Two people start the same language the same week. Same teacher, same app, same two hundred hours logged by spring. One of them ends up trading jokes with strangers. The other is still making the exact handful of mistakes she made in month one, the ones everybody understood her through just fine.
Here's the strange part: the learner who got stuck had the easier ride. People always caught her meaning, nodded, moved on. Nobody ever made her fix anything. The errors that hardened into permanent fixtures were the ones that worked. That isn't bad luck or a missing talent gene. It's a feature of the hidden system every learner is quietly running, the one linguists call your interlanguage (Selinker, 1972): not your first language, not the language you're aiming at, but a third language wedged in between, with its own private grammar that you built and never noticed. Those repeated mistakes you file under "I'm sloppy" aren't noise. They follow rules. Your rules.
Once you can see the system underneath your mistakes, you can aim your practice at the actual rule instead of flinching at the symptom. Below we'll unpack what interlanguage is, what the research shows about how it grows, and the practical tips to overcome interlanguage errors that actually move you forward. We understand the system so we can work with it instead of fighting it.
What Is Interlanguage?
Interlanguage is the language system a learner actually produces at any given stage of learning: not their first language, not the target language, but a third system in between, with its own internal rules. Those rules generate consistent, predictable errors rather than random ones. The term was coined by Selinker (1972), who argued that the language between your native tongue and your goal is coherent enough to study in its own right.
The word landed in 1972, when Selinker (in a paper in the journal IRAL) made a claim that sounded almost backwards at the time: the broken-sounding language a learner produces isn't a degraded copy of the target. It's a genuine system. It has rules the learner follows consistently, even when those rules don't match the language they're aiming for. Every learner has one. It shifts as you learn. And, as we'll see, it can also get stuck.
That reframing matters because of what it does to the feeling of making mistakes. If your errors were random, there would be nothing to do but try harder and hope. If they come from a system, they're readable. They tell you which rule you're currently running, which is why second language acquisition research treats learner errors as evidence, not noise.
Is interlanguage the same as "broken" language?
No, and the difference is the whole point. "Broken" implies something defective, a smashed-up version of the real thing. Interlanguage is more like a house built to its own blueprint: not the architect's plan, but standing, weatherproof, and fully lived in. It does its job, communication, using rules that are internally consistent even where they're wrong by target-language standards.
Take a classic example. English learners across the world routinely drop the third-person -s: "she go," "he want." That lonely little -s carries almost no meaning, and the brain knows it. The dropping isn't forgetfulness in the moment. It's because their interlanguage hasn't yet built in that rule, so the system that's actually running produces the form without it, every time, predictably. The error is the output of a rule, not the absence of one.
This is why it helps to separate two very different things. A one-off production failure, where you know the form, you've used it correctly before, and you just garbled it because you were tired, is the kind of thing that fixes itself when you slow down. A systematic interlanguage error reappears across different sentences and contexts because the rule behind it is consistent. The first is a small, normal stumble. The second is a signal worth acting on.
What makes interlanguage "systematic"?
The strongest evidence that interlanguage is a real system, and not just a personal collection of bad habits, comes from a finding that surprised researchers: learners with completely different first languages tend to make the same errors in the same target language, in roughly the same order.
In one landmark study, Dulay and Burt (1974) tracked 55 Chinese-speaking and 60 Spanish-speaking children learning English and looked at how they acquired a set of common grammar features (grammatical morphemes, things like plural -s, articles, and verb endings). Spanish and Chinese share almost nothing structurally. If errors came purely from each learner's native language, the two groups should have stumbled in different places. They didn't. Both groups acquired those features in nearly the same order.
That result, repeated many times since, is hard to explain unless something universal is steering the process. Picture two drivers on the same rough road. One shows up in a Spanish car, the other in a Mandarin one, totally different machines, and they still hit the same potholes in the same order. Your first language matters, but it isn't the only author of your interlanguage. There's a shared developmental logic underneath, which is exactly why a Japanese speaker and an Arabic speaker can sit in the same class and trip over the same English structures despite having nothing in common in their native grammars.
What Drives Interlanguage Development?
If interlanguage is a system, the obvious question is what rewrites it. Selinker pointed to several forces, and later research filled in how they actually play out when you're sitting there trying to learn.
The five things shaping your interlanguage
Selinker (1972) named five processes that feed into the system you build. In plain terms:
- Your first language leaks in. You borrow sounds, word order, and structures from the language you already have. This is language transfer, and it's loudest early on, especially in pronunciation.
- You apply a new rule too widely. You learn that English past tense adds -ed, and suddenly you're saying "goed" and "runned." More on this below, because it's weirdly encouraging.
- How you were taught shows up. Drills, textbooks, and a teacher's favorite examples leave fingerprints on what you produce.
