CELPIP Speaking score chart: the 4 criteria and what each band means
If you came here looking for a raw score to convert, there is good news and bad news. Speaking has no raw score at all. Every one of the 8 tasks is rated on four official criteria, and this page is the chart that actually decides your level: what each criterion measures, and the band descriptors that separate a CLB 5 from a CLB 9 from a CLB 11.
There is no raw score in CELPIP Speaking
Most people arrive at a score chart holding a number. In Reading that works beautifully: 38 questions, 1 mark each, count your correct answers and read the row. We built a calculator for exactly that, because reading is arithmetic.
Speaking is not arithmetic, and no amount of searching will produce the conversion table you are looking for, because it does not exist. There are no questions with right answers. There is nothing to add up. Instead, a rater listens to each of your 8 tasks and rates it on the same four criteria, and your reported level from CLB 1 to 12 emerges from all 8 tasks together.
That sounds vaguer than it is. The four criteria are published, they are specific, and each one names exactly what it measures. Once you know them, feedback stops being mysterious and your own recordings become readable. That rubric is the real CELPIP speaking score chart, and it is what the rest of this page lays out.
The 4 CELPIP Speaking criteria
These four are what your speaking level is actually made of. Every one of the 8 tasks is rated on all four, so they apply just as much to describing a picture as to persuading your supervisor. The sub-points and the guiding question under each are Paragon's own.
Content/Coherence
"How well are the ideas organized and developed?"
- •Number of ideas
- •Quality of ideas
- •Organization of ideas
- •Examples and supporting details
Note that number and quality are separate points. Six shallow ideas and one deep idea both fall short, and the examples line is why "I think it is better" without a reason costs you here even when the opinion is perfectly sensible.
Vocabulary
"What is the range of the test taker's vocabulary and can they use it naturally?"
- •Word choice
- •Suitable use of words and phrases
- •Range of words and phrases
- •Precision and accuracy
Read the guiding question again: range and whether you use it naturally. Memorised advanced words dropped where they do not belong fail the suitability and precision points, so they can lower this score rather than raise it.
Listenability
"How easy is it to listen to and understand the response?"
- •Rhythm, pronunciation and intonation
- •Pauses, interjections and self-correction
- •Grammar and sentence structure
- •Variety of sentence structure
Speaking's own criterion, and the most misread. It is not an error count and it is not an accent test. It asks how much work the listener has to do. Grammar sits here because a broken sentence is hard to follow, not because it is untidy.
Task Fulfillment
"How well did the test taker follow the instructions? Did they use an appropriate tone?"
- •Relevance
- •Completeness
- •Tone
- •Length
The cheapest marks in the test live here. Tone is scored, so your supervisor and your friend do not get the same sentences. Length is scored, so stopping at 20 seconds of a 60 second window costs you even when everything you said was right.
Yes, "Task Fulfillment" really is spelled with two Ls. It is the official name Paragon gives the criterion, so we keep their spelling even though the rest of our writing is Canadian English. Listenability is speaking's own criterion and has no equivalent in the reading or listening sections.
CELPIP Speaking band descriptors: CLB 5, 9 and 11 compared
Knowing the four criteria tells you what is measured. The band descriptors tell you what each level looks like. Put three levels side by side and the ladder becomes obvious: it is not about making fewer mistakes, it is about what language you are entitled to reach for at all.
What a speaker at each level can do
| What is measured | CLB 5 | CLB 9 | CLB 11 |
|---|---|---|---|
| Vocabulary you can reach for | Common words and phrases only. No abstract language at all | Common and context-specific vocabulary, with some idiom | A broad range of concrete and abstract language, and a range of figures of speech |
| Sentence structures | Some control of simple structures | Some control of complex structures | Good control of a broad range of complex and diverse structures |
| Your listener | Conveys information on a familiar topic | Adjusts style and tone to a range of audiences | Adapts to the situation, the purpose, and their relationship to the listener |
Source: Paragon Testing Enterprises, the test developer for CELPIP. These are the entitlements at three points on the CLB 1 to 12 scale, chosen because they mark the practical steps most candidates are climbing between.
