What is actually going wrong
People who lose marks on CELPIP Reading can nearly always read the passage perfectly well.
They lose to the clock, to the blanks, or to a distractor built to catch someone who read carefully. The standard advice — read more English — is the slowest possible route, because it treats a reading-speed problem, a task problem and a trap problem as one thing.
The failure: the undiagnosed retake
The four Reading parts come to 38 questions in 43–56 minutes, and a count out of 38 tells you nothing about which failure took your marks. So someone scores 28, books a retake, studies the same way for six weeks, and barely moves — the practice was aimed at the average of four separate problems. Below are the four. Find yours, with your own practice results in front of you.
1. Running out of time in one part
Recognise it: your wrong answers cluster at the end of a part. Not spread evenly — bunched in the last two or three items, often with a guess you do not remember making.
What it is: the section is timed per part, and the test moves on by itself — "the test will automatically move forward to the next part", and "you can't go back to a previous part." So finishing Part 1 with four minutes to spare does not buy you four minutes in Part 4. Each part is a closed box.
The fix is per-part discipline, not faster reading. Practise one part at a time against a clock, and treat its last item as a deadline you hit deliberately. Because there are no points deducted for wrong answers, never let a part close with anything unanswered.
One warning: the free official sample test lets you go back and change answers. The real test does not. If you have only practised there, you may never have felt this failure.
2. Losing the blanks
Recognise it: you do well on the comprehension questions and badly overall. The gap is the tell.
What it is: Parts 1, 2 and 4 each have a second, shorter text with blanks to fill from dropdowns — 5 in Part 1, 3 in Part 2, 5 in Part 4. That is 13 of the 38 marks on a task that is not comprehension at all. That second text is a reply or a comment written by someone who has read the main text, so a blank tests consistency with the main text, not grammar.
Every option in a well-built blank is grammatical. If you choose by ear — which one sounds right — you are answering a question the test is not asking. Go back to the main text and ask which option the writer of the reply could actually have written, given what they read.
3. Distractors
Recognise it: when you check the key, the right answer is obvious the moment you see it. That reaction means your reading was fine and your selection was not.
What it is: the wrong options are built from true statements that answer a different question, details attached to the wrong person, and scope shifts — some for most, may for must, this year for last year.
The fix is a checking habit, not more vocabulary. Before you commit, name what the question asked about, and confirm the option is about that same one.
4. The Part 2 miss
Recognise it: Part 2 is the part where your accuracy drops, even though it is the shortest.
What it is: in Reading to Apply a Diagram the main text is "a diagram with point form text and visuals" — not prose — and the task is to apply it to a person's situation. So the standard trap is a true reading of the diagram belonging to the wrong person. Every distractor can be verified somewhere in the diagram.
A worked example
(original AlmiWorld practice item)
Riverbend Community Centre — Winter Registration
| Stream | Who it is for | Register by | Note |
|---|---|---|---|
| Waitlist transfer | On last winter's waitlist | 8 Nov | Place automatic, confirm by email |
| Early Bird | Members who joined before 1 September | 15 Nov | Last winter's fee held |
| Household rate | Two or more at one address, at least one a member | 15 Nov | Early Bird stream only |
| Open registration | Anyone, member or not | 1 Dec | Full fee |
Priya became a member in July. She lives with her brother, who has never joined. Neither was on a waitlist last winter. She wants both of them in the program at the household rate.
What does Priya need to do?
- A. Register both of them under Early Bird by 15 November.
- B. Register both of them by 1 December.
- C. Have her brother become a member first.
- D. Confirm their places by email.
A is correct. The household rate is Early Bird only, and Priya joined before the 1 September cut-off.
B is true of a different question. 1 December is a real deadline in the diagram — the open registration one, at full fee. The date is accurate; it does not go with the rate she wants.
C is a scope shift. The diagram says at least one a member. C reads that as "each".
D is a detail from another row. Confirming by email belongs to waitlist transfer — and neither of them was on a waitlist. A true detail attached to the wrong person is the standard Part 2 trap.
The shape that works: run a session so it diagnoses
- Do ten items under a clock, one part at a time. You want detail, not a score.
- Mark them, but do not stop at the score. The score is what hides the diagnosis.
- For each wrong answer, write which of the four it was: clock, blank, distractor, Part 2.
- Next to it, write whether you still had time left. Wrong-with-time and wrong-with-none are different illnesses.
- If it was a distractor, add one word: other question, wrong person, scope.
- Ten items give you a distribution, and the distribution is your study plan — four marks in one column means you have one problem, not four.
What actually costs you the mark
- Practising whole sections and only ever recording the total.
- Choosing blanks by what sounds fluent instead of by what the main text established.
- Reading Part 2's diagram thoroughly, then answering about the wrong row.
- Leaving anything unanswered when a part is about to close.
- Rehearsing only where going back is allowed, so you never rehearse the deadline.
- Treating "read more English" as a plan. It is a habit — it is not a diagnosis.
Try one
[Try a Reading task →] — one full task, marked, no card required.