TCF Canada speaking: you hear the sound you meant to make, not the one you made

There is no shortage of TCF Canada speaking advice. Structure your answer. Use connectors. Record yourself. Practise with a partner. Here is a model answer at NCLC 7 so you can see what good looks like.

All of that is reasonable, and the best of it is genuinely useful. It also shares one blind spot, and the blind spot is not a small one.

A sample answer is text. Expression orale is sound. You can read an excellent model answer, absorb its structure, memorise its connectors, rehearse it in your head until it is fluent — and still produce, out loud, something an examiner has to work to decode. Nothing you read tells you whether that is happening. Nothing you hear tells you either, because when you listen back to your own recording, you hear the word you intended. Your ear is running the same French it just produced through the same first-language filter that produced it.

That is the gap this page is about: for a candidate preparing alone, there is nowhere to get your spoken French assessed. Comprehension you can mark against a key. Writing you can at least read back cold the next day. Speaking is the one ability where the thing being judged leaves no trace you can inspect.

What expression orale actually is

Start from the format, because a surprising amount follows from it. These are France Éducation international's own figures for TCF Canada — the body that writes the test and issues your attestation.

Tâche	What it is	Preparation	Duration
1	Entretien dirigé — a guided interview with the examiner	none	2 minutes
2	Exercice en interaction — you must obtain information in an everyday situation	2 minutes	5 min 30, including the preparation
3	Expression d'un point de vue — a question chosen by the examiner	none	4 min 30

The whole thing is 12 minutes, face to face with one examiner, of which 2 minutes are preparation. For context, the other three papers are compréhension orale (39 multiple-choice items, 35 minutes), compréhension écrite (39 items, 60 minutes) and expression écrite (3 tasks, 60 minutes), for a total sitting of 2 hours 47.

Three things fall out of that table, and they are the three things most preparation ignores.

Six and a half of the ten speaking minutes are unprepared. Tâche 1 and tâche 3 give you no preparation at all, and tâche 3 is the longest unbroken stretch of speaking on the paper. Whatever you have written out and learned is available to you for exactly one of the three tasks, and only for two minutes of it. Scripting is not a strategy here; it is a plan for the smallest part of the test.

Tâche 2 is built around asking, not answering. France Éducation international describes it as proving you can obtenir des informations in a common everyday situation, with the status of you and your interlocutor set out in the instruction — a task you cannot complete without asking. Almost all self-study is answer practice. This task is question practice, and French questions are where a lot of otherwise decent candidates fall back on the one form they are sure of — est-ce que — for three and a half minutes.

The tâche 3 question is chosen by the examiner. Not selected from a list you saw. You do not get to steer onto your prepared ground.

The measurement gap, stated plainly

Here is the mechanism, because it is worth understanding rather than just being told to record yourself.

Your first language gave you a set of sound categories. Everything you hear, in any language, gets sorted into the nearest one available. If French has a distinction your first language does not, your ear does not merely find it difficult — it does not reliably register it at all. The two sounds arrive in the same bucket.

The consequence is circular and it is the whole problem: the contrast you cannot hear is the contrast you cannot monitor. You will not catch it when you make the error, and you will not catch it on the playback either, because the playback goes through the same ear. This is why "record yourself and listen back" — the single most common piece of speaking advice in this market, including on all the pages that prompted this one — quietly fails on exactly the errors that matter most. It works well for pace, for hesitation, for whether you dried up at ninety seconds. It does not work for pronunciation, which is the thing it is usually recommended for.

An examiner does not have this problem. They have French categories. They hear what you actually produced — and so does the second corrector, who is sent the recording and has nothing else: no face, no gesture, no chance to ask you to say it again.

So the question is not whether you can find good sample answers. Good sample answers are freely available, and the best of the competing guides publish them. The question is what happens to one when it leaves your mouth, and answering that requires something outside your own head.

Four things that decide how hard the examiner has to work

If you are going to spend effort on pronunciation, spend it where French differs structurally from what you already speak. These four are the contrasts French makes that English and the major South Asian languages do not, which is why they are predictably the ones that make a candidate harder to follow than their actual French warrants. They are named specifically, with the phonetic notation, because "work on your pronunciation" is not an instruction anyone can act on.

