Speaking is not only the result of learning a language but also a way of learning it: when they have to speak, learners notice what they cannot yet say, try out ways of expressing it and get feedback. This is the core of two hypotheses that complement Krashen: Michael Long's interaction hypothesis, which stresses the back-and-forth of trying to understand each other, and Merrill Swain's output hypothesis, which stresses being pushed to express things accurately.
These two hypotheses explain something parents often see: a child who has studied for years, reads and listens fairly well, but speaks slowly and without confidence. They are also the scientific basis for small classes, task-based activities and target-language-only rules in many language programmes. This article sets out each hypothesis, the evidence, its limits, and how to count how much your child actually speaks during a course.
Long and the negotiation of meaning
From the early 1980s, Michael Long studied conversations between native speakers and learners. He saw that when learners did not understand, both sides adjusted: the native speaker slowed down, repeated, changed words and checked whether the learner had understood; the learner asked again, asked for an explanation, guessed and checked. He called this process negotiation for meaning.
According to Long, negotiating meaning turns difficult input into comprehensible input, at the moment and at the point where the learner needs it. Unlike a pre-recorded listening exercise, a conversation adjusts itself to the learner. He presented an updated version of the hypothesis in a 1996 book chapter, which put more weight on the role of attention and feedback: while negotiating, learners are shown the gap between how they said something and how it should be said.
An illustration: your child says "Yesterday I go to beach". A roommate asks "You went to the beach yesterday? Which one?". The question contains exactly the forms "went" and "the beach", straight after your child's mistake, at a moment when your child is paying attention because they want to carry on with the story. According to the interaction hypothesis, this is a good moment to notice the correct form, far better than meeting "went" in a random reading passage.
Swain and the three functions of output
Where it came from. As described in the research on Canadian immersion classes, Swain found that pupils understood spoken French very well but still made many mistakes when speaking. In 1985 she proposed that comprehensible input is necessary but not sufficient, and that learners need "pushed output", that is, being required to express themselves accurately and coherently. In 1995 she set out three functions of output in detail.
The noticing function. When trying to say something, learners realise they do not know how to express it. This gap makes them pay more attention when they later meet that expression in input. Listening can be understood thanks to context without processing the grammar; speaking forces you to choose words and structures.
The hypothesis-testing function. Every sentence spoken is a test: "can I say it like this?". The listener's reaction, whether they understand, whether they ask again, is data the learner uses to adjust. Without speaking, there are no tests.
The reflective function. When pupils work together on a task, such as rewriting a story, they talk about the language itself: this word or that one, which tense is right. Swain and her colleagues called these exchanges language-related episodes and found they were linked to learning the forms being discussed.
Schmidt and the noticing hypothesis: the link between interaction and acquisition
In 1990, Richard Schmidt put forward the noticing hypothesis, partly based on his own diary of learning Portuguese in Brazil. The main points:
- The core idea: a language form has to be noticed by the learner in the input before it has a chance of being acquired. Input that is not attended to largely passes by.
- Evidence from the diary: Schmidt recorded that the structures he began using in speech were often ones he had noticed in input earlier, often after being taught them in class.
- Link with interaction: negotiation of meaning and feedback create moments when learners are more likely to notice form, because they are paying attention and have a reason to care.
- Debate: how much conscious awareness is needed, and whether anything can be learned without noticing, are points on which researchers such as John Truscott and Nick Ellis disagree. The weak version of the hypothesis, that attention helps learning, is more widely accepted than the strong version.
Put the three hypotheses together: input provides the material, interaction makes input comprehensible and generates feedback, output forces learners to process grammar and notice gaps, and attention is the bridge between them. This is the framework many second language acquisition textbooks use today.
The evidence: does interaction really help?
In 2007, Alison Mackey and Jaemyung Goo published a meta-analysis of experimental studies on interaction in Conversational Interaction in Second Language Acquisition, edited by Mackey. The overall conclusion was that interaction has a positive effect on learning vocabulary and grammar, and that part of the effect shows up more clearly in delayed tests, that is, after some time has passed. This suggests interaction helps learning last, not just short-term recall.
Meta-analyses of corrective feedback, such as Roy Lyster and Kazuya Saito's in 2010, also show that feedback during speaking works, and that the effect was stronger for younger learners in some comparisons. Feedback that lets learners correct themselves usually gives good results.
Alongside this is task-based language teaching, developed by Long, Rod Ellis, Peter Skehan and many others. The idea is to