When you make a mistake in English, when should a teacher fix it? Some teachers stop you right away. Others wait until you finish.

Scientists have tested both ways, and the results are mixed. In a 2025 study, 186 students did a task two times. Students who got help between the tasks did better on a grammar test than students who got help at the end. Maybe they could use the right form again in the second task.

"I go to Kyoto yesterday." Should a teacher stop the student and say "went" right away, or wait until the conversation is over? A review of 20 studies found no clear answer, partly because researchers define "right away" and "later" in different ways.

A 2025 study tried a third option. Vocational college students, 186 in total, did two tasks that used the English passive voice. Some were corrected during the tasks, some between the two tasks, and some only after both tasks.

The between-task group did better on a grammar-rule test given soon after the lessons than the after-both-tasks group, and about as well as the group corrected during the tasks. The researchers think the second task mattered. It gave students a chance to try the corrected form right away.

Teachers have long argued about whether to correct an error the moment it happens or save it for later. A systematic review of 20 studies published between 2006 and 2021 could not pick a winner, in part because "immediate" and "delayed" were defined so differently from one experiment to the next. A 2025 replication in Language Learning changed the question.

Instead of asking when, it asked what comes after the correction. Among 186 vocational college students practicing the English passive, those corrected between two tasks outperformed those corrected after both tasks on a test of explicit grammar knowledge, and matched those corrected during the tasks. No group showed real gains in the automatic knowledge used in fast speech. Other studies add conditions.

When 145 seventh graders practiced the past tense, early correction beat correction that came only in the final session, perhaps because their errors had not yet hardened into habits. Anxiety plays a part too. In an experiment with 82 learners, on-the-spot correction helped calm learners more than nervous ones. And in a semester-long trial with an AI chatbot, 66 learners improved about equally either way, though those given instant feedback rated the experience more highly.