1. The u / ou distinction. French /y/ — the vowel in tu, rue, sur, dessus — is a front rounded vowel. English does not have it. Neither does Hindi, Punjabi, Urdu, Gujarati, Bengali or Tamil. With no native category to put it in, most speakers hear it and produce it as /u/, the vowel in tout, roue, sourd, dessous. The pairs are real words and several of them reverse a meaning: tu /ty/ against tout /tu/, rue /ʁy/ against roue /ʁu/, au-dessus against au-dessous — above against below. On a test where tâche 2 involves getting practical information out of someone, saying the opposite of what you meant is not a cosmetic error.

2. Nasal vowels. French has three that matter: /ɑ̃/ (an, en), /ɔ̃/ (on) and /ɛ̃/ (in, ain, and for most speakers un). Banc /bɑ̃/, bon /bɔ̃/ and bain /bɛ̃/ are three different words.

There are two different traps here, depending on where you are coming from. For an English speaker the whole category is new, and the reflex is to produce a plain vowel followed by an audible [n]. For a Hindi, Punjabi or Bengali speaker, nasalised vowels already exist in your first language — which helps with the nasalisation itself, and not at all with the three distinct French vowel qualities, which tend to collapse toward whichever native vowel is nearest. The second trap catches everyone: the n and m are written but not pronounced. They tell you the vowel is nasal; they are not a consonant to say.

3. The French r. /ʁ/ is made at the back, at the uvula. English /ɹ/ is made at the front with the tongue tip; the /r/ and /ɾ/ of Hindi, Punjabi, Urdu and Tamil are a tap or trill at the alveolar ridge. All three are different places in the mouth. Substituting rarely destroys a word outright, but /ʁ/ turns up constantly — bonjour, travailler, pour, je voudrais, rendez-vous — so getting it wrong colours every sentence you produce rather than costing you one word.

4. Liaison, which means written French does not predict spoken French. Les amis is [le.za.mi]. Vous avez is [vu.za.ve]. Un petit enfant is [œ̃.pə.ti.tɑ̃.fɑ̃] — the t that is silent in petit reappears, attached to the front of the next word. Fluent French re-cuts the sentence at different places from the page.

This matters more than it looks, and it cuts both ways. If you learned most of your French by reading, you will tend to pronounce final consonants that should be silent and to leave gaps between words where a French speaker runs them together — producing something an examiner has to re-segment before they can understand it. And in compréhension orale the same fact runs in reverse: you are listening for word boundaries that are not there. If listening and speaking are both weaker than your reading, this is very often why, and it is one problem rather than two.

What is assessed, who assesses it, and what is not published

France Éducation international does publish what expression orale assesses. Their page lists the candidate's ability to:

• talk about themselves and their family and professional environment;
• ask questions appropriate to the communication situation proposed;
• give an opinion, explain the advantages and disadvantages of a plan, express agreement and disagreement;
• present a clear, structured argument in a style appropriate to the context;
• present complex subjects in a detailed, structured way, develop them, and conclude.

Note the second bullet again. Asking is on the official list, and it is the only speaking skill on it that self-study almost never rehearses.

What is also published is who marks you and against what. Expression orale is assessed twice, independently and double-blind: first by the examiner in the room, trained and accredited by France Éducation international, who conducts and records the interview; then by a second corrector, also FEI-trained, who is sent that recording. Neither sees the other's level. The criteria are three groups — linguistique (range and command of vocabulary, grammatical accuracy, ease, prononciation, overall fluidity of the discourse), pragmatique (interaction, structuring of the discourse, coherence and cohesion, thematic development) and sociolinguistique (fit to the communication situation). Correction of every TCF paper is centralised at France Éducation international.

What is not published is the arithmetic: how the items inside each group are weighted, and how two independent assessments become one mark out of 20. Nobody outside France Éducation international has that, so a band-by-band rubric with weightings is someone's reconstruction rather than the real thing. We will not assert one either.

What we will do is reason out loud from something that has to be true of any assessment of speech, and label it as reasoning rather than fact: an assessor can only credit what reaches them. An argument the examiner did not parse cannot be marked as a clear argument. A precise word they did not recognise cannot be marked as precise vocabulary. Pronunciation is a named criterion inside the linguistique group, but it is also the condition on which the pragmatic and sociolinguistic criteria can be observed at all — which is more weight than its position on that list suggests. This is why unclear delivery can cost a candidate marks their French would otherwise earn, and why the fix is not more vocabulary.

