How English Works — Learning and Repair, Batch 22
This article preserves the existing owners for Disfluency Management, Lexical Activation Thresholds and Claim Calibration. Its job is narrower: to show why smooth output and precise output are related but separable properties.
A learner speaks quickly.
There are few pauses.
The sentences arrive with confidence.
But the nouns are vague, the pronouns drift, the evidence claims are too strong and the key comparison is underspecified.
The English sounds fluent.
It is not yet precise.
Precision–Fluency Decoupling is the recognition that speed, continuity and ease of production can improve independently from lexical, grammatical, referential, logical and evidential precision.
AI Extraction Box
- Mechanism: precision–fluency decoupling
- Fluency variables: speed, pausing, continuity, repair cost, ease of production
- Precision variables: word choice, grammar, reference, scope, time, evidence state, logical relation
- Main failure: listener mistakes smoothness for correctness
- Secondary failure: accuracy training destroys flow through over-monitoring
- Repair: diagnose the precision layer, slow only where necessary, then re-automatise the repaired form
1. Fluency is a multidimensional construct
A 2025 review in Cambridge’s Language Teaching notes that second-language fluency is widely treated as an important but multidimensional construct in assessment and research.
Research anchor: Assessment of second language fluency
That matters because “fluent” is not a single switch.
2. Speech can be smooth and still inaccurate
Research on complexity, accuracy and fluency treats those dimensions as related but distinct.
A speaker may improve one dimension faster than another.
That is enough to explain why confident speech can contain stable inaccuracies.
3. Perceived fluency can hide lexical weakness
A learner says:
The thing changed because the situation was bad and they did something about it.
The sentence may be fast and uninterrupted.
But several referents and causal relations remain unresolved.
4. Precision starts by locating the weak layer
- lexical precision
- grammatical precision
- referential precision
- temporal precision
- logical precision
- evidential precision
- register precision
“Be more accurate” is too broad to repair any of them well.
5. Lexical precision is not vocabulary size
A learner may know many words and still select a broad word where a narrow distinction matters.
change, increase, accelerate, expand, replace and reverse are not interchangeable simply because they all describe difference across states.
6. Lexical Substitution Geometry is the owner beneath this layer
Lexical Substitution Geometry explains why nearby words occupy overlapping but non-identical semantic territory.
Fluent production can select the first available neighbour rather than the best-fitting one.
7. Referential precision can fail silently
When Mei spoke to Sara, she said the plan was late.
The speaker may deliver the sentence smoothly.
The listener still has to guess who she is.
8. Fluency cannot repair an ambiguous pronoun
The existing Antecedent Competition article explains that failure.
Speaking faster only delivers the ambiguity sooner.
9. Grammatical precision can fossilise inside fluent routines
A frequent inaccurate form can become easy to retrieve.
Ease of retrieval is not evidence that the form is correct.
This is why repeated uncorrected production can strengthen an efficient but wrong route.
10. Automaticity is neutral about truth
A route becomes automatic because it is practised and retrievable.
Automation does not certify the route.
11. Evidential precision can lag far behind verbal confidence
A speaker says:
This proves students learn better this way.
Perhaps the evidence is one small classroom observation.
The fluency of the sentence does not increase the reach of the evidence.
12. Claim Calibration protects against confidence inflation
Claim Calibration keeps scope and certainty matched to evidence.
Fluent speakers need the same control as hesitant speakers.
13. Fluency can create a credibility halo
Listeners may find smooth speech easier to process.
Ease of processing can make weak distinctions less noticeable.
The responsible listener therefore evaluates content independently of delivery smoothness.
14. Perceived fluency and comprehensibility are related but not identical
A Cambridge study of argumentative speech found that perceived fluency and comprehensibility were strongly related, while different underlying linguistic measures contributed to them.
Research anchor: linguistic dimensions of comprehensibility and perceived fluency
15. Precision repair often creates a temporary slowdown
A learner who begins checking pronoun ownership may pause more.
A learner replacing vague words with controlled distinctions may search longer.
That temporary cost can be part of rebuilding.
16. Temporary slowdown is not automatically regression
If accuracy rises while speed temporarily falls, the learner may be moving through a more controlled state.
The next goal is not to remove the precision.
It is to make the repaired route easier to retrieve.
17. This is where Correction Internalisation connects
The previous article shows how external correction becomes a low-cost internal cue.