- You invent ways to get the message across. When you don't have the exact word, you talk around it, gesture, simplify. Those workarounds become part of the system.
- You develop habits for learning itself. The strategies you lean on to absorb the language shape what ends up sticking.
Most of your recurring errors trace back to one of these. Naming the source is the first step toward changing it, which is the whole premise of treating an error as information.
Why "goed" is actually good news
There's a moment that looks like getting worse and is really getting better. A learner who has been correctly saying "went" suddenly starts saying "goed." A child does the same thing in their first language, drifting from "broke" to "breaked" before landing back on "broke" (Marcus et al., 1992).
Researchers call this a U-shaped learning curve: good performance, then a dip into a new kind of error, then good performance again at a higher level. The best-documented case is in children's first language, the went to goed and back-to-went swing, and the same worse-before-better shape turns up in second-language development too (Shirai, 1990). Picture a kid who just figured out that one master key, add -ed, opens every past-tense door in the language. Naturally they start jamming it into the doors that have their own locks: goed, runned, breaked. The dip isn't decay. It's the sign that the learner just grabbed a powerful general rule and is now trying it everywhere, exceptions included. The brain has decided the rule is worth running by default. The irregular forms get re-sorted afterward.
So when your own output suddenly sprouts a new, confident-sounding error, that's frequently a system upgrade in progress, not a regression. Knowing that takes a lot of the sting out of it.
Why some errors survive for years
Not every part of your interlanguage keeps moving toward the target. Some features lock in and stay. The mechanism is almost embarrassingly practical: once an error communicates well enough, the pressure to fix it drops to nearly zero.
Think of your brain as a contractor with a long to-do list and zero appetite for overtime. "She go to work" gets understood. Nobody asks you to repeat it. The message passed inspection. From an efficiency standpoint, there's no reason to tear open a wall and rewire a rule that already works. So the brain waves it through and moves on to the next job, and the form sets in place. That's a leading explanation for language fossilization, the point where a specific form stops responding to more exposure even as the rest of your language keeps improving. The good news, covered in depth at that link, is that fossilized forms aren't truly frozen; they just need a different kind of attention to shift.
What the Research Shows About Interlanguage Stages
Here's the finding that should change how you study: interlanguage doesn't develop in whatever order you teach it. It develops in its own order, and you can't shortcut that order by drilling a later form early.
You move through stages in a fixed order
One of the most durable results in the field comes from Pienemann's Teachability Hypothesis (Pienemann, 1984), which maps out the developmental stages learners pass through for certain grammar structures. The core claim: these stages come in a set sequence, and learners move through them in the same order regardless of what they're taught, because each stage depends on the mental processing the previous one built.
He tested it directly. In a study of Italian-speaking children learning German, instruction only worked for the learners who were already at the stage right before the target form. The ones further back didn't pick it up, no matter how it was taught. The form was two stages ahead of where their system could process it, so it didn't register as learnable yet.
The practical translation is liberating. If a structure won't stick no matter how many times you study it, you may not be failing. The structure may simply be ahead of where your interlanguage currently is. Pitch your input and practice at the form just past your current level, and it lands. Aim two stages ahead, and you're pouring water on a sealed jar.
Why correction sometimes lands and sometimes bounces off
This stage logic also explains a frustration every learner has felt: a teacher corrects you, you nod, and the correction does nothing. The research says timing is the difference. Correction and feedback can speed you through a stage, but only for forms you're developmentally ready to process. Feedback on something two stages out is largely wasted; your system isn't listening for it yet.
This is also where noticing comes in. For input to reshape your interlanguage, you generally have to consciously notice the gap between what you said and what the target form is. Correction that lands at the right stage, paired with you actually clocking the difference, is what gets the old rule swapped out.
Why you backslide under pressure
You've had this happen too: in a calm exercise you produce the correct form fine, then in a real conversation you fall back to the old version. That's normal, and it has a tidy explanation. Under the pressure of real-time talking, your brain grabs whatever tool is already in its hand, not the better one still sitting in the drawer. The accurate form is there. It just isn't fast enough yet to win the grab when the clock is running.
That's not a regression in your knowledge. It's a sign the newer form hasn't become automatic. The fix isn't to relearn the rule; it's to get enough low-pressure reps that the better form becomes the one that fires by default.
