Kit

TCF Canada listening and reading kit

Your kit for the two multiple-choice papers of TCF Canada: listening (compréhension orale, CO) and reading (compréhension écrite, CE). The target is NCLC 7 (Canada's language level 7): 458 in listening and 453 in reading, out of 699.

What it is built from. exam-analysis.md sections 2.1, 2.2 and 3, and three research reports: tcf-co-ce-items.md, gap-2.md and tcf-format-scoring.md. Word ranks come from gap-1.md.

How to read it.

• FEI (France Éducation international) writes and runs the TCF.
• Confidence tags, as in the analysis: [H] an official document, or a count made directly from one; [M] counted, but resting on one coder's rules or a mapping assumption, or several unofficial sources that agree; [L] unofficial, a single source, or my own calculation.
• A word marked #124 is rank 124 in the exam's own frequency list (vocab/master-list-ranked.tsv), which your Anki decks (vocab/) are built from. Numbers are given for the main words only, so a word without one may still be in your decks: search Anki before you add it.
• Words this kit uses:
  • an item is one question; a band is a block of items at one level (for example Q20–29, B2);
  • the options are the 4 answers A–D; the key is the right one; a wrong option (prep sites say "distractor") is built to tempt you;
  • a paraphrase says the same thing in other words; an inference is an answer that no single sentence states, so you combine sentences to get it;
  • a trap code (E, S, R, U, P, C, O, W) names the way a wrong option misleads you (section 3.3).
• Example items in this kit are study examples written to show a pattern. They are not exam items, and nothing here is a script to memorise. FEI treats reproducing memorised texts as fraud, which matters for writing and speaking; in these two papers the matching habit is to learn patterns, not answer keys.

The five rules that matter most

1. Never leave a blank. A wrong answer costs nothing.
2. Read the options before the audio, and the question before the text.
3. Listen and read to the end: plans change late.
4. An option that copies the text's words is often a trap; the right answer usually says the same thing in other, more general words.
5. Questions 20–29 decide your score: get 9 of 10 there.

ID index

Suggested order (the calendar sets the days)

This follows the ranked list in exam-analysis.md section 3: numbers and question words first, everyday documents next, then the Q20–29 band, which becomes the largest daily block from week 3.

1. The score maths (unofficial grid)

1.1 What FEI says, and the grid prep sites use

What is official [H] (FEI candidate manual, Version J, pp.18–23):

• Each paper has 39 questions with 4 options and one right answer. Questions get harder from 1 to 39. Levels run from A1 (beginner) to C2 (mastery).
• Listening lasts 35 minutes and each recording is played once. Reading lasts 60 minutes, and you manage the clock.
• Only right answers score. A wrong answer and a blank both score 0.
• FEI weights each right answer by the question's difficulty and converts the total into a score from 100 to 699. It does not publish the weights and does not tell you how many answers you got right.

What prep sites assume: the grid below. It appears on 8 sites, with a variant on 1 more, and none of them cites an FEI document [M that it is the common standard; L that it is exact]. Use it as a planning tool, not as the real formula.

The variant (HiTCF's listening page) gives 21 points to Q20–30 and 27 to Q31–35. It also adds up to 699.

Your score under the grid = 3 × (right in Q1–4) + 9 × (Q5–10) + 15 × (Q11–19) + 21 × (Q20–29) + 26 × (Q30–35) + 33 × (Q36–39).

What the cut-offs mean [H]: NCLC 7 is 458–502 in listening and 453–498 in reading; NCLC 8 is 503–522 and 499–523; NCLC 9 is 523–548 and 524–548. B2 starts at 400, so "B2" on a certificate is not enough by itself: one candidate's listening score of 448 was B2 but NCLC 6.

1.2 Your targets

• 31 of 39 in each paper. Under the grid, any 31 right answers score at least 463 (459 under the variant), which clears both 458 and 453 whichever 8 you miss. With 30 right you can score anywhere from 437 to 642, depending on which you miss. [L]
• Band floors:
  • Q1–19: at most 1 miss.
  • Q20–29: at least 9 of 10.
  • Q30–39: at least 4 of 10 (3 only if Q1–19 are all right), mostly by elimination (LR-P5).
• Either way the floors add up to 31: 18 + 9 + 4, or 19 + 9 + 3 with a clean Q1–19. The table in section 1.3 shows why: 1 miss in Q11–19 with 9 of 10 on Q20–29 needs 4 of the last 10 for listening.
• Two full timed mocks in a row at 31 or more (LR-MOCK).

Why 9 of 10 on Q20–29, not 8 [L, grid arithmetic re-done for this kit]:

• 1 miss on Q1–19, 8 of 10 on Q20–29 and 3 of 10 on Q30–39 gives 432–465, usually 439–451: below 458, and usually below 453.
• The same with 9 of 10 on Q20–29 gives 453–486, usually 460–472.
• With 1 miss in Q11–19 and only 8 of 10 on Q20–29, you need 4 of the last 10.
• Each Q20–29 item is worth 21 points, and the band holds 210 points, the most of any band.
• In reading, 30 can be enough if the misses are cheap: 4/6/9/8/2/1 right (Q1–4 / Q5–10 / Q11–19 / Q20–29 / Q30–35 / Q36–39) scores 454, enough for reading but not for listening.

1.3 How many Q30–39 answers you need

Read across from your Q1–19 result to your Q20–29 result. Each cell gives the right answers you need in Q30–39: listening (458) / reading (453). The table assumes the worst case: a miss in Q1–10 costs 9 points (a Q5–10 item), and your right answers in Q30–39 are the cheaper C1 items (26 points each). [L, my calculation under the grid]

1.4 Why elimination pays

Expected points from Q30–39 under the grid [L]:

• With all of Q1–29 right (411 points), blind guessing on the last 10 reaches 458 about 76% of the time.
• With all of Q1–19 right and 9 of 10 on Q20–29 (390 points), you need 3 right in the last 10 to be sure of 458. Blind guessing gets there about 47% of the time, removing one option per item about 70%, and removing two about 95%. [L: my calculation, treating each guess as independent]

1.5 Pacing for reading (60 minutes)

Sources: the 15 / 25 / 18 / 2 split is suggested in tcf-format-scoring.md §3.4 (not official); 45 seconds for Q1–10 is LanguageNext's target; the split of the first 15 minutes into 7 min 30 + 7 min 30 is mine [L]. HiTCF suggests 8 / 14 / 18 / 18 + 2 [L].

