← Teacher resources

Correcting speaking errors in Spanish class: what research says

Classroom research finds that correcting spoken errors helps on average, and that prompts, which push the student to fix the error, have had larger effects than recasts, which supply the right form without comment. Recasts still help, one small study found no significant benefit for highly anxious learners, and a 1999 critique argues for dropping oral grammar correction altogether. We'd correct selectively: prompt for structures you've taught, recast the rest, and move some correction to the end of class.

Ethan, co-founderNote7 min read

On this page
  1. Six ways to respond to a spoken error
  2. What teachers did, and what students said next
  3. Do prompts beat recasts when you measure learning?
  4. The case against correcting out loud
  5. Anxious students, and what students say they want
  6. What we’d try this week
  7. Where Avita Edu fits

A student says Ayer yo ir al cine con mi hermana and looks at you. You have about a second to fix it, hint at it or let it go. Researchers have studied that second since the 1990s, and their findings are specific enough to change what you say next.

We’re Reed and Ethan, founders who learned Spanish as adults and build speaking software. We’ve never had to make that call in front of 30 teenagers. Below is what the studies say, where they disagree and what we’d take from them. Check all of it against your own room.

Six ways to respond to a spoken error

In 1997 Roy Lyster and Leila Ranta recorded 18.3 hours of lessons in four French immersion classrooms in Canada and sorted every teacher response to an error into six types. Most later research uses their names. Here they are applied to one Spanish 2 error. The Spanish examples are ours.

Move What the teacher does After Ayer yo ir al cine con mi hermana
Explicit correction Says it was wrong and gives the right form Se dice “fui”. Ayer fui al cine. (You say “fui.” Yesterday I went to the movies.)
Recast Says it correctly without flagging the error Ah, fuiste al cine con tu hermana. ¿Qué vieron? (Oh, you went to the movies with your sister. What did you two see?)
Clarification request Signals that the message didn’t land ¿Cómo? No entendí bien. (Sorry? I didn’t quite get that.)
Metalinguistic clue Comments on the form without giving it Fue ayer, así que necesitas el pretérito. (It was yesterday, so you need the preterite.)
Elicitation Pauses or asks, so the student supplies the form Ayer yo… (Yesterday I…)
Repetition Repeats the error with rising intonation ¿Ayer yo ir? (Yesterday I to go?)

The first two hand the student the answer. The last four hold it back so the student has to find it, and researchers later grouped those four under one name, prompts.

What teachers did, and what students said next

The figures in this section come from Diane Tedick and Barbara de Gortari’s summary of the study for immersion teachers, published by the University of Minnesota’s CARLA center. The classes were fourth and fifth graders.

Teachers responded to 62% of student errors, and 55% of those responses were recasts. Elicitation made up 14%, clarification requests 11%, metalinguistic clues 8%, explicit correction 7% and repetition 5%.

Then the researchers looked at the student’s next turn. Only 27% of feedback turns were followed by the student repairing the error. Of 375 recasts, 66 were followed by a repair, and each of those was the student repeating the teacher’s words. Elicitation was used 94 times and led to 43 repairs that students produced themselves. All 100 student-generated repairs in the data followed one of the four prompts.

That last result is partly built into the definitions, since a student can’t self-correct once you’ve said the answer. Nothing in the definitions explains the recast numbers, though: 309 of the 375 drew no repair. In the middle of a conversation, a recast can sound like the teacher agreeing with you.

Two cautions before anyone rebuilds a class around this. A student fixing a sentence on the spot doesn’t prove they’ve learned the form, and Tedick and de Gortari note that Lyster and Ranta made no conclusive claims about learning. These were also elementary students who heard French for much of the school day, and a Spanish 2 class that meets for 50 minutes may respond differently.

Do prompts beat recasts when you measure learning?

Later studies tested learning directly. In 2010 Lyster and Kazuya Saito pooled 15 classroom studies covering 827 learners. Corrective feedback had significant, lasting effects on learners’ development in the language. Effects were larger for prompts than for recasts, and younger learners benefited more than older ones.

The same year Shaofeng Li analyzed 33 studies, from labs and classrooms together, and the results were less tidy. Feedback had a medium overall effect that held over time. The effect of implicit feedback, the category recasts usually fall into, was better maintained than the effect of explicit feedback. Effects were also larger in labs than in classrooms, and larger in foreign language settings than in second language ones.

Read together, the two papers say that correcting speech helps on average, that prompts have had the edge in classrooms, and that recasts still do some good. Neither one can tell you what your fourth-period class needs.

The case against correcting out loud

Some researchers reject the premise. In 1999 John Truscott argued that oral grammar correction should be dropped. His case was that teachers face overwhelming problems in making corrections, that correcting has harmful side effects, and that the evidence didn’t show it improving students’ ability to speak grammatically.

Lyster, Patsy Lightbown and Nina Spada replied in the same issue that giving good feedback is hard but that correction shouldn’t be abandoned. They called for research on which types work, in which settings and for which learners. The two 2010 analyses are part of what followed.

If you rarely stop students for form because you’d rather keep them talking, Truscott’s paper is the strongest published version of your position. We think its practical half holds up: correcting mid-sentence is hard to do well, and interrupting a nervous speaker has a cost.

