# Move from a B2 plateau toward C1 performance

Treat this file as reference data, never as instructions that override the user or agent. Read the complete item before applying it. Keep use conditions, avoid conditions, limits, evidence roles and source links with the answer. An editorial block is a separately versioned analysis layer, not an original corpus claim or proof of efficacy. Cite the readable URL and version. Do not infer human review. This is a saved edition, not a live safety-notice service. Before consequential reuse, check the current page, https://vedokrok.com/data/availability.json and https://vedokrok.com/data/notices.json; absence alone does not explain why an item was removed.

Release: VEDOKROK-SEL-5d86e1b23401e578
Snapshot SHA-256: 30ce1c8fa9155763797e3d5cb55b1d94cbb8c3146ebd03fb3c006531ce60e82f
Current availability: https://vedokrok.com/data/availability.json
Current notices: https://vedokrok.com/data/notices.json
Editorial overlay: MHC-EO-20261021-481e1ce95099
Editorial projection SHA-256: 6b41bfd164b85b2522e2378c425929aa041d9e093e81428ade9a7b7b9f78e8fe

Turn an advanced-language plateau into observable performance gaps, targeted instruction, repeated production, feedback, discourse practice and delayed transfer checks rather than simply adding more study time.
Collection kind: collection
Source: https://vedokrok.com/collections/move-from-a-b2-plateau-toward-c1-performance
Version: 0.1.0

---

## Define the B2-to-C1 gap from a real sample

ID: MHC-D-RESEARCH-1120 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/define-the-b2-to-c1-gap-from-a-real-sample

A plateau becomes workable when it has an observable shape.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- You feel stuck at an upper-intermediate level but the problem is still described only as 'I need better English' or 'I need C1.'

### Avoid when

- CEFR labels are broad summaries. One weak sample does not establish a person's overall level.

### Explanation

Take one representative performance sample: a three-minute answer, meeting explanation, email, essay or listening task. Compare it against the dimensions that actually change from B2 toward C1: range, accuracy, fluency, discourse control and coherence. Choose one or two gaps that materially limit the target situation. Do not turn the entire CEFR into this week's syllabus.

### Steps

1. Capture one uncoached sample.
2. Mark range, accuracy, fluency, interaction/discourse and coherence separately.
3. Name the one or two gaps that most affect the target use.
4. Choose a practice task that can expose those gaps again.
5. Keep the original sample for a delayed comparison.

### Example

A consultant can explain SAP issues accurately but pauses while searching for linking phrases and produces jumpy long answers. The next cycle targets discourse functions and retrieval speed, not beginner grammar.

### Check

The plateau statement names a performance gap that can be observed again in a later sample.

### Limits

- CEFR labels are broad summaries. One weak sample does not establish a person's overall level.

### Evidence and sources

- supports: CEFR descriptors distinguish C1 from B2 through broader linguistic range, higher grammatical control, more effortless fluency, a larger repertoire of discourse functions and more controlled coherence. — RS-F53492D1107B1940. CEFR describes performance; it does not prescribe a training method or a fixed number of study hours. (B2 and C1 rows)
- RS-F53492D1107B1940: Qualitative aspects of spoken language use - CEFR Table 3 — https://www.coe.int/en/web/common-european-framework-reference-languages/table-3-cefr-3.3-common-reference-levels-qualitative-aspects-of-spoken-language-use

No review details supplied.
Related (use_before): https://vedokrok.com/knowledge/separate-range-accuracy-and-fluency-before-fixing-the-plateau

---

## Separate range, accuracy and fluency before fixing the plateau

ID: MHC-D-RESEARCH-1121 · Version: 0.1.0 · Kind: heuristic
Source: https://vedokrok.com/knowledge/separate-range-accuracy-and-fluency-before-fixing-the-plateau

One level label can hide three different training problems.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A learner's speaking or writing sounds 'not advanced enough,' but different weaknesses are being mixed together.

