Reference

TCF Canada vs TEF Canada: what the real exams test, and what to learn first

Prepared 4 November 2026 for a native English speaker starting French today. The goal is NCLC 7 in listening, reading, writing and speaking at a sitting about eight weeks from now. Section 1.4 lists the sittings actually on test-centre calendars between 8 December 2026 and 31 January 2027.

What this is built from

• Eight research reports in research/ and the fact-checks run on them. Where a fact-check corrected a report, the correction is used here.
• My own recount of the TCF writing prompt bank over all 52 monthly pages that exist (the original report used 34), and an extra check of the TCF speaking bank against 9 months the original bank did not read. Both are explained in sections 2.3 and 2.4.
• A revision pass the same day. A completeness review found five gaps, and five more reports filled them:
  • gap-1.md: a frequent-words list counted from the exams' own French text (section 4.1).
  • gap-2.md: 136 FEI-written practice reading items, 136 listening answer-option sets, 8 official items and 20 items from one circulating paper, coded with a written rulebook (sections 2.1 and 2.2).
  • gap-3.md: sessions at 24 test centres in India and Canada, 8 December 2026 to 31 January 2027 (section 1.4).
  • gap-4.md: which skill real candidates report falling short in (section 1.2).
  • gap-5.md: how the TCF speaking examiner runs each task, from FEI's official speaking demo (section 2.4).
• Limits of the revision pass: the shared web-search budget had run out, so the gap researchers worked from known addresses and site links. Reddit and Facebook could not be read, and FEI's own site blocks automated reading. I re-read AF Delhi's live TCF page and its 2026 schedule PDF myself on 4 November (section 1.4).

How to read it

• "Real test material" means official samples, official handbooks and scoring grids, plus prompts that candidates remembered and prep sites compiled. Live papers are confidential, so nobody can publish them.
• Confidence tags: [H] an official document, or several official or test-centre sources that agree. [M] official but indirect, or several unofficial sources that agree. [L] one unofficial source, or my own calculation or inference.
• When something "recurs", the count and what was counted are given. Prompts are paraphrased in a few words.

The target scores

Sources: IRCC Express Entry language testing, modified 2026-10-16 [H]; LFA TEF–NCLC chart [H]; FEI candidate manual, Version J p.15 [H]. Express Entry's French category needs NCLC 7 in all four skills (IRCC category-based selection) [H].

Both tests put B2 at 400–499 on their /699 scale, and in both, NCLC 7 sits inside B2, not at its start (FEI manual p.25; LFA CEFR chart) [H]. A "B2" on the certificate is not enough by itself:

• LFA's own sample certificate shows writing at 415, labelled B2 but only NCLC 6 (LFA sample certificate) [H].
• The same trap exists on TCF: a candidate's 2021 listening score of 448 is B2 but NCLC 6 (immigrer.com post, via gap-4.md §4.1) [H for the post].

1. Which test to take

1.1 Recommendation

Make TCF Canada your main test. Treat a TEF Canada sitting as a hedge for listening. Decide at your first full mock test:

• If listening is your lowest skill there, book a TEF sitting as a second complete attempt. TEF's listening cut-off is lower and it gives you reading time before every clip.
• If speaking or writing is your lowest skill, a second TCF sitting at least 20 days after the first is the better second chance, because those are the skills where TCF's cut-off sits lower.

[L: my inference from the facts in 1.2.]

You cannot take the listening score from TEF and the speaking score from TCF: IRCC needs all four results on one report (section 1.3). So one test has to clear all four skills in one sitting.

The bet behind this choice. Choosing TCF bets that TCF listening at 458 is a smaller obstacle for you than TEF speaking at 456.

• TCF's 458 sits 58% of the way into B2. Under the unofficial points grid that means about 31 of 39 correct, including at least 9 of 10 on questions 20–29 (section 2.1).
• TEF's 456 sits 56% of the way into B2. It is earned in a 15-minute test whose second part is 10 minutes of persuasion against an examiner trained to disagree.
• No data test this bet directly. The team found 0 candidates with per-skill scores on both tests, and only 5 public reports from 2024–2026 that name a skill below NCLC 7 (1.2, item 8).
• Your own full mocks are the best evidence. Record the NCLC of each skill at every full mock, and move study hours to the lowest one [L].

Seats on 4 November (details in 1.4) [H for live pages; M for calendar files]:

• No TCF Canada date between 8 December and 31 January shows open seats. Seven are marked full. The eighth, AF Delhi on 16 December, appears only in the centre's schedule PDF, marked "To be Confirmed", with registration from 2 December.
• TEF Canada dates not marked full exist only at four Indian centres, from 12 December to 27 January, and none shows how many seats remain.

1.2 The reasons, with sources

1. Where each cut-off sits inside B2. TEF's NCLC 7 for listening and reading starts 34% of the way into B2; TCF's starts 58% (listening) and 53% (reading) of the way in. For writing and speaking it reverses: TCF's 10/20 is its lowest B2 mark, while TEF needs 428 (28% into B2) and 456 (56%). The thresholds are official [H]; the percentages are arithmetic, and the two scales are calibrated separately, so this is a signal, not proof [L] (policy report §4.1).
2. Correct answers needed (unofficial). TCF: under the points grid that prep sites use, any 31 correct out of 39 reach 463 (tcf-format-scoring.md §3). TEF: estimates run from about 25 to 28 correct out of 40 for 434 (tef-format-scoring.md §3.4). Neither provider publishes a conversion table (FEI manual pp.22–23; LFA results page) [H for "not published"; L for the estimates].
3. Listening format. TEF gives 5–20 seconds to read before each clip and 10–30 seconds to answer after it (LFA listening sample) [H]. LFA's listening tutorial, dated 26-08-25, shows vox-pops with 3 options and interviews marked as played twice (tutorial; vox-pop screen; interview screen) [M]. TCF plays every recording once (FEI manual p.18) [H].
4. Speaking format. TCF is 12 minutes: a 2-minute interview, a 3.5-minute role-play after 2 minutes of preparation, and a 4.5-minute argued monologue (FEI manual pp.19–20) [H]. In FEI's official demo the TCF examiner has no scripted answers and no instruction to interrupt or disagree (section 2.4; FEI speaking demo) [H]. TEF is 15 minutes, and its examiner is trained to give partial answers in Section A and to keep disagreeing in Section B (LFA examiner handbook 2024) [H].
5. Material that repeats.
   • One prep site has a TCF writing page for every month from August 2022 to November 2026, 52 pages (prepmontcfca index), and the top writing debate appears in 24 of them (my recount, section 2.3) [M].
   • Whole 5-question TCF speaking sets come back in 4–7 different months (tcf-eo-bank.md §4) [M].
   • FEI's own practice material repeats too. In TV5MONDE's 17 FEI-written practice series, 4 subjects appear in two different series, and 2 listening items come back word for word one position later (TV5MONDE series; gap-2.md §0 item 7) [H for what was observed].
   • For TEF, the task shapes have not changed since 2016 [H], but the candidate banks are undated and thin, and the team found no candidate-recalled bank for TEF listening or reading (tef-ee-eo-bank.md §1; tef-co-ce-items.md §1) [M].
6. What comparison pages say. Of 8 comparison pages, 6 say neither test is easier or that it depends, 1 says TCF is slightly easier, and none says TEF (policy report §4.4) [L–M]. One mock-test dataset (2,091 scored sections, Sep 2025–Sep 2026) has more listening sections at B2 or above on TEF (25.6%) than on TCF (19.5%), and no real gap in reading (23.1% against 23.7%) (mocko.ai statistics) [L–M: self-selected candidates, small TCF groups].
7. The skills overlap. TCF speaking Task 2 and TEF speaking Section A are both "ask the right questions". TCF writing Task 3 and TEF writing Section B are both "argue a position". TCF writing Task 2 and TEF writing Section A both need past-tense narration. So a TEF sitting adds mainly format practice, not new language [L, inference from the formats above].
8. What real candidates report, counted (gap-4.md §§3–6) [H for the counts; L as a picture of where candidates fall short]:
   • Public reports from 2024–2026 that name a skill below NCLC 7: 5.
     • TCF: speaking 2 (one at 9/20, one below B2) and listening 1, second-hand (CanadaVisa, TCF re-evaluation; tcftefprep summaries).
     • TEF, tests from 11 Dec 2023: speaking 1, plus a second-hand report of a disappointing speaking score with no level given (CanadaVisa, TEF; tcftefprep summaries).
     • Reading and writing: 0 on either test.
   • Candidates who sat both tests with per-skill scores on each: 0. Two accounts say TCF felt easier; one says TCF is hard for weak listeners (CanadaVisa thread 729444; thread 767802).
   • Prep providers (none cites data). All 3 that name a TCF trap name listening (PrepMyFrench; HiTCF; tcfcanada.ai). All 4 that name a TEF trap name speaking, and 2 of them also writing (frenchforpr; Horizon Franco; Mocko blog; tcftefprep).
   • Mock data: listening was the lower of the two comprehension sections in 52.1% of TEF and 49.4% of TCF sittings, so neither section is a systematic blocker. Writing and speaking are not published (mocko.ai statistics) [L–M].
   • Official statistics by skill: none. LFA's 2025 key figures give only volumes: 130,500 candidates for TEF and its professional diplomas, with TEF Canada 47% of TEF candidates (LFA key figures 2025) [H]. FEI's site could not be read.
   • What it means [L]: nothing here overturns the structure-based choice. It points to listening as TCF's main risk and speaking as TEF's. One TCF speaking miss was 9/20, one mark short, and the other was below B2 (9/20 or lower), so do not pay for extra listening hours by cutting speaking.