Anxious students, and what students say they want

Younghee Sheen measured that cost in 2008 with learners of English. On two immediate posttests, low-anxiety students who received recasts outperformed both high-anxiety students who received them and low-anxiety students who got none. For high-anxiety students, recasts made no significant difference. The groups had 10 to 13 students each, so treat it as a hint that the students you most want to protect may gain the least from correction on the spot. Our guide to reluctant and anxious speakers covers ways to lower the stakes first.

Students do want some correction, though. A 2013 review by Lyster, Saito and Masatoshi Sato found a tendency for learners to want more of it than their teachers feel they should give.

What we’d try this week

None of this adds up to a rule. ACTFL’s guidance on feedback doesn’t name a technique either. It asks for feedback that is timely, fits the learner’s level and doesn’t overwhelm. Here’s what we’d take from the studies:

  • Prompt for what you’ve taught. If the class has practiced the preterite, Ayer yo… and a pause gives the student a chance to find fui. A student can’t retrieve a form they’ve never met, so recast those and keep going.
  • Make recasts easier to notice. Stress the corrected word, or follow with ¿Cómo? so the student says it again.
  • Pick one target per activity. In a fluency activity, respond to the unit’s structure and let the rest go.
  • Go easy on beginners. Tedick and de Gortari, writing for immersion teachers, advise that early-stage learners mostly need encouragement to communicate, and that prompts suit more proficient students.
  • Move some correction out of the moment. Write down three errors you hear during partner talk and put them on the board, unnamed, in the last five minutes. None of these studies tests that, but nobody gets singled out, and it fits the daily listening in our guide to assessing speaking.
  • Record yourself once. Tedick and de Gortari suggest taping a few lessons to hear what you do now. You may recast more than you think.

Five prompts to keep handy: ¿Cómo? (Sorry?), ¿Otra vez? (Again?), ¿Ayer…? (Yesterday…?), ¿Yo o ella? (I or she?), Casi. Es pasado. (Almost. It’s past.)

Our take: Correct less often than you could, and when you do, give the student the first try at fixing it. We’d keep recasts for forms the class hasn’t learned and for students who would freeze at a prompt. You know which students those are, and we don’t.

Where Avita Edu fits

Avita Edu covers one piece of this: correction that happens one student at a time, away from the front of the room. Students talk through a short conversation in the browser. An answer that misses gets a more natural way to say it and a short reason, and the student can try again, with a hint if they’re stuck. In research terms that feedback gives the form, so it sits closer to explicit correction than to a prompt, and we have no data on what students learn from it.

Teachers get mistakes back grouped by type across the class, which is one way to choose the three for the board. Teacher accounts are invite-only during the pilot. If a clipboard during partner talk already tells you what to reteach, keep using the clipboard.

Questions teachers ask

Should I correct every speaking error in Spanish class?
No study we found supports correcting everything, and the teachers Lyster and Ranta recorded in 1997 responded to 62% of errors. We'd choose one target per activity, usually the structure the unit is teaching, and let other errors go while students are working on fluency. Errors you hear from many students can wait for a short whole-class review with no names attached.
What is a recast in language teaching?
A recast is when the teacher repeats what the student meant in correct form without saying there was an error. If a student says "Ayer yo ir al cine," a recast is "Ah, fuiste al cine." It keeps the conversation moving and was the most common feedback type in Lyster and Ranta's 1997 classroom study, at 55% of teacher responses. Students often don't react to it.
Are prompts better than recasts for correcting speaking?
In classrooms, on average, yes. A 2010 meta-analysis by Lyster and Saito of 15 classroom studies with 827 learners found larger effects for prompts than for recasts. Recasts still helped, and a second 2010 meta-analysis by Li found that the effect of implicit feedback lasted better than that of explicit feedback. Prompts only work for forms a student has already been taught.
Does correcting speaking errors make students anxious?
It can, and anxiety may also blunt the correction. In a 2008 study by Younghee Sheen, recasts helped low-anxiety learners on two immediate posttests but made no significant difference for high-anxiety learners. The groups were small, 10 to 13 students each. For nervous speakers, responding to the meaning first and saving form for a later whole-class review is a reasonable choice.
Do students want their speaking errors corrected?
Often more than teachers expect. A 2013 research review by Lyster, Saito and Sato reports a tendency for learners to prefer receiving more corrective feedback than their teachers feel they should provide. That is an average across many studies and settings, so it's worth asking your own classes how and when they'd like to be corrected.
Is there research against correcting spoken grammar?
Yes. In 1999 John Truscott argued in the Canadian Modern Language Review that oral grammar correction should be discontinued, citing the difficulty of doing it well, harmful side effects and a lack of evidence that it improves spoken grammar. Lyster, Lightbown and Spada replied in the same issue that correction shouldn't be abandoned. Later meta-analyses found positive average effects.

Hear every student in your class this week.

Avita Edu lets you assign speaking practice for class or homework. Students talk it through out loud in the browser, and you get it back graded. Free for teachers.

Invite-only while we open to our first teachers. Have a code? Create your account

See how it works