What Your Errors Are Actually Telling You
Here's the table version of the consensus, so you can match an error to what it means and what to do about it.
| What you notice | What's going on underneath | What it's telling you |
|---|---|---|
| The same mistake across many different sentences | A rule in your interlanguage, consistently applied | This is the rule to target. It won't fix itself. |
| A mistake that vanishes when you slow down | A one-off production failure, not a rule | Low priority. Your system already has the right form. |
| A new, confident-sounding error ("goed") | You just absorbed a broad rule and are over-applying it | A system upgrade in progress. Let it settle. |
| A form that won't stick no matter how you study it | The form is ahead of your current stage | Pitch input to the level just past you, not two ahead. |
| Correct in practice, wrong under conversation pressure | The accurate form isn't automatic yet | Needs reps under lighter load, not relearning. |
| A years-old error that ignores all your input | A fossilized form; communication succeeds, so the brain stopped | Needs targeted attention: noticing, feedback, pushed output. |
The throughline: an error is evidence about your current stage. Cataloguing the ones that repeat is far more useful than cringing at them, because the repeats point straight at the rule your system has locked in, which is the one worth your study time.
Practical Tips to Overcome Interlanguage Errors That Actually Work
You don't fight interlanguage. You read it, then feed it the right thing at the right time. Here's how to do that without a linguistics degree.
Catalogue your repeat offenders instead of cringing at them
The single highest-leverage move is to stop treating each mistake as a fresh confession and start treating your errors as a dataset. Keep a running record of what you get wrong. Give it a few weeks and the systematic errors sort themselves out from the one-off stumbles by sheer repetition, the way the same name keeps surfacing on a suspect list: the same wrong verb ending, the same dropped article, the same backwards word order showing up across totally different sentences.
Those repeats are your interlanguage's actual rules made visible. That short list, not the textbook's chapter order, is your real study agenda. Writing is where this shows up most clearly, because written output sits still long enough to spot the pattern. Atlas Runa's Writing Product is built around exactly this: it turns your writing into a record of recurring patterns, not a scolding list of individual mistakes, so the rule to update is easy to see.
Feed your system the level just past where you are
The stage research is blunt about this: input pitched too far ahead doesn't restructure anything. The content that moves your interlanguage forward is the content that sits just past your current level, close enough that your system can process it, far enough that it's actually teaching you something new (comprehensible input).
In practice that means most of what you read and listen to should feel slightly hard, not brutal. If you're looking up half the words, it's too far ahead and your interlanguage can't use it yet. The trick is finding a steady supply of material at that just-past-comfortable level, which is hard to do by hand. Atlas Runa's Reader calibrates content to your level for this exact reason, so the language you're meeting is close enough to drive change instead of just washing over you.
Pair production with feedback, every time you can
Reading rebuilds your interlanguage, but reading alone leaves the old automatic forms in charge under pressure. To swap them out, you have to produce the form yourself, get a signal that it didn't work, and try again. Production plus feedback together is reliably stronger than either alone.
The reason is the noticing piece. When you produce something and the feedback shows you the gap, you consciously register the difference, and that's the moment the rule has a chance to update. So don't just consume. Write the sentence, say the thing, and get evidence about whether the form landed.
Run these moves mid-session
When you're actually in a practice block or a conversation, a handful of small adjustments keep your interlanguage moving instead of just exercising the forms it already has:
- Warm up on yesterday's repeat offenders before doing anything new, so the rules you're trying to update are the ones getting fresh reps.
- Reframe errors as data the moment they happen: note the pattern, don't relive the embarrassment. The goal is the rule, not the feeling.
- Aim your effort at the form just past your level, and let the structures two stages ahead wait. They aren't learnable yet.
- Lower the stakes when you practice production, since the accurate form only becomes automatic with low-pressure reps, not high-pressure performances.
- Get evidence over weeks, not minutes. A rule update shows up as an error fading across sessions, not as one clean attempt.
Expect the dips and the backslides
Build in the expectation that getting better will sometimes look like getting worse. The "goed" phase, where a new error appears because you just absorbed a powerful rule, is progress wearing a disguise. The conversation backslide, where the right form deserts you under pressure, means the form simply isn't automatic yet. Neither one means you've lost ground.
Knowing the shape of the curve keeps you from quitting in the dip. The learners who pull through aren't the ones who never make these errors; they're the ones who read the errors correctly and keep feeding the system.
How stuckness can help, strangely
The reframe at the heart of all this is small but it holds up the whole roof: your interlanguage is a real system, your repeated mistakes are its rules showing through, and rules can be rewritten when the right input and feedback land at the right stage. The error that's nagged you for months isn't proof you're bad at languages. It's a coordinate, pointing straight at the one rule worth your next hour.
Atlas Runa is built to do the two things this article keeps circling back to: keep your input sitting just past your current level so your interlanguage always has something to chew on, and turn your own output into a record clear enough to read the patterns instead of flinching at the blur. The Reader holds the input at your edge. The Writing tool turns your errors into a map of what repeats instead of a scolding list. The Progress Log shows those patterns fading week over week, the slow proof a rule really did get rewritten. You bring the willingness to look; it makes the looking pay off.
So pick one mistake you've made a dozen times this month and treat it as your first coordinate. That's the spot your interlanguage is quietly asking to grow.