Every speaking teacher struggles with the same moment - and the challenge of deciding when to correct. A student is mid-sentence, clearly engaged, building towards something interesting - and they use the wrong tense. Do you correct them? Let it go? Write it down for later? Signal it without breaking the flow?
The fluency versus accuracy debate has been running in language teaching for decades. And while it's often presented as a philosophical tension, the research is actually quite direct about when correction helps and when it actively makes things worse.
The short answer: during speaking practice, fluency should almost always win. YapYapGo is a classroom speaking practice tool for ESL and EFL teachers built around this principle - it creates space for extended, uninterrupted pair work where students speak without being corrected mid-sentence.
What the research says
Merrill Swain's Output Hypothesis established that speaking is a driver of acquisition, not just a demonstration of it. When students try to say something and struggle, they notice a gap in their language knowledge. That noticing is what triggers learning. But it only happens when students are actually attempting extended production - not when they're waiting for feedback.
Studies on corrective feedback timing consistently show that immediate correction during fluency activities produces two effects: a short-term improvement in the corrected item, and a significant reduction in the amount of language the student produces in the rest of the session. Students become more cautious. They self-censor. They say less. The immediate gain in accuracy is outweighed by the loss in practice time.
This is especially pronounced for students with high language anxiety - roughly a third of all language learners, according to research by Horwitz and colleagues. For these students, being corrected mid-sentence is not a learning moment. It's a shutdown signal.
The case for delayed correction
Delayed error correction - noting errors during an activity and addressing them afterwards - consistently outperforms immediate correction in speaking contexts. Students speak more, take more risks, and develop faster over time.
The key to making delayed correction effective is specificity. Vague feedback ("some of you were making mistakes with past tenses") wastes the teaching opportunity. Specific feedback ("I heard three different ways people tried to express the idea of something being temporary - here are two that worked well and one that needs adjusting") is actionable and connects directly to what students just did.
The debrief is where your expertise as a teacher adds the most value. During the activity, your job is to listen and take notes. After the activity, your job is to turn those notes into targeted instruction.
When immediate correction does make sense
There are specific circumstances where immediate correction is appropriate even during speaking activities.
Phonological errors that impede communication. If a student's pronunciation is making them actively hard to understand - not just accented, but incomprehensible - quiet, immediate clarification is warranted. High-stakes exam practice. When students are specifically practising for an IELTS or Cambridge speaking exam, targeted immediate feedback on examiner-facing behaviours (extended response length, discourse marker use) can be useful when delivered carefully and non-interruptively. Student request. If a student explicitly asks you to correct them as they go, respect that preference. Factual errors about language. If a student states a grammar rule incorrectly during a discussion about language, brief correction is appropriate.In all other cases during a fluency-focused speaking activity, write it down and address it after.
A practical correction framework
During the activity: Circulate, listen, and note. Use shorthand. You're looking for patterns (errors that multiple students made) rather than individual mistakes. Immediately after: Address the two or three most important patterns you noticed. Connect them to what students just said. "When you were arguing about whether social media is harmful, several pairs struggled to express the idea of 'it depends on how you use it.' Here are three ways to say that..." For persistent errors: If the same error appears week after week, it needs a focused teaching moment, not just feedback. Build a short grammar mini-lesson around the actual error from real student speech. For strong language: Note good things too. "I heard someone say 'on balance, the benefits outweigh the risks' - that's a great phrase for expressing a nuanced position. Can you share how you used it?" Positive examples are as instructive as corrections.Tool tip: Because YapYapGo handles pairing, timing, and question delivery automatically, your hands are free to circulate and take notes from the first minute. No stopwatch, no deciding who speaks next - just listening. A classroom timer, activity timer, and random student picker are useful for calling on individuals to share during the debrief. They mark the end of each round so the transition to debrief is clean.
The recasting technique
For teachers who find it genuinely difficult not to respond to errors in real time, recasting offers a middle path. Recasting means reformulating what the student said correctly without explicitly marking it as a correction.
Student: "Yesterday I go to the market."
Teacher (naturally): "Oh, you went to the market? What did you buy?"
The student hears the correct form embedded in a natural response. No interruption, no shame, no shutdown. Whether recasting produces uptake - whether the student notices and incorporates the correction - depends on the student, the error type, and the context. Research results are mixed. But it's consistently better than explicit mid-sentence correction during fluency activities.
The real question
The fluency-accuracy question is sometimes framed as "which matters more?" That's the wrong question. Both matter. The better question is: which should you prioritise during speaking practice?
The answer the research supports is unambiguous: during pair work and free speaking activities, protect fluency. Create space for extended production without interruption. Take notes. Debrief well. Your students will develop both fluency and accuracy faster than they would under a regime of constant correction.
For more on maximising pair practice time, see the pair work multiplier in your students are not speaking enough. For students who worry that pair practice without correction will reinforce errors: the evidence suggests the opposite. Students who speak more, and who feel safe speaking, develop accuracy over time because they're producing more language and getting more feedback at the right moments. The students who never speak because they're afraid of being corrected are the ones who plateau.
See also our post on reducing teacher talking time for the broader case on why less teacher intervention during speaking activities produces better outcomes.
Sources:
- Swain, M. (1985). Communicative Competence: Some Roles of Comprehensible Input and Comprehensible Output. Input in Second Language Acquisition. - The Output Hypothesis: speaking drives acquisition.
- Sheen, Y. (2004). Corrective Feedback and Learner Uptake. Studies in Second Language Acquisition. - Timing and type of correction affect uptake and production volume.
- Horwitz, E., Horwitz, M., & Cope, J. (1986). Foreign Language Classroom Anxiety. The Modern Language Journal. - Immediate correction significantly increases anxiety for high-anxiety learners.
- Lyster, R. & Ranta, L. (1997). Corrective Feedback and Learner Uptake. Studies in Second Language Acquisition. - Recasting vs. explicit correction: effects on uptake.
- Long, M. (1996). The Role of the Linguistic Environment in Second Language Acquisition. Handbook of Second Language Acquisition. - Interaction and negotiation of meaning as drivers of acquisition.