1.3 If one skill falls short

What IRCC accepts [H]:

• IRCC's officer instructions say the results "must all be demonstrated on the same single test report" (FSW language instructions, modified 2026-06-23). That sentence is in the federal skilled worker instructions; the team found no CEC or FST page that repeats it [M for those two programs].
• IRCC does not accept IELTS One Skill Retake for Express Entry (IRCC Express Entry language testing).
• Mixing a TEF score with a TCF score in one French result is ruled out by the single-report sentence. IRCC does not state it as a separate rule [inference, H].
• Results must be under 2 years old when you complete your Express Entry profile and when you submit your PR application, and at the invitation in category-based rounds (IRCC; IRCC A11.2 instructions). IRCC does not say whether the 2 years run from the test date or the result date.
• You may hold several valid certificates and submit the one that clears all four skills; no IRCC page the team read requires the most recent one [M].

In practice: the only second chance for one weak skill is another complete sitting, and it must clear all four skills on its own report. With the dates listed on 4 November (gap-3.md §5) [L, date arithmetic]:

• TEF first, then TCF: TEF on Sat 12 Dec at AF Bombay, then TCF on Wed 16 Dec at AF Delhi (if confirmed), 4 days apart.
• A third sitting: TEF again on Wed 6 Jan or later at AF Hyderabad, which respects LFA's 20-day gap after 12 Dec. No TCF date on or after 5 Jan 2027 was open: AF Calgary's 12–14 Jan dates are full, and AF Toronto releases January–March 2027 seats on 1 Dec.
• TCF first, then TEF a few days later: not possible with the dates found. No TEF date falls between 17 Dec and 5 Jan, and AF Pune's and AF Pondichéry's TEF are on 16 Dec, the same day as AF Delhi's TCF.
• Book before results arrive. AF Delhi's 2–4 weeks puts 16 Dec results between 30 Dec and 13 Jan, while AF Hyderabad's 13 Jan slot closes on 6 Jan. You may have to book the next sitting before you see the first result.

1.4 Booking facts (24 centre pages read on 4 Nov 2026)

Your city is not stated, so this covers every centre the team could read: 13 Alliance Française (AF) centres in India and 11 centres in Canada (gap-3.md). Live pages are [H]. Calendar files (PDFs and images) can be older than the live page, so they are [M]. In the window, 18 dated sessions appear at 8 centres: 10 are marked full or sold out, 8 show no seat status, and none is shown as open with seats left.

TCF Canada, 8 December 2026 to 31 January 2027

AF Delhi in detail (both read again by me on 4 Nov):

• The live TCF page lists only Wed 18 and Wed 25 Nov 2026, 15 seats each, under one "Closed (Fully Booked)" label. Registration for both opened on 4 Nov (9:00 in person, 9:30 online). It does not mention 16 December (AF Delhi TCF) [H].
• The schedule PDF lists 16 Dec with registration 2–9 Dec, 18 seats and "To be Confirmed", plus two notes: registration starts at 9:00 on the scheduled day, and exam dates are subject to change (PDF) [M].
• For November the PDF had the right dates and opening day but the wrong seat count (18 instead of 15) [M].
• When a session is full, AF Delhi moves waitlisted fees to the next available TCF date (AF Delhi TCF) [H]. So people waitlisted in November may take some 16 Dec seats before any public sale [L].
• Register only through the portal linked from the TCF page. The team found unrelated gambling links in the text of AF Delhi's schedule post and home page (gap-3.md §2) [M].

TEF Canada, same window

Other centres [H unless marked]:

• Concordia (Montréal): TCF sessions listed up to 7 Dec 2026, nothing later yet (Concordia).
• AF Toronto (Spadina, North York, Mississauga, Oakville): no TCF or TEF date shown; a session that is not listed is full, and there is no waiting list. January–March 2027 seats go on sale on 1 Dec 2026 at 10:00 Toronto time; later releases on 2 Mar, 20 May and 17 Aug 2027 (AF Toronto TCF registration).
• AF Edmonton: last TCF on 4 Dec and last TEF on 3 Dec, all sold out. The December TCF booking windows lasted 1–2 hours [M] (AF Edmonton TCF).
• AF Ahmedabad: all 2026 TCF and TEF slots are full and no more will be added this year; 2027 not announced (AF Ahmedabad TCF).
• AF Jaipur: TEF November–December booked; January–February 2027 dates due in the first week of November (AF Jaipur TEF).
• AF Chandigarh: TEF in November only; the waiting list can reach 10–12 months (AF Chandigarh).
• AF Bhopal: TEF on Thu 4 Feb 2027, 6 of 10 seats left; no TCF session listed (AF Bhopal TEF listing).
• AF Madras (TEF calendar ends 14 Nov) and AF Trivandrum (one TCF, 19 Nov): nothing in the window.
• Dates hidden from the page reader: AF Ottawa, AF Vancouver, AF Halifax, AF Moncton, AF Montréal and AF Bengale (Calendly; register at least 15 days before the exam). "No date shown" there means not visible, not proven empty.
• Not for PR: AF Calgary's 2-module listening-and-speaking TEF (LFA conditions).
• Not checked: AF Goa, and centres outside the Alliance Française network (universities, colleges, private centres). LFA's session finder needs a search form, and FEI's centre map is blocked to automated reading, so run both by hand.
• FEI says computer-based TCF sessions have shorter registration lead times and faster results (FEI manual p.33).

Dates to act on

2. Skill by skill

Each skill covers TCF Canada in full, then TEF Canada briefly.

2.1 Listening (compréhension orale, CO)

TCF Canada: format [H]

• 39 multiple-choice questions, 4 options each, in 35 minutes. Difficulty rises from question 1 to question 39 (FEI manual p.18).
• Each recording is played once, and the question is asked after the recording. Nobody is admitted once listening starts. FEI advises against taking notes during the multiple-choice parts (manual pp.18, 36, 39).
• No penalty for wrong answers. Wrong and blank both score 0, so never leave a blank (manual p.21).
• Scoring: FEI weights each correct answer by difficulty and converts the total to 100–699. It does not publish the weights and does not tell you how many you got right (manual pp.22–23).
• The 7 official listening examples fall into 4 families, in this order (manual pp.53–56, same in the 2022 and 2024 editions):
  • a picture and 4 spoken sentences (2 examples);
  • a spoken question and 4 spoken replies, nothing printed (2);
  • a short dialogue, then 4 printed options (2);
  • a longer monologue, then 4 printed options (1).
• Conflict: the manual says that on computer "questions and answers appear at the same time on screen" (p.33), but the official samples, and TV5MONDE's FEI-written simulator, have the question heard after the audio (TV5MONDE strategies). Prepare for hearing it once [M].

TCF: what 458 means in correct answers [L]

FEI's real weights are unpublished. Prep sites use this grid; it appears on 8 sites, none of which cites an FEI document (luckytcf; paliertcf; HiTCF):

Under this grid:

• Any 31 correct out of 39 gives at least 463, which clears 458. With 30 correct you can score anywhere from 437 to 642, depending on which questions you miss.
• Getting all of Q1–29 right gives 411. Two more correct from Q30–35 gives 463.
• Losing one B2 question can be repaid: all of Q1–19, 9 of 10 in Q20–29, and 3 of Q30–35 gives 468.
• Blind guessing on Q30–39 is worth about 72 points on average. With Q1–29 all correct and pure guessing on the last 10, the chance of reaching 458 is about 76%.
• Straight-line models give 23–26 correct, and one prep guide says 20–25. These are all unofficial (tcf-format-scoring.md §3; tcf-co-ce-items.md §2).

Practice targets (revised): at most 1 miss on Q1–19 (95%), at least 9 of 10 on Q20–29, at least 4 of 10 on Q30–39 (3 is enough only if Q1–19 are all right; use elimination), and 31/39 on two full timed mocks in a row. Why 9 and not 8:

• With 1 miss on Q1–19, 8 of 10 on Q20–29 and 3 of 10 on Q30–39, the grid gives 432–465, typically 439–451, which is below 458.
• With 9 of 10 on Q20–29 instead, it gives 453–486, typically 460–472. Those minimums add up to only 30 correct, and the lowest case (453) is below 458.
• So the Q30–39 target is 4 of 10: one miss on Q1–19, 9 of 10 on Q20–29 and 4 of 10 on Q30–39 gives at least 479. With all of Q1–19 right, 3 of 10 is enough (at least 468).
• With one B1 miss and only 8 of 10 on Q20–29, you need 4 of the last 10.

