How to Fix Fossilised Errors in Long-Term Students
You fix fossilised errors by treating them as habits rather than knowledge gaps: record the student, audit their five most frequent mistakes, then attack one error at a time for two to three weeks through noticing drills, self-correction and production pressure. Correcting everything at once fails, because fluent speech always reverts to the strongest stored habit. These errors survive years of lessons precisely because they have been reinforced thousands of times — so they need thousands of correct repetitions, focused and deliberate, to displace.
What is a fossilised error?
A fossilised error is a recurring mistake that has survived years of exposure and correction because it has been produced, understood and socially rewarded thousands of times. "I am agree." A missing third-person -s in every anecdote. "He explained me the problem." The Spanish speaker's dropped articles, the Italian speaker's "people is". The student usually knows the rule — they can ace a gap-fill on it — but under real-time pressure their automatic system outputs the old pattern, and their explicit knowledge never gets a vote.
That is the defining feature: fossilised errors are not ignorance, they are competition. The wrong form is stored as the default, retrieved faster than the right one, and every fluent use strengthens it further. The term comes from Selinker's interlanguage work in the 1970s; the practical meaning for you is simpler — this error has thousands of rehearsals behind it, and one correction per lesson will never outbid that.
Why does fluency practice cement these errors?
Fluency practice cements errors because automaticity rewards whatever is already stored, right or wrong. Every time the student speaks fluently, they rehearse their habits at speed. If the message lands — and it usually does, because listeners are forgiving — the error receives no negative feedback, and the habit strengthens.
Communicative teaching made this worse by accident. When the goal is getting the message across, "I am agree" succeeds: the listener understands, the conversation continues, and the error just earned itself another repetition. Multiply that across five years of twice-weekly lessons and you have a groove, not a slip.
This is also why long-term students hit the intermediate plateau and stay there: their fluent output is fluent, their known rules are intact, and the gap between the two never closes because practice keeps feeding the wrong side.
How do you diagnose which errors to attack first?
Diagnosis means recording real, unscripted speech and auditing it for the errors that appear most often and cost the student most. You are not hunting every mistake. You are looking for the top five, from which you will pick one.
The procedure:
- Record ten minutes of unscripted talk. Warm storytelling — "tell me about your week", a picture description, a rant about their commute. Nothing prepared, nothing read aloud.
- Transcribe a sample. Even two minutes transcribed by hand reveals the patterns. If your platform records and transcribes lessons — Tuton's feature set includes classroom recording and transcripts — the evidence is already sitting there.
- Tally recurring errors. Count, don't impression. "'He don't' appeared nine times" is a fact; "his grammar is a bit off" is not.
- Rank by frequency, damage and willingness. Frequency: occurrences per 100 words. Damage: does it block understanding, or mark the student as less fluent than they are? Willingness: does the student actually want it fixed? A student who does not care about articles will not do article drills, no matter how important articles are.
- Pick one. One error, one burst, then the next.
A realistic top five for a Spanish-speaking B2: missing third-person -s, "explain me the problem" instead of "explain to me", "actually" used for "currently", dropped articles before singular countable nouns, and a /b/–/v/ merger. Start with frequency: the error they make twenty times a day pays back fastest.
What is a correction burst?
A correction burst is two to three weeks in which one error becomes the student's personal project, attacked from a different angle each week while everything else is deliberately left alone. It runs in three stages, and the order matters: noticing, then self-correction, then production pressure.
- Week 1 — noticing. The student learns to catch the error, not yet to fix it. They highlight every instance in their own transcripts, tally occurrences, and hunt the pattern in podcasts and articles. The goal is a brain that flags the error automatically, because self-monitoring is the only correction that scales to real life without you in the room.
- Week 2 — self-correction. You stop correcting entirely. The student repairs their own slips within five seconds of making them, helped by an agreed cue — a raised finger, a word in the chat. Slow is fine; the constraint is that their own system does the repair.
- Week 3 — production pressure. Tasks that force the pattern repeatedly at rising speed: retell a story entirely in the third person, describe a picture stuffed with countable nouns, explain a schedule using "currently" and "at the moment". Speed climbs across the week, and accuracy must hold as it climbs. That is the whole test.
Keep the burst visible between lessons. A simple tally sheet the student fills in themselves, a shared document where you both log the weekly count, a named goal on the lesson plan — the mechanics matter less than the fact that the target is written down and reviewed every single lesson. Bursts that live only in the teacher's head quietly evaporate by week two.