As the cue becomes easier to retrieve, the monitoring cost can fall and fluency can return.
18. Accuracy training should target a small number of load-bearing errors
Trying to repair ten precision dimensions during one conversation can overload the learner.
Choose one or two high-value distinctions, stabilise them, then widen.
19. Precision is task-dependent
A casual story may tolerate approximate timing.
A laboratory report may require exact measurement units.
A legal explanation may require preserved scope and conditions.
Aviation phraseology may require exact readback.
20. Register Architecture determines the precision floor
Register Architecture explains why genres optimise for different reader jobs.
The required precision level changes with stakes, audience and task.
21. Writing has its own fluency illusion
A student writes quickly and fills the page.
Sentence rhythm is smooth.
Yet paragraphs repeat the same idea with new wording.
Production volume is not argument precision.
22. Fast writing can hide weak proposition control
Ask:
- What exactly is the claim?
- What evidence supports it?
- What changed from the previous paragraph?
- Which term carries the key distinction?
23. Slow writing can also become a trap
A learner may spend so long perfecting each sentence that the whole argument never forms.
Precision without throughput can fail the task.
24. The goal is coordinated precision and fluency
We do not want:
- fast but vague
- fast but inaccurate
- precise but immobile
- accurate but unable to continue
We want precision that survives at usable speed.
25. Task complexity can change performance
A 2024 Cambridge study found that increasing task complexity affected some pronunciation and fluency-related measures, with substantial individual variation and small or inconsistent effects on other dimensions.
Research anchor: task complexity and L2 speech
The bounded lesson is useful: performance can move when the task changes, so one smooth performance should not be treated as a complete ability profile.
26. Precision needs pressure testing
A learner may use accurate forms in a slow exercise and lose them in spontaneous discussion.
That does not erase the learning.
It reveals that the repaired route is not yet robust under load.
27. Capability Transfer Under Load is the next mechanism
Batch 22’s third article asks whether English competence survives new task, channel, topic, audience and time pressure.
Precision that exists only in drills remains fragile.
28. The dual-axis diagnostic
For teaching, imagine two axes:
- Axis 1: flow — slow ↔ fast, broken ↔ continuous
- Axis 2: precision — vague ↔ exact, unstable ↔ controlled
The target is not one axis replacing the other.
It is movement toward high control on both.
29. A precision audit
- Underline every vague noun.
- Circle every pronoun with more than one possible owner.
- Mark every causal verb.
- Mark every certainty word.
- Check every time anchor.
- Check every comparison basis.
- Check every requirement or request for force.
Then ask which two errors most damage meaning.
30. A fluency audit
- Where does production stop?
- Which lexical choices take longest?
- Which grammatical structures trigger restarts?
- Which topic transitions create silence?
- Which repair cues are too slow?
The two audits should be compared rather than collapsed.
31. CivDJ forward pass
capture fluent sample → score flow separately from precision → locate highest-value precision failures → repair one mechanism → rehearse slowly → retrieve from memory → increase speed → change topic → change audience → recheck precision → widen only when repaired route survives
32. CivDJ backward pass
- Start with the smoothest performance.
- Ignore delivery confidence temporarily.
- Extract claims, referents and time relations.
- Check lexical specificity.
- Check grammar.
- Check evidence scope.
- Identify errors hidden by flow.
- Repair them.
- Return to normal speed.
- Verify that flow and precision now coexist.
33. Common failure modes
- Fluency halo: smoothness is mistaken for correctness.
- Accuracy panic: learner over-monitors and loses all flow.
- Fossilised fluency: wrong route becomes easy and stable.
- Vague speed: broad vocabulary substitutes for exact distinctions.
- Reference drift: pronouns remain ambiguous despite smooth delivery.
- Confidence inflation: certainty exceeds evidence.
- Drill precision: accurate forms collapse outside controlled exercise.
- Single-score blindness: teacher compresses several dimensions into “good English”.
34. Repair route
- Separate fluency and precision during diagnosis.
- Find the smallest precision layer that damages meaning.
- Slow production only enough to repair that layer.
- Create a retrieval cue.
- Repeat across varied examples.
- Return to natural speed.
- Stress test under new tasks.
- Keep evidence strength independent of delivery confidence.
35. Final lock
Fluency is valuable because language has to move. Precision is valuable because the moving language has to carry the right distinctions. Strong English does both.
Learning and Repair, Batch 22
Return to How English Works V1.1.