What Happens When You Try to Fix Errors in a Language Class
I've watched teachers spend twenty minutes on a worksheet only for students to produce the exact same mistake the next day. Corrective Feedback In Second Language Acquisition is one of those topics where the theory sounds straightforward and the practice falls apart in about three weeks. There are different kinds of feedback, they work differently, and most of what you read about them is either too optimistic or misses the part where nothing changes. Let me start with what actually happens in a classroom. A student says "he go to the store yesterday" and the teacher responds. That response can take several forms, and the form matters more than most people realize. An explicit correction tells the student directly what was wrong. "He goes" or "You should say 'went'." An implicit correction might involve recasting — repeating the sentence back correctly without drawing attention to the error. "Oh, he went to the store?" The student may or may not notice the difference. A metalinguistic clue gives a hint about what kind of error was made without fixing it. "That verb tense doesn't match your time marker." Each type demands different cognitive work from the learner, and they land differently depending on proficiency level, age, and the individual's willingness to engage with correction at all. The research on this goes back decades. Long, Segalowitz, and others mapped out the landscape, and the general consensus isn't as clean as textbooks suggest. Feedback works, but only under certain conditions that don't always hold in real classrooms. The conditions are things like the learner being ready to acquire that particular feature, the feedback being noticeable, and the learner actually processing it rather than just absorbing it passively. Noticeability is the tricky one. A recast might be so embedded in natural conversation that the learner walks away thinking they said nothing wrong.
A Specific Problem I Hit Head-On
I ran into this with an intermediate learner who had a persistent article error — using "the" where "a" was needed and dropping articles entirely in spontaneous speech. We tried explicit correction for six sessions. The written tests improved. The speaking didn't budge. The gap between controlled production and free communication was wide enough to drive a truck through. What worked was combining focused input Floods with a retrieval practice protocol. Instead of correcting every error in real time, I flagged recurring patterns and built short exercises where the student had to actively choose the correct form from memory rather than recognize it when presented. It took about eight weeks before the oral errors dropped by roughly forty percent. The written improvement happened in two weeks. The lag between recognizing the right form and producing it under pressure is something most feedback frameworks don't address adequately. That lag is the difference between knowing grammar rules and using them when you're trying to communicate something you actually care about saying. Feedback that only targets accuracy in controlled tasks often fails to close this gap because the conditions under which the learner practices are completely different from the conditions they face in real interaction.
Things Most People Get Wrong About This
More correction does not equal better acquisition. This is the biggest mistake I see, and it's not limited to new teachers. It shows up in textbook design, in policy documents, in well-meaning tutoring programs. There is a threshold where correction becomes noise. The learner starts predicting the correction rather than attending to the meaning, and the corrective event itself becomes part of the background chatter. I've seen this in adult learners who develop a kind of correction fatigue — they stop registering the feedback after a certain density because the cognitive cost of processing every single correction outweighs the benefit. Recasts are not universally effective. They work best when the learner is in a zone where they are tentatively attempting the target form but missing it. If the learner hasn't developed any internal representation of the structure yet, a recast just sounds like normal speech with no signal that anything was wrong. The SLA literature has known this for a while but classroom practice moves slowly. The type of feedback should match the type of error. Some errors are system-level — the learner hasn't acquired the underlying rule. Others are performance-level — the learner knows the rule but slipped. Performance errors respond better to explicit feedback because the knowledge is already there and just needs activation. System errors need input and time, and feedback alone won't accelerate acquisition much beyond what natural exposure provides. Mixing up these two categories and treating them the same way is a common source of wasted effort.
Get the Full Details
How to Actually Use This Without Wasting Time
Prioritize which errors to correct. Not all errors are equal. Some interfere with comprehensibility and deserve immediate attention. Some are developmental and will resolve on their own with sufficient exposure. A good rule of thumb is to focus on errors that are persistent across multiple attempts and that the learner has been exposed to instruction on before. If the learner has never seen the structure, correcting it in isolation is low yield. Space your corrective events. Batching corrections into a single session or correcting every utterance during a conversation creates diminishing returns. Distribute feedback across multiple encounters with the same target structure. Each encounter should be spaced far enough apart that the learner has to retrieve the rule rather than rely on recent exposure. Combine feedback with opportunities for modified output. Correction without the chance to produce the corrected form is largely inert. After giving feedback, create a task that requires the learner to use the target structure in a way that carries meaning. The meaning component is non-negotiable. If the task is purely mechanical, the learner is practicing form recognition, not acquisition.
Track what changes and what doesn't. Keep a simple log of error types and their frequency over time. This takes maybe ten minutes per week and gives you data that most teachers and tutors don't have. You'll quickly see which feedback strategies are actually moving the needle and which are just creating the illusion of progress.
The Honest Part
Corrective feedback has limits. It cannot compensate for insufficient input. It cannot accelerate acquisition beyond what the learner's developmental sequence allows. It is least effective for features that are not salient in the input and most effective for features the learner is developmentally ready to acquire. It works better for some learners and worse for others based on working memory capacity, motivational orientation, and prior exposure. There is no universal protocol that fits every classroom or every student. When feedback doesn't seem to be working, the issue is usually one of three things: the learner isn't ready for that particular structure yet, the feedback type doesn't match the error type, or there simply isn't enough meaningful exposure to the target form outside of corrective events. Addressing any of these three tends to matter more than tweaking the feedback mechanism itself.
