What TOPIK I reading actually tests, compared to listening: 12 rounds, measured

I sat down with a TOPIK I reading set once, the year I was still counting -습니다 endings on my fingers to feel like I was making progress. It ran longer than I expected — paragraph after short paragraph — and it still felt easier than the listening section I’d done the week before, which was shorter and left me lost twice. At the time I read that as my ear being worse than my eyes. Turns out that’s not quite the right explanation, and the past papers say why.

What grammar does TOPIK I reading actually use, compared to listening?

Less grammar per character, packed into longer text, and a different set of sentence endings almost entirely. We took the reading section out of all 12 released TOPIK I exam booklets — the same rounds already tagged for listening — and ran the identical 88-point grammar inventory over it. Reading passages run 76% longer per paper than the listening script from the same sitting, but they pack fewer grammar hits into every thousand characters, not more.

	Listening	Reading	ratio (reading ÷ listening)
Papers	12	12	—
Characters per paper (mean)	1,266	2,224	1.76×
Grammar hits per 1,000 characters	94.6	79.8	0.84×
Distinct grammar points seen	56	53	0.95×
Share of hits that are first-year forms	88.9%	90.5%	—

That last row is worth sitting with: the two sections land at almost the same grammatical level — both are around nine-tenths first-year material — and see nearly the same number of distinct points. What differs isn’t difficulty. It’s density and register, and register is the bigger story.

The core forms — and where reading and listening pull apart

Twenty grammar points show up in at least nine of the twelve TOPIK I reading papers. The rightmost column is how often the same point appears in the listening scripts from the same rounds — and a few of the gaps are large enough to change how you’d study for one section versus the other.

Form	What it does	Reading papers	Hits in reading	Hits in listening
-(스)ㅂ니다	formal polite declarative	12/12	1,215	103
-아/어서	so, and then	12/12	97	78
-(으)ㄹ 수 있다 / 없다	can, cannot	12/12	87	49
-는 것 / -(으)ㄴ 것	turns a verb into a noun (“the act of,” “the fact that”)	12/12	73	28
-(으)ㄹ 때	when	12/12	72	18
-고 싶다	want to	12/12	51	31
-(으)ㄴ 후에 / -기 전에	after, before	12/12	41	27
-지 않다	not, don’t (formal negation)	11/12	40	25
-(으)ㄹ 것이다 / 거예요	will, probably	12/12	38	37
-지만	but	11/12	38	10
-아/어 주다	do something for someone	12/12	38	25
-(으)려고	intending to	10/12	32	20
-고 있다	-ing, in the middle of doing something	9/12	23	15
-(으)면서	while (simultaneous)	10/12	23	6
-아/어지다	become, get	9/12	21	2
-기 때문에	because (written/formal)	10/12	21	3
-는데 / -(으)ㄴ데	background, contrast	9/12	19	96
-(으)려고 하다	be about to, plan to	9/12	19	13
-게 되다	come to, end up	9/12	15	5
-(으)러	purpose of going, coming	9/12	12	10

Look at -(스)ㅂ니다 and -는데 / -(으)ㄴ데: they move in opposite directions. The formal declarative ending shows up nearly twelve times more often in reading than in the listening script from the same rounds — reading passages are almost entirely narrated in formal register, the flip side of the everyday casual -아/어요 ending that dominates spoken Korean generally (not itself a point tracked in this inventory, but consistent with the pattern this table shows). And -는데, which sets up a contrast or gives background, appears five times more in listening — it’s doing conversational work (softening a request, leading into a question) that written narration mostly doesn’t need.

The endings that never cross over

Six sentence-final endings appear reliably in TOPIK I listening — -네요 in all 12 papers, -지요 / -죠 in 10, -는데요 / -(으)ㄴ데요 in 10 — and not once in the reading section of the same 12 rounds:

Form	What it does	Listening papers	Hits in listening
-네요	noticing, mild surprise	12/12	47
-는데요 / -(으)ㄴ데요	softened statement	10/12	28
-지요 / -죠	seeking agreement	10/12	24
-군요 / -구나	realisation	6/12	8
-(으)ㄹ래요	I’ll…, want to…? (casual, stating what you plan to do)	5/12	8
-나요 / -(으)ㄴ가요	softened question	4/12	5
-잖아요	as you know	4/12	4

Every one of these is a spoken-register ending — the kind that marks noticing something out loud, softening a question, or checking agreement with whoever you’re talking to. Written narration doesn’t produce them because there’s no listener in the room to soften a statement for. If you’ve been drilling these from a general “TOPIK grammar list” as if they mattered equally everywhere, the reading section is where that time is wasted: none of the 12 papers use them.