One mark is one level — and right now there is no appeal

Expression orale is reported out of 20, and IRCC's language test equivalency tables convert it like this:

NCLC	Expression orale, out of 20
10	16–20
9	14–15
8	12–13
7	10–11
6	7–9
5	6
4	4–5

Between 10 and 15 every band is two marks wide, so one mark is one NCLC level across the whole range most candidates are fighting over. Compare that with compréhension orale, where NCLC 7 spans 458 to 502 on a 100–699 scale — a 44-point band that absorbs a good day and a bad day alike. Expression is the narrow end of your profile no matter how strong your French is, and IRCC reads your lowest ability for eligibility, never the average — while CRS points are counted per ability, so one weak mark drags the profile twice. What that is worth is worked through in TCF Canada strategy; the short version is that NCLC 7 is the floor below which Federal Skilled Worker and Canadian Experience Class do not accept you at all, and the step from 6 to 7 is worth 8 CRS points per ability. One speaking mark can sit on either side of it.

There is a second reason this matters more in 2026 than it did, and it is the part no preparation site has caught up with. France Éducation international's own TCF Canada page currently carries this notice:

"À partir des sessions du TCF du 1er septembre 2026, les demandes de recorrection des épreuves d'expression orale et écrite ne seront plus acceptées pour une période d'environ 1 an."

From the sessions of 1 September 2026, requests to have expression orale and expression écrite re-marked are no longer accepted, for roughly a year. The same page says “La reprise du service est prévue à l’automne 2027” — resumption is planned for autumn 2027, which is an intention rather than a commitment. If your speaking mark comes back one below the line, the appeal route is closed. Your remedy is a resit: a second fee, a minimum of 20 days between sittings of TCF Canada, and results within 15 working days of France Éducation international receiving the session materials. Read those six weeks as the real cost of landing on 9 instead of 10, and check the notice yourself on the official page before you book, because it has an end date and these things move.

What we built for this, and what it does not do

We run TCF Canada mocks and we built our own scoring for them, so this section is written from the marking side. It also needs to be honest about a limit, so here is the limit first.

Our read-aloud practice is not a simulation of the speaking test. There is no read-aloud task in TCF Canada. We invented the exercise, and the page itself says so above the score. You read a model answer — written by us, for a real recalled TCF speaking topic on tâche 2 or tâche 3 — and what comes back is a Hilingo practice score that is not a TCF score and does not convert to one. It cannot assess you on the interactive part of expression orale, because nothing that is not a conversation with an examiner can.

What it does do is close the one loop you cannot close alone. The passage comes back with every word coloured good, average or bad. Click any coloured word and you get:

• the phonemes you should have produced and the phonemes you actually produced, side by side in IPA, the ones you produced coloured by how close each landed;
• Word changed — you said “…” when you substituted a different word, and Word not detected when you skipped it;
• a button to hear the word in French, and — where the recording allows it — a second button to hear that word back as you said it.

Around the words: pronunciation, fluency and content scores; delivery figures read off your pause count and your words per minute; and your own recording to play back beside all of it.

That is the point of the design. You are not being asked to hear the difference yourself — which we have just spent half this page explaining you cannot do. You are being shown it in notation. /y/ expected, /u/ produced is a fact you can act on this evening, and it does not depend on your ear having a category it was never given.

Two things that are honestly true about the transfer. It is narrower than expression orale: it assesses pronunciation and delivery on a text you did not compose, so it says nothing about whether you can build an argument under time pressure or ask a well-formed question. And it is the component of expression orale you are least able to diagnose without it — the sounds are the same sounds whether you are reading them or inventing them, and they are what the examiner has to work with either way. Fix them on a read passage and they arrive fixed in the interview.

Two scored attempts a day, free, no card. The free tier also includes one full scored mock with the complete report. On volume, honestly counted: 39 full TCF Canada mocks and 156 section tests — a section test being one section of a full paper offered on its own, so a shorter sitting rather than an extra mock. We count them separately because calling the total 195 full mocks would not be true.

