CELPIP Reading rewards three kinds of processing: global (the big picture), local (one precise detail), and inferential (a logical conclusion the text points to but doesn’t say outright). This bootcamp wires those habits into muscle memory so you stop guessing and start verifying.
Goal: capture the passage’s core idea, structure, and stance fast—so later choices have context.
How to skim efficiently
- Lead lines first: Read the first sentence of each paragraph. They usually announce the topic or a shift.
- Flag contrast markers: however, although, whereas, on the other hand, still. These flip direction—note them.
- Name the sections: In the margin (paper) or mind, tag P1–P4 with one word each (e.g., Problem / Evidence / Counterpoint / Recommendation).
- Check the edges: Look at any title, caption, or footnote; in diagrams, scan legend + units (m, km, $, am/pm).
- One-line gist: Say it to yourself: “This is mostly about ___, arguing that ___ because ___.”
Skim pitfalls to avoid
- Detail drag: During skimming, skip examples and numbers; you’ll verify later.
- Scope drift: Your gist should cover the whole passage, not one paragraph you liked.
Goal: jump straight to the line that answers a Specific Information item and prove it with the text.
What counts as an anchor (won’t paraphrase)
- Proper nouns: names of people, streets, programs, agencies.
- Numbers & formats: 12:30 pm, $14.50, June 21, 2019, 3 lanes, Unit 5B.
- Exact labels: FoodSafe Level 1, Permit 27-B, Line 980.
Micro-routine
- Extract the constraint string from the stem: e.g., “≥3 lanes, after 7 pm, weekdays.”
- Hunt the anchor: scan vertically for the number/label/time window.
- Read ±1 sentence around the anchor (or the whole cell/row in a table).
- Verify every constraint. One mismatch = reject the option.
High-frequency traps
- Digit flips: 12 vs 21; $13 vs $31.
- Time windows: before/after/until/from; check am vs pm and weekday vs weekend.
- Unit swap: km vs m; per day vs per week.
- Scope creep: a rule that applies to “temporary staff” only becomes “all staff” in a wrong option.
Goal: recognize when a choice is the same idea in new words vs. a meaning change disguised as a synonym.
Safe paraphrase families (usually equivalent)
- buy ↔ purchase, help ↔ assist, residents ↔ people who live here, because ↔ due to.
Meaning-changing pairs (not equivalent)
- must / required ≠ should / recommended
- will ≠ may / might / could
- free ≠ discounted / reduced
- extend hours ≠ hire more staff (different solutions)
- few / a few / fewer than—watch the quantifier precisely.
How to test an option for paraphrase fidelity
- Swap it back: Mentally replace the option’s phrase with the passage’s phrase. Does the sentence still mean the same thing?
- Check modality & strength: If the text says may help, reject an option that says will solve.
- Check polarity: not uncommon means “fairly common.” Double negatives hide reversals.
Goal: hear how the writer sounds (cautious, supportive, skeptical) and what they’re doing (informing, requesting, persuading, complaining).
Tone signals
- Hedges: may, might, tends to, appears to—tone softens.
- Boosters: clearly, definitely, certainly—tone strengthens.
- Attitude verbs: appreciate, regret, recommend, oppose—purpose and stance in one word.
- Register cues: Hi Alex… Thanks! (informal, friendly) vs. Dear Sir/Madam… (formal, neutral).
Purpose quick-labels
- Inform (explain a process, announce a change)
- Request (ask for action, information, confirmation)
- Persuade (argue for a policy, recommend a decision)
- Complain (report a problem + request remedy)
Two-step check
- Function test: “If I reply to this writer, what would I do?” If you’d perform an action, it’s likely a request; if you’d form an opinion, it’s persuade.
- Strength match: Align option strength with the writer’s language (hedged vs. absolute).
Infer when the text points to a conclusion through clues; Not stated when the statement isn’t supported anywhere, even indirectly.
Three filters to separate them
- Text-only rule: If you need outside knowledge to make it true, it’s not safe.
- One small step: Inference takes a tiny logical step. If you’re stacking assumptions, stop.
- Global sweep for “Not stated”: Check all relevant paragraphs for any supporting line. If none exists, then choose Not stated—not because it feels unlikely.
Examples of tiny steps vs. leaps
- Tiny step: Text says “evenings are often crowded”; safe inference: crowding is common after work.
- Leap: Text says “a pilot will run for two weeks”; unsafe inference: the city will expand it next year.
Part 1 — Reading Correspondence
- Skim salutation + closing lines for relationship and purpose.
- Scan for constraints: times, dates, budgets, addresses.
- Paraphrase check on reply drop-downs: make sure tone and facts still match.
Part 2 — Apply a Diagram
- Skim title → legend → footnotes (units/“except” notes).
- Scan the grid or rows for your constraint string.
- Paraphrase check the option text against labels exactly—singular/plural and “at least” vs “exactly” matter.
Part 3 — Information
- Skim first lines to map A–D.
- Scan to the most likely paragraph, then verify wording.
- Infer vs Not stated test before choosing E.
Part 4 — Viewpoints
- Skim to build an opinions map (who thinks what, and why).
- Paraphrase check for stance strength (cautious vs strong).
- Infer from context, not from your world knowledge.
(No sample questions here—just self-contained actions that mimic the test.)
- Gist sprint (40s): Read first lines + contrast markers, then write a one-line summary. If your line could be a title for the whole passage, you did it right.
- Anchor hunt (30s): Pick a fake constraint string (e.g., “≥3 lanes, after 7 pm, weekdays”). On any timetable, prove where it’s satisfied, and where it isn’t, using only labels and notes.
- Paraphrase A/B (45s): Take any sentence you see today. Rewrite it twice: once safe (meaning unchanged), once shifted (change strength or polarity). Train your ear to spot the shift.
- Tone toggle (30s): Turn a neutral line into request, then into complaint, by swapping 3–4 words. Notice which words flip the purpose.
- Inference filter (45s): Read a short claim. List the clues that support it. If you can’t list clues from the text, it’s not an inference—it’s a guess.