Reading isn’t empty-handed in return. Eight points appear only in reading, though thinly — -지 못하다 (7 papers, 9 hits) is the only one with any real spread; the rest, including -기 위해(서) and -(으)ㅂ시다, show up in one or two papers each and shouldn’t be read as reading-exclusive material worth specifically drilling.

What this changes about how you read a TOPIK I text

If reading and listening tested the same grammar the same way, the extra effort you put into one would carry straight over to the other. The numbers say it partly does — the same first-year forms, at a similar level of difficulty — and partly doesn’t: reading gives you more text to lean on for the same idea, in a register built almost entirely around -습니다, while listening compresses the same content into fewer, denser lines and adds spoken-register endings reading never asks you to recognize.

So a resource built for one doesn’t fully transfer to the other, not because the grammar is different, but because the register is. That’s the part a printed grammar list flattens away, and it’s the part up103 tries to keep: we read your study notes and write Korean texts and dialogues that draw on what’s actually in them, in whichever register the practice calls for. Founding Learners get 6 months free, then 50% off for life, at app.up103.com.

Method

Corpus. All 12 TOPIK I exam booklets already used for the listening study — rounds 35, 36, 37, 41, 47, 52, 60, 64, 83, 91, 96 and 102, 2014 through 2025 — from the official past-paper archive at topik.go.kr (학습하기 > 학습 자료실). TOPIK I administers listening and reading in one continuous booklet, so the same OCR’d text that already carried the listening script also carries the reading section; no new download was needed for these 12 papers. TOPIK II reading is a separate booklet from round 41 onward and needed fresh acquisition; only 2 of the target 12 rounds (36, 37) turned out to already be on hand, bound into an all-in-one booklet from an earlier printing convention, and the network this study ran on could not reach the archive to fetch the rest (verified live, not assumed — the connection reset rather than answered). That 2-paper TOPIK II sample is reported as exploratory above, not as a level comparison.

Extraction. The reading section was separated from listening by item number for TOPIK I (reading is items 31–70, a fixed feature of the exam’s public structure) and by the booklet’s own running page header for TOPIK II (reading prints under a distinct “2교시, 읽기” header once the booklet moves past listening and writing). Instruction boilerplate (“read the following and answer,” point values, “choose the most appropriate”) and the printed multiple-choice options were excluded the same way listening’s answer options were — left in, a single repeated phrase like “choose the most appropriate” would inflate the command ending it happens to end with far past what a candidate actually reads. That yielded 1,079 reading-text blocks and 35,719 characters across all 14 papers.

Tagging. The identical 88-point grammar inventory used for the listening study, copied unchanged rather than re-registered, specifically so the two studies’ numbers are directly comparable rather than measuring against two different yardsticks. A hand-judged sample of 120 tagged hits, read with surrounding context, scored 95.8% correct (115 of 120) — 96.9% on the ten most frequent points, which account for 77.3% of every hit. The errors that did turn up are the same kind the listening study’s own check found: a word-boundary collision once OCR spacing is removed, for example 오후에 (“in the afternoon”) tagged as the connective -(으)ㄴ 후에, or 위해서 (“in order to”) tagged as the separate connective -아/어서 because it ends in the same two syllables. Both studies share this weakness because they share the inventory; patching one copy and not the other would break the comparison the shared copy exists to protect.

Completeness. Reading items carry no independent marker the way listening utterances carry a speaker label, so extraction completeness here is measured against NIIED’s published item counts instead — a fixed 40 items for TOPIK I reading, 50 for TOPIK II. By that measure the extractor attached a distinct item number to 60.9% of expected items overall (64.6% at TOPIK I, lower at the thin TOPIK II sample). This undercounts how many item numbers were individually resolved, not how much text was captured: a shared passage introduced under one range header keeps its text even when a sub-question’s own number isn’t recognized after OCR, which is also why no figure above is published per item, only aggregated by paper or level.

— 준호, up103