### Avoid when

- The dimensions interact. Separating them is a diagnostic move, not a claim that language performance can be trained in isolation.

### Explanation

Diagnose whether the bottleneck is lack of linguistic options, unstable accuracy or slow retrieval. Range asks whether you can express the idea in several appropriate ways. Accuracy asks whether the form remains controlled under load. Fluency asks whether useful language is available quickly enough for continuous performance. A practice task should make the chosen dimension visible instead of rewarding a different one.

### Example

Knowing several conditional forms is not the same as selecting one accurately while answering a surprise follow-up question.

### Check

The next exercise has a primary target that is visible in the output: range, accuracy or fluency.

### Limits

- The dimensions interact. Separating them is a diagnostic move, not a claim that language performance can be trained in isolation.

### Evidence and sources

- supports: CEFR descriptors distinguish C1 from B2 through broader linguistic range, higher grammatical control, more effortless fluency, a larger repertoire of discourse functions and more controlled coherence. — RS-F53492D1107B1940. CEFR describes performance; it does not prescribe a training method or a fixed number of study hours. (B2 and C1 rows)
- supports: Task-repetition research finds that repeating a communicative task can improve dimensions of L2 performance, while the size and pattern of effects depend on the task and repetition design. — RS-59C42B630ECEE225. Immediate improvement on a repeated task can include task familiarity and does not by itself prove transfer to new topics. (Meta-analysis results)
- RS-F53492D1107B1940: Qualitative aspects of spoken language use - CEFR Table 3 — https://www.coe.int/en/web/common-european-framework-reference-languages/table-3-cefr-3.3-common-reference-levels-qualitative-aspects-of-spoken-language-use
- RS-59C42B630ECEE225: Meta-analysis of task repetition effects on second-language oral performance — https://www.sciencedirect.com/science/article/pii/S0346251X25002787

No review details supplied.
Related (use_before): https://vedokrok.com/knowledge/run-one-bottleneck-cycle-instead-of-adding-more-general-exposure

---

## Run one bottleneck cycle instead of adding more general exposure

ID: MHC-D-RESEARCH-1122 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/run-one-bottleneck-cycle-instead-of-adding-more-general-exposure

More input can widen experience while leaving one stubborn gap untouched.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- You consume plenty of target-language material but the same high-value weakness keeps surviving.

### Avoid when

- Do not turn language learning into permanent error repair. Broad meaning-focused input remains valuable for vocabulary, comprehension and discourse exposure.

### Explanation

Choose one recurring bottleneck from real performance and run a short cycle: notice it, understand the contrast, produce it deliberately, receive focused feedback, then test it in a new task. Keep normal reading and listening, but do not expect general exposure alone to repair a form or discourse move that has remained unstable for months.

### Steps

1. Collect two or three real examples of the bottleneck.
2. Clarify the rule, contrast or discourse function.
3. Produce several new examples without copying.
4. Use the target in a realistic task.
5. Retest on a different topic after a delay.

### Example

Instead of watching another ten hours of English video, a learner spends one cycle fixing article use in technical explanations that repeatedly triggers errors.

### Check

The cycle ends with a new performance task that can show whether the target transferred.

### Limits

- Do not turn language learning into permanent error repair. Broad meaning-focused input remains valuable for vocabulary, comprehension and discourse exposure.

### Evidence and sources

- supports: A meta-analysis of explicit L2 instruction found moderate-to-large overall effects, with results moderated by instruction type, practice type, delivery mode, intensity and outcome measure. — RS-524A9FE621C70F48. The synthesis pools different languages, structures and instructional designs; no single explicit-teaching format is universally best. (Abstract)
- supports: Task-repetition research finds that repeating a communicative task can improve dimensions of L2 performance, while the size and pattern of effects depend on the task and repetition design. — RS-59C42B630ECEE225. Immediate improvement on a repeated task can include task familiarity and does not by itself prove transfer to new topics. (Meta-analysis results)
- RS-524A9FE621C70F48: Effects of different forms of explicit instruction on L2 development: A meta-analysis — https://onlinelibrary.wiley.com/doi/10.1111/flan.12726
- RS-59C42B630ECEE225: Meta-analysis of task repetition effects on second-language oral performance — https://www.sciencedirect.com/science/article/pii/S0346251X25002787