Not "uses abstract language poorly". None at all. This is why a CLB 5 answer can be entirely correct and still sound thin: every idea has to be carried by concrete, common words.
Adjusting style and tone to a range of audiences is a CLB 9 entitlement. If you speak to everyone in one register, that alone is holding you below 9 no matter how accurate you are.
Not just who your listener is, but what you are to them. A CLB 11 speaker sounds different asking a favour of a stranger than of a colleague they have known for years.
What actually separates the bands
Here is the most useful thing on this page, and it took reading the descriptors side by side to see it. The CLB 5 candidate and the CLB 12 candidate described the same picture. Same farmers market, same eight people, same dog pulling toward the same dropped muffin. Both of them saw it. The gap between them is not what they noticed.
The gap is the descriptive machinery they had to render it with. One had common words and simple structures and got the information across. The other had abstract language, figures of speech, and complex clauses that could hold two ideas in the right relationship to each other. Same eyes, different equipment.
That reframes what practice is for. If the gap were observation, you would practise looking harder at pictures. It is not, so you should not. What separates the levels is precision instead of vagueness, hedging when you are unsure, clauses that subordinate one idea to another, and tone tuned to who is listening. Every one of those transfers to all 8 tasks, which is exactly why they are worth building and why memorising topics is not.
The same logic explains Listenability. The ladder there is control first, then range, not a tally of errors. A simple sentence you finish cleanly outscores a complex one that collapses in the middle, because the criterion asks how easy you are to listen to. Reach for the complex structure only once you can land it. That is what "some control of complex structures" at CLB 9 actually means, and what "good control of a broad range" at CLB 11 sounds like.
How your CELPIP Speaking level is built
Eight tasks, four criteria, one level. Every one of the 8 tasks is rated on all four criteria, and your reported level is a product of all 8 together. That last part is worth sitting with, because it cuts both ways. A disastrous Task 5, the one where your mind went blank and you talked about the option you had already rejected, cannot sink you on its own. Neither can a brilliant Task 1 carry the seven that follow it.
What that structure rewards is consistency. Eight solid tasks beat two excellent ones and six adequate ones, which is why drilling your favourite task type is a poor use of practice time and why sitting the full set of 8 is the honest test. It is also why our speaking tests always run all 8 tasks rather than letting you cherry-pick.
There is no percentage either, and no mark out of anything, because there is no raw score to take a percentage of. Your speaking result is a level from CLB 1 to 12 that appears on your CELPIP score report beside your listening, reading and writing levels. The four levels stay separate and are never averaged into one overall band, so a strong reading level cannot rescue a weak speaking one. Immigration programmes read each section on its own.
Each task is rated
All four criteria are applied to each of the 8 tasks, from advice to the unusual situation.
All 8 combine
The level is a product of the whole set, not a pass or fail on any single task.
One level is reported
A single CLB 1 to 12 speaking level, listed beside your three other section levels.
The CELPIP Speaking levels, band by band
Speaking levels group into the same bands as the rest of CELPIP. Here is what each band sounds like in a speaking test, described through the criteria on this page rather than through a raw score.
The band names are Paragon's, the plain English summaries are ours. The right-hand column is our reading of what each band sounds like on a speaking task, written to be useful when you listen back to your own recording. It is a study aid, not an official CLB descriptor.
| CLB level | Band | What it sounds like on a speaking task |
|---|---|---|
| 10 to 12 | Highly Effective | Concrete and abstract language both available, figures of speech used naturally, good control of complex and diverse structures, and the register shifts with the listener and the relationship |
| 7 to 9 | Effective | Context-specific vocabulary and some idiom on top of common words, some control of complex structures, and style and tone adjusted across a range of audiences. The level most competitive PR applicants target |
| 5 to 6 | Developing | Common words and phrases carrying every idea, some control of simple structures, information conveyed on familiar topics. Accurate but thin, and no abstract language to reach for |
| 3 to 4 | Limited | Short, predictable responses on very familiar topics, with the listener doing much of the work to follow |
| M | Level M | Minimal proficiency, reported when there is not enough evidence to place the response on the CLB 1 to 12 scale |
Level M means minimal proficiency. It is not a failure of the test, it simply means the result sits below the measurable range. Only Paragon can issue an official CELPIP level.