• Drill faster than the exam. Q20–29 drills run at 90 seconds an item (LR-P4), so the exam's 2 min 30 leaves a margin.
• If you fall behind, protect Q20–29 (21 points each), but reach Q30 by 45:00 at the latest and run LR-P5 at about 1 minute an item. [L]
• Moving around: in reading you can move back and forth between items [H], so skip a hard one, note its number, and return to it if time allows. On paper you may correct an answer only once, in the round box [H]. Fill in the answer sheet as you go, not at the end [L].

1.6 Pacing for listening (35 minutes, set by the machine)

• You cannot control the clock, and each recording plays once [H].
• The gap between items is short. Prep sites report about 10–15 seconds between items, or about 10–12 seconds to read the options and 10–12 to answer [M/L]. Spend it on step 1 of LR-P1 (reading the next options), never on the item you just missed.
• In FEI's samples the question is spoken after the recording. The manual also says that on computer "questions and answers appear at the same time on screen", but the samples and TV5MONDE's FEI-written simulator put the question after the audio. Train for hearing the question once [M].
• FEI advises against taking notes in the multiple-choice papers. The research's compromise: digits only, if anything [H/M].
• Nobody is admitted once listening has started [H]. Be early.

1.7 Never leave a blank

• Wrong and blank both score 0, so a guess can only help [H].
• Listening: if you are lost, pick a letter at once and turn to the next options.
• Reading: at 58:00, fill every blank with one letter you chose before the test. It takes 30 seconds.
• Answer every item in drills too, so the habit is automatic. A blank in a mock goes in the error log as code X (LR-P6).

1.8 Score sheet

Copy one row per mock or timed set. Points per right answer: 3, 9, 15, 21, 26, 33.

2. The bands

2.0 Before you start

• Bands are approximate. Sources put the picture items at Q1–4 or Q1–7, and the B1/B2 border at Q19/20, Q20/21 or Q23/24 [M]. Two TV5MONDE listening items come back word for word one position later in a later series, and the circulating paper's recording at position 25 announces itself as question 24 [H for what was observed]. Learn the item types, not the positions.
• Where the counts come from (gap-2.md, one coder with written rules) [M]:
  • 136 reading items and 136 listening option sets from TV5MONDE's 17 FEI-written series. "TV5 positions" are places in a 15-item series; section 5.1 says which positions stand in for which band.
  • The 7 official listening and 4 official reading examples, plus 2 listening examples from FEI's Québec manual.
  • Q20–29 of one circulating paper (InfinLume Paper 01, IDs INF-L01 and INF-R01), provenance unknown [L].
• Topics in FEI's sample items (all papers, writing included): food 3, holidays and travel 3, shopping 2, work 2, administration 2, family and childcare 2 [H]. An earlier, unverified coding of 65 prep and official items put work (about 10), administration (about 9), shopping (about 6), transport and appointments (about 5 each) on top.
• Wrong-option codes (E, S, R, U, P, C, O, W) are defined in LR-P3 (section 3.3).
• Listening item families [H]: (1) a picture and 4 spoken sentences; (2) a spoken question and 4 spoken replies, nothing printed; (3) a short dialogue, a spoken question and 4 printed options; (4) a longer monologue, a spoken question and 4 printed options. FEI's 7 examples: 2, 2, 2 and 1, in that order.

2.1 Listening Q1–10 (A1–A2): 66 points, 9%

What you hear

• About Q1–4: a picture and 4 spoken sentences. Pick the sentence that fits the picture. Official Ex. 1 (a family scene) and Ex. 2 (a cinema). Nothing is printed.
• About Q5–10: a spoken question and 4 spoken replies. Pick the reply that fits. Official Ex. 3 and 4. Nothing is printed.
• Some sites run the pictures to Q7; one puts very short announcements with printed options at Q8–10 [M].
• TV5MONDE squeezes this block into positions 1–4. Across its 17 series, those positions hold 20 picture items, 41 question–reply items and 7 short documents [H, counted].

What is asked: which sentence someone in the scene would say; which reply fits both the question word and the situation.

What the counts show

• In both official question–reply examples, every wrong reply answers a different question word [H]:
  • "What do you have for breakfast?": the wrong replies give a place, a companion and a time.
  • "How much is that red dress?": the wrong replies give a size, "no, this one" and "yes, it's red". The right reply says it is not expensive: it answers the price without a number.
• The picture examples test the things people say in a scene, often imperatives (venez, allez, regardez, asseyez-vous, attendez) and routine verbs: eat, sleep, sit, pay, buy, wait, call, watch, finish, come, go [H for the examples; M for the word list].
• 9 of 10 sources that describe listening by band put pictures and everyday exchanges first [M]. Topics: home, meals, leisure, cinema, shops, the street; invitations, prices, greetings, directions, weather, times.

Exam protocol

1. Picture item: before the sound, look at who, where, how many, doing what, at what time of day. Guess the line someone would say.
2. Question–reply item: catch the question word in the first two or three words (LR-T1) and hold on to it.
3. For each reply ask: "does this answer that question word?" Drop replies that repeat words from the question but answer something else (code W).
4. Accept an indirect reply: Elle n'est pas chère (said of a dress) answers a price question.
5. Answer at once. Never leave it blank.

Québec note: in Québec, bienvenue can be the reply to merci (France: de rien, je vous en prie) [M].

Drills: VL-01, VL-02, LR-T1; TV5 positions 1–4 with LR-P1.

2.2 Listening Q11–19 (B1): 135 points, 19%

What you hear: a short dialogue, voicemail or announcement, then a spoken question, then 4 printed options (family 3; official Ex. 5 and 6). Service dialogues (shop, payment, booking), workplace scheduling, phone messages, and station, airport or shop announcements. 9 of 10 sources describe the middle band (about Q11–29) as dialogues, voicemails and announcements rising to interviews and radio reports; 5 name voicemails and 10 name public announcements [M]. TV5 clips at the matching positions (5–7) last about 55 seconds (44–70) [M].

What is asked: a detail (time, place, number, means), what someone must do, why someone calls, what changes, the final decision. Typical questions: Que doit faire Marc ? Comment le client va-t-il payer ? Pourquoi Julie appelle-t-elle ? Qu'est-ce qui change ?

What the counts show

• Official Ex. 5, a checkout: cheques are refused, the customer has no card and too little cash, then the cashier takes a cheque as an exception. The answer is "by cheque". All 3 wrong options are the alternatives offered under the old rule (code P) [H].
• Official Ex. 6: a boss asks an assistant to set up a meeting "any time except Thursday afternoon", and another meeting is not cancelled. The answer is a paraphrase ("arrange a professional appointment"); the option "cancel a meeting" is contradicted (code R) [H].
• In TV5 positions 5–7 (51 option sets), the options are aims written as infinitives in 29%, names of things or people in 25%, a person's actions in 24% and statements in 20%; 1 set is attitudes [M]. Options that signal a change: 4 of the 51 [M].
• Prep sources: number or time swaps are described by 7 sources (near-miss times such as 13 h 30, 15 h 30, 16 h 30; the planned time against the real time), and changes or corrections by 6 [M].