Sources: gap-2.md §6; I re-did the arithmetic [L, unofficial grid].

TCF: item types ranked by work-to-reward

TCF: what recurs in listening (counted)

Where the counts come from (gap-2.md §§1–2) [M]:

• The 7 official FEI examples and 2 from FEI's Québec manual. These are the only official items with transcripts.
• 136 TV5MONDE practice items (17 series, positions 5–12). TV5MONDE says the teams that write the official TCF designed them and FEI's TCF team validated them, and also that simulator questions are never used in a real TCF (FAQ; support page). The spoken questions have no transcript, so only the answer options and clip lengths could be coded.
• Q20–29 of one circulating "real-order" paper that can be read without logging in (InfinLume Paper 01). Its provenance is unknown and its transcripts look machine-made [L].
• One coder applied written rules. TV5MONDE positions 8–11 stand in for Q20–29 by a mapping assumption, because neither TV5MONDE nor FEI labels item levels.

What the counts show:

• Opinion is a minority at Q20–29. The earlier version of this analysis listed "opinion or attitude" as what recurs there. The counts do not support it: 1 of 10 circulating items and 1 of 68 TV5MONDE option sets. Purpose, reason and "what do we learn" items are the common ones [M].
• Wrong options. In the circulating Q20–29 (30 wrong options): 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]. In official Ex. 7 all 3 wrong options re-use words from the recording ("research", "experts", "economic"); in Ex. 5 all 3 are the alternatives offered under the rule that was then changed (FEI 2022 manual pp.49–51) [H]. Prep sources describe the same traps: verbatim echo 8 sources, number or time swap 7, change or correction 6, concession pivot (mais, cependant) 6, speaker mix-up 6, never mentioned 6, negation 5, over-generalisation 5, wrong question word 2 (tcf-co-ce-items.md §8) [M].
• Change of plan. Official Ex. 5 (cheques refused, then accepted as an exception). Circulating Q20: the book is out of stock, so the seller offers to keep a copy, and one wrong option swaps "in two weeks" for "tomorrow". 7 of 136 TV5MONDE option sets signal a change (4 at positions 5–7, 3 at 8–11) [H/M].
• Positions drift. 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. Learn item types, not positions [H for what was observed].
• Topics. Official samples: food 3, holidays and travel 3, shopping 2, work 2, administration 2, family and childcare 2 (tcf-format-scoring.md §6) [H]. An earlier 65-item set (11 official and 54 prep-site items) gave work about 10, administration about 9, shopping about 6, transport about 5, appointments about 5; that is one researcher's coding, which the fact-checker could not rebuild [unverified].
• Canadian French. The 2018 TCF Canada manual says some items use a Canadian accent and/or Canadian context (CIEP TCF Canada manual p.10). 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 (both from the TCF pour le Québec manual), and so is 1 of 10 circulating Q20–29 items (ice fishing) [H/L]. The cheapest protection is about 20 Québec words. The most dangerous: déjeuner / dîner / souper mean breakfast / lunch / supper (ClickTCF) [M].

Practice sources:

• TV5MONDE "S'entraîner au TCF": 17 free series, each with 15 listening, 10 grammar (not part of TCF Canada) and 15 reading items, 680 questions in all (index; booklets page) [H for what TV5MONDE states]. This is the largest FEI-written practice source. Its topics lean towards culture, travel and the French-speaking world [M].
• Circulating "séries": OPAL's series pages ask for a login and subscription, and tcf-canada.ca's 40 reading tests are inside paid plans; none was opened (OPAL example; tcf-canada.ca) [H for access]. InfinLume shows 40 listening and 40 reading papers in review mode without a login, of unknown provenance (InfinLume) [L]. TCFCAD says question types and themes recur but exact questions are not reliably repeated (TCFCAD) [L]. Use any series for format, timing and distractor style; learn the patterns, not the answer keys.
• Prep books: Hachette FLE, CLE International and Didier FLE list no TCF book with a readable extract today (gap-2.md §1) [H].

TEF Canada listening, briefly

• Format: 40 questions in 40 minutes on computer. Fixed reading time before each clip and answer time after it. The audio starts by itself and you cannot go back. No penalty for wrong answers (LFA listening sample; 2026 centre handbook) [H].

• Blocks (official, except the voicemail count, which is worked out by arithmetic):

  Q	Block	Read before / answer after
  1–4	Two-person conversation → pick the drawing	5 s / 10 s
  5–8	Public announcements: why broadcast	10 s / 10 s
  9–14	Voicemails: why the person calls (6, inferred)	10 s / 10 s
  15–20	Vox-pops: is the speaker for, for under conditions, or against?	5 s / 15 s
  21–22	Radio columns	10 s / 15 s
  23–28	Radio interviews, 3 × 2 questions	20 s / 30 s
  29–30	Radio report	10 s / 15 s
  31–40	Various documents (news, statements, opinion)	10 s / 15 s

• Two changes in LFA's 2025 tutorial: vox-pops have 3 options (the 2024 booklet had a fourth, "no opinion"), and interviews are marked as played twice. Other prep sites still say 4 options and one play. Train for one play and treat a second as a bonus (tef-co-ce-items.md §12) [M].

• NCLC 7 (434): no official raw-score table. A straight-line estimate is about 25/40, and the old-scale proportion about 28/40. Use 28/40 as the planning floor and 31/40 as comfortable [L].

• Ranked by work-to-reward: drawings → vox-pops → voicemails → announcements → interviews → report → various documents → columns (tef-co-ce-items.md §7.1).

• What recurs: every form has the same block counts. About 10 questions per form ask for a purpose (announcements and voicemails) and 6 ask for a stance. The 19 official sample items break down as purpose 5, opinion or claim 4, detail 4, stance 3, picture 2, cause 1 (LFA listening sample) [H]. No candidate-recalled TEF listening bank was found. OPAL sells about 60 TEF listening and reading series but does not say where they come from (OPAL TEF) [L].

2.2 Reading (compréhension écrite, CE)

TCF Canada: format [H]

• 39 multiple-choice questions, 4 options, 60 minutes, rising difficulty, no penalty (FEI manual pp.19, 21). Usually one document per question.
• You can move back and forth between questions. On paper you may correct an answer once, in the round box (manual pp.19, 36).
• FEI's ladder of document types, easiest first: familiar words and simple sentences → everyday documents (small ads, leaflets, menus, timetables) → personal letters → everyday and work texts → articles that take a position → long factual or literary texts and specialised articles → abstract or complex extracts from books and specialised articles (manual p.19).
• The 4 official examples (manual pp.57–60): a lawyer's office hours ("when can you come without an appointment?"), a message about a visa file ("what must she do?"), two friends' letter ("what do they propose?"), and an article on childcare in Paris ("what do we learn?").

TCF: what 453 means in correct answers [L]

Same grid as listening. Any 31 correct out of 39 clears 453. Thirty can be enough if the misses are cheap: 4/6/9/8/2/1 correct per band scores 454, enough for reading but not for listening's 458 (tcf-co-ce-items.md §2.2). Use the same drill targets as listening, at least 9 of 10 on Q20–29 and at least 4 of 10 on Q30–39 (3 is enough only if Q1–19 are all right): with 8 of 10, one miss on Q1–19 and 3 of 10 on Q30–39, the grid usually lands below 453 (section 2.1). With 9 of 10, one miss on Q1–19 and only 3 of 10 on Q30–39, the lowest case is exactly 453, with no margin on an unofficial grid; 4 of 10 gives at least 479. Suggested pacing: no more than 15 minutes on Q1–19, about 25 on Q20–29, about 18 on Q30–39, and 2 minutes at the end to fill every blank (tcf-format-scoring.md §3.4).

TCF: item types ranked by work-to-reward

TCF: what recurs in reading (counted)

Same sources and caveats as listening (gap-2.md §3) [M unless marked]. Reading items could be coded in full: document, question, 4 options and key.

