Input builds familiarity. Output tests capability.
Merrill Swain’s Output Hypothesis says that producing language forces deeper processing and reveals what is missing.
That is why breakdown moments are high-value learning data.
What a real breakdown looks like
A breakdown is not only a grammar error.
It can be:
- Lexical missing: you cannot find the word.
- Precision collapse: you use vague language instead of exact meaning.
- Register mismatch: tone is wrong for the context.
- Collocation failure: words are correct alone but unnatural together.
The practical loop
- Write or speak without stopping.
- Mark one breakdown point.
- Repair the sentence to match your real intent.
- Retrieve it later under time pressure.
One repaired point per session is enough for steady progress.
Why this beats over-correction
If you stop at every error, fluency collapses.
If you ignore errors, precision stalls.
The right sequence is fluency first, repair second.