Exam protocol

1. Read the 4 options. Name the answer type: time, place, person, action or reason. Infinitives mean "what to do" or "what for" (predictor table in LR-P1).
2. Listen to the very end for the final decision: finalement, en fait, plutôt, au lieu de, à la place de, sauf, désolé, ce n'est plus possible, reporté, annulé (LR-T5).
3. For Que doit faire… ? listen for the request: pourriez-vous, il faut, n'oubliez pas, merci de.
4. Voicemails follow one frame: who is calling, why, what you must do, how to call back.
5. Two times or prices in one clip: the one still standing at the end, after any correction, is usually the answer. The first one is a ready-made trap (P or S).

Québec note: in Québec dîner is lunch and souper is the evening meal, and fin de semaine is the weekend. A meal name can flip a time question [M].

Drills: LR-P1 on TV5 positions 5–7 and on Q11–19 of the INF drill papers; VL-03; LR-T2 and LR-T5.

2.3 Listening Q20–29 (B2): 210 points, 30%, the deciding band

What you hear: mostly interviews and everyday dialogues, some radio reports.

• Circulating Paper 01, Q20–29: 5 interviews, 4 everyday dialogues, 1 radio report. Several (a bookshop, the weather, a travel agency) sound like B1 service dialogues, which fits Q19–20 being a seam [L/M].
• TV5 clips at positions 8–11 average 61 seconds (43–78), and about 65 at positions 10–12 [M].

What is asked: mostly purpose and "what do we learn". Opinion is rare.

• Circulating Q20–29: purpose 4, detail 3, inference 2 (no single sentence says it: you combine sentences), opinion 1 [L/M].
• TV5 positions 8–11 (68 option sets): statements about a topic 34%, names of things, people or topics 28%, aims as infinitives 28%, a person's actions 9%, attitudes 1 of 68 [M].
• Typical questions: Que propose le vendeur ? Quel est l'objectif de… ? Pourquoi… ? Qu'apprend-on sur… ? De quoi parle-t-on ?

What the counts show

• Wrong options in circulating Q20–29 (30 of them): echo a word that was heard 9, swap a real element (time, place, person) 9, contradicted 6, never said 5, the old plan 1 [L/M].
• Official Ex. 7 (a presentation, C1 level): all 3 wrong options re-use words from the recording ("research", "experts", "economic") [H]. Official Québec Ex. 4: a wrong option re-uses the speaker's own verb (délaissent, "abandon") but attaches it to package holidays, whose sales actually rose [H].
• Change of plan: circulating Q20 (a book is out of stock, so the seller offers to keep a copy; one wrong option swaps "in two weeks" for "tomorrow"); 3 of the 68 TV5 option sets signal a change [M].
• Speaker mix-ups (one person's view given to the other) are described by 6 prep sources [M].
• Québec: 1 of the 10 circulating items is set in Québec (ice fishing), and so are 2 of the 5 official listening samples at B1 or above [H/L].
• All 3 prep providers that name a TCF trap name listening [H for the count].

Exam protocol

1. Read the options and predict the question from their form: infinitives → purpose; names → identity or topic; statements → "what do we learn"; a person + verb → what someone did or plans (LR-P1).
2. Catch the topic in the first sentence.
3. Listen for intent markers (je peux vous…, je vous propose, l'objectif est de, afin de, pour; LR-T4) and for the pivot (mais, pourtant, cependant; LR-T6). Keep track of who says what.
4. When the question comes, run LR-P3 fast: echo? swap? contradicted? never said? old plan? the conceded half?
5. If two options survive, take the one that says it in different, more general words.

Drills: LR-P1 on TV5 positions 8–12 and Q20–29 of the INF drill papers, 10 clips a day of 60–70 seconds; LR-P2 on RFI-01 and RC-01; VL-04. Target 9 of 10. From week 3 this is the largest single block of daily time (exam-analysis.md §3, item 15).

2.4 Listening Q30–39 (C1–C2): 288 points, 41%

What you hear: presentations of organisations or projects (official Ex. 7), expert analysis, radio debates and round tables, then lectures and talks on abstract subjects. One site says 45–120 seconds. 8 of 10 sources describe this end band as debates, lectures, academic talks and RFI-style radio [M].

What is asked: the role or purpose of something, the main argument, what is implied; then the thesis, the hidden meaning, the speaker's real position. Quel est le rôle de… ? Quel argument… ? Que sous-entend… ? Quelle est la thèse… ?

What the counts show: options that copy the audio word for word, or use absolutes (toujours, jamais, tous, totalement, unanime), are usually wrong [M, prep sources]. Elimination is worth 24 to 72 expected points over blind guessing (section 1.4). This band gives the lowest return per study hour in an 8-week course, so use technique.

Exam protocol (LR-P5): read the options during the pause; lock in the opening sentence (the theme); wait for the pivot (cependant, toutefois, or, en réalité); strike copies and absolutes; take the nuanced paraphrase; answer before the next clip starts; reset.

Drills: LR-P5 on TV5 positions 13–15 and Q30–39 of the INF drill papers.

2.5 Reading Q1–10 (A1–A2): 66 points, 9%

What you read: about Q1–4, very short notes, signs, labels, postcards and text messages; about Q5–10, small ads, flyers, menus, timetables, opening-hours and closure notices, forms and invitations. 11 of 11 sources that describe the bands put these documents first [M].

What is asked: what someone is doing, where someone is, what a sign means; a time, date, price, deadline or document to bring; what is compulsory; where you would read the text. Quand… ? Jusqu'à quand… ? Que faut-il apporter ? Où peut-on lire ce texte ?

What the counts show

• Official Ex. 11: a legal advice service's timetable. "When can you see a lawyer without an appointment?" The answer is "in the afternoon": you link the condition (sans rendez-vous) to the right time slot [H].
• Numbers matter here, and almost only here: no key among 136 FEI-written reading items at TV5 positions 5–12 depended on reading or comparing a number [M].
• Three traps: a start date against an end date; the start of a closure against the reopening time; optional against compulsory [M].