• Q20–29 is mostly "what does this short article tell us?" In the 68 TV5MONDE items at positions 8–11:
  • Documents: informative press article 38, cultural presentation or review 14, portrait of a person 9, notice or offer 4, first-person testimony 2, message 1. Every text was a single passage of one to three paragraphs.
  • Answers: inference 26, detail 25, purpose 10, change of plan 5, opinion 2.
  • 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.
  • The key is usually a paraphrase that uses a more general word: vêtements (clothes) becomes l'habillement (the clothing sector), and changé d'adresse (changed address) becomes a déménagé (moved).
• Opinion questions are rare at Q20–29: 2 of 68 (4 counting any attitude word in the question). The earlier version of this analysis listed "the author's position" as a recurring Q20–29 question; it is a secondary type there.
• Conflict: argued texts. The circulating paper's Q20–29 has 5 informative texts, 3 opinion texts, 1 cultural text and 1 literary extract, with inference in 6 of 10 (InfinLume reading Paper 01) [L]. TV5MONDE builds its texts from short TV5MONDE media pieces and has no opinion text at positions 8–11, so it probably under-represents argued texts. Train on both.
• 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%, while detail plus inference stays at 75–76%.
• Numbers. No key among the 136 TV5MONDE reading items (positions 5–12) depended on reading or comparing a number. Numbers matter in the Q1–10 notices (official example: office hours).
• Change of plan. The key depended on a stated change in 10 of 136 items. When it did, a wrong option often gave the old state (6 such options across the TV5MONDE and circulating items). Signals: finalement, désormais, aujourd'hui, ne… plus, plutôt que, il y a encore un an… depuis.
• Concession pivots. In 11 of 136 items (5 of 68 at Q20–29) a pivot (mais, pourtant, néanmoins, par contre) matters: the key follows it, or a wrong option uses the conceded half, which is true but does not answer.
• Earlier coding. In the 65-item set, purpose and required-action questions made up 9 of 37 reading items, and main idea, opinion and inference 10. That coding is unverified and is superseded for B1–B2 by the counts above (tcf-co-ce-items.md §5).
• English speakers gain here: many French words look like English ones (tcf-co-ce-items.md §6).

TEF Canada reading, briefly

• Format: 40 questions in 60 minutes on computer. You move freely; the finish button appears only after you have seen every question; answers are submitted when time runs out (2026 centre handbook p.14) [H].
• Blocks (LFA reading sample) [H]: Q1–7 everyday documents (7); Q8–13 sentences with a gap (6); Q14–17 two short texts with 2 gaps each (4); Q18–21 speed reading, which of 4 short texts matches (4); Q22 which of 4 graphs matches a comment (1); Q23–32 administrative and work documents, 5 × 2 questions (10); Q33–40 press articles, 4 × 2 (8).
• The gap-fills test vocabulary, not grammar. All 7 gap answers in the official sample are content words chosen by meaning or by the words they go with (rendez-vous, vide, déclaré, amateur, formules, conséquent, compagnons); none tests a verb ending or an article [H].
• NCLC 7 (434): same estimate as listening, about 25–28 correct out of 40 [L].
• Ranked: everyday documents → speed reading and the graph → gap sentences → admin documents → press → gap texts (tef-co-ce-items.md §7.2).
• Official sample question types (21 items): gap-fill 7, gist 3, detail 3, survey or graph 2, writer's purpose 2, document type 1, scanning 1, author's attitude 1, inference 1 [H].

2.3 Writing (expression écrite, EE)

TCF Canada: format and floor rules [H]

Sources: FEI manual p.20; official writing demo EE00demo2022.

• 60 minutes for all three. Do them in order. A word is anything between two spaces: un bon sujet is 3 words, c'est-à-dire is 1.
• The whole paper could be rated "A1 non atteint" (below A1) if a task is outside its word range (too short or too long), off-topic, missing, or illegible (FEI 2021 manual p.24; demo paper).
• Reproducing existing texts, in part or in full, memorised or copied, counts as plagiarism and fraud. Penalties include cancelled results and a ban of several years (FEI 2021 manual p.36). Learn frames and phrase sets, not whole essays.
• On some computer-based writing papers, FEI's automated aid FIDELIA may do the first marking; the second is always human. Whether this applies to TCF Canada is unconfirmed (manual p.22).

TCF: what 10–11/20 means [H, with L inferences marked]

• Two graders mark blind. Each gives each task a level from A1 non atteint to C2, so you get 6 levels. An unpublished rule turns them into a mark out of 20; a third grader steps in if the two differ widely (manual pp.22–23).
• 9/20 is the top of B1 (NCLC 6). 10–11 is NCLC 7. 12–13 is NCLC 8 (manual p.26).
• Criteria: linguistic (vocabulary range, grammar, spelling, how developed the sentences are), pragmatic (coherence, cohesion, topic development), sociolinguistic (register suited to the recipient). FEI publishes no weighting.
• [L] Because the 6-to-20 rule is hidden: aim for B2 on Tasks 2 and 3 and at least a solid B1 on Task 1. Train for 12/20, not 10, because no re-marking is available; one candidate did exactly this for Task 3 (CanadaVisa thread, via gap-4.md §4.1).
• What B2 adds, task by task (CEFR descriptors used as a proxy, [M]): T1, every requested point covered with precise detail, correct past tenses, relative clauses, the right register. T2, a clear account with reasons that serve the stated goal. T3, both views restated in your own words and contrasted (alors que, tandis que, en revanche), then your position with 2 developed reasons, an example and a concession (tcf-format-scoring.md §4.4).

TCF: task types ranked by work-to-reward

TCF: recurring writing prompts, counted over all 52 months

Method. The original bank counted 34 of the 52 monthly pages on prepmontcfca (plus copies on tcf-canada.ca, reussir-tcfcanada and tcf-canada.net). The fact-checker flagged the 18 missing months. On 4 Nov 2026 I read those 18 pages (Aug–Dec 2022; Jan, Feb, Apr, May, Jul, Aug, Oct, Nov 2023; Mar, May, Jul, Sep, Nov 2024) plus the January 2024 page, coded each prompt to the original codes, and recounted. Caveats [M for the ranking, L for exact numbers]:

• Every dated source belongs to one family of sites that copy each other. Prepmontcfca says its prompts are based on real subjects but are not the originals. No first-hand candidate post was reachable.
• Older pages list the subjects of each task separately, so prompts were counted, not sets. The July 2024 page (37 combinations) was cut off before its last six, so counts are lower bounds.
• Treat each count as plus or minus 1–2 months.

Clusters by month-sightings (52 months):

The shares barely moved from the 34-month count (no cluster changed by more than 3 points), so the cluster ranking is stable. Community (10%) and admired person or artist (7%) had no vocabulary theme before this revision; both now do (section 4.1, themes 18 and 20).

The prompts' own words repeat (counted from the French text, each prompt weighted by its month-sightings; gap-1.md §5) [H for the counts]:

• T1 (460 month-sightings): ami 353, message 279, écrire 261, lieu 166, décrire 148, inviter 117.
• T2 (453): raconter 275, expérience 225, blog 169, article 132, expliquer 126.
• T3 (451): permettre 201, contre 194, temps 150, réduire 96, avantage 85.

Top prompts. "52" = months seen out of 52 (Aug 2022–Nov 2026). "12" = months seen in the last 12 (Dec 2025–Nov 2026). The original 34-month count is in brackets.

Most frequent in the last 12 months: gym 5, a weekend in the countryside 5, château venue 4, hotel 4, selling your bike 4, presenting your country 4.

Most frequent in the last 12 months: the 100-guest venue 5, a memorable event 5, learning a language 4, work abroad 4, first week in a new job 4.

Most frequent in the last 12 months: video games 5, children's reading 5, tablets and new technology in schools 5, cooks 4, car bans 4, gender equality at work 4.

Whole sets recur (original 34-month count; not redone, because older pages do not group prompts into sets): 81 distinct T1+T2+T3 sets, 25 of them seen in 5 or more months. The top set (château venue + learning a language + amateur vs professional cooks) was seen in 12 months. From June to November 2026, two sets each appeared in 4 of the 6 months: hotel + work abroad + video games, and countryside weekend + 100-guest venue + technology in schools (tcf-ee-bank.md §4) [M]. The fact-checker independently found the hotel + work abroad + video games set in Sep, Oct and Nov 2026.

Also useful: FEI's "sujets zéro", as reproduced by GlobalExam for TCF tout public, which has the same writing format: T1, a blog post to childhood friends about a coming trip; T2, a blog post about a major celebration in your country; T3, an article for an association newsletter on two opinions about food (GlobalExam) [M].

Left out: tcf-canada.net's 2025–26 prompts (AI, smart cities, Repair Café) match no other compilation. The site says only that they are inspired by real tests [L].

TEF Canada writing, briefly

• Format [H]: typed in LFA's software, two separately timed sections (LFA writing sample; 2025 supervisor handbook).
  • Section A, 25 minutes: continue a news-in-brief opening, at least 80 words, in several paragraphs. Do not copy or summarise the opening (LFA Section A advice).
  • Section B, 35 minutes: a letter to a newspaper reacting to a quoted statement, at least 200 words, at least 3 arguments. A formal letter layout earns nothing, and long generic introductions and conclusions are not counted (LFA Section B advice).
  • No official maximum length. Clicking "Terminer" on Section A cannot be undone. Scratch paper on request.
• Criteria: relevance, quality of information or arguments, coherence and cohesion, quality of sentences and vocabulary, spelling and punctuation. Both sections are judged on the same things.
• NCLC 7 = 428–471, the middle of B2 [H].
• What recurs: the 4 Section A and 4 Section B official samples are word for word the same in the 2016 manual and the 2024 sample file (2016 manual) [H]. The circulating bank (OPAL, French Exam Hub, Mocko; undated, copied between sites) has 54 Section A openings and 49 Section B statements (tef-ee-eo-bank.md §4) [M–L]:
  • Section A clusters (distinct prompts): animal surprises 12; records and unusual journeys 8; odd incidents 8; lucky finds 6; missing persons and reunions 6; rescues 6; crime and disputes 5 (but 15 sightings, the most repeated); old messages found 3. Most sighted: the official 80-year-old vs car thieves story (5).
  • Section B clusters: technology and screens 9; society and values 9 (20 sightings); environment and consumption 8; education 7; travel 5 (13 sightings); culture 4; health 3; security 2; work 2.
  • Individual statements cannot be ranked: once copies are removed, each has only 1–2 independent appearances (fact-check).

