TOEIC Link Part 7 — Inference and NOT-Stated Questions in Single Passages: Answering What the Text Implies and Isolating What It Never Said

Two Part 7 question types demand the opposite reading behavior from every other item on the module, and candidates who use one general "find the answer in the text" strategy get both of them wrong. The first is the inference question — "What is suggested about…", "What can be inferred…", "What does the writer imply…" — where the correct answer is deliberately not printed in the passage; you must construct it. The second is the NOT-stated question — "What is NOT mentioned…", "Which is NOT true…" — where three options are printed in the passage and your job is to find the one that is not. One type rewards adding information the text forces; the other rewards refusing to accept information the text never gave. Confuse the modes and you will look for a printed answer on the inference item (and pick a trap) and construct an unprinted answer on the NOT-stated item (and pick a trap).

This guide separates the two reading modes cleanly and gives each its own evidence standard. For the adjacent skill of catching answer choices that are directionally right but wrong in scope, pair this with the Part 7 answer-choice scope trap: too broad versus too narrow guide, and for the listening-side version of the same inference muscle, see the listening inference and implication questions guide.

Inference: the answer is unprinted but not unlimited

An inference answer is not stated in the text, but it is forced by the text. That distinction is the whole game. A correct inference is a conclusion you could defend to a skeptical reader by pointing at specific lines: "The passage says X and Y, and X plus Y can only mean Z." An incorrect inference is a conclusion that is merely consistent with the passage — it could be true, but the passage does not compel it. TOEIC Link inference distractors are almost always "could be true" statements, and they feel right precisely because nothing in the passage contradicts them.

Take a memo that says: "Because the third-floor conference rooms are being renovated through the end of the month, all recurring meetings have been moved to the annex." An inference question might ask what is suggested about the annex. The forced conclusion is that the annex has rooms available for meetings — the memo relocates meetings there, so it must be usable. That is defensible from the text. A trap option might say the annex is larger than the third floor. Could be true — but the memo says nothing about relative size. The evidence standard for inference is: can I point to the specific lines that make this the only reasonable conclusion? If you cannot draw the arrow from a line to the option, it is a "could be true" trap, not an inference.

The reading protocol for inference:

1. Read the stem to locate the target (what is being inferred about — the annex, the writer's attitude, a product, a deadline).
2. Find the lines that mention that target, plus one line on either side for context.
3. Ask "what must be true given these lines?" — construct the forced conclusion in your own words before looking at the options.
4. Match your constructed conclusion to an option and reject any option you cannot tie to a specific line. "Consistent with" is not "supported by."

The step-3 habit — building the answer before reading the choices — is the same anti-anchoring discipline that governs paraphrase items; see the answer-option paraphrase recognition without keyword matching guide for how to keep the options from steering you.

NOT-stated: three are printed, one is not

The NOT-stated question inverts everything. Here three of the four options are supported by the text — often as direct paraphrases of specific lines — and the answer is the single option the text never supports. This is a search-and-eliminate task, not a construct task. You are not looking for the "best" answer; you are looking for the one that fails to appear.

The correct behavior is to treat each of the four options as a claim to be verified against the passage, one at a time, and to physically locate the supporting line for each. Three options will yield a line; one will not. That one is the answer. The discipline is to keep hunting for the supporting line and only conclude "not stated" after you have genuinely failed to find it — not after a quick skim makes it feel absent.

The trap architecture is subtle. NOT-stated distractors (the three that are stated) are paraphrased, not quoted, so a candidate scanning for exact keyword matches will fail to find the line and wrongly flag a stated option as "not stated." Conversely, the true answer — the genuinely unstated option — is often written to sound like the topic of the passage, using the passage's own vocabulary, so it feels present even though no line supports it. The defense against both is the same: verify by locating a line, not by vibe. A stated option has a line; an unstated option does not, no matter how on-topic it reads.

The reading protocol for NOT-stated:

1. Read all four options first — unusually, here you should, because you need the full claim set before scanning.
2. Scan the passage once, tagging where each option's support appears. Most items let you confirm two or three options quickly.
3. The unconfirmed option is your candidate. Before locking it, do one deliberate second pass looking specifically for its support — the paraphrase may be buried.
4. Lock the option that survives the second pass with no supporting line.

Note the reversal: on inference you construct the answer and avoid the options until step 4; on NOT-stated you read the options first and let them drive a targeted search. Using the wrong protocol on the wrong type is the core failure this item pair is designed to catch.

The evidence standard, side by side

The two types differ on exactly one axis — what counts as a right answer — and holding that difference explicitly is what keeps them separate under time pressure.

• Inference — the right answer is unprinted but forced. Wrong answers are "printed" (too literal, just restating a line) or "unforced" (consistent-but-not-compelled). Your test: can I draw an arrow from specific lines to this conclusion?
• NOT-stated — the right answer is the one that is unprinted and unforced. Wrong answers are the three that are printed, usually as paraphrases. Your test: did I fail to find a supporting line after a genuine second pass?

A useful way to remember it: inference punishes you for staying too literal; NOT-stated punishes you for accepting too much. Inference wants you to add exactly one forced step beyond the text. NOT-stated wants you to add nothing and accept only what has a line.

The two-question trap you must anticipate

Test-writers frequently place an inference item and a NOT-stated item in the same single-passage set, sharing the passage. The shared context makes it easy to carry the wrong reading mode from one to the next. If you just did an inference item and built a forced conclusion, you are primed to construct — and if the next item is NOT-stated, that construction habit will make you invent support for the unstated option. Reset your mode explicitly between items. Read the stem, name the type ("this is NOT-stated — search and eliminate"), and switch protocols before touching the options.

A three-day drill

• Day 1 — inference arrow-drawing. Do ten inference items. For each correct answer, write the specific line numbers that force it and one sentence explaining the forced step. For each trap you were tempted by, label it "printed" or "unforced." This makes the evidence standard concrete.
• Day 2 — NOT-stated line hunting. Do ten NOT-stated items. For each of the three stated options, physically write the supporting line reference. Prove to yourself that three always have a line and one never does. This kills the "it felt absent" guessing habit.
• Day 3 — mixed sets, mode switching. Do five single-passage sets that each contain both types. Before every item, say the type and the protocol out loud. Track whether any error came from carrying the wrong mode across items — that is the highest-value error to eliminate.

The one-sentence rule to carry into the exam

Inference asks you to add the one step the text forces; NOT-stated asks you to accept only what has a line — name the type before you read the options, because the two demand opposite reading modes. Every trap in this pair exploits a candidate running one general strategy across both. Separate the modes, hold each evidence standard, and the traps lose their grip.