How CELPIP Reading Works: 4 Parts, 38 Questions
Four parts, 38 questions, about 39 minutes, and a timer that resets every time you advance. Before any technique is worth learning, you need a map of what is coming and how long you get for it.
The section in one page
CELPIP Reading is four parts, 38 scored questions and roughly 39 minutes of clock. That is the whole section. Nothing is typed, nothing is recorded, and every answer is chosen from a drop-down of four options. There is no penalty for a wrong choice, which is a structural fact worth remembering long before test day.
The four parts are not four versions of one task. Part 1 hands you a personal letter. Part 2 hands you a chart. Part 3 hands you an informational passage chopped into labelled paragraphs. Part 4 hands you a website article in which several named people disagree with each other. A candidate who reads all four the same way will feel comfortable in two of them and lost in the other two, and will usually blame vocabulary for a problem that is really about method.
The question counts are uneven, and so are the minutes. Part 1 carries 11 questions in about 10 minutes. Part 2 carries 8 in about 8. Part 3 carries 9 in about 9. Part 4 carries 10 in about 12. Those add to 38 questions in about 39 minutes, which sounds comfortable until you notice that two of the four parts require you to hold two documents against each other rather than read one.
What each part actually asks
Part 1, Reading Correspondence. One personal message, usually 300 to 400 words, from one named person to another. Questions 1 to 6 are sentence stems that walk down the message from top to bottom. Questions 7 to 11 show you something new: a short reply written back to the sender, with five blanks in it. Same source text, entirely different reading job.
Part 2, Reading to Apply a Diagram. On the left, a chart comparing four similar options, each with a short list of features and a line giving a price or a term. Four transit passes, four campsites, four storage units. On the right, five blanks inside an email that applies the chart to one person's circumstances, then three sentence stems about that email's reasoning.
Part 3, Reading for Information. One informational passage split into four labelled paragraphs, A to D, and nine statements. For each statement you choose the paragraph that contains that information, or E, meaning none of them does. Nobody asks what the passage means or what the author thinks. The question is always the same one: where is this?
Part 4, Reading for Viewpoints. An article from a website, roughly 450 to 600 words, about a decision some Canadian town is arguing over, with three or four people quoted by name. Questions 1 to 5 are sentence stems. Questions 6 to 10 are five blanks inside a comment posted underneath the article by a reader, which brings a fifth voice into the argument.
The clock resets, and that changes your planning
The reading timer is set per part and restarts when you advance. Minutes you save in Part 1 are not banked for Part 4, they are simply gone. This is the opposite of how most candidates plan, because most candidates arrive with a single global sense of being ahead or behind, and that sense is meaningless here.
What replaces it is a per-part budget you should already know before you sit down. Look at the ratios rather than the totals. Part 1 gives you about 55 seconds a question. Part 2 gives you about 60. Part 3 gives you about 60. Part 4 gives you about 72, the most generous in the section and also the longest text.
Part 2 is the part where those numbers mislead. Sixty seconds sounds like plenty until you realise the minute has to cover two trips across the screen: one into the email to find what the person needs, one into the chart to find the fact that settles it. The reading is not slow. The cross-referencing is, and no amount of raw reading speed fixes it.
Learn the four per-part budgets as numbers before test day: 10 minutes for 11 questions, 8 for 8, 9 for 9, 12 for 10. Because the clock resets, a vague sense of being ahead is worth nothing, and a specific one is worth marks.
Two answer formats, and only one of them is obvious
The format you expect is the sentence stem: half a sentence with a drop-down at the end, four options, pick the one the text supports. Parts 1, 2 and 4 all use it, and so does most reading practice material on the internet. It behaves the way you think it behaves.
The format that costs marks is the blank inside a second text. Five in Part 1 inside the reply, five in Part 2 inside the email, five in Part 4 inside the comment. That is 15 of the 38 questions, nearly forty per cent of the section, all working on the same principle: you are reading a document written by somebody who has already read the source text, and the facts they relied on have been cut out and replaced with drop-downs.
The rule that follows applies in all three parts. The second text never contains its own answers. It contains the shape of an answer, a sentence with a hole in it, and the hole can only be filled by something the source text will support. If you can fill a blank without going back to the letter, the chart or the article, you have probably been caught.
Part 3 is the exception to both formats. It is a matching task with five possible answers rather than four, and its fifth answer, E, is a real conclusion rather than a safety valve.
Where the time actually disappears
Three places, reliably. The first is Part 2, for the reason above. Candidates read all four columns of the chart closely before they know what they are looking for, discover that the four options are near-identical, and emerge four minutes later with no idea which details mattered.
The second is the reply and comment blocks at the end of Parts 1 and 4. Those blocks are slower per question than the stems that precede them, and they sit at the end of the part, where any overspend earlier has already eaten the margin. Candidates who reread the letter once per question in Part 1 arrive at the reply with three minutes left for five of the harder questions in the section.
The third is a single question that will not resolve. In a part with a fixed clock and no penalty for guessing, a question you have already spent ninety seconds on is not an investment, it is a leak. Choose the best of the four, move, and come back only if the part ends early.
Never leave a drop-down on its default. There is no penalty for a wrong answer in CELPIP Reading, so an unanswered question is a mark handed back for nothing. If a part is about to end, spend the last ten seconds filling every blank you skipped.
How 38 questions become a level
Your raw score out of 38 is converted to a CELPIP level from 1 to 12, aligned to the Canadian Language Benchmarks. The conversion is not linear, and knowing where the boundaries sit changes what you should aim for.
Near the top, the bands are narrow. Thirty-one or 32 correct is CLB 9. Twenty-eight to 30 is CLB 8. Twenty-four to 27 is CLB 7. That means a single question can move you a whole band at the level most immigration applications care about, while three questions in the middle of the CLB 7 range move nothing at all.
Two things follow. First, nobody needs 38 out of 38, and chasing it wastes preparation on question types you will meet once. Work out the level you need, find the raw score that corresponds to it, and aim two or three questions above that so one bad passage on the day does not cost you the band. Second, every question is worth the same. Part 4 is the hardest part of the section and is worth no more per question than Part 1. The marks are not where the difficulty is, they are where the questions are.
What this section is not testing
It is not testing the size of your vocabulary in the abstract. The passages use ordinary Canadian English about ordinary Canadian subjects: a bylaw, a transit pass, a beaver, a library that lends tools. Words you do not know will appear, and almost all of them are survivable from context.
It is not testing reading speed for its own sake either. Very few candidates fail Reading because their eyes move too slowly. They fail it because they read the wrong thing first, reread what they had already understood, or answered a question adjacent to the one they were asked.
What it is testing is the ability to hold a text and a question against each other and notice the difference. Almost every wrong option you will meet in this section is true. It is a real detail, from the real text, attached to the wrong person, the wrong option or the wrong question. Learning to ask does this answer the stem rather than is this true is most of the work of the next eighteen lessons.
When an option feels right because you remember reading it, stop. That is the exact sensation the distractors are built to produce. Go back and check that the detail belongs to the person, option or paragraph the question is actually about.
Practise it
Five questions on the shape of the section itself, which is worth a mark or two before you read a word of any passage.