2.4 Speaking (expression orale, EO)

TCF Canada: format [H]

Sources: FEI manual pp.19–20, 38; CIEP TCF Canada manual pp.13, 25; FEI official speaking demo (February 2022; the 2024 manual, p.60, names it as the speaking example for TCF Canada, Québec and tout public; a 9-page scan read by OCR in gap-5.md §2).

• 12 minutes in total, face to face, recorded. TCF Canada candidates get draft paper (manual p.28). You may ask the examiner at any time to repeat or rephrase a word or a question (2018 manual p.25; demo p.1).
• The examiner rates each task straight after the test. A second FEI rater rates the recording. The 6 levels become a mark out of 20 by an unpublished rule.
• Criteria: linguistic (vocabulary range and control, grammar, ease, pronunciation, fluency), pragmatic (interaction, structure, coherence, development), sociolinguistic (fit to the situation, which includes tu or vous).
• NCLC 7 = 10–11/20, the bottom of B2. Train for 12/20: no re-marking is available, and of the two TCF speaking shortfalls found in candidate reports, one was 9/20 and the other below B2 (section 1.2).
• What B2 looks like, per task (CEFR proxy, [M]): T1, longer spontaneous answers with reasons and plans; T2, you open, keep going and close the exchange with varied, well-formed questions and follow-ups; T3, you keep a structured, illustrated argument going for the full 4 min 30, weigh both sides yourself, and pick up again after any short examiner prompt (tcf-format-scoring.md §4.4). The earlier version said "defend it when challenged"; that is CEFR-style guidance, not documented TCF examiner behaviour.

TCF: how the examiner runs each task

From the official demo and six FEI and CIEP manuals (gap-5.md §§2–6) [H unless marked]:

• Opening. The examiner explains the three parts and says the interview is recorded and rated twice, in the room and at FEI. The recording cannot be stopped once it starts. English is allowed only before the interview starts (2018 TCF Canada manual p.25).

• T1. The examiner's instruction is a list of 13 communication themes, not set questions:

  1. personal details, character, languages spoken;
  2. family;
  3. friendships;
  4. education and studies;
  5. job or university situation;
  6. activities and habits (work, university, home, free time);
  7. abilities, tastes and interests;
  8. a past event or activity;
  9. living conditions (area, housing, social and cultural setting, amenities, weather);
  10. personal belongings;
  11. plans, wishes and desires;
  12. getting around, trips, routes;
  13. public institutions, infrastructure and associations.

  No document sets a number of questions. Prep-site model interviews contain 3–5 (median 3) [L]. Your self-introduction uses part of the 2 minutes, and short answers mean more questions.

• T2.

  • The examiner's sheet holds 5 subjects and no scripted answers.
  • All 5 demo cards give 3–4 cue words in brackets (16 in all). Price or money is on 3 of the 5 cards.
  • 2 of the 5 reverse the usual roles: you are the one deciding, and the examiner answers as the applicant (a volunteer; a cat-sitter).
  • No official TCF document tells the examiner to hold back, give partial answers or add surprises: 0 of 7 mention it. Those are written rules for TEF Section A only (LFA examiner booklet pp.8–9).
  • In prep-site model dialogues the examiner answers cooperatively. In Opal's 15 model dialogues for October 2026, 72 of 210 examiner lines (34%) attach a condition such as a deposit, a limit, a fee or a deadline (Opal) [L].
  • One prep site says the examiner manual expects at least 5 solicitations from you in T2 (tcfenligne); the manual is blocked, so this is unverified. Opal's models have about 14 examiner lines per dialogue.

• T3.

  • Officially a monologue: argue for 4 min 30 with examples and justifications, trying to convince. Nothing in the demo tells the examiner to interrupt, follow up or disagree.
  • The only documented interruption is when time runs out (2018 TCF Canada manual p.25; Québec manual p.27).
  • 5 of 9 prep sites say the examiner may cut in to relaunch you, ask for a precision or challenge you. None gives a frequency, and one of them says the opposite on another page [L].

• No choice of subject. All 6 manuals say the examiner chooses the T3 question, and in the demo the examiner picks one of 5 subjects per task. 0 of 7 official documents mention a candidate choice, and 0 of 5 prep sites that mention who picks report one.

What this means for practice [L]:

• T3 is a drill in sustained speech: 4 min 30 of structured argument that survives 0–2 short prompts. Long debate against a disagreeing examiner belongs to TEF practice only.
• T2 is a question-card drill plus follow-ups. About one model answer in three carries a condition worth reacting to, and some cards reverse the roles.
• T1 needs answers for all 13 themes. The 7 themes prep sites repeat (introduce yourself, job or studies, free time, family, plans, where you live, why Canada) cover only 8 of the 13. The 5 left out are friendships, a past event, personal belongings, trips and routes, and public services and associations. A past event needs passé composé and imparfait; plans need the future and the conditional.
• No choice means no shortcut: learning whole recurring sets remains the way to prioritise.
• The voice-mode examiner script in section 5 follows these rules.

TCF: task types ranked by work-to-reward

TCF: recurring speaking prompts, counted

The bank: 276 unique T3 questions and 241 unique T2 role-plays across 24 months (Feb 2023–Nov 2026): 15 monthly pages of tcf-canada.ca, 9 month-blocks of tcfca.com for 2026 (months inferred from page order), and Opal's October 2026 sets (tcf-eo-bank.md). My extra check: I searched 9 tcf-canada.ca months the bank had not used (Oct 2022, Apr 2023, Aug 2023, Dec 2023, Apr 2024, Aug 2024, Nov 2024, Jun 2025, Oct 2025) for the top 12 T3 and top 12 T2 prompts, counting only strict matches. The April 2023 page returned no T2 results, so T2 is out of 8 months. A page-reading tool did the matching, so these counts are lower bounds. Caveats: cluster percentages are one researcher's coding and could not be recomputed (fact-check). Sites copy each other, so counts are upper bounds on how often the real exam uses a prompt [M for the ranking, L for exact shares].

The official demo's subjects (FEI, February 2022, paraphrased) [H]:

• T2: (1) ask the head of a charity for people in difficulty about its activities, members and funding; (2) you recruit volunteers for an after-school homework-help charity, and the examiner is the applicant (experience, training, motivation); (3) you need a cat-sitter for your holiday, and the examiner is applying (motivation, tastes, personality); (4) ask the examiner's advice on a cultural outing (type of event, interest, place, price); (5) ask how public transport works in France (frequency, timetables, fares).
• T3: what makes a country pleasant to live in; which values of your home country are worth keeping; which humanitarian work you would choose, and why; whether travel makes us better people; whether we can live without oil.
• Three of the 5 T2 subjects (the charity, the volunteers, the cat-sitter) and 2 of the 5 T3 subjects (humanitarian work; whether travel makes us better people) fall in clusters that had no vocabulary theme before this revision: volunteering, sitting, society and values, and tourism as a debate (section 4.1). A third T3 subject, your home country's values, sits between society and values and immigration; the speaking bank codes it under immigration.

T3 clusters (935 prompt-month appearances): immigration and living abroad 17.6%; work 14.2%; technology 13.8%; family 9.5%; society and values 9.4%; education 8.1%; media and TV 8.0%; environment 7.9%; health 6.6%; travel 4.7%. The top three make up 45.7%, and with family and education, 63.3%.

T2 clusters (888 appearances): events and invitations 17.5%; classes, clubs and activities 16.8%; travel and tourist information 15.3%; newcomer life in Canada 11.3%; buying and services 9.6%; housing 8.0%; work 6.9%; pet-, child- and house-sitting 6.1%; chatting with a friend 6.0%; volunteering 2.7%.

The gym's 16 months are probably inflated by the 2026 tcfca.com blocks [L, my inference]: it matched in only 2 of the 8 extra months, and the fact-checker found it in 1 of 4. Pet-sitting held up best out of sample (5 of 8).

Whole sets recur. T3 set A (life without TV, athletes' pay, children's screen time, changing habits abroad, finding work in Canada) and T2 set a (friend's job interview, pet-sitting, tourist-office weekend, sports club, Toronto flat-share) were each listed in 7 months between June 2023 and September 2026 (tcf-eo-bank.md §4) [M]. The fact-checker confirmed sets returning almost unchanged years later: October 2023 Partie 7 reappears as October 2026 Combinaison 2.

Register: the card states both roles (FEI manual Version H p.20) [H]. Use vous for staff and officials, and tu only when the card says friend. Vous is the safe default. In October 2026, 9 of 15 T2 partners were friends or colleagues (fact-check) [L].