Then measure. Record the same three-minute task you used for the baseline, count errors per 100 words, and compare. The error counts as gone when it stays at roughly one per 100 words or fewer across two consecutive recordings on pressure tasks. Only then do you open the next burst on error number two.
Which burst technique fits which error?
Different habits need different drills. Match the burst to the error:
| Error type | Burst technique | How you know it's gone |
|---|---|---|
| Missing third-person -s | Two weeks of third-person retells — describing a friend's daily routine; the student tallies every verb as they speak | Zero misses in a three-minute retell, twice in a row |
| Dropped articles | Picture description dense with countable nouns; a sort drill of a / an / the / zero before speaking; a final "article check" pass | One slip or fewer per 100 words in a timed description |
| False friend ("actually" for "currently") | Meaning-contrast work: twenty sentences choosing between the two words, then using both correctly in free talk | All twenty correct in a spaced quiz, plus correct use in unrehearsed conversation |
| "I am agree" verb-structure errors | Contrast cards with the correct form; self-correction on cue within five seconds, cue withdrawn in week three | Self-corrects without any cue by the end of the burst |
| /b/–/v/ merger | Minimal-pair listening first — they must hear the difference before they can say it — then shadowing pairs, then words under pressure | Produces and distinguishes the pair in a recorded word list and in free speech |
How do you keep the rest of the lesson alive during a burst?
The rest of the lesson stays normal. The burst is a lens, not a curriculum: you still read, discuss, argue and laugh about whatever the week's topic is. You simply funnel any natural opening toward the target pattern during the production stage.
What you must not do is build every lesson around the error. A student who spends three weeks talking about nothing but their own mistakes stops enjoying lessons, and enjoyment is what keeps them booking. One drill, one measured task, one tally — then back to real English.
Why does correcting everything at once fail?
Full-spectrum correction fails because working memory cannot monitor five new habits while speaking. Give a student five targets and one of three things happens: their fluency collapses as they check themselves on every clause, their motivation collapses along with their fluency, or they quietly stop hearing your corrections at all. Correction fatigue is earned, and it is expensive to reverse.
One error at a time sounds slow until you do the arithmetic. A burst every three weeks is roughly seventeen focused errors a year, each one actually displaced rather than nagged at. "Fix everything" fixes nothing; "fix one thing per burst" clears the list in about eighteen months. This is also the honest answer to the student who asks why you have stopped correcting their prepositions: you have not given up, you have scheduled them. For the in-the-moment mechanics of when and how to correct, see error correction in online lessons.
How do you handle "this is just how I talk"?
Identity resistance is real, and after five years of lessons an error can feel like part of the student's voice. Attack it bluntly and it feels like an attack on them, which kills the burst before week one ends.
Separate identity from intelligibility, and let the evidence argue. Play their baseline recording, then a clean re-record of the same story with the error fixed, and ask which version they want in a job interview. Offer the choice explicitly: "Do you want this on the list?" A student who chooses their own target will do the noticing homework; a student who was assigned it will skip it.
And accept that some students genuinely do not care. "This is how I talk" is sometimes a settled decision rather than resistance — plenty of successful professionals live comfortably with a fossil or two, and it costs them nothing. Agree that the error is parked, keep it off your correction list, and protect the relationship. Not every fossil is worth mining. The ones that block understanding or undercut the student's credibility in meetings usually are.
Frequently asked questions
What is a fossilised error?
A recurring mistake that has survived years of lessons because it has been repeated fluently thousands of times. The student usually knows the rule but cannot beat the old habit in real time.
How long does it take to fix one?
Two to three weeks of focused work for most grammar and vocabulary habits, longer for pronunciation. The error is gone when it stays near zero under pressure, not just in careful speech.
Why can't I just correct every mistake as it happens?
Because nobody can monitor five new habits while speaking. Fluency and motivation drop, or the student tunes out. One error per burst gets fixed; everything at once fixes nothing.
What if the student refuses to change?
That is their call. Show the evidence, ask whether they want the error on the list, and accept a genuine no. Some errors cost the student nothing and are not worth the fight.
Can old errors come back?
Yes — under stress, fatigue or after a long break. Record a short audit once a quarter and run a refresher burst on anything that has crept back past one error per 100 words.