Exam protocol: question first; underline its keyword (when? what to bring? without an appointment?); scan the document for the matching line; check the condition words (sauf, à partir de, jusqu'au, sur / sans rendez-vous, obligatoire, conseillé); answer in under 45 seconds.

Drills: LR-P4 on TV5 positions 1–4 and Q1–10 of the INF drill papers; LR-T2.

2.6 Reading Q11–19 (B1): 135 points, 19%

What you read: personal letters, official, work or customer emails, short news items, blogs, forum posts (10 of 11 sources) [M]. In TV5 positions 5–7 (51 items): portraits of a person 33%, informative articles 29%, notices 16%, cultural presentations 14%, first-person accounts 8% [M].

What is asked: in TV5 positions 5–7, what someone wants or plans to do 22%, "what do we learn" 20%, purpose 16% [M]. The official examples ask "What must she do?" (a message about a visa file: post a document) and "What do the two friends propose?" (to come on holiday with them) [H].

What the counts show

• The answer is one detail, said in other words, in 57% of the TV5 items at positions 5–7; inference 20%, purpose 12%, change of plan 8% [M].
• Wrong options: echo 33%, not mentioned 29%, swap 22%, contradicted 7%. 69% of items have at least one echo [M].
• In official Ex. 13, all three wrong options recombine the letter's own details (July, August, Rome, renting a house): code S [H].

Exam protocol: first find who writes, to whom, and why (LR-T3). Find the one sentence that answers. Then check who and when: most traps here attach a real detail to the wrong person or time. Plan verbs: vouloir, compter, avoir l'intention de, prévoir (LR-T4). About 50 seconds an item.

Drills: LR-P4 on TV5 positions 5–7 and Q11–19 of the INF drill papers; LR-T3 and LR-T4.

2.7 Reading Q20–29 (B2): 210 points, 30%, the deciding band

What you read: mostly short informative press articles.

• TV5 positions 8–11 (68 items): informative article 38 (56%), cultural presentation or review 14 (21%), portrait of a person 9 (13%), notice or offer 4, first-person account 2, message 1. Every text was a single passage of one to three paragraphs [M].
• Circulating Paper 01, Q20–29: 5 informative texts, 3 opinion texts, 1 cultural text, 1 literary extract, with inference in 6 of 10 [L]. TV5MONDE builds its texts from its own short media pieces and has no opinion text at positions 8–11, so it probably under-represents argued texts. Train on both.

What is asked: "what do we learn / what is it about" 31%, purpose 19%, "what can you find or do" 12%, "what is special" 10%. Answers: inference 38%, detail 37%, purpose 15%, change of plan 7%, opinion 3% [M].

What the counts show

• Wrong options (204): echo a word, name or picture from the document but say something it does not say 90 (44%); swap real elements 42 (21%); plausible but not mentioned 39 (19%); contradicted 24 (12%). 54 of the 68 items (79%) have at least one echo [M].
• The key is usually a paraphrase that uses a more general word: vêtements (clothes) becomes l'habillement (the clothing sector); changé d'adresse becomes a déménagé (moved) [M].
• Difficulty rises through inference, not length: from positions 5–7 to positions 8–11, detail answers fall from 57% to 37% and inference rises from 20% to 38% [M].
• A change of plan decides 5 of the 68, and a concession pivot matters in 5. Opinion questions: 2 of 68 [M].

Exam protocol: read the question first and name its family (question-frame table in LR-P4). Find the sentence or sentences. Run LR-P3 on every option. Prefer the general-word paraphrase. Budget 2 min 30 an item in the exam and 90 seconds in drills.

Drills: LR-P3 with LR-P4 on TV5 positions 8–11 and Q20–29 of the INF drill papers (these add the argued texts); 10 items a day at 90 seconds each; target 9 of 10; the LR-T7 notebook.

2.8 Reading Q30–39 (C1–C2): 288 points, 41%

What you read: opinion pieces, reports, essays, specialised or scientific articles; then literary extracts, philosophical or academic essays, legal or economic texts, up to about 500 words. 11 of 11 sources agree [M].

What is asked: the author's main criticism, thesis or final position, what a sentence implies, a study's conclusion, the meaning of a word in context, what a pronoun refers to, irony.

What the counts show: opinion questions start to appear around Q30: 3 of the 17 TV5 items at position 12, where contradicted wrong options also rise (14 of 51) [M]. Certes… mais sets up a rebuttal, or marks a turn in the argument, and the last paragraph usually holds the author's view [M].

Exam protocol (LR-P5): about 18 minutes for the ten. Question first; find the paragraph; strike word-for-word and extreme options; take the nuanced paraphrase; fill every blank.

Drills: LR-P5 on TV5 positions 12–15 and Q30–39 of the INF drill papers.

2.9 Exam-day routine

Listening

1. Arrive early: nobody is admitted once listening starts.
2. Every item: options first → predict → listen to the end → answer within seconds → let it go.
3. Q1–10: catch the question word. Q11–19: catch the final decision. Q20–29: topic, purpose, pivot, option check. Q30–39: eliminate and guess.
4. Never leave a blank.

Reading

1. Clock checks: Q20 by 15:00, Q30 by 40:00 (45:00 at the latest), every blank filled at 58:00.
2. Every item: question first → options → find the sentence → option check → answer.
3. On paper, fill in the answer sheet as you go; you may correct an answer once, in the round box.
4. Never leave a blank.

3. Drill protocols

3.1 LR-P1 Single-play listening

Use it for every listening item, in drills and in mocks. Rules: one play; no pause; no notes except digits; answer before you review.

1. Read the options (about 10 seconds). If nothing is printed, look at the picture, or get ready to catch the question word.
2. Predict the question from the options' form (table below). Name it to yourself in one word: purpose? who? what happened? what do we learn? when?
3. Play once and listen to the end.
4. Answer within about 10 seconds of the question. If unsure, strike what you can (LR-P3) and guess. Never leave a blank.
5. Mark your confidence next to the answer: ✓ sure, ~ half-sure, ? guess.
6. Review, now as many times as you like, with the transcript if there is one. Find the sentence that holds the answer. Name the trap behind each wrong option (codes in LR-P3).
7. Log every miss and every lucky "?" in LR-P6. Take 1 or 2 clips a session into LR-P2.

Predict the question from the options' form (shares from 51 TV5 option sets at positions 5–7 and 68 at positions 8–11) [M]:

Daily volumes from the research [M/L]: 20 audio-only exchanges a day for Q1–10; 10 single-play clips a day for Q11–19; 10 clips of 60–70 seconds a day for Q20–29.

