JLPT N2 common mistakes
JLPT N2 · Understand the traps and how to avoid them.
For kanji-reading items, the candidate reads each kanji with the on-yomi they memorised for that character in isolation and ignores sound changes inside the compound - they answer しゅっぴ as しゅつひ, いっぽう as いちほう, or miss rendaku in words like 物語（ものがたり）.
Vocabulary is usually studied as single-character flashcards, so the character-to-reading link is stored without the compound-level phonology (sokuon, rendaku, handakuten, dropped kana) that N2 items are built to test.
What to do instead
Learn vocabulary as whole written words, not characters. Keep a list of the four sound-change patterns and, when you miss an item, write the compound plus two or three other compounds using the same change (出費・出発・一遍・発足) so the pattern, not the single character, is what gets stored.
On 用法 (usage) questions, the candidate judges each option by translating the sentence into their native language and asking 'does this make sense?', never checking whether the word takes the particle, transitivity or collocation shown in the sentence.
Word lists are built with one translation per word, so the candidate stores a meaning and not a usage frame; translation-equivalence feels like enough to decide.
What to do instead
Store each new word with its frame: particle, whether it is transitive/intransitive, and the noun it typically takes (影響を与える、感心する vs 感動する). In the exam, read only the option sentence and the target word's immediate neighbours, and eliminate options where the particle or verb type is wrong before considering meaning.
On 文法形式判断, the candidate picks the option that fits the meaning of the sentence but is not licensed by the form of the word before or after the blank - for example choosing わけがない after a ます-stem, or つつ after a noun.
The habit is to read the whole sentence first and 'feel' which grammar point belongs, so the three-word window around the blank, which is where the item is actually decided, is never inspected.
What to do instead
Before reading for meaning, cover everything except the blank and the words immediately before and after it, and write the required form in the margin (plain form / ない-stem / noun+の / ます-stem). Only then check meaning. Time spent is seconds, and it removes the most common distractor type.
On 文の並べ替え, the candidate assembles the sentence from the top down according to what 'sounds like a story', then places the starred item wherever the remaining fragment lands, instead of using formal anchors.
The sentence is treated as a comprehension task, but it is a grammar task: the correct order is fixed by connectives, demonstratives and paired patterns, not by plausibility.
What to do instead
Work backwards. Find the clause-final predicate, then locate (a) こそあど words and 接続詞 that must follow a specific clause, (b) fixed pairs such as ～のは…からだ、～ば～ほど、～てはじめて、～どころか, (c) modifiers that must sit directly before a noun. Build around those and check the starred position last.
On 内容理解, the candidate selects the option that repeats the largest number of words from the passage, or the option that is true in general knowledge, without locating a line that actually supports it.
Distractors are deliberately written by copying passage vocabulary and then changing one element (who, when, cause, or logical direction), so surface overlap feels like evidence.
What to do instead
For every option, go back to the text and point at the sentence that supports or contradicts it; if you cannot point, the option is wrong. Expect the correct answer to be a paraphrase with different vocabulary, and expect the 'most quoted' option to be the trap.
On 主張理解 (author's argument), the candidate chooses the option stating an opinion the passage quotes and then rebuts, because it appears earlier and in stronger wording than the author's own conclusion.
N2 argument passages are structured as cite-and-rebut (「確かに～という考えもある。しかし…」), and the candidate reads the cited view as the content of the passage rather than as a foil.
What to do instead
Physically mark 逆接 and 主張 signals while reading (しかし、だが、むしろ、～のではないだろうか、筆者は～と考える). Treat everything before the first しかし as set-up. Answer with the marked final claim, and check that the option does not contain a reversed causal or concessive relation.
The candidate works the reading section strictly front to back at a uniform pace, so 統合理解 and 情報検索 - the items with the highest marks-per-minute - are reached with two minutes left and answered by guessing.
School habit says do the paper in order; the candidate does not know the per-item time budget, so a single 長文 can silently consume fifteen minutes.
What to do instead
Set a fixed order and budget before exam day: 情報検索 and 統合理解 first (about 10–12 minutes combined), then 短文・中文, then 長文 with whatever remains. Practise with a visible timer and a hard rule to abandon an item after its allotted minutes and move on.
In 課題理解, the candidate writes down the content of every utterance and then chooses the option matching the first thing the speakers discuss, rather than the task set by the final question.
The audio gives no second listening, so anxiety drives exhaustive note-taking, which leaves no capacity to track the constraints (who must do it, by when, under what condition) that the question actually asks about.
What to do instead
Use a three-column memo (だれ／いつ・条件／すること) and write only keywords. Do not decide while listening; wait for the question sentence, then scan the memo against the conditions. Cross out options as each is explicitly rejected in the dialogue - rejections are stated far more clearly than acceptances.
In ポイント理解, the candidate starts listening cold and hears the options for the first time in the audio, so numbers, places and reasons blur together and they choose on tone of voice.
The option-reading window at the start of each item is treated as dead time, and the candidate has not trained listening with printed options in view.
What to do instead
Use the example-item window to read ahead: underline the part that distinguishes the options (the number, the place, the reason) and predict which Japanese words will signal each one. While listening, listen for the paraphrase of your underlined part, since the correct option almost never reuses the audio's exact wording.
In 即時応答, the candidate translates the short utterance literally and rejects the natural reply because it does not 'answer the question', then runs out of time and leaves items blank.
These items test conversational formula and ellipsis, not logic: replies such as それはよかったですね or そうでもないですよ answer the speaker's attitude rather than the literal content, and a translating reader never hears that.
What to do instead
Memorise formulaic responses by function (reassurance, hedging, deflection, agreement with reservation) and drill them at natural speed so recognition is automatic. Judge by the utterance's ending and intonation. Never leave a blank - an immediate guess costs nothing before the next audio starts.