Why the speaking score chart looks nothing like the reading one
If you have already seen our reading score chart, this page will have felt strange. That one is a calculator: you tell it how many of the 38 questions you got right and it tells you your CLB level. It can do that because reading is objective. Every question has one correct answer, the marking is mechanical, and a raw score genuinely exists.
Speaking has none of those properties. There are no questions, no options and no answer key, so there is no number to feed a calculator. Building a speaking converter would mean inventing a raw score that the test does not produce, and any site showing you one is showing you fiction. What speaking has instead is a rubric, and a rubric answers a different and frankly more useful question: not "what level is 29 out of 38", but "why is my level what it is, and what would move it".
The two sections do share the scale. Both report CLB 1 to 12, both use the same band names, and both sit as separate levels on your score report with no averaging. If it is the raw score conversion you actually came for, that is the reading one: the CELPIP reading score chart and calculator turns a score out of 38 into a CLB level, with the official conversion table and an explanation of why the published ranges overlap.
How to move up the CELPIP Speaking bands
Read the descriptors as instructions rather than as verdicts and each one names something you can build. Here is where the levels are actually won.
Trade vague words for precise ones
Vocabulary marks precision and accuracy explicitly. "A thing for cutting" and "a pair of shears" describe the same object at two different levels. Hunt for the vague nouns in your transcript, "thing", "stuff", "place", and replace them one at a time.
Land one complex structure per task
CLB 9 needs some control of complex structures, not a display of them. Pick one, a because clause, a relative clause, a conditional, and land it cleanly. One that works beats three that collapse, because Listenability rewards control before range.
Name your listener before you speak
Tone is scored under Task Fulfillment, and adjusting to your audience is a CLB 9 entitlement. In your prep window, say who you are talking to in your head. Supervisor, cousin, stranger on the phone. Let that set the register before the first word.
Fill the window with content, not filler
Length is scored, but padding is not. If you finish early, add an example or a reason, which also feeds the supporting details point under Content/Coherence. Two criteria improve from the same habit, and it is the easiest one on this list.
Self-correct briefly, then move on
Self-correction sits inside Listenability, which means it is measured but not forbidden. A quick fix and onward costs little. Restarting the whole sentence twice costs a lot, because the listener has to rebuild your thought each time.
Diagnose by criterion, not by feeling
"That went badly" is not actionable. "My ideas were fine but I had no words for the shape" is. Score your own recording against the four criteria separately and you will usually find one is dragging the rest, and that one is where your practice belongs.
Hear yourself against these criteria
Each test is a full 8 task speaking section on the official clock, recorded, transcribed, and returned with expert feedback on all four criteria and a CLB estimate.
CELPIP Speaking score chart FAQs
Put the criteria to work
A rubric only tells you what is measured. Practice is what moves you up it. Every section is free and built for the 2026 format.
Speaking
All 8 tasks recorded on the real clock, transcribed and rated on the four criteria.
Start speaking →Reading score chart
The raw score conversion: turn a score out of 38 into a CLB level instantly.
Convert a raw score →Writing
Email and survey tasks in a timed editor, returned with expert feedback.
Start writing →Listening
Problem solving, daily life, information, news, a discussion and viewpoints across all 6 parts, plus an unscored practice task.
Start listening →Find out where you land on these criteria
Stop guessing at your band. Sit a free CELPIP Speaking test built for the 2026 format, record all 8 tasks, and get expert feedback on Content/Coherence, Vocabulary, Listenability and Task Fulfillment with a CLB estimate.