Notes by source

• TV5MONDE's training mode lets you replay. Don't, until step 6. It has no transcripts: in review, replay up to three times and write down the sentence that holds the answer.
• InfinLume's review mode lets you replay and shows a transcript after you answer. Its transcripts look machine-made, so distrust odd words (one says bédières for rivières).
• Voice-mode drills (section 6): keep the screen out of sight until you have answered.

3.2 LR-P2 Transcript and shadowing

Use it for 1 or 2 clips a session after LR-P1, mostly RFI-01, the InfinLume review transcripts, your VL drills, and the official transcripts in FEI-02 and FEI-03. Allow 10–15 minutes per minute of audio [L]. Goal: to turn "I knew the word but didn't catch it" into instant recognition, and to get used to the speed.

1. Listen again without the transcript. Note where you got lost.
2. Listen while reading the transcript. Underline everything you did not catch.
3. Sort each underlined bit:
   • V, you did not know the word: add it to Anki (check its # rank first);
   • H, you knew it but did not hear it: name the sound rule (list below).
4. Shadow in three passes: (a) pause after each sentence and say it aloud, copying the rhythm and the linking; (b) speak along with the audio about half a second behind it, reading the transcript; (c) the same without the transcript.
5. Final listen, no transcript, no pause. You should now catch nearly everything.
6. Harvest: add 1–3 items to the paraphrase notebook (LR-T7) or the marker tables (LR-T4 to LR-T6).

Sound rules behind most H misses

• Dropped sounds in speech: je suis → "j'suis", il y a → "y a", tu as → "t'as". The ne of a negative usually disappears (je sais pas), so listen for pas, plus, jamais, rien.
• Liaison: ils ont (with a z sound, "they have") against ils sont (with an s sound, "they are").
• Number pairs [L, my list]: deux / douze, treize / trente, quatorze / quarante, quinze / cinquante, seize / soixante, soixante-dix / quatre-vingt-dix, cent / cinq cents.
• Québec [M]: t and d before i or u hiss (tu sounds like "tsu", dire like "dzire"); je suis → "chu", il → "y", elle → "a"; a question can end in -tu (Tu viens-tu ?).

3.3 LR-P3 The option check, then the general-word rule

Use it for every reading item from Q11 up, every listening review, and in your head during listening Q20–29.

The wrong-option codes

Steps

1. Read the question and know what it asks (question-frame table in LR-P4).
2. Test each option in this order, and drop it at the first "yes":
   • Echo: does it re-use a word from the text? Then check the relation: who does what, when, to whom. A changed relation means it is out.
   • Swap: are the pieces real but wrongly paired?
   • Contradicted: does the text say the opposite?
   • Not mentioned: can you point to the sentence that says it? If not, it is out.
   • Old plan: is it the state before finalement, désormais, ne… plus, plutôt que?
   • Concession: is it the conceded half before mais or pourtant, or a true detail that does not answer this question?
   • At Q30–39, also strike absolutes (O).
3. One option left: answer it.
4. Two left: apply the general-word rule. The key usually restates the text in different, more general words (vêtements → l'habillement; changé d'adresse → a déménagé). Distrust the option that copies the text's words. This is a tie-breaker, not a law: keys sometimes re-use a text word.
5. None left: you struck the key. Re-read the question word; the key is usually the paraphrase you struck as "not mentioned".
6. Drill version: after answering, write a code beside every wrong option. Reading drill: 10 items a day at 90 seconds each; target 9 of 10.

Study example (written for this kit, not an exam item; a real item has 3 wrong options, this one shows every code)

Depuis septembre, la bibliothèque municipale de Rouen ne prête plus seulement des livres : les habitants peuvent aussi y emprunter des outils, comme des perceuses ou des échelles. Au départ, l'association de quartier qui a eu l'idée voulait ouvrir son propre local. Finalement, la mairie a préféré installer le service à la bibliothèque, qui est ouverte six jours sur sept. Le prêt est gratuit pour les habitants inscrits, mais les autres paient deux euros par objet.

Question: Qu'est-ce qui a changé à la bibliothèque ? (What has changed at the library?)

3.4 LR-P4 Timed reading

Use it for reading band drills and full papers.

Steps for each item

1. Question first. Name its family (table below).
2. Read the 4 options. What makes them different from each other? That difference is what you hunt for.
3. Find the place in the text. Notice: scan for the keyword. Letter: who writes, to whom, why. Article: the first and last sentence of each paragraph, then the sentence that matches; mark the claim, the evidence and any concession.
4. Run LR-P3 on each option.
5. Answer, mark ✓ ~ ?, and move on when the budget runs out (guess and flag it).
6. After the set: check, then log every miss in LR-P6.

For full papers, use the clock checks in section 1.5 and LR-MOCK.

Question frames and what to hunt for (counts: the 68 TV5 items at positions 8–11, from gap-2's codebook) [M]

3.5 LR-P5 Elimination on Q30–39

Use it for every Q30–39 item, in both papers. The aim is not to understand everything but to raise your odds from 1 in 4 to 1 in 3 or 1 in 2 (section 1.4).

Reading (about 1 min 45 an item)

1. Read the question and underline its key noun.
2. Find the paragraph. The theme is in the first lines; the author's view is usually in the last paragraph; certes… mais sets up a rebuttal; or marks the turn.
3. Strike: (a) options that copy a sentence of the text word for word; (b) absolutes (toujours, jamais, tous, totalement, unanime, uniquement); (c) options about something the text does not discuss; (d) options that contradict the last paragraph.
4. Of what is left, take the most nuanced paraphrase.
5. Time up: choose among what is left. Never leave a blank.

Listening

1. Read the options during the gap.
2. Lock in the opening sentence: the theme.
3. Wait for the pivot: cependant, toutefois, or, en réalité.
4. As you listen, strike copies and absolutes.
5. Take the nuanced paraphrase, answer before the next clip starts, and reset.

3.6 LR-P6 Error log

Log every miss, plus every right answer you marked "?" (a lucky guess hides a gap). Keep it in a notebook or a spreadsheet.

Codes: the trap codes E, S, R, U, P, C, O, W (section 3.3), or a cause code when no trap fooled you:

• V vocabulary: you did not know a key word.
• H hearing: you knew the word but did not catch it.
• Q question: you misheard or misread the question.
• X time: you ran out of time or left a blank.

Study examples (invented rows, to show the format)

Weekly review (end of each week): count the codes. The most frequent one sets next week's extra drill.

