Inference & Implied Meaning

What you'll learn: the difference between a valid inference and a guess, and a two-question test you can run on any answer choice before selecting it.

Stated vs. implied

Most CELPIP Reading questions ask what a passage says. A smaller but consistent subset — appearing across all four parts, most often in Parts 3 and 4 — asks what a passage implies: the writer's attitude, an unstated reason, what probably happens next, or how someone would likely react to something not directly discussed. These questions feel harder because there's no exact sentence to point to, which pushes anxious test-takers toward one of two failure modes: over-reading (inventing a conclusion the passage doesn't support) or under-reading (rejecting a correct inference because "it's not literally stated," even though CELPIP explicitly tests inference).

The two-question validity test

For any inference answer choice, ask:

1. Is there a specific detail in the passage this answer depends on? A valid inference is always traceable to something concrete in the text — a fact, a word choice, a juxtaposition of two statements — not a general impression.
2. Does the answer require adding information the passage never gives, even indirectly? If yes, it's over-reading, no matter how plausible it sounds in the real world.

A correct inference answer passes both: grounded in the text, but not itself stated word-for-word.

Worked example

Passage excerpt: "The company announced record quarterly profits on Tuesday. By Thursday, its stock price had fallen 8%. Analysts pointed to the earnings call, in which the CEO gave vague answers about the company's plan to address rising material costs next year."

Question: "What can be inferred about investor reaction to the earnings call?"

(A) Investors were pleased with the record profits.

(B) Investors were reassured by the company's plan for next year.

(C) Investors were likely concerned by the lack of a clear plan, despite strong current profits.

(D) The CEO was fired as a result of the stock drop.

Run the test: (A) fails question 2 — it ignores the stock drop entirely and contradicts the overall pattern the passage sets up. (B) fails outright — the passage says the CEO's answers were "vague," the opposite of reassuring. (D) fails question 1 — nothing in the passage discusses the CEO's employment status; it's invented. (C) passes both: it's traceable to two concrete details (record profits + stock drop right after a vague earnings call) without requiring any fact the passage doesn't already imply through that sequence. (C) is the answer.

Inference checklist

• Named the specific detail(s) the inference is built on before selecting an answer
• Rejected any option that requires a fact the passage never mentions, directly or indirectly
• Rejected any option that's simply the literal restated fact (that's a detail question, not an inference)
• Checked tone words carefully — "vague," "reluctant," "enthusiastic" often signal the inference the question wants
• Didn't default to the most "dramatic" or extreme-sounding option out of anxiety

Common mistakes

• Over-reading: picking an answer that would make sense in real life but that the passage gives no actual basis for (like assuming the CEO was fired).
• Under-reading: dismissing the correct inference because it isn't a direct quote, even though the question specifically asked what's implied.
• Missing tone-carrying words (adjectives and adverbs like "vague," "reluctantly," "surprisingly") — these are usually where the inference evidence is hiding.
• Confusing an inference question for a detail question and re-scanning for an exact quoted sentence that will never exist.

Recap

Inference questions reward answers that are grounded in specific textual evidence but go one small logical step beyond it — never further. Run every candidate answer through the two-question test (traceable to a detail? free of invented information?) and you'll consistently separate the correct inference from the over-read and under-read distractors sitting next to it.