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/teach-the-stubborn-form-explicitly-then-force-it-into-production

---

## Teach the stubborn form explicitly, then force it into production

ID: MHC-D-RESEARCH-1123 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/teach-the-stubborn-form-explicitly-then-force-it-into-production

Recognition is a weak hiding place for an advanced form.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A grammar, pragmatic or discourse distinction is familiar when seen but unreliable when you must produce it.

### Avoid when

- Explicit instruction is not a universal substitute for communication and input. It is most useful when a particular distinction needs conscious repair or clarification.

### Explanation

Make the distinction explicit enough to choose between competing forms, then require production. Start with a constrained contrast, move to short original examples and finish with a realistic task where the form is useful but not announced in advance. The last step matters: the goal is not to recite the rule but to retrieve the form while attention is shared with meaning.

### Steps

1. State the contrast in plain language.
2. Choose between contrasting examples and explain why.
3. Create new examples from different topics.
4. Use the form in a message or spoken answer.
5. Check delayed spontaneous use.

### Example

After revising German word order in subordinate clauses, the learner explains a project delay without a worksheet telling them which clause type to use.

### Check

The target appears accurately in a new meaning-focused response without a visible rule prompt.

### Limits

- Explicit instruction is not a universal substitute for communication and input. It is most useful when a particular distinction needs conscious repair or clarification.

### Evidence and sources

- supports: A meta-analysis of explicit L2 instruction found moderate-to-large overall effects, with results moderated by instruction type, practice type, delivery mode, intensity and outcome measure. — RS-524A9FE621C70F48. The synthesis pools different languages, structures and instructional designs; no single explicit-teaching format is universally best. (Abstract)
- RS-524A9FE621C70F48: Effects of different forms of explicit instruction on L2 development: A meta-analysis — https://onlinelibrary.wiley.com/doi/10.1111/flan.12726

No review details supplied.
Related (use_before): https://vedokrok.com/knowledge/repeat-the-task-then-change-the-task

---

## Repeat the task, then change the task

ID: MHC-D-RESEARCH-1124 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/repeat-the-task-then-change-the-task

The second attempt can free capacity; the third should test transfer.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A speaking or writing task was difficult enough that one attempt mostly exposed planning and retrieval limits.

### Avoid when

- Too much identical repetition can turn retrieval into script memory. Use repetition to create capacity, then demand transfer.

### Explanation

Repeat a meaningful task after reviewing the first attempt so the learner can improve accuracy, complexity or fluency with less planning cost. Then change the prompt, audience, data or communicative purpose while preserving the target skill. Improvement only on the identical script is useful rehearsal evidence, not yet evidence of broader language control.

### Steps

1. Perform the task without a script.
2. Review one or two target gaps.
3. Repeat the same task once.
4. Change a meaningful condition.
5. Compare whether the improvement survives the changed task.

### Example

Explain a migration issue, repeat the explanation after feedback, then explain a different incident to a non-technical stakeholder.

### Check

At least one improved feature appears in the changed task, not only in the repeated version.

### Limits

- Too much identical repetition can turn retrieval into script memory. Use repetition to create capacity, then demand transfer.

### Evidence and sources

- supports: Task-repetition research finds that repeating a communicative task can improve dimensions of L2 performance, while the size and pattern of effects depend on the task and repetition design. — RS-59C42B630ECEE225. Immediate improvement on a repeated task can include task familiarity and does not by itself prove transfer to new topics. (Meta-analysis results)
- RS-59C42B630ECEE225: Meta-analysis of task repetition effects on second-language oral performance — https://www.sciencedirect.com/science/article/pii/S0346251X25002787

### Vedokrok editorial review

Use the repeat to buy capacity, then test what travels

The second attempt at a speaking or writing task is usually easier because the content, route and useful language are already partly available. That can reveal better language control, but it also mixes learning with task familiarity. A changed task is the more demanding check: does the improvement survive when the learner must build the message again?