3.7 LR-MOCK Full timed mock and scoring

1. Paper: the next unopened InfinLume mock paper (INF-L25 onwards and INF-R25 onwards, in order), in exam mode (section 5.3).
2. Conditions: headphones, phone away, no dictionary, no pause. Listening (35 minutes), a 10-minute break, then reading (60 minutes).
3. Method: LR-P1 steps 1–5 on every listening item; LR-P4 clock checks for reading; LR-P5 on Q30–39.
4. Seen items: if you recognise an item from earlier practice, still answer it, but tick R beside it. About 43% of InfinLume's question slots repeat across its own papers, so this will happen.
5. Score with the grid (section 1.1) and fill in the score sheet (section 1.8). The site's own estimated 0–699 score uses its own method: note it, but use the grid.
6. If you ticked more than 3 items R, also score the unticked items only, band by band, as percentages, and compare them with the targets: Q1–19 95%, Q20–29 90%, Q30–39 40% (30% only if Q1–19 are all right). To turn them into one number: 12 × (Q1–4 %) + 54 × (Q5–10 %) + 135 × (Q11–19 %) + 210 × (Q20–29 %) + 156 × (Q30–35 %) + 132 × (Q36–39 %). For example, 6 right of 7 unticked Q20–29 items is 86%, worth 210 × 0.86 = 180.
7. Targets: 31 of 39; at most 1 miss on Q1–19; at least 9 of 10 on Q20–29; at least 4 of 10 on Q30–39 (3 only if Q1–19 are all right); then two mocks in a row at 31 or more.
8. Same day: log every miss (LR-P6), and take the Q20–29 listening misses through LR-P2.
9. Record the NCLC of listening and reading at every full mock, alongside writing and speaking: the analysis moves study hours to the lowest skill, and START-HERE.md explains the Day 21 decision about a TEF sitting.

4. Tables

4.1 LR-T1 Question word → answer type

Use it for listening Q5–10 and for catching every spoken question. The example pairs are study examples.

The trap, shown on one study example. Qu'est-ce que vous voulez boire ?

• A. Au café d'en face. → answers où (W)
• B. Avec plaisir. → answers an invitation (W)
• C. Un jus d'orange, s'il vous plaît. → right
• D. Dans dix minutes. → answers quand (W)

4.2 LR-T2 Notice and admin words

For reading Q1–10 and the announcements and messages in listening Q11–19.

Opening and times

Appointments and registration

Documents

Prices and payment

Rules

Transport and announcements

Phone messages

4.3 LR-T3 Letter purpose formulas

Reading Q11–19 asks "why does X write?", "what does X propose?", "what must the reader do?". The purpose sits in a formula, usually at the start. Options are often written as infinitives, as in the last column.

4.4 LR-T4 Intent markers

Purpose is 4 of 10 circulating listening Q20–29 items, 15% of the B2-band reading answers and 19% of their questions; aims written as infinitives make up 28% of the B2-band listening option sets [M]. These words carry the purpose.

Purpose and offers

Plans (the most common question at reading Q11–19: what someone wants or plans to do, 22%)

4.5 LR-T5 Change-of-plan markers

The key depended on a stated change in 10 of 136 FEI-written reading items, and when it did, a wrong option often gave the old state (6 such options across the FEI-written and circulating items) [M]. In listening: official Ex. 5, circulating Q20, and 7 of 136 TV5 option sets [H/M]. Rule: the new state is the answer; the old state is a ready-made wrong option (P).

Careful: reporter also means "to report", and the noun un reporter is a journalist.

4.6 LR-T6 Concession pivots

A pivot matters in 11 of 136 FEI-written reading items (5 of 68 at Q20–29): the key follows it, or a wrong option uses the conceded half, which is true but does not answer [M]. Six prep sources name this trap [M]. Rule: the answer usually follows the pivot.

4.7 LR-T7 Paraphrase-pairs notebook (seed)

The key usually restates the text with a more general word [M]. Start a notebook with these columns, and add 1–3 pairs from every LR-P2 and LR-P3 session.

Seed pairs. The first four are from the research [M]; the rest are common patterns added here [L].

4.8 LR-T8 Attitude words

Opinion items are rare at Q20–29: 1 of 68 TV5 listening option sets and 1 of 10 circulating listening items; 2 of 68 B2-band reading items (4 counting any attitude word in the question). They begin to appear around Q30 (3 of 17 at TV5 position 12) [M]. Learn these to recognise them; do not build daily drills around them.

In the options

In the question (these words signal an opinion item; gap-2 counted one only when the answer was also a judgement or a feeling): penser, trouver, juger, apprécier (#469), aimer, plaire (#604), préférer (#307), regretter (#973), reprocher, critiquer, dénoncer (#304), déplorer, se plaindre, s'opposer (#894 opposer), réagir; l'avis, l'opinion, le point de vue, la position, l'attitude, le sentiment.

Québec attitude expressions are in LR-T10.

4.9 LR-T9 False friends

The first ten are those named by the prep sources the research counted (3 sources name false friends as a trap) [M]; the rest are common ones added here [L].

4.10 LR-T10 Québec words

Why they matter: FEI's 2018 TCF Canada manual says some items use a Canadian accent and/or Canadian context; the 2022 and 2024 manuals do not repeat the sentence [M]. 2 of the 5 official listening samples at B1 or above are set in Québec, and so is 1 of 10 circulating Q20–29 items (ice fishing) [H/L]. The cheapest protection is about 20 words; the most dangerous are the meal names (ClickTCF) [M].

Sounds to recognise (practise on RC-01): t and d before i or u hiss (tu sounds like "tsu", dire like "dzire"); i, u, ou sound more open in closed syllables; je suis → "chu", il → "y", elle → "a"; the question particle -tu (Tu viens-tu ?).

5. Practice source map

5.1 TV5MONDE "S'entraîner au TCF" (TV5-01 to TV5-17)

• What it is: 17 free practice series. TV5MONDE says the teams that write the official TCF designed them and FEI's TCF team validated them [H for what TV5MONDE states]. Each series has 15 listening, 10 grammar and 15 reading items, 680 questions in all. Skip the grammar block: TCF Canada has no grammar paper.
• What it is not: the live test. TV5MONDE says its simulator questions are never used in a real TCF, and its practice follows the shorter TCF tout public format, so use the series for band drills, not for timing 39-item papers. Topics lean towards culture, travel and the French-speaking world [M].
• Index: https://apprendre.tv5monde.com/fr/tcf (also: strategies, FAQ, PDF booklets, support page).
• Address of one series: https://apprendre.tv5monde.com/fr/tcf/test-dentrainement-au-tcf?tcf_lot_id=LOT&competence=CO for listening, and the same with competence=CE for reading. The lot number identifies the series.
• Listening has no transcripts. The training mode allows replays; don't replay before answering.