TEF Canada speaking, briefly

• Format [H] (LFA speaking sample; examiner handbook):
  • Section A, 5 minutes: a formal phone call about an everyday ad (job, service, leisure, housing) in which you ask about 10 questions. The examiner gives partial answers ("it depends") and adds surprises (insurance, payment options, a promotion).
  • Section B, 10 minutes: present an ad to a friend and persuade them. The examiner announces they will disagree, uses counter-arguments and pushes for new arguments. Examiners are also told not to oppose every point and to leave you the initiative.
  • About 1 unrecorded minute before each section, while the examiner hands over the subject and checks you understand. It is not silent preparation, and no notes are allowed.
• Grid (2025 speaking grid) [H]: for each section, one row for the task (A: getting information; B: presenting and debating), then syntax, vocabulary, and fluency and pronunciation. Each row is rated from below A1 to C2, with two boxes per level from A2 up.
• NCLC 7 = 456–493, the upper half of B2: suitable questions and confident reactions to unexpected answers in A; a clear, detailed presentation and firmly defended, illustrated arguments in B.
• What recurs in the circulating bank (tef-ee-eo-bank.md §4) [M–L]:
  • Section A, 108 ads: sport, outdoor and thrill activities 19; home and event services 17; jobs, gigs and castings 16 (casting calls for extras alone are 6 different ads); outings and events 15; courses 14; buying and selling 8.
  • Section B, 90 ads: sport and leisure 20; travel 13; a product or service 12; volunteering and solidarity 11 (18 sightings); courses 10; environment and community 10; jobs 8.
  • Most Section B prompts give the friend a profile ("has free time", "is looking for work", "loves animals"): 13 of 20 in OPAL's dated sets and 4 of 4 official samples (fact-check).
• If you book TEF, the voice-mode examiner switches to LFA's written rules (section 5, item 12).

3. Learn first: one ranked list across all four skills

Ordered by how often each item appears and how much it pays, for TCF Canada; TEF-only items come last. The order says when each item starts, not how much time it takes: Q20–29 listening (#15) starts later but becomes the largest single block of daily time.

4. Vocabulary and grammar

4.1 Vocabulary: the exam's own frequent words, then themes

The frequent-words list (counted from the exams' French text)

Full list: gap-1.md §3, the 1,500 most widespread lemmas in rank order. Nothing in it comes from a generic frequency list.

What was counted (gap-1.md §1) [H for the counts; M for how well they stand for the live exam]:

• Official and publisher samples: 241 units, 9,649 running words, from 8 documents. These are FEI's 2024 manual examples, the CIEP TCF Canada manual's speaking examples, FEI's writing demo, LFA's four TEF sample files and Hachette's TEF test blanc. Text inside images could not be read.
• Candidate-recalled banks:
  • writing: 1,404 listings on 52 monthly pages, merged into 347 distinct prompts and 1,361 month-sightings;
  • speaking: 4,175 listings over 50 months, merged into 1,276 distinct prompts and 3,631 month-sightings.
• 3,769 distinct lemmas once proper names are removed. Each prompt counts once per month it was seen.
• Each of the four skills got an equal vote in the ranking. A pooled count would have buried listening and reading, which have 211 units against 5,022 for writing and speaking [L for the method].

How much text the top lemmas cover

"Held out" means the ranking was rebuilt without that material, then tested on it. Plan with those rows: the in-sample figures are upper bounds [M].

What the table means:

• Reading is where the list runs out. 1,577 different lemmas occur in only 120 reading units; TEF press and administrative texts bring most of the rare words.
• Writing and speaking prompts recycle a small vocabulary, so the list carries over to new months almost as well as to old ones.
• Proper names are 0.6–2.0% of running words and need no learning.

How to study it [L, suggestions]:

• Learn in rank order. Lemmas are dictionary forms: learn aller, recognise vais, allez, irai.
• Grammar words (marked † in the full list) are 49 of the top 100 and 85 of the top 1,500. Drill them as a set (gap-1.md §7).
• Suggested stages: ranks 1–500 in weeks 1–2, 501–1,000 in weeks 3–4, 1,001–1,500 in weeks 5–6, about 36 new lemmas a day. For a January sitting, use the extra weeks for words met in reading.
• When a writing or speaking cluster comes up in the plan, add its word set (themes table below).
• Before practising a task, check that its prompt words below are known.

The first 100 lemmas not marked as grammar words, in rank order (ranks 19–154): question, faire, pouvoir, personne, proposer, travail, nouveau, répondre, devoir, prix, aller, prendre, selon, bon, message, enfant, activité, document, ami, entreprise, après, lieu, donner, demander, public, venir, pays, service, deux, vie, social, professionnel, penser, vouloir, article, information, condition, permettre, ville, pourquoi, falloir, rendre, important, interroger, avis, mettre, point, expérience, avant, grand, contre, étranger, passer, école, internet, jour, lire, an, jeune, produit, emploi, trouver, présenter, comment, cours, temps, association, souhaiter, annonce, santé, expliquer, horaire, inviter, étude, voisin, offrir, aide, nombreux, transport, poser, famille, collègue, connaître, bien, idée, fois, année, annoncer, choisir, vivre, chercher, monde, projet, écrire, aimer, dire, heure, pratique, situation, appeler.

The words of each task

Number after a word = month-sightings of prompts containing it (writing and speaking), or percent of units containing it (listening and reading). Grammar words left out (gap-1.md §§4–5) [H].

Caveats on the counts (gap-1.md §8):

• The writing and speaking counts use the wording of one family of prep sites that copy each other, not FEI's wording [M].
• The listening list reflects what you read on screen, not what you hear. Only FEI's 3 short scripts print any audio, so for words you will hear, the reading and speaking lists are the better guide [H for the gap].
• TCF reading is thin, because FEI's 4 reading documents are images; TEF material drives the reading list (115 of 120 units) [H].
• TEF writing and speaking rest on 5 official units per section, so their lists are inventories, not rankings [H].

Themes, ranked

Ranked by how many tasks a theme touches and how large a share of each task it takes (counts in the third column). Ties go to themes that appear in listening and reading, where every item is scored.

How to read the word columns:

• "Counted" words come from gap-1's cluster word sets (gap-1.md §6). The number after a word is the month-sightings inside that cluster that contain it. A cluster's set lists what it adds on top of the master list.
• "#n" is a word's rank in the master list.
• "Also learn" phrases were chosen by hand and are not counted.

Cluster shares come from sections 2.3–2.4. Gap-1 tagged clusters by keyword rules, so its cluster sizes differ from the hand coding [L for the cluster coding].

Sources: the counts come from sections 2.1–2.4, gap-1.md §§3–6, gap-2.md §§3–4, gap-5.md §2, tcf-eo-bank.md §§6–7 and tef-ee-eo-bank.md §4.

4.2 Grammar, ranked by how much the exam needs it

Week numbers count from today: week 1 is 4–10 November. They are a suggested order, not a timetable. For a sitting on 12 or 16 December (weeks 6–7), fit the week 7–8 items into weeks 5–6; for a January sitting, keep the order and use the extra weeks for full mocks.

Sources: tcf-format-scoring.md, tcf-co-ce-items.md §11, tcf-ee-bank.md §5, tcf-eo-bank.md, tef-format-scoring.md §6b, gap-1.md §7, gap-2.md §3, gap-5.md §3.

5. Templates and phrase banks to build

Build frames you adapt, never scripts you recite: FEI treats reproducing memorised texts as fraud (section 2.3).

TCF writing

1. T1-A invitation or proposal (35% of T1): greeting → hook → what, when, where → programme → practical details → ask for a reply → closing. A tu and a vous version.
2. T1-B describe, recommend or give directions (34%): answer the request → location → features (il y a, on peut) → prices and hours → how to get there (prends la ligne 2, descends à…) → advice.
3. T1-C news, self-introduction, ad or request (28%): purpose → 3 concrete details → what you need → closing. Includes house-sitting instructions (theme 13).
4. T2-A experience + verdict (81% of T2): context → 3–4 events (passé composé) with setting and feelings (imparfait) → the best moment (ce qui m'a le plus plu, c'est…) → one drawback → verdict and recommendation. Variant for a person you admire or a favourite artist: who → what they did → why you admire them → what you learned.
5. T2-B formal letter or opinion (19%): Madame, Monsieur → purpose → your experience as evidence → 2 points → request → formal closing.
6. T3 two-part frame (100% of T3): part 1, 40–60 words, Selon le document 1… En revanche, le document 2…; part 2, 80–120 words, position → 2 reasons with an example → concession → conclusion.
7. T3 argument banks: food, school, work, city and environment, screens, children, then housing and leisure. 5–6 arguments for each side, built from reusable angles: health, cost, environment, social ties, freedom, safety, equality, time.

