Lesson 9, Speed, Elimination & Guessing
What to do when the clock is winning. The tactics here assume the part methods (Lessons 5–8); this is the layer on top.
Elimination is the primary tool, not the fallback
Most candidates treat elimination as what you do when you can’t find the answer. Reverse that:
Work by removing wrong options first, even when an option already looks right.
Three reasons this wins on CELPIP specifically:
- The traps are catalogued. After Lesson 4 you can often name why an option exists, “keyword echo,” “absolute,” “reattribution.” A named trap is an eliminated option, and elimination by trap-type is faster than confirmation by re-reading.
- It defuses the “two good options” deadlock. The question is never “which feels right?” but “which one can I kill?” One of them always fails the pointing test or carries an unsupported absolute.
- It converts guesses into odds. Four options blind = 25%. Two eliminated = 50%. Across the handful of questions you’ll genuinely guess on, that difference is often two or three raw marks, real levels, for seconds of work.
The elimination pass, in practice
For any question where the answer isn’t immediate:
- Kill the trap options first: echo, absolute, reattribution, assumption. Usually two die in ten seconds.
- Apply the pointing test to the survivors: which one has an actual line behind it?
- Still tied at 90 seconds? Pick the more supported (not the more interesting), flag it, move.
The guessing hierarchy
Not all guesses are equal. When the 90-second rule fires, guess in this order of preference:
| Situation | Guess |
|---|---|
| Two options eliminated | Either survivor, 50% |
| One option eliminated | Avoid absolutes and echoes among the rest |
| Nothing eliminated, blank-filling question | The option matching the tone and polarity of the surrounding text |
| Nothing eliminated at all | The moderate, hedged option over the extreme one |
That last row isn’t superstition, it follows from Trap 2: unsupported absolutes are disproportionately wrong, so under total uncertainty, the measured option is the better bet.
In Parts 1 and 4 blanks, you’re never really guessing blind. By the response letter or the comment section you know the writers, their side, their warmth, their register. A “guess” informed by character and polarity is closer to a 70% shot than a 25% one. That’s why the part methods front-load understanding the people.
Speed tactics by part
The methods are the speed. This is where each part’s minutes hide:
| Part | Where time is lost | The recovery move |
|---|---|---|
| 1 | Reading options before the question; early summary questions | Cover the options; defer the summary |
| 2 | Absorbing the whole diagram | Email first, retrieve, don’t read |
| 3 | Re-reading paragraphs per statement | Prime statements; one pass; E assigns itself |
| 4 | Re-reading the article per question | The speaker map, jump to clusters only |
The shared thread: every recovery move replaces re-reading with structure. If you notice yourself scrolling, that’s the signal that you’ve dropped the method, pause two seconds, name the method step you’re on, resume.
The last-two-minutes protocol
At 55 minutes, whatever remains unfinished:
- Fill every blank first. Thirty seconds. Guesses per the hierarchy, a blank is the only unforgivable answer.
- Then revisit flags, newest first, the passage is freshest for the most recent part.
- Don’t reopen settled answers. Late-test second-guessing without new evidence degrades scores more than it repairs them. Change an answer only if you can name the specific error you made, “that was Wendell’s claim, not James’s”, never on drift.
The mindset piece
Two failure patterns live entirely between the ears:
The sunk-cost question. You’ve spent two minutes; leaving feels like waste. It isn’t, the two minutes are gone either way, and the next question is fresh marks. The 90-second rule is precisely a pre-commitment against this trap.
The wobble after a hard part. A brutal Part 3 bleeds confidence into Part 4, and shaken candidates start second-guessing method itself, re-reading everything, marinating on options. The fix is mechanical: each part is a new test. Run the method. The method doesn’t know you had a bad ten minutes.
Grade your own speed discipline
Audit any timed part:
- I eliminated before I confirmed
- I named the trap in at least half the options I killed
- No question exceeded ~90 seconds on first pass
- Every guess followed the hierarchy, no blind picks with an elimination available
- Zero blanks at the buzzer
- I only changed flagged answers with a nameable reason
- I caught myself scrolling and returned to the method step
6+ ticks = your minutes are going where the marks are.
Your practice
Open one set on Celpora and put today’s lesson to work.
Practice Reading. Say the reason each wrong option fails before you pick the correct one.