Which positions stand in for which band (gap-2.md §2.1) [M]

In listening, position 5 already behaves like a TCF Canada Q11+ item [H, counted].

The 17 series

Keep TV5-12, TV5-16 and TV5-17 unseen for timed checks [L, my choice]. They have no known overlap with another series in the research, so training on the other 14 cannot leak their content; 16 and 17 also have the highest lot numbers. A fourth repeated subject, the Expolangues language fair, is in two series the research does not name; there the text or question differs, so it does little harm. If the calendar needs a fourth check, use TV5-11.

How to run a kept-unseen series as a timed check. Use one at your first full mock (Day 21), one about halfway between that and your test, and one in the last week.

• Listening: all 15 items, one play each, no pause.
• Reading: the 15 items in 23 minutes (TCF Canada's rate of 60 minutes for 39 items).
• Read the result by position, not as a grid score: any miss at positions 1–7 is a warning; at positions 8–11, 4 of 4 is on target and 3 of 4 is the 8-of-10 warning zone; positions 12–15 are a bonus.

Supply. The 14 drill series give, per skill, 56 items at positions 1–4, 42 at positions 5–7, 56 at positions 8–11, 14 at position 12 and 42 at positions 13–15. At 10 a day, the Q20–29 items last under a week per skill, so the InfinLume drill papers and the voice-mode drills carry most of the daily volume.

Do not use TV5MONDE's simulator (80 random questions in 90 minutes, listening played once) until your last kept-unseen series is done: it probably draws from the same bank and would show you those items [L, inference]. The same caution applies to the TV5MONDE TCF podcast (17 episodes, December 2022; listing), whose contents the research did not check [M].

5.2 Official FEI samples (FEI-01 to FEI-04)

The 2018 TCF Canada manual (https://www.afsf.com/exams/tcf/Manuel_du_candidat_TCF_Canada.pdf) has no sample items; skip it.

5.3 InfinLume full papers (INF-L01 to INF-L40, INF-R01 to INF-R40)

• What it is: 40 listening and 40 reading papers of 39 questions each, free, in review mode without a login. Library: https://www.infinlume.com/tcf-canada. These are your full-length timed mocks.
• IDs: INF-Lnn is listening Paper nn, INF-Rnn is reading Paper nn. Paper 01: https://www.infinlume.com/tcf-canada/listening/test-101?mode=review and https://www.infinlume.com/tcf-canada/reading/test-101?mode=review. The papers are numbered 101–140, so Paper nn should be test-1nn (INF-L27 → .../listening/test-127). That pattern is inferred from Paper 01's address; if a link fails, open the paper from the library.
• Two modes (listening Paper 01 checked on 4 November 2026) [L, one page checked]:
  • Exam mode: a 35-minute timer, each recording plays once, no answers until you submit, an estimated 0–699 score at the end. Use it for LR-MOCK.
  • Review mode: no timer, recordings can be replayed, right or wrong straight away, then an explanation, a translation and the transcript. Use it for drills, keeping to LR-P1 (one play before you answer).
  • No login is needed; signing in only syncs your progress. The reading papers were not checked: if their exam mode does not time 60 minutes, time yourself.

Allocation [L, my suggestion]

Provenance caveat [L]

• Not FEI material. InfinLume calls its papers "original… in their original question order"; where they come from is unknown, and it is not affiliated with FEI.
• Its listening transcripts look machine-made, with errors such as bédières for rivières.
• Positions can be off by one: Paper 01's recording at position 25 announces itself as question 24. Its level labels differ by skill (listening: B1 = Q12–18 and B2 = Q19–29; reading: B1 = Q11–19 and B2 = Q20–29).
• About 43% of its question slots repeat across its own papers: 885 distinct listening and 890 distinct reading questions in 1,560 slots per skill [M, a fact about its bank]. That is why LR-MOCK has the R tick.
• Its estimated score uses its own method. Score yourself with the grid.
• Learn the patterns, not the answer keys. One prep site (TCFCAD) says types and themes recur but exact questions are not reliably repeated [L].

5.4 RFI (RFI-01, RFI-02)

FEI's manual recommends RFI's listening training for the TCF, at https://francaisfacile.rfi.fr (manual p.39) [H]. The research could not open the site itself, so its current layout is unchecked.

5.5 Radio-Canada audio (RC-01)

• What: Radio-Canada's audio, on its OHdio app and website. The research gives no address; OHdio is at https://ici.radio-canada.ca/ohdio [L: added here, not from the research files].
• Why: the 2018 TCF Canada manual's Canadian-accent sentence, and the Québec settings in the official and circulating samples (LR-T10) [M/H]. ClickTCF advises about 15 minutes a day of Radio-Canada audio (accent guide) [M].
• How: choose news or interview segments of 1–5 minutes. Use LR-P1 for the gist (topic, purpose, pivot), then replay twice and note Québec words (LR-T10) and sounds (the hissed t and d, chu, Tu viens-tu ?). Most segments have no full transcript, so treat this as accent training rather than LR-P2 material.

5.6 Skip, or handle with care

• TCF tout public (29 questions) and TCF Québec (29 on paper, 34 on computer) mocks: the wrong length for timing practice [H].
• OPAL's series (login and subscription) and tcf-canada.ca's 40 reading tests (paid): the research did not open them [H for access]. If you use them, the same rule applies: patterns, not keys.
• LearnFrenchCanada's preparation guide describes a 29-question listening paper, which is wrong for TCF Canada.
• Pages announcing "2026 TCF reforms" (adaptive listening, new thresholds): unsourced, and contradicted by IRCC's table updated on 16 October 2026 [L].
• Prep books: Hachette FLE, CLE International and Didier FLE list no TCF book with a readable extract today [H].

6. Voice-mode listening drills (VL-01 to VL-04)

How to run them

• Run them in the TCF Examiner project: start a new chat there and type the calendar's DRILL line, for example Jour 9 | DRILL | VL-03 L1 (voice-mode/session-types.md). The prompts below are a fallback for an ordinary chat only: paste the drill's line as your first message, replace LEVEL with 1, 2 or 3, switch on voice mode and say Commence.
• If your app shows the text of what Claude says, keep the screen out of sight until you have answered. Afterwards, that text is your transcript for LR-P2.
• Levels change the length, the vocabulary and how natural the speech is. Start every drill at level 1 and move up a level after two rounds in a row at 80% or better.
• Claude speaks with one voice per chat and cannot reliably produce a Québec accent, so use RC-01 for the accent.
• For VL-03 and VL-04 the four options are read aloud before the audio, because there is no printed sheet. Jotting one word per option is allowed here; in the exam you simply read them.
• Log misses in LR-P6 with the source "VL-0n".
• These are practice items written by Claude, not FEI items. They train the mechanics (numbers, question words, changes of plan, the option traps); TV5MONDE and the InfinLume papers give you the exam's own style.