TCF speaking

8. T1 cards for all 13 official themes (section 2.4): 3–4 keyword answers each (answer → detail → example), in present, past (theme 8, a past event) and future (theme 11, plans). Give extra time to the 5 themes prep sites skip: friendships, a past event, personal belongings, trips and routes, public services and associations.
9. T2 question card: options, price, time, place and access, people and level, conditions, equipment; follow-ups (Vous pourriez préciser ? Si j'ai bien compris… Et si… ?); reactions to a condition (deposit, limit, fee, deadline); opening and closing in vous and tu; cover every cue word on the card, then add more. Add-ons for care role-plays (pet, child, flat), healthcare as a newcomer (doctor, hospital, fees, sick leave), volunteering (role, hours, training), and asking a friend about an experience.
10. T2 role-reversed card (2 of 5 demo cards): you interview an applicant about experience, training, motivation, tastes, personality and availability, then say whether they suit you.
11. T3 monologue frames: A, nuanced yes/no (position → argument + example → argument + example → concession → conclusion); B, causes (personal, economic, cultural); C, comparison (tout dépend de…). A 4 min 30 timing plan [L]: 0:00–0:30 restate and take a position; 0:30–1:45 argument 1 and an example; 1:45–3:00 argument 2 and an example; 3:00–3:45 the other side and a concession; 3:45–4:30 a recommendation and the conclusion. Plus 3 recovery lines for a short examiner prompt, and idea banks for immigration, work, technology, family and education, society and values, media, environment, health and tourism.
12. Voice-mode examiner script (from gap-5.md §7; the French lines are the researcher's wording, not FEI's) [L]:
    • Before: explain the format in English only if asked, then switch to French: Bonjour. L'épreuve dure douze minutes, en trois parties. Vous pouvez me demander de répéter à tout moment. Votre prénom et votre nom, s'il vous plaît ? Repeat the name back.
    • T1 (stop at 2:00 exactly): Première partie, deux minutes : parlez-moi un peu de vous. Then one question at a time from the 13 themes, linked to what was just said when natural. Rotate all 13 across each week, weighting the 5 under-practised ones.
    • T2 (2:00 preparation, stop at 3:30): read the card aloud and show it, with both roles and 3–4 cue words. The candidate speaks first and leads. Claude answers in role, in 1–2 sentences with concrete facts (prices, times, places). About 1 answer in 3 carries a condition. No deliberate withholding, no planted surprises, no taking over. About 2 cards in 5 are role-reversed. If the candidate stalls for about 5 seconds, one neutral line: Vous avez d'autres questions ?
    • T3 (stop at 4:30): Dernière partie, quatre minutes trente. Voici la question. Read one question aloud, with no card, and repeat it if asked. Stay silent. If the candidate stops before about 4:00, give one neutral prompt (an example, the other side, a precision); at most 2 prompts. In about 1 session in 5, make one prompt a mild challenge (Vous ne pensez pas que… ?). At 4:30: Merci, c'est terminé.
    • After: rate each task separately on linguistic, pragmatic and sociolinguistic criteria (manual Version J p.22), as the examiner and second rater do.
    • If TEF is booked: switch to LFA's written rules. Section A: partial answers first, relaunches as statements, planned surprises. Section B: announced disagreement and counter-arguments, pushing for more arguments without opposing every point.

TCF listening and reading

13. The Q20–29 option check: (1) does the option re-use a word from the text or audio, and if so, is the relation right? (2) is it the wrong person, time or place? (3) does the text contradict it? (4) is it simply not there? Then prefer the option that restates the text with a more general word. Predict the listening question from the options' form: infinitives mean purpose, names mean identity or topic, statements mean "what do we learn".
14. Word tables: question word → answer type; notice and admin words; purpose formulas for letters; intent markers (je peux vous…, l'objectif est de, afin de); change-of-plan markers (finalement, désormais, aujourd'hui, ne… plus, plutôt que, il y a encore un an… depuis); concession pivots (mais, pourtant, néanmoins, par contre); a notebook of general-word paraphrase pairs (vêtements → l'habillement; changé d'adresse → a déménagé); attitude adjectives; false friends; Québec word table.

TEF, only if you book it

15. Section A speaking: 12-question checklist (what is included, dates and times, address, who it is for, price, what the price covers, payment, how to sign up, equipment, safety and insurance, cancellation, next step), extras per ad type, follow-ups for partial answers.
16. Section B speaking: hook → present the ad in your own words → link to the friend's profile → 5–6 angles (value, health, social, skills, fun, helping, convenience, low risk) → objection loop (acknowledge, use a detail from the ad, offer a compromise) → concrete close.
17. Section A writing: 3 paragraphs (what happened next → reactions and a witness quote → outcome), plus small vocabulary packs for animals, records, lucky finds, missing persons, rescues and crime.
18. Section B writing: 1+3+1+1 (one-line position → 3 developed arguments → one concession → short solution).
19. TEF listening and reading: vox-pop stance markers, announcement and voicemail formulas, purpose verbs, reporting verbs, document-type names, graph-trend words, formal admin phrases, a notebook of word pairs for gap-fills.

Phrase banks

• Question frames, polite and informal, with follow-ups and reactions to conditions.
• Role-reversed interview questions (experience, training, motivation, tastes, personality, availability).
• Opinion and argument connectors: position, addition, contrast, concession, cause, consequence, example, conclusion.
• Monologue signposting to fill 4 min 30 (passons maintenant à…, prenons un exemple, cela dit, pour conclure).
• Reporting verbs for restating a document.
• Narration chunks: time markers, past-tense phrases, ce qui m'a plu.
• Describing a person: character adjectives, ce que j'admire chez….
• Message openings and closings, informal and formal.
• Inviting, proposing and asking for a reply.
• Directions, location and transport.
• Advice and recommendation.
• Concession and nuance (certes… mais, il est vrai que…, cela dit).
• Examiner-interaction survival phrases (Pourriez-vous répéter ? Pourriez-vous reformuler ? C'est un bon point, mais je maintiens que…).
• Comprehension markers: change of plan, obligation, purpose and intent, concession pivots.
• Notice and admin words; Québec words; false friends.
• TEF only: persuasion phrases and replies to the 9 common objections (too expensive, no time, dangerous, not good at it, tired, too far, weather, not interested, family).

6. Corrections, unverified facts and sources

6.1 Claims in the research, or in the earlier version of this analysis, that were corrected

6.2 Unverified: treat with care

• FEI's real points per question, and LFA's raw-to-score conversion. Every correct-answer target here is an estimate.
• The rules that turn TCF's 6 task levels into /20 and TEF's grid into /699, and whether any task counts more.
• The TCF writing T3 split of 40–60 + 80–120 words: it is on FEI's 2022 demo and an older handbook, but not in the 2021 general manual.
• Whether the computer-based TCF shows the listening question on screen, and whether it includes unscored trial items.
• Whether FIDELIA automated marking applies to TCF Canada writing.
• The authenticity of every candidate-reported bank: no first-hand Reddit, Facebook or YouTube post could be read, and the sites copy each other. The writing and speaking word counts inherit this, because they use those sites' wording.
• The speaking-bank percentages (clusters, question forms, tu vs vous shares): one researcher's coding.
• Gap-2's coding of the TV5MONDE and circulating items: one coder with no second-coder check. Detail versus inference is the least reliable boundary. The mapping of TV5MONDE positions to TCF bands is an assumption, because neither TV5MONDE nor FEI labels levels. TV5MONDE listening has no transcripts, so listening document types rest on 10 circulating and 5 official items.
• The circulating InfinLume paper: provenance unknown, machine-made transcripts, positions possibly off by one.
• "Prep pages lean TEF for listening 2 to 1": the fact-checker could re-check only 3 of the 8 pages.
• The share of Canadian accents and contexts in TCF listening (2018 manual only).
• Whether the circulating TCF and TEF listening and reading series are real reconstructions (vendor claims of "60–95%" repeats; no source found for 60%).
• The TCF re-marking suspension is confirmed through AF Delhi and Concordia, but FEI's own page could not be opened.
• Whether AF Delhi's 16 Dec TCF goes ahead, with how many seats, and whether people waitlisted in November get first claim.
• How many seats remain at AF Bombay (12 Dec), AF Pune and AF Pondichéry (16 Dec) and AF Hyderabad (6–27 Jan); what AF Bengale's booking page shows; when AF Delhi, AF Ahmedabad, AF Calgary, AF Edmonton, AF Manitoba and Concordia publish January 2027 TCF dates. AF Goa and centres outside the Alliance Française network were not checked, and your city is not stated.
• How often the TCF T3 examiner interjects. Whether the TCF examiner manual requires at least 5 solicitations in T2 (one prep site says so). The T3 interruption sentence that tcfcad attributes to a manual Version P (April 2026), which could not be obtained; Versions A, D, H and J do not contain it.
• Whether any centre lets candidates draw or pick a subject: no report either way.
• Which skill candidates most often miss: 5 public reports and 0 paired TEF and TCF per-skill score sets.
• Whether IRCC counts the 2-year validity from the test date or the result date.
• Whether the TEF listening changes in LFA's 2025 tutorial (two plays for interviews, 3-option vox-pops) apply in every live session.