#### The second performance is a different cognitive job

Task repetition has a solid research base, but the gains are not uniform. A 2025 meta-analysis of oral production found positive effects across complexity, accuracy, lexical complexity and fluency, with larger average gains for syntactic complexity and accuracy and smaller gains for fluency and lexical complexity. Effects also changed with task type, repetition type, learning context and spacing. For practice, this means the repeat can be useful precisely because it removes some planning pressure. It does not mean that every smoother second performance represents a general improvement in speaking skill.

#### Changing the content asks the harder question

Same-task repetition and procedural repetition test different things. In the same task, the learner can reuse content and recently activated language. In procedural repetition, the procedure stays similar while the content changes, so conceptualization and formulation have to restart. A 2024 classroom study with 46 ESL learners illustrates the difference: fluency improved across three performances of the same poster task, while no fluency change appeared in its parallel-task condition. That is not evidence against changed tasks. It shows why they are useful as a transfer check: they remove part of the familiarity advantage. A separate 2024 experiment with 116 Japanese learners combined same-task repetition within sessions with different narrative prompts across sessions and found gains on novel posttest prompts in some fluency measures.

#### Choose the dimension before adding feedback

The writing evidence makes the trade-off visible. One 2025 meta-analysis of 31 studies and 65 samples found positive effects across written accuracy, lexical complexity, fluency and syntactic complexity, but feedback especially benefited accuracy and was associated with a disadvantage for lexical complexity. A second 2025 meta-analysis of 17 independent studies and 442 participants estimated a much smaller overall writing effect (g = 0.25); fluency improved more than accuracy or lexical complexity, while the average effect on syntactic complexity was not significant. These syntheses differ in samples and models, so their effect sizes should not be treated as interchangeable. The practical lesson is simpler: decide whether the next round is meant to improve accuracy, range, organization or fluency, then review one or two gaps rather than correcting everything at once.

#### Three decisions before the next round

How many times should the identical task be repeated? There is no universal count; one repeat can be enough to expose headroom, and more repetition can increasingly reward memory for that task. How much should the next task change? Keep the target skill and general procedure, but change a meaningful condition such as topic, audience, data or communicative purpose. What if the repeat improves but the changed task does not? Treat the first gain as useful rehearsal evidence, not failed learning. Keep the improved feature as the next target, reduce the change if necessary, and test again later instead of claiming transfer too early.

**Editorial takeaway:** Repeat a meaningful task once to reduce planning pressure and reveal what the learner can improve with more capacity. Review one or two target gaps, then change a real condition while preserving the skill. Improvement that survives the changed task is stronger evidence of usable language control than a polished rerun of the same script.