VL-01 Number, time and price dictation

VL-01, level LEVEL. Act as my French listening coach for numbers. Give me 20 items, one at a time. Each item is one short, natural French sentence containing one number: a price in euros or dollars, a clock time (both the 24-hour form such as 18 h 45 and spoken forms such as sept heures moins le quart), a date, a year, a phone number said in pairs, a percentage, or a room, flight, bus or platform number. Mix in near-misses: deux/douze, treize/trente, quatorze/quarante, quinze/cinquante, seize/soixante, soixante-dix/quatre-vingt-dix, cent/cinq cents, and times such as 13 h 30, 15 h 30 and 16 h 30. Say each sentence once only, then stop and wait. I will answer with the number written in digits. Then tell me in English whether I was right; if not, say the French number again slowly and tell me which sound I missed. Level 1: slow, numbers up to 100, simple sentences. Level 2: normal speed, any number, prices with cents. Level 3: natural speed, two numbers in one sentence where the second corrects the first (for example a time that changes), and I must give only the final one. At the end, give my score out of 20 and list the numbers I missed for my error log.

VL-02 Question and reply, Q5–10 style

VL-02, level LEVEL. Act as a TCF Canada listening examiner for the question-and-reply items (questions 5 to 10). Give me 10 items, one at a time, all in French. For each item, say one everyday question (meals, shopping, prices, invitations, directions, transport, times, weather, appointments, work), then four short spoken replies labelled A, B, C and D. Exactly one reply fits both the question word and the situation. Build the three wrong replies the way the official examples do: each one answers a different question word (where, when, with whom, which one, yes or no) or repeats words from the question without answering it. Sometimes make the right reply indirect, for example "it isn't expensive" for a price question. Include invitations, requests, permission and thanks, in both tu and vous. Say the question and the replies once only, then stop and wait for my letter. Then tell me in English the right letter, which question word each wrong reply answers, and one short tip. Level 1: slow, very common words. Level 2: normal speed. Level 3: natural speed with spoken forms such as t'as, y a and a dropped ne. In one of the 10 items, make the question a thank-you and the right reply the Québec bienvenue. After 10 items, give my score and list my misses.

VL-03 Short dialogue with a change of plan, Q11–19 style

VL-03, level LEVEL. Act as a TCF Canada listening examiner for questions 11 to 19. Give me 5 items, one at a time, all in French. For each item: first read four short answer options labelled A to D, slowly, so that I can jot one word for each (in the real exam they are printed). Then play one everyday dialogue between two people, or a voicemail, or an announcement: a shop, a booking, an appointment, a work schedule, transport, a neighbour, a school. In a dialogue, say each speaker's first name before their line. Make the plan change before the end, or add an exception, with words such as finalement, en fait, plutôt, au lieu de, sauf, reporté, annulé or déplacé. Then ask one question in French, for example "Que doit faire Marc ?", "Quand aura lieu le rendez-vous ?" or "Comment le client va-t-il payer ?", and stop and wait for my letter. Make one wrong option the old plan, one a swap of time, place or person, and one an option that repeats a word from the audio but says something else. Never repeat the audio before I answer. Then tell me in English the right letter, the sentence that decided it, and the trap behind each wrong option (old plan, swap or echo). Level 1: about 20 seconds, slow. Level 2: about 30 seconds, normal speed. Level 3: about 40 seconds, natural speed, and in one item out of five the first plan stands, so that I cannot simply pick the last thing said. In about one item in five, set the scene in Québec and use dîner for lunch, souper for the evening meal or fin de semaine for the weekend. After 5 items, give my score and list my misses.

VL-04 A 60-second interview, Q20–29 style

VL-04, level LEVEL. Act as a TCF Canada listening examiner for questions 20 to 29. Give me 3 items, one at a time, all in French. For each item: first read four answer options labelled A to D, slowly, so that I can jot one word for each (in the real exam they are printed). Then play a radio interview: a journalist talks with a guest (a worker, a student, a volunteer, an organiser or a researcher) about everyday life, such as work, housing, food, transport, studies, travel, volunteering or a local project. The guest explains what they do and why, and one mais or pourtant matters. Then ask one question in French and stop and wait for my letter. Most questions should ask about purpose ("Quel est l'objectif de… ?", "Que propose… ?") or what we learn ("Qu'apprend-on sur… ?"); ask about an opinion only rarely. Make the right option say what the guest said in different, more general words. Make the wrong options: one that repeats a word from the interview but says something the guest did not say, one that swaps a person, time or place, and one that the interview contradicts or never mentions. Never repeat the audio before I answer. Then tell me in English the right letter, the sentence that gave the answer, the trap behind each wrong option, and one pair "word in the interview → word in the option" for my paraphrase notebook. Level 1: about 40 seconds, slow, simple words. Level 2: about 60 seconds, normal speed. Level 3: about 75 seconds, natural speed. In about one round in five, set the interview in Québec and use one Québec word. After 3 items, give my score.

Sources

• Analysis: exam-analysis.md §§2.1, 2.2 and 3.
• Research: tcf-co-ce-items.md (official items, prep-source counts, traps, Québec words, false friends); gap-2.md (rule-based coding of 136 TV5MONDE reading items, 136 listening option sets, 8 official items and InfinLume Paper 01's Q20–29; the 9-of-10 target); tcf-format-scoring.md (format, the grid, the 31-correct rule, pacing); gap-1.md (the # ranks).
• Official: FEI candidate manual, Version J: https://medellin.alianzafrancesa.edu.co/manuel_du_candidat_tcf ; Version D: https://www.unicaen.fr/wp-content/uploads/2023/02/Manuel_du_candidat_TCF-1.pdf ; CIEP TCF Québec manual: https://elf.uqac.ca/wp-content/uploads/2018/10/manuel-candidat-tcf-quebec1.pdf ; CIEP TCF Canada manual: https://www.afsf.com/exams/tcf/Manuel_du_candidat_TCF_Canada.pdf
• Community grid (unofficial): https://paliertcf.com/tcf-canada ; https://luckytcf-www.pages.dev/en/tcf-canada ; https://hitcf.com/en/guide/tcf-listening
• Checked for this kit: the modes of InfinLume listening Paper 01, on 4 November 2026.