CELPIP score comparison chart: how to read the levels

“CELPIP score comparison chart” is what many people search for when they want to see what each CELPIP level means. It is not a conversion table. In our reading, the closest thing Paragon, which runs CELPIP, publishes is its level descriptors: its Scoring Levels page (Writing for levels 12, 11 and 10, Speaking for level 12) and a PDF with a row for each level down to 3. Below are what raters look at, the Listening and Reading raw score charts, and how Paragon's and IRCC's IELTS tables differ.

You may know this as the “CELPIP score comparison chart”. In our reading, people who search for this want to see what each CELPIP level means. It is not a table that converts CELPIP to another test. Paragon, the company that runs CELPIP, shows what each level means with level descriptors: short lists that each start with “At this level, I can”.

Paragon publishes the descriptors in two places that this page uses. Its Scoring Levels page has a Speaking Scores tab and a Writing Scores tab, each with sections for levels 12, 11 and 10. Only the Writing sections differ from one level to the next: the Speaking sections for levels 11 and 10 repeat the level 12 text. Its level descriptors PDF has one Speaking table and one Writing table, with a row for each level from 12 down to 3, and a final row for the lowest levels.

The Scoring Levels page also has tabs headed Sample Responses and Response Analysis next to the descriptors.

We do not copy Paragon's sample responses. You can read them on Paragon's own page. This page explains what the descriptors measure and how to use them.

Listening and Reading are not in the descriptors. A computer marks those two, so there are no responses for raters to judge.

Paragon says its raters score each response on four dimensions and that each dimension has five performance levels. Here are the dimensions and the factors Paragon lists for each.

Paragon's results page also says each response is rated by several people who work alone. Its Contact CELPIP page, in its Frequently Asked Questions, describes an AI-human hybrid system for Writing, with a human rating panel. Neither page says how the two fit together. The results page says: “Each CELPIP speaking performance is rated by three to five speaking raters, and each CELPIP writing performance is rated by four to six writing raters.” It adds: “Each dimension is divided into five performance levels.”

The descriptors all start with “At this level, I can”. We picked levels 12, 9, 7, 5 and 3 so you can see how the wording climbs. Level 12 is from Paragon's Scoring Levels page. Levels 9, 7, 5 and 3 are from the level descriptors PDF, which sets each level out as a table row with columns for Content/Coherence, Vocabulary, Listenability (Speaking) or Readability (Writing), Task Fulfillment and When. Each cell holds several lines. We took one line from the Content/Coherence column and the grammar line from the Listenability or Readability column, and joined the broken lines of each back into one sentence. After the Level column, the next two columns are Speaking and the last two are Writing.

Our reading: two things change as the level goes up. The ideas move from saying what you like (level 3), to describing something you know (level 5), to concrete and some abstract information backed by reasons (level 7), to supporting ideas with reasons (level 9), to complex, clear and precise detail (level 12). And the grammar moves from sometimes using correct grammar (level 3), to some control of simple structures (level 5), to good control of simple structures in Speaking and adequate control of complex ones in Writing (level 7), to control of complex structures (level 9), to very good control of a very broad range (level 12). If you are stuck at a level, ask which of these two is holding you back, and fix that one.

The top three levels in Writing

The Scoring Levels page gives Writing descriptors for levels 12, 11 and 10, and they differ from one level to the next. Here are three lines, one from each of the Content/Coherence, Readability and Task Fulfillment dimensions.

Our reading: at the top, the levels differ in how much range you use, how much control you have and how well your tone fits the situation. Level 10 asks for “some complexity”, level 11 for “a range of purposes” and level 12 for “a full range of purposes”.

Because a computer marks Listening and Reading, Paragon gives an approximate chart that turns the number of correct answers into a level. The ranges overlap. Paragon says the chart is approximate because the raw score required for a level may vary slightly from one test to another.

Our reading: the Reading chart asks for a slightly lower raw score than the Listening chart at the same level, for example 24 at level 7 against 27. So the same number of correct answers can mean different levels in the two skills. Compare each skill with its own chart.

Paragon does not report your raw score, and it explains why: “a raw score of 30 will not have the same meaning across different forms of a test.” For the level-by-level detail see our CELPIP listening score chart and CELPIP reading score chart.

Each CELPIP level sits next to a Canadian Language Benchmark (CLB) level, a European (CEFR) level and, for some levels, an Australian Department of Home Affairs English level. This is Paragon's table. Paragon's CELPIP versus IELTS page prints a different Australian table (Superior: CELPIP 8 to 10, Proficient: CELPIP 6 to 8, Competent: CELPIP 1 to 5). It does not match Paragon's results page or the Home Affairs pages, which ask for at least CELPIP 7 in each component for Competent English, so use the table below and check the Home Affairs page for your visa.

Our CELPIP score chart and CLB levels page sets this table out in full, and CELPIP for Australian visas covers the Australian side.

Level titles: CELPIP's pages disagree

CELPIP's own pages do not word the level titles the same way. Its results page, its guidebook and its level descriptors PDF name level 12 with “high-stakes social, educational, or workplace contexts”. Its Scoring Levels page names it with “workplace and community contexts”. We show both and do not pick one.

The Scoring Levels page covers levels 12, 11 and 10 only. For levels 9 and below, CELPIP's results page, guidebook and level descriptors PDF are the places that give a level title, and they use the older wording.

How long a CELPIP score is valid: Paragon's pages differ

Two Paragon pages give different answers. Its test format page says: “CELPIP scores are valid for two years from the date your score report is issued.” Its results page says: “CELPIP scores are valid for five years from the date your score report is issued.” Whichever is right, the body that receives your score sets its own rule. IRCC says for Express Entry: “Your test results must be less than 2 years old”, when you complete your profile and when you apply.

Paragon's CELPIP versus IELTS page prints a conversion table. IRCC prints its own tables for immigration points. They give different answers, so we show both and say which counts.

Paragon's table

IRCC's table for Express Entry (Federal Skilled Worker Program)

Our reading: look at IELTS 7.0. Paragon's table shows CELPIP 8. IRCC's table puts IELTS 7.0 in Speaking, Reading and Writing at CLB 9, and Listening needs 8.0 for CLB 9. For immigration points, IRCC's own table is the one the government applies to an IELTS result, so use it. For a CELPIP result, IRCC's table lists the same number as the CLB level: CELPIP-G 7 for CLB 7, 8 for CLB 8 and 9 for CLB 9. Treat Paragon's table as a rough guide for choosing a test. See our IELTS to CLB page and our CELPIP vs IELTS page.

This is our method, not Paragon's.

1. Choose the level you need. Check the level your application requires and read the matching row on the CLB table above.
2. Open the descriptors for Speaking or Writing at that level and read the line for each dimension. The descriptors PDF has every level from 12 down to 3. The Scoring Levels page has Writing for levels 12, 11 and 10 and Speaking for level 12.
3. Record yourself, or write, on a similar task. Use a timer.
4. Mark yourself dimension by dimension against the descriptors. Do not give one overall mark. Find the weakest dimension.
5. Read the descriptor for the level above yours and note the words that change, such as how much control of grammar it asks for.
6. Redo the task with one change: fix the weakest dimension only. Repeat with a new task a few days later.
7. Ask someone else to mark you if you can. People mark their own work too kindly, or too harshly.

Do not memorise a sample response from any source. Paragon says “providing responses that are not original and/or spontaneous may result in score cancellation”. The point of a sample is to see the features, not to copy the words.

Common questions