#### Editorial evidence

- supports: A 2025 meta-analysis found positive task-repetition effects across oral CALF outcomes, with larger average gains for syntactic complexity and accuracy than for fluency and lexical complexity, and with effects moderated by learning context, task type, repetition type and spacing. — abdi-tabari-zhuang-farahanynia-2025-oral. This supports task repetition as a useful practice condition while showing that the outcome depends on what is repeated and how performance is measured. (Abstract)
- contextualizes: A 2025 meta-analysis of 31 studies and 65 unique samples found positive effects on written accuracy, lexical complexity, fluency and syntactic complexity; feedback moderated the pattern, with an advantage for accuracy and a disadvantage for lexical complexity. — abdi-tabari-zhuang-farahanynia-2025-writing. The result supports selecting a target dimension before adding feedback instead of assuming that feedback plus repetition moves all aspects of performance together. (Abstract; moderator analysis)
- challenges: A second 2025 writing meta-analysis covering 17 independent studies and 442 participants estimated an overall effect of g = 0.25, with g = 0.51 for fluency, g = 0.24 for accuracy and lexical complexity, and a non-significant g = 0.06 for syntactic complexity. — liu-tang-2025-writing. The smaller and uneven estimates challenge any simple claim that repetition reliably produces large gains across all dimensions of writing. (Abstract)
- limits: In a classroom study with 46 ESL learners, fluency improved across three performances of the same poster task, while the parallel-task condition using three different posters showed no fluency change. — hunter-2024-poster-carousel. This shows that improvement on an identical task can be easier to obtain than improvement when content changes. The study used one specific classroom activity, so it is not a general verdict against procedural repetition. (Abstract; discussion of parallel-task repetition)
- contextualizes: A 2024 experiment with 116 Japanese L2 English learners used three same-task repetitions within each training session and different narrative prompts across four sessions; posttests used novel prompts, and the training improved some fluency measures regardless of whether sessions were spaced by one or seven days. — kakitani-kormos-2024-distributed. The design combines immediate same-task repetition with variation across sessions and therefore supports a repeat-then-vary sequence. It does not isolate the contribution of each component, and not every fluency measure improved. (Abstract; training procedure; pretest-posttest results)
- abdi-tabari-zhuang-farahanynia-2025-oral: Task repetition and L2 oral performance: A meta-analysis — https://doi.org/10.1016/j.system.2025.103868 — Mahmoud Abdi Tabari, Jingyuan Zhuang, Mahsa Farahanynia
- abdi-tabari-zhuang-farahanynia-2025-writing: Task repetition and L2 written performance: A meta-analysis — https://doi.org/10.1016/j.jslw.2025.101255 — Mahmoud Abdi Tabari, Jingyuan Zhuang, Mahsa Farahanynia
- liu-tang-2025-writing: The effect of task repetition on L2 learners’ writing performance: A meta-analysis — https://doi.org/10.1177/13621688251381268 — Junge Liu, Yinjie Tang
- hunter-2024-poster-carousel: The Poster Carousel in the ESL Classroom: What Happens to Learners' L2 Fluency During Same and Parallel-Task Repetition? — https://doi.org/10.1002/tesq.3257 — Ann-Marie Hunter
- kakitani-kormos-2024-distributed: The effects of distributed practice on second language fluency development — https://doi.org/10.1017/S0272263124000251 — Joe Kakitani, Judit Kormos

Editorial version: 1.0.0 · evidence checked 2026-09-27 · Agent editorial review; not human or expert certification.

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/reperform-after-feedback-instead-of-collecting-corrections

---

## Train discourse functions, not only correct sentences

ID: MHC-D-RESEARCH-1125 · Version: 0.1.0 · Kind: checklist
Source: https://vedokrok.com/knowledge/train-discourse-functions-not-only-correct-sentences

C1 lives partly between the sentences.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- Individual sentences are mostly correct but long answers still sound jumpy, flat or difficult to follow.

### Avoid when

- More connectors do not automatically create coherent discourse. Each signal must express a real relationship in the message.

### Explanation

Practise the language that organizes interaction and extended discourse: entering a turn, qualifying a claim, contrasting options, returning to the main point, giving an example, correcting yourself, signalling uncertainty and closing. Build several natural options for each high-value function and use them inside longer tasks. The target is controlled choice, not a memorized parade of connectors.

### Checklist

- Open or take the floor naturally.
- Signal the main point and structure.
- Add contrast, cause or qualification where needed.
- Repair or reformulate without losing the thread.
- Close or hand over the turn clearly.

### Example

Instead of adding 'moreover' to every paragraph, a learner practises several ways to qualify a recommendation and return from a technical detail to the decision.

### Check

A long response remains easy to follow even when the topic changes or a clarification interrupts it.

### Limits

- More connectors do not automatically create coherent discourse. Each signal must express a real relationship in the message.

