Most candidates can recite the pass mark and almost none can explain how they would lose it. The NAATI CCL is not scored on a general impression of how well you did — a marker moves through your rendering unit by unit, taking marks off for specific, named problems. Once you know what those problems are called and what they sound like in Tamil, your practice stops being vague self-assessment and becomes genuine error-hunting.
- The test is marked out of 90 across two dialogues of 45 marks each, and the two are marked separately.
- You need 63 out of 90 overall AND at least 29 out of 45 on each dialogue — both conditions, not either.
- Because the per-dialogue floor is checked first, a strong dialogue cannot rescue a weak one.
- Marks come off for named problems: omission, distortion, unjustified addition, register error and language quality.
- One segment per dialogue can be repeated with no penalty — a second repeat in the same dialogue costs marks.
- A word-for-word rendering is often the inaccurate one — accuracy means the meaning unit, said naturally.
The arithmetic: 90, 63, and the 29 floor
Everything else in this guide hangs off the structure. Your test is two dialogues, each worth 45 marks, so the whole test is marked out of 90. To pass you need 63 out of 90 overall, and separately at least 29 out of 45 on each of the two dialogues.
Notice what the two conditions do together: the 63 is a bar on your total performance, the 29 a bar on your worst. You must clear both, and they are not the same test.
A strong dialogue cannot rescue a weak one
This is where candidates get caught. The instinct is to treat the test as one pool of 90 marks and assume that banking a big score on the dialogue you find easy gives you room to struggle on the other. Suppose you have a very strong first dialogue — a health topic you have drilled to death — and you score 40 out of 45. Then the second dialogue is a legal one and it goes badly: 28 out of 45. Your total is 68. That is five marks clear of 63. And you have not passed, because 28 is below the 29 minimum for that dialogue, and the minimum is applied before the total is ever consulted.
| Dialogue 1 | Dialogue 2 | Total | Result |
|---|---|---|---|
| 40 / 45 | 28 / 45 | 68 / 90 | Fail — dialogue 2 is under the 29 floor |
| 34 / 45 | 30 / 45 | 64 / 90 | Pass — 63 cleared, both dialogues above 29 |
| 29 / 45 | 33 / 45 | 62 / 90 | Fail — both floors cleared, but the total misses 63 |
| 44 / 45 | 18 / 45 | 62 / 90 | Fail — both conditions missed at once |
The one free repeat per dialogue
There is one mark-protecting mechanism built into the test itself, and most candidates either forget it or misunderstand how it is rationed. In each dialogue you may ask for one segment to be repeated with no penalty at all; ask for a second repeat within that same dialogue and it costs you marks.
Read the allowance carefully, because the wording does the work: it is one per dialogue, not one per test, and it does not carry over. Sail through the first dialogue without using your repeat and it is simply gone. Spend it on an early segment that felt slightly quick, and you have nothing left when a genuinely dense one arrives later.
How marks actually come off
A marker is not scoring an overall impression. They work through your rendering unit by unit against the meaning the speaker expressed, and where a unit is missing, wrong, invented, or delivered in language that damages the message, marks come off — and the problems have names.
The deductions fall into two families. Accuracy problems are about the meaning: what failed to arrive, what arrived wrong, and what arrived that should never have been there. Language quality problems are about the delivery: grammar, vocabulary, register and intelligibility. Both cost you marks, and a rendering can be hit by both at once. Practically, every fact, condition, instruction, name, number and qualifier in the source is a unit that has to survive the crossing — so a rendering that sounds fluent but quietly drops one is still short a unit.
Deduction 1 — omission
An omission is meaning that was in the source and never appeared in your rendering. It is the most common deduction and the easiest to miss in your own practice, because it does not sound like an error — it sounds like a fluent sentence that happens to be short one fact. The source says: “The appointment is on Thursday at ten, and please bring your Medicare card.” The rendering comes back as:
சந்திப்பு வியாழக்கிழமை பத்து மணிக்கு.Sandhippu viyaazhakkizhamai paththu maNikku.The appointment is on Thursday at ten.
Nothing in that sentence is wrong: the day is right, the time is right, the Tamil is clean. And the instruction to bring the Medicare card has vanished. That is a full omission of a meaning unit, scored as one no matter how good the rest was — and the patient turns up on Thursday without their card.
Deduction 2 — distortion
A distortion is meaning that arrived, but arrived changed. The unit is present in your rendering; it has simply been turned into something the speaker did not say. Numbers, dates, negation and conditions are the classic casualties, because each one has a near neighbour that sounds almost the same and means something entirely different.
Worked example
The source says: “You do not need to fast before this blood test.” The rendering comes back as:
இந்தப் பரிசோதனைக்கு முன் நீங்கள் சாப்பிடக் கூடாது.Indhap parisodhanaikku mun neengaL saappidak koodaadhu.You must not eat before this test.
Here the speaker removed an obligation — you do not need to fast — and the rendering has imposed one: you must not eat. Every word is real Tamil and the sentence is grammatical. It is also the opposite instruction, and the patient arrives hungry for no reason. That is a distortion, and it is treated more seriously than an omission, because the listener is now confidently misinformed rather than uninformed.
An omission leaves the listener not knowing something. A distortion leaves them knowing something false.
Negation is the highest-risk category here, followed closely by numbers and dates — our guide to the most common mistakes drills the patterns that produce them.
