above: 何がありますか, 誰が知っていますか, and 私が知っています when you are the one.

Subordinate and relative clauses. The subject inside them takes が: 私が作った料理, 彼が来る前に, 雨が降ったら. The reason follows from what は does. A topic is set for your listener, at the level of the whole utterance — a clause buried inside a noun phrase has no listener of its own to set one for. When は does appear inside such a clause it is doing contrast, not topic.

Predicates that govern が. A set of very common words takes が where an English speaker reaches for an object: 日本語がわかる, 水がほしい, 猫が好き, ピアノが上手, 時間がある. These are not exceptions to the topic rule, they are ordinary subject marking — in Japanese the thing liked is the subject of being likeable.

Negation pulls は, and it says more than you meant

The mirror case: some sentences pull は even when が looks structurally fine, and negation is the strongest pull.

コーヒーを飲みません is grammatical and oddly flat. コーヒーは飲みません is what a speaker actually says — and it carries an implication, because a frame invites the question of what lies outside it. Coffee, I don't drink leaves tea hanging in the air unspoken.

That is contrastive は, and it is the same particle doing the same job. Set something aside as the frame and you have implicitly set everything else outside it. 私は行きません says you are not going, with a faint suggestion that others might.

This is worth knowing before you use は to sound natural in a negative sentence and accidentally imply a comparison you did not intend. It is also the reason a flat denial often drops the particle debate entirely and restructures the sentence.

Japanese has a dedicated grammar for that same half-open denial, one level up the JLPT ladder: わけではない states outright that only part of a claim is being denied, rather than leaving the rest to は's implication.

象は鼻が長い is the system stated twice, not an exception

The sentence every learner meets and nobody explains well: 象は鼻が長い. The elephant's nose is long — apparently two subjects in one clause.

There is one subject. 鼻 is it, marked with が, doing exactly what が always does inside the comment. 象 is the topic, marked with は, setting the frame: we are talking about elephants now. The comment made inside that frame happens to be a complete little sentence of its own.

Read it as as for elephants — the nose is long and it stops being strange. The pattern is everywhere once you have the shape:

• 私は頭が痛い — as for me, the head hurts.
• 日本は物価が高い — as for Japan, prices are high.
• この店はラーメンがおいしい — as for this shop, the ramen is good.

Each is a frame plus a self-contained observation. This is also the construction that makes Japanese feel like it is describing situations rather than performing actions, and it is built entirely out of the two particles you are trying to tell apart.

When both are correct, you are choosing what you claim

The genuinely hard cases are the ones where nothing is ungrammatical and the particle carries the meaning by itself.

私は田中です introduces you. As for me, Tanaka. It is what you say walking into a room.

私が田中です identifies you out of a set. I am the one who is Tanaka. It is what you say when someone at reception has asked which of the three of you that is — and it is faintly strange when nobody asked, the way I'm the one who ordered the coffee would be if no coffee had arrived.

Same fork elsewhere:

私がやります	I'll do it — I'm the one, as opposed to anyone else
私はやります	I'll do it — whatever the rest of you decide

This is sometimes called exhaustive listing が, and the label is accurate: が picks its noun out of the available candidates and excludes the others. When a sentence with が sounds unaccountably emphatic to a Japanese ear, that exclusion is why.

The choice depends on something no table contains

A rule sheet cannot settle は versus が because the deciding information is not in the sentence. It is in what your listener is currently holding in mind, and that changes between one sentence and the next. Keigo has the same problem — who else is in the room is not a column you can print — and both resist study by grid for the same reason.

What does transfer is the same move that works for readings: make the unit you learn bigger than the piece confusing you. A particle drilled on its own is a coin flip with a rule attached. A sentence you actually heard arrives with its context still on it — who was speaking, what had just been said, what the listener already knew — and that context is the missing column.

So collect sentences, not particles. When a line surprises you with は where you expected が, that sentence is the one worth keeping, because the surprise marks the exact gap in your model. Capture it where you meet it — the sign, the message, the line of dialogue — because by evening you will remember that something was odd and not what it was.

Kikusho puts an example sentence on every card by default, alongside the word, its reading, romaji, meaning and a register note, and lets you select any word inside that sentence to turn it into a card of its own. For particles that matters more than for vocabulary: the word is not what you are learning here, the arrangement around it is, and a card without a sentence on it has thrown away the only evidence you had.

は and が are not two ways of marking the subject. One of them is not marking the subject at all. Once the pair is a question about your listener rather than about your sentence, most of the ambiguity turns out to have been in the explanation.