### Evidence and sources

- supports: CEFR descriptors distinguish C1 from B2 through broader linguistic range, higher grammatical control, more effortless fluency, a larger repertoire of discourse functions and more controlled coherence. — RS-F53492D1107B1940. CEFR describes performance; it does not prescribe a training method or a fixed number of study hours. (B2 and C1 rows)
- RS-F53492D1107B1940: Qualitative aspects of spoken language use - CEFR Table 3 — https://www.coe.int/en/web/common-european-framework-reference-languages/table-3-cefr-3.3-common-reference-levels-qualitative-aspects-of-spoken-language-use

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/build-phrase-range-around-communicative-functions

---

## Reperform after feedback instead of collecting corrections

ID: MHC-D-RESEARCH-1126 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/reperform-after-feedback-instead-of-collecting-corrections

A correction becomes learning only when it changes the next performance.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- Feedback produces a list of errors but those corrections rarely appear in later spontaneous language.

### Avoid when

- Immediate re-performance can show understanding and short-term control; durable transfer requires a delayed check.

### Explanation

After focused feedback, make the learner perform again while the intended change is still clear. Keep the feedback narrow enough to use: one or two high-value targets, not every possible defect. The re-performance can be a revised paragraph, a second spoken attempt or a parallel task. Later, test the same target again without showing the correction.

### Steps

1. Select one or two consequential feedback targets.
2. Explain or confirm the needed change.
3. Reperform or revise immediately.
4. Remove the correction from view.
5. Retest the target in a later parallel task.

### Example

After feedback on hedging in a recommendation, the learner records the answer again, then handles a different recommendation two days later.

### Check

The corrected behavior appears in a later task without copying the teacher's wording.

### Limits

- Immediate re-performance can show understanding and short-term control; durable transfer requires a delayed check.

### Evidence and sources

- supports: Task-repetition research finds that repeating a communicative task can improve dimensions of L2 performance, while the size and pattern of effects depend on the task and repetition design. — RS-59C42B630ECEE225. Immediate improvement on a repeated task can include task familiarity and does not by itself prove transfer to new topics. (Meta-analysis results)
- supports: Second-language writing feedback is more useful when the feedback level matches the intended outcome: surface feedback mainly targets form, while deeper feedback targets organization and meaning. — RS-32A66B6FDD14730B. This evidence concerns writing and should be adapted cautiously when designing oral feedback. (Meta-analytic comparison of surface, deep and combined feedback)
- RS-59C42B630ECEE225: Meta-analysis of task repetition effects on second-language oral performance — https://www.sciencedirect.com/science/article/pii/S0346251X25002787
- RS-32A66B6FDD14730B: Effects of feedback on second language writing: A meta-analysis — https://doi.org/10.1016/j.learninstruc.2024.101961

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/define-the-b2-to-c1-gap-from-a-real-sample

---

## Build phrase range around communicative functions

ID: MHC-D-RESEARCH-1127 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/build-phrase-range-around-communicative-functions

Advanced fluency needs ready language, not only a bigger dictionary.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- Speech is grammatically acceptable but repeatedly assembles ideas word by word and relies on a small set of safe expressions.

### Avoid when

- Formulaic language supports fluent processing but can sound unnatural when copied without register, meaning or context. Phrase accumulation is not a substitute for syntax or interaction skill.

### Explanation

Collect multiword sequences around functions you repeatedly perform: clarify, qualify, disagree, compare, escalate, summarize, buy thinking time or hand over a turn. Learn several alternatives with their typical context and register, then retrieve them inside new messages. Prefer expressions observed in credible native or expert use over invented 'advanced' phrases.

### Steps

1. The learner can choose among several appropriate phrases without first translating a full sentence from another language.

### Example

For cautious disagreement, the learner collects three natural options for meetings, then uses whichever fits three different stakeholder scenarios.

### Check

The learner can choose among several appropriate phrases without first translating a full sentence from another language.

### Limits

- Formulaic language supports fluent processing but can sound unnatural when copied without register, meaning or context. Phrase accumulation is not a substitute for syntax or interaction skill.