Deduction 3 — unjustified addition
An unjustified addition is meaning in your rendering that was never in the source. Candidates find it the hardest deduction to accept, because additions almost always come from a genuine wish to help — you explain a term, you soften bad news, you add the advice the listener needs.
Worked example
The source says: “Your application has been refused.” The rendering comes back as:
உங்கள் விண்ணப்பம் நிராகரிக்கப்பட்டது, ஆனால் நீங்கள் மீண்டும் விண்ணப்பிக்கலாம்.UngaL viNNappam niraakarikkappattadhu, aanaal neengaL meeNdum viNNappikkalaam.Your application has been refused, but you can apply again.
The first half is accurate. The second half is advice the speaker never gave and which may not even be true: you have stepped out of the interpreter's role into the caseworker's, and the listener now believes an official told them something no official said.
Deduction 4 — register
Register is how formal or respectful your language is, and Tamil marks it structurally rather than optionally: the pronoun you choose and the verb ending that agrees with it announce your stance on every sentence. There is no neutral option to hide behind. Here a doctor is speaking to an adult patient they have never met, and the rendering comes back as:
நீ நாளைக்கு வா.Nee naaLaikku vaa.Come tomorrow. (familiar / intimate)
The information is accurate; the relationship it encodes is wrong. நீ and the matching verb form are what you use with a child or a close friend, so what the listener hears is not a doctor being warm but a doctor being inappropriate. The accurate version keeps the meaning and repairs the stance:
நீங்கள் நாளைக்கு வாருங்கள்.NeengaL naaLaikku vaarungaL.Come tomorrow. (polite / respectful)
That decision is made in the first second and then has to hold. Drifting between நீ and நீங்கள் mid-answer is worse than choosing the wrong one consistently, because it reads as loss of control.
Deduction 5 — language quality
The final family covers control of the language you speak into: grammatical accuracy, word choice, clear pronunciation and delivery fluent enough to follow. Not a style assessment — a judgement about whether your output works as language.
| Problem | What it sounds like | Why it costs marks |
|---|---|---|
| Grammatical error | Wrong case marking, a verb that does not agree, a tense that contradicts the sentence | Small slips are minor; repeated ones force the listener to reconstruct your meaning |
| Vocabulary choice | A word that exists but is wrong for the domain, or an English word dropped in where Tamil has a normal everyday equivalent | The listener either mishears the concept or is excluded from it |
| Intelligibility | Mumbling, trailing volume, sounds that blur the difference between two words | A meaning unit the marker cannot make out is a meaning unit that did not arrive |
What “accurate” means when word-for-word is wrong
Candidates often treat literal rendering as the safe option — stay close to the words and nobody can accuse you of changing anything. In interpreting that is backwards: the unit of accuracy is the meaning, and a rendering that tracks the words while losing the meaning is scored as inaccurate. Take an English speaker saying “we'll keep an eye on it.” Word by word you produce a sentence about a literal eye placed on an object, which no Tamil speaker would ever say. The accurate rendering carries the meaning — that the matter will be monitored — in language they actually use:
நாங்கள் அதைக் கவனித்துக்கொண்டே இருப்போம்.Naangal adhaik kavaniththukkoNdE iruppOm.We will keep monitoring it.
The test is not “does this match the source words?” but “would a Tamil speaker, hearing only my version, understand exactly what the English speaker meant, and nothing more?” Where literal and natural diverge, the natural one is accurate. The guide to idioms and literal versus meaning-based rendering works through many more cases of exactly this.
How to mark your own practice
Vague self-review — “that felt okay, my Tamil was a bit shaky” — names no error type and so suggests no fix. Marking yourself by category does.
- 1Write the source units down before you listen backList every fact, number, date, condition and instruction in the segment. This is your checklist, and building it first stops your memory of what you said from contaminating it.
- 2Play your rendering, tick units off, then hunt for extrasAnything left unticked is an omission; anything ticked but altered is a distortion. Be strict — 'close enough' is a category the marking does not have. Then look for meaning that is not on your source list at all: explanations, advice, softeners and intensifiers are the usual suspects.
- 3Count your error types, not your feelingsOver a week, the pattern tells you what to train. Six omissions and one grammar slip is a listening and note-taking problem, not a Tamil problem — and the fix is completely different.
Quick check: how the marking works
After the result: what your raw score lets you do
Your result arrives by email, typically after around four to six weeks, and what comes back is a raw score against the 90 marks. NAATI offers a review of a result, sometimes called a remark, in which your recorded performance is assessed again — but it is not open to everyone. A review is only available where the raw score is 58 or above, and it costs $187. Below that threshold a review is simply not on the table: a review corrects a marking discrepancy, it does not compensate for a performance that fell a long way short.
Knowing the deduction types changes what practice is for: you are now hunting five specific, nameable failures in your own output, and each has a different fix. Omissions are usually a listening and note-taking problem, distortions a processing-speed problem, additions a discipline problem. Register is a decision you make in the first second, and language quality is the long, slow one. Run full-length practice under exam conditions so the per-dialogue floor becomes real rather than theoretical — a timed mock test puts both dialogues back to back the way the real thing does, and shorter practice tests let you drill one deduction type at a time. If you have not yet read how the two dialogues are actually structured and delivered, start with our breakdown of the test format.