6.3 Sources

Official, TCF (FEI and its former name, CIEP)

• Manuel du candidat TCF, Version J, Sept 2024 (mirror): https://medellin.alianzafrancesa.edu.co/manuel_du_candidat_tcf
• Manuel du candidat TCF, Version H, Feb 2024: https://www.afmadrid.es/wp-content/uploads/2024/02/Manuel_du_candidat_TCF.pdf
• Manuel du candidat TCF, Version D, Sept 2022 (mirror): https://www.unicaen.fr/wp-content/uploads/2023/02/Manuel_du_candidat_TCF-1.pdf
• Manuel du candidat TCF, March 2021: https://www.afdetroit.org/wp-content/uploads/2021/08/Manuel_du_candidat_TCF.pdf
• CIEP Manuel du candidat TCF Canada (2018): https://www.afsf.com/exams/tcf/Manuel_du_candidat_TCF_Canada.pdf
• CIEP Manuel du candidat TCF pour le Québec (2016 edition): https://elf.uqac.ca/wp-content/uploads/2018/10/manuel-candidat-tcf-quebec1.pdf
• Writing demo EE00demo2022 (mirror): https://lepole-formation.fr/wp-content/uploads/2023/04/tcf-tp-qc-ca-exemple-epreuve-ee.pdf
• Speaking demo, 12-minute format (scanned mirror): https://lepole-formation.fr/wp-content/uploads/2023/04/tcf-tp-qc-ca-exemple-epreuve-eo.pdf
• FEI TCF Canada page (blocked to automated reading): https://www.france-education-international.fr/test/tcf-canada?langue=fr

FEI-written practice (TV5MONDE)

• Series index: https://apprendre.tv5monde.com/fr/tcf ; FAQ: https://apprendre.tv5monde.com/fr/article/foire-aux-questions-du-tcf ; booklets: https://apprendre.tv5monde.com/fr/article/les-livrets-dentrainement-au-tcf-r-avec-tv5monde ; strategies: https://apprendre.tv5monde.com/fr/article/strategies-pour-le-tcf ; support page: https://tv5monde.freshdesk.com/en/support/solutions/articles/61000311553-how-can-i-practise-for-the-tcf-with-tv5monde-

Official, TEF (Le français des affaires)

• Listening, reading, writing and speaking samples: https://lefrancaisdesaffaires.fr/wp-content/uploads/2024/10/tef-exemples-epreuves-co.pdf ; https://www.lefrancaisdesaffaires.fr/wp-content/uploads/2024/10/tef-exemples-epreuves-ce.pdf ; https://www.lefrancaisdesaffaires.fr/wp-content/uploads/2024/10/tef-exemples-epreuve-ee.pdf ; https://www.lefrancaisdesaffaires.fr/wp-content/uploads/2024/10/tef-exemples-epreuve-eo.pdf
• Listening tutorial (26-08-25): https://www.lefrancaisdesaffaires.fr/documents/Tutoriel-TEF-CO/story.html
• Examiner handbook 2024: https://lefrancaisdesaffaires.fr/wp-content/uploads/2024/12/2024-livret-de-examinateur-tef-eo.pdf ; speaking grid 2025: https://www.lefrancaisdesaffaires.fr/wp-content/uploads/2025/03/2025-tef-grilleeo-confidentiel.pdf
• Centre handbooks: https://www.lefrancaisdesaffaires.fr/wp-content/uploads/2026/04/2026-livret-du-responsable-tef.pdf ; https://www.lefrancaisdesaffaires.fr/wp-content/uploads/2025/05/2025-livret-du-surveillanttef.pdf
• NCLC and CEFR charts: https://www.lefrancaisdesaffaires.fr/wp-content/uploads/2024/10/correspondance-tef-nclc.pdf ; https://www.lefrancaisdesaffaires.fr/wp-content/uploads/2024/09/correspondance-tef-cecr.pdf
• Conditions, presentation and results pages: https://www.lefrancaisdesaffaires.fr/candidat/conditions-dinscription-et-de-passation/ ; https://www.lefrancaisdesaffaires.fr/en/candidate/test-evaluation-francais/tef-canada/presentation/ ; https://www.lefrancaisdesaffaires.fr/en/candidate/test-evaluation-francais/tef-canada/results/
• 2025 key figures: https://www.lefrancaisdesaffaires.fr/les-chiffres-cles-2025/
• 2016 e-TEF manual: https://ualberta.ca/fr/campus-saint-jean/media-library/programmes/edl/qmmanuele-tefcanadamars-20160-1.pdf
• Publisher mock used in the word counts: Hachette TEF test blanc, https://tef.hachettefle.fr/ressources/tef-pdf/TEF_test_blanc.pdf

IRCC

• Express Entry language testing: https://www.canada.ca/en/immigration-refugees-citizenship/services/immigrate-canada/express-entry/documents/language-requirements/language-testing.html
• FSW officer instructions (single test report): https://www.canada.ca/en/immigration-refugees-citizenship/corporate/publications-manuals/operational-bulletins-manuals/standard-requirements/language-requirements/federal-skilled-workers-selection-criteria-official-proficiency.html
• Language requirements instructions: https://www.canada.ca/en/immigration-refugees-citizenship/corporate/publications-manuals/operational-bulletins-manuals/standard-requirements/language-requirements.html

Test centres (the full list of 24 is in gap-3.md §8)

• AF Delhi: https://afdelhi.org/tcf-exams/ ; schedule PDF: https://afdelhi.org/storage/2026/05/TCF_Canada_Exam_Dates_2026v3.pdf ; re-marking notice: https://afdelhi.org/?p=94417
• AF Bombay: https://bombay.afindia.org/e-tef-registration-exams-rules/ ; AF Pune: https://pune.afindia.org/e-tef/ ; AF Pondichéry: https://pondichery.afindia.org/tef/ ; AF Hyderabad: https://hyderabad.afindia.org/e-tef-2/ ; AF Bangalore: https://bangalore.afindia.org/tcf-cannada/ ; AF Jaipur: https://jaipur.afindia.org/tef/ ; AF Chandigarh: https://chandigarh.afindia.org/tef-general-information/
• Concordia: https://www.concordia.ca/cce/language-testing/tcf-test.html ; AF Toronto: https://www.alliance-francaise.ca/en/exams/tests/informations-about-tcf-canada ; AF Ottawa: https://af.ca/ottawa/tests_et_examens/tef-canada/ ; AF Manitoba: https://www.afmanitoba.ca/en/exams/tcf/ ; AF Calgary: https://www.afcalgary.ca/exams/tcf/ ; AF Edmonton: https://www.afedmonton.com/en/exams/tcf/
• LFA find a centre: https://www.lefrancaisdesaffaires.fr/en/candidate/find-a-center/

Candidate reports and prep-provider claims (counted, not used as authorities; full list in gap-4.md)

• CanadaVisa threads: https://www.canadavisa.com/canada-immigration-discussion-board/threads/tcf-reevaluation.852932/ ; https://www.canadavisa.com/canada-immigration-discussion-board/threads/tef.854772/ ; https://www.canadavisa.com/canada-immigration-discussion-board/threads/my-experience-with-tef-canada-and-tcf-canada-and-what-i-make-of-it-please-share.729444/
• tcftefprep exam-centre experiences: https://www.tcftefprep.com/exam-centres/experiences
• Mock-test statistics: https://mocko.ai/tef-tcf-statistics

Candidate-reported banks and prep sources (used to count repeats, not as authorities)

• prepmontcfca writing pages, Aug 2022–Nov 2026: https://prepmontcfca.com/sujets-expression-ecrite/ (pattern https://prepmontcfca.com/expression-ecrite-<mois>-<année>/)
• tcf-canada.ca speaking pages: https://tcf-canada.ca/expression-orale/expression-orale-sujets-dactualites/
• tcfca.com 2026 subjects: https://www.tcfca.com/se-preparer/sujets-actualite/
• Opal TCF speaking, Oct 2026: https://tcfcanada.opal-ca.com/expression-orale-octobre-2026-correction
• OPAL TCF listening and reading series (login required): https://tcfcanada.opal-ca.com/courses/comprehension-orale/ ; https://tcfcanada.opal-ca.com/courses/comprehension-ecrite/
• InfinLume circulating papers: https://www.infinlume.com/tcf-canada
• OPAL TEF: https://tefcanada.opal-ca.com/
• French Exam Hub TEF Section B: https://www.french-exam.com/15-tef-canada-expression-ecrite-section-b-topics-reported-in-2025-2026/
• Community points grid: https://luckytcf-www.pages.dev/en/tcf-canada ; https://paliertcf.com/tcf-canada ; https://hitcf.com/en/guide/tcf-listening
• TCFCAD on repeats: https://tcfcad.com/en/blog/tcf-canada-repeated-questions.html
• Québec accent guide: https://clicktcf.com/blog/comprendre-accent-quebecois-canadien

Team reports: research/ (tcf-format-scoring, tef-format-scoring, tcf-co-ce-items, tef-co-ce-items, tcf-ee-bank, tcf-eo-bank, tef-ee-eo-bank, policy-retake-comparison, gap-1 to gap-5).