备考 HSK 到底需要多少词汇？

"How many words do I need?" is usually the second question a learner asks about the HSK, right after "how long will it take?". Both questions want the same thing: a finish line. A number you can divide by the weeks you have left, so that studying becomes arithmetic instead of uncertainty.

The honest answer is that no single number can carry that weight — not because the information is secret, but because the thing being counted is not stable enough to be counted usefully.

First, a boundary

This article does not reproduce any official vocabulary list, and no figure here should be read as an official count. Vocabulary inventories for the examination, their exact sizes, and the revisions they go through are published by the examination authority. If you need the inventory for the level you intend to sit, get it from that authority or from your registered test centre, and check the publication date on whatever you download — third-party lists circulating online are frequently out of date, and a list that was accurate three years ago may not describe the examination you are about to take. Everything below is about how to use a vocabulary target, not what the target is.

Why raw counts mislead

A word is not a character, and neither is a syllable. Chinese builds enormous amounts of vocabulary from a small set of reusable morphemes. Once you know 电 (diàn, electric), you are a short step from 电话 (diànhuà, telephone), 电脑 (diànnǎo, computer), 电影 (diànyǐng, film), 电视 (diànshì, television) and 电梯 (diàntī, lift). A list would count those as six items. Your brain, if you learned them well, stores something closer to one idea plus five combinations. This is why learners routinely report that vocabulary gets easier per item as they advance, even as the totals grow.

One list entry can be several words. 打 (dǎ) is glossed "to hit", and that gloss will not get you through 打电话 (dǎ diànhuà, make a phone call), 打车 (dǎchē, take a taxi), 打篮球 (dǎ lánqiú, play basketball), 打扫 (dǎsǎo, to clean) or 打折 (dǎzhé, give a discount). Counting 打 as one word implies one unit of learning. In practice it is a small family of them.

Counts hide the direction of use. 会 (huì) is a modal verb meaning "know how to", a marker of future likelihood, and a noun meaning "meeting". You can pass a multiple-choice item recognising any of these and still be unable to produce the right one when you speak.

Function words punch far above their weight. A handful of items — 了 (le), 的 (de), 就 (jiù), 才 (cái), 还 (hái), 把 (bǎ), 被 (bèi) — carry more of the load in an exam paper than dozens of nouns, and they take much longer to master because their meaning is structural rather than lexical. On a list they are one line each. In your study time they are not.

Lists carry no collocations. Knowing 穿 (chuān) and 戴 (dài) both mean "to wear" does not tell you that 穿 takes clothing and shoes while 戴 takes hats, glasses, watches and jewellery. Knowing 书 (shū, book) does not give you 一本书 (yì běn shū) rather than the wrong measure word. This information is where a large share of your errors will come from, and it is invisible in a total.

What "knowing" a word actually means

It helps to stop treating knowledge of a word as a binary and split it into layers, roughly in the order they are acquired:

1. Recognise it in context. You meet 认识 (rènshi) in a sentence and the meaning arrives without effort. This is the cheapest layer and the one that inflates self-assessments.
2. Recall it out of context. You see 认识 alone on a card and know it means "to be acquainted with someone". Harder, because the sentence is no longer doing half the work.
3. Recognise it by ear. Someone says rènshi at normal speed, inside a phrase, and you catch it. For many learners this lags well behind reading, because the character was carrying the recognition.
4. Retrieve it for production. You are speaking, you need it, and it arrives in time. This is a different memory operation from recognition and it is trained differently.
5. Use it correctly. You know that 认识 is for people and things you are acquainted with, while 知道 (zhīdào) is for facts you know — so 我认识他 and 我知道他的名字, not the reverse.

Any claim about vocabulary size is meaningless until it says which layer it means. A learner with three thousand words at layer 1 and three hundred at layer 4 will read comfortably and speak painfully. A learner with the opposite profile will chat happily and stall on a reading passage. Both could describe themselves as "having three thousand words".

The gap between reading vocabulary and speaking vocabulary is large in every language, including your own — you recognise far more words than you would ever use. In Chinese the gap widens further, because reading is supported by characters that carry meaning directly while speaking gives you no such scaffolding.

What examination levels actually ask of you

Levels are usually described in terms of what a candidate can do: handle personal information and simple daily exchanges lower down; sustain a conversation on familiar topics in the middle; follow argument, handle abstraction, and read longer connected text higher up. Vocabulary size is a consequence of that capability, not the definition of it. Two candidates with the same word count and different task ability do not perform the same.

This is also why "I finished the list" is a poor readiness signal. Finishing a list means you have met the items. Readiness means you can deploy them under time pressure, in listening at natural speed, and in writing without a dictionary.

Targets that are actually usable

Set targets by function, not by total. Some that work:

• Function words first, and deeply. Twenty structural items understood properly will do more for your comprehension than two hundred nouns.
• Learn in chunks. Store 不好意思 (bù hǎoyìsi, excuse me / sorry), not 不, 好, 意 and 思 separately. Store 有意思 (yǒuyìsi, interesting) and 什么意思 (shénme yìsi, what does it mean) as whole units.
• Keep an active core. Choose a subset — the words that describe your own life, work and study — and drive those all the way to layer 5. Everything else can sit at layer 1 or 3 without shame.
• Measure coverage, not inventory. The useful question is not "how many words do I have?" but "when I read a text at my level, how often do I stop?" If you stop more than a few times per paragraph, the text is above you regardless of what any count says.
• Train listening separately. If you can read a word but not hear it, you do not have it for the listening section. Add audio to your reviews rather than assuming transfer.

The practical answer

If you want a plan rather than a number: work from the official inventory for your level, obtained from the examination authority, and treat it as a checklist of coverage rather than a queue. Sort it by whether each item is layer 1, layer 3 or layer 5 for you. Push everything to layer 3 — you must be able to hear it. Push your personal core to layer 5. Then spend the remaining time on past-paper-style practice under time limits, because that is what converts a vocabulary you own into marks on a paper.

A word count is a description of a state you will pass through. It was never the target.