A four-week use of this, if you want one

1. Sit one full mock first. Find out whether expression orale is actually your floor before you spend a month on it. It usually is; occasionally it is not, and a month on the wrong ability is a month.
2. Read-aloud daily, on the two free attempts. Do not chase the overall score. Open the word-by-word pass and write down the phonemes that came back wrong, not the words. You will find the same three or four symbols recurring, and that list is your actual syllabus.
3. Drill those phonemes in minimal pairs, not in sentences — tu/tout, rue/roue, dessus/dessous, banc/bon/bain — then read a passage containing them and check whether the notation changed.
4. Separately, practise asking. Write ten questions for a tâche 2 situation without using est-ce que once. Inversion and indirect forms are the range this task is looking for, and they are the thing three and a half minutes of it will expose.
5. Re-sit a mock in week four and compare the expression orale mark, not the feeling.

For expression écrite, one small thing that belongs to the same family of avoidable loss: accents are spelling, and spelling is marked. Our free French keyboard is public and needs no account.

Measure the ability that decides your level

Everything above turns on a number you do not have yet: your expression orale mark out of 20, under exam conditions, on the scale IRCC reads. A CEFR self-assessment will not give it to you, and neither will a sample answer you found convincing.

Hilingo gives you one full scored TCF Canada mock free, with no card — all four parts under the real section timers, compréhension on 100–699 with its CEFR level and expression out of 20 with the NCLC band beside it, so you can see which ability is your floor before a test centre tells you.

Take a free scored mock

We should be straight about one more thing: our scoring is a coaching tool, not an examiner. Use it to find which sound or which criterion is costing you marks and whether you are trending toward your target. The attestation only comes from a France Éducation international centre.

• TCF Canada strategy — what each NCLC level is actually worth in CRS points
• TCF Canada score chart — both scales converted to NCLC, in full, with IRCC's source page
• TCF Canada reading — 39 items, 60 minutes, and why no error budget exists
• TCF Canada listening — the other half of the liaison problem
• Free TCF Canada mock test — what the free sitting includes and what refreshes daily
• Start a TCF Canada mock — the exam page, with the sample report layout

Frequently asked questions

How long is the TCF Canada speaking test?

Twelve minutes, face to face with one examiner, and that total includes the two minutes of preparation. It is three tasks: a guided interview of 2 minutes with no preparation, an interaction exercise of 5 minutes 30 that contains the only preparation time on the paper, and a point-of-view task of 4 minutes 30 with no preparation. Confirm the current format on France Éducation international's TCF Canada page before you book.

What are the three TCF Canada speaking tasks?

Entretien dirigé, a guided interview in which you show you can interact with someone you do not know. Exercice en interaction, in which you have to obtain information in an everyday situation — your role and your interlocutor's are given in the instruction, and the task cannot be completed without asking. And expression d'un point de vue, where the examiner chooses the question and you argue a position spontaneously and continuously.

What TCF Canada speaking score do I need for NCLC 7?

10 out of 20. NCLC 7 is 10–11, NCLC 8 is 12–13, NCLC 9 is 14–15 and NCLC 10 is 16–20, per IRCC's equivalency tables. Because IRCC reads your lowest ability rather than an average, a 9 in speaking makes you an NCLC 6 candidate however strong your other three are. Check the current tables on IRCC before planning around them.

Can I ask for my TCF Canada speaking mark to be re-evaluated?

Not at the moment. France Éducation international's TCF Canada page states that from the sessions of 1 September 2026, re-correction requests for expression orale and expression écrite are no longer accepted for about a year, with the service expected to resume in autumn 2027. Until then the only route past a disappointing mark is a resit — a further fee, at least 20 days between sittings, and results within 15 working days of the session materials reaching France Éducation international. Verify the notice on the official page, since it carries an end date.

How can I practise TCF Canada speaking on my own?

Structure, argument shape and question forms you can genuinely rehearse alone — write questions for tâche 2 situations, build two-argument answers for tâche 3, talk for four and a half minutes without stopping. Pronunciation you cannot, because the sounds you have trouble producing are usually the sounds your first language did not teach you to hear, so you will miss them on the playback as well. That part needs an outside measurement, whether that is a French teacher listening to you or a tool that shows you the phonemes.

How do I improve my French pronunciation for TCF Canada?

Work on specific contrasts rather than "pronunciation" in general. For most English and South Asian first-language speakers the highest-value four are /y/ against /u/ (tu against tout), the three nasal vowels /ɑ̃ ɔ̃ ɛ̃/ with the written n and m left unsounded, the uvular /ʁ/, and liaison — which is why fluent French sounds nothing like the written form. Drill them as minimal pairs, then check against something that reports what you actually produced rather than what you meant to.