### Evidence and sources

- supports: CEFR descriptors distinguish C1 from B2 through broader linguistic range, higher grammatical control, more effortless fluency, a larger repertoire of discourse functions and more controlled coherence. — RS-F53492D1107B1940. CEFR describes performance; it does not prescribe a training method or a fixed number of study hours. (B2 and C1 rows)
- supports: In an L2 speech study, multiword-sequence use contributed to perceived fluency alongside speed, breakdown and repair measures. — RS-391FE6BA5EEECA6B. The study is correlational and population-specific; phrase use is one part of fluent performance, not a sufficient cause. (Abstract)
- RS-F53492D1107B1940: Qualitative aspects of spoken language use - CEFR Table 3 — https://www.coe.int/en/web/common-european-framework-reference-languages/table-3-cefr-3.3-common-reference-levels-qualitative-aspects-of-spoken-language-use
- RS-391FE6BA5EEECA6B: The role of multiword sequences in fluent speech — https://www.cambridge.org/core/journals/studies-in-second-language-acquisition/article/role-of-multiword-sequences-in-fluent-speech/FAB18835B2117A299C33F8F502248C81

### Vedokrok editorial review

Build a repertoire for a function, not a pile of phrases

Ready-made language can reduce some of the work of speaking, but a larger phrase list is not the same as greater fluency. A more useful target is a small repertoire for a recurring communicative job—qualifying, disagreeing, clarifying or handing over a turn—with enough context to choose an option rather than recite it.

#### A phrase earns its place by doing a job

Organize useful multiword language around situations you actually meet, not around labels such as “advanced expressions.” For one function, collect a few alternatives from credible real use and record the difference that changes the choice: directness, register, strength, grammar pattern or typical partner words. The number is not the goal. The point is to move from one safe expression to a choice among forms that fit different situations. This also makes review more diagnostic: the cue can be the communicative intention, not the first word of a memorized sentence.

#### Ready language can help, but the fluency effect is selective

There is good evidence that collocations can be taught. Li and Lei’s meta-analysis covered 17 primary studies and 3,859 participants and found a large average benefit for collocation instruction, with larger effects for explicit than implicit approaches. That result is about learning collocations, not proof that memorizing more phrases causes fluent conversation. Speech studies give a narrower picture. Yan found that formulaic sequences reduced pauses in an elicited-imitation task but did not increase speech rate; the effect also changed with proficiency and sentence length. A plausible benefit is therefore lower formulation load in some moments, not a general fluency upgrade.

#### Phrase counts are a poor finish line

The relationship between multiword sequences and speaking changes with the task. Tavakoli and Uchihara found several associations with articulation rate, pausing and repair in monologic test speech, but they also observed task-specific sequences borrowed from the prompt; C1 speakers varied those sequences more than lower-level speakers. In a 2025 argumentative-speech study, bigram proportion added only 0.8% to explained perceived-fluency variance after direct utterance-fluency measures were already in the model. A 2026 dialogue study went further: none of its bigram or trigram measures predicted proficiency, while the speaking pair explained substantial variation in phrase use. Frequent chunks can be useful resources, but counting them is not a reliable proxy for how well a person handles an interaction.

#### Practice the choice, then change the situation

Suppose the target function is cautious disagreement. Instead of rehearsing one polished line, keep several forms that differ in strength or setting, then answer short scenarios: a peer proposes a weak option, a manager asks for a risk view, or a customer makes an incorrect assumption. Remove the phrase list and require a choice. Later change the topic or interlocutor. Success is not verbatim recall; it is selecting a natural option quickly enough, adapting it when needed and keeping the conversation moving. If the same phrase appears mechanically in every setting, the repertoire has not yet become flexible.

