On this page
Avita asks you to speak from the first lesson. A common objection is that this is backwards: shouldn’t you listen for months first and let speech come on its own? It’s a serious position with serious researchers behind it, so here is what the research says, including the parts that argue against us.
The case for input first
Stephen Krashen’s input hypothesis, developed in the late 1970s and 1980s, holds that we acquire a language by understanding messages in it. Speaking is the result of acquisition, not the cause. In a 1998 paper titled “Comprehensible output?” in the journal System, Krashen argued that learners produce relatively little output, that the kind of output supposed to drive learning, where a learner fails to get a message across and reworks it, is rarer still, and that people can reach high levels of competence without much speaking at all.
A lot of this is right. You can’t say what you’ve never heard or read, and most of what you’ll eventually know will come from input. Anyone who has watched a heritage speaker understand everything at a family dinner knows comprehension can run far ahead of speech.
Speech running behind comprehension is the problem many learners have, though, and it’s where the other side of the research comes in.
What Merrill Swain saw in Canadian immersion classrooms
In the 1980s, Merrill Swain studied students in Canada’s French immersion programs: English-speaking kids who had been taught school subjects in French for years. They got enormous amounts of comprehensible input, and their comprehension was excellent. Their spoken and written French still had persistent grammatical errors that years of understanding hadn’t fixed.
Swain’s explanation, first set out in 1985, was that understanding a sentence doesn’t require you to process its grammar closely. You can follow quiero hablar con Mateo (I want to talk with Mateo) from the words quiero, hablar and Mateo alone. Producing it forces decisions: is it quiero hablar or quiero hablo? Con or a? She called this comprehensible output, and in a 1995 chapter she described three things it does:
- Noticing: trying to say something shows you the gap between what you want to say and what you can say.
- Hypothesis testing: every sentence you produce is a small guess about how the language works, and the response tells you whether the guess held.
- Reflection: talking about your own language use helps you control it.
Swain didn’t claim output does most of the work. Her argument was that input alone leaves gaps that producing the language helps close.
Pulling it out of memory is practice
A separate line of research, from memory psychology rather than linguistics, points the same direction.
In 2006, Henry Roediger and Jeffrey Karpicke had students read short prose passages and then either reread them or try to recall them. A week later, students who had practiced recalling remembered 61% of the passage, against 40% for those who had only reread it. Rereading looked better five minutes after studying. A week later the advantage had flipped.
Two years later in Science, Karpicke and Roediger had students learn 40 Swahili–English word pairs, like mashua (boat). Some kept being tested on every pair until the end of the session; others stopped being tested on a pair once they’d recalled it correctly, even if they kept studying it. A week later, the first group recalled about 80% of the pairs. The second recalled 36% and 33%, depending on the condition. Every group predicted they’d remember about half, so the students had no idea which method was working.
Answering a question out loud in a new language is a retrieval attempt. Rereading the dialogue isn’t. These studies measured memory for words and texts rather than speaking skill, so they don’t prove that speaking makes you fluent. They do show that pulling language out of memory strengthens it more than looking at it again.
Saying it out loud adds something
Colin MacLeod and colleagues reported in 2010 that people remember words they read aloud better than words they read silently. They called it the production effect. In their experiments it appeared when the same person read some words aloud and others silently, not when separate groups were compared, so it’s a narrow effect. Still, it’s a small memory boost that comes along every time you speak.
Comparing the two approaches directly
The cleanest test of input versus output is a study that teaches the same thing both ways. In 2013, Natsuko Shintani, Shaofeng Li and Rod Ellis published a meta-analysis of 35 such comparisons for grammar teaching in Language Learning. Both approaches worked well. Comprehension-based teaching had an early edge for understanding that faded on later tests. For producing the grammar, the two were similar at first, and production-based teaching came out ahead on the delayed tests.
That’s about as balanced as research gets: input builds understanding, and producing the language builds the ability to produce it.
Where the input side has a point
Forced speech has a cost. Krashen has long argued that anxiety blocks acquisition, and in Horwitz, Horwitz and Cope’s 1986 study of language anxiety, difficulty speaking in class was the concern anxious students raised most (we wrote a guide on speaking anxiety). Speaking without any input to draw on also produces guesses rather than practice.
So speaking early works best when it’s built on something you’ve just read or heard, when it’s private and low stakes, and when you get feedback on each attempt. That’s the order an Avita lesson follows: you read a short text that teaches the grammar and words, listen to it read aloud, then answer a character from the story out loud, about 49 times a lesson. Each answer gets a corrected version and a short reason why, and you can ask for a hint or a translation when you’re stuck.
What to do with this
- After any input (a podcast, a lesson, a dialogue), say something with what you just learned, out loud.
- Recall instead of rereading. Cover the answer and produce it first, then check.
- Keep practicing what you already got right. Karpicke and Roediger’s students lost most of the words they stopped retrieving.
- Keep the input coming too. Read and listen to far more than you can say.
Our take: The research supports both camps more than either camp likes to admit. Listening and reading build understanding, and speaking from early on builds the ability to speak. If your goal is conversation, start saying things out loud from the first week, on top of plenty of input.
Questions people ask
What is Swain's output hypothesis?
Merrill Swain proposed in 1985 that producing language, not only understanding it, helps learners acquire it. She based it on Canadian French immersion students whose comprehension was excellent but whose speaking and writing kept grammatical errors. In 1995 she described three functions of output: noticing gaps in what you can say, testing guesses about how the language works, and reflecting on your own language use.
Is comprehensible input enough to learn to speak?
Input is necessary and does most of the work for understanding, but the evidence suggests it isn't the whole story for speaking. Swain's immersion students had years of input and still made grammar errors when they spoke and wrote. A 2013 meta-analysis by Shintani, Li and Ellis found production-based grammar teaching did better than comprehension-based teaching on delayed tests of producing the grammar.
What is the testing effect?
It's the finding that trying to recall something strengthens memory more than studying it again. In Roediger and Karpicke's 2006 study, students who practiced recalling a passage remembered 61% of it a week later, against 40% for students who reread it. In a 2008 study with Swahili vocabulary, students who kept retrieving the words recalled about 80% after a week; those who stopped recalled about a third.
Should beginners stay silent before they start speaking?
Some input before speaking helps, since you can't say what you've never heard or read. The research doesn't require a long silent period, though. Reading or listening to something short and then speaking about it right away combines both: the input gives you something to say, and saying it turns recognition into something you can produce. Keep early speaking private and low stakes if it makes you anxious.