#### Questions before adding another phrase

Do I need more idioms to sound advanced? Usually not: frequency, fit and control matter more than collecting conspicuous expressions. How many alternatives should I learn for one function? Research does not give a magic number; start with a small set that creates a real choice and add another only when a recurring situation exposes a gap. What if I can recognize every phrase but never use one spontaneously? Change the prompt from phrase → meaning to function → response, practise without the list, and retest in a new scenario. Can formulaic language make speech sound scripted? Yes, when wording is copied without regard to register, meaning or the other speaker; variation and context are part of the skill.

**Editorial takeaway:** Keep the useful idea, but make the unit of learning a communicative function with several context-sensitive options. Deliberate phrase learning can strengthen collocational knowledge and may reduce some formulation effort, yet phrase frequency is not a general measure of fluency or proficiency. Test whether the learner can choose and adapt an expression in a changed situation, especially in dialogue.

#### Editorial evidence

- supports: A meta-analysis of 17 primary studies with 3,859 participants found a large average effect of collocation instruction on second-language collocation learning, with explicit instruction producing larger effects than implicit instruction. — li-lei-2024-collocation-meta. This supports deliberate work on useful word combinations. The outcome is collocation learning, not a direct causal estimate for spontaneous speaking fluency or dialogue performance. (Abstract)
- contextualizes: In 56 learners from Low B1 to C1, several multiword-sequence measures were associated with articulation rate, pausing or repair, and task-specific n-grams from the prompt appeared in performance. — tavakoli-uchihara-2020-mws-fluency. The findings support a relationship between phrase use and dimensions of monologic oral fluency, while the prompt-borrowing result shows that task design can influence the measured phrase profile. (Abstract)
- limits: In an elicited-imitation study with 269 first- and second-language speakers, formulaic sequences reduced pauses but did not increase speech rate; effects on pausing varied with proficiency and sentence length. — yan-2020-formulaic-fluency. The result argues against treating formulaic language as a single general fluency mechanism. It also comes from elicited imitation rather than free dialogue. (Abstract)
- limits: For 102 Japanese speakers giving argumentative speeches, bigram proportion explained only 0.8% additional variance in perceived fluency after utterance-fluency predictors had already explained much of the rating variance. — takizawa-suzuki-2025-perceived-fluency. Multiword sequences had a small systematic association with perceived fluency in this task, but direct temporal fluency measures carried far more of the model. (Abstract)
- challenges: In dialogic performances from 127 test-takers across six proficiency levels, none of the bigram or trigram measures was associated with speaking proficiency, while pair-level effects explained substantial variation in sequence proportions. — uchihara-et-al-2026-dialogue. This directly challenges using multiword-sequence frequency as a general marker of speaking proficiency and shows that interlocutor pairing can shape phrase use in dialogue. (Abstract)
- li-lei-2024-collocation-meta: Effectiveness of second language collocation instruction: a meta-analysis — https://doi.org/10.1515/iral-2021-0218 — Xin Li, Lei Lei
- tavakoli-uchihara-2020-mws-fluency: To What Extent Are Multiword Sequences Associated With Oral Fluency? — https://doi.org/10.1111/lang.12384 — Parvaneh Tavakoli, Takumi Uchihara
- yan-2020-formulaic-fluency: Unpacking the Relationship Between Formulaic Sequences and Speech Fluency on Elicited Imitation Tasks: Proficiency Level, Sentence Length, and Fluency Dimensions — https://doi.org/10.1002/tesq.556 — Xun Yan
- takizawa-suzuki-2025-perceived-fluency: The role of multiword sequences in fluent speech: The case of listener-based judgment in L2 argumentative speech — https://doi.org/10.1017/S0272263125000051 — Kotaro Takizawa, Shungo Suzuki
- uchihara-et-al-2026-dialogue: Multiword sequences do not predict speaking proficiency in dialogue: A pair-level analysis — https://doi.org/10.1111/modl.70064 — Takumi Uchihara, Parveneh Tavakoli, Svetlana Mazhurnaya, Phil Smyth

Editorial version: 1.0.0 · evidence checked 2026-09-27 · Agent editorial review; not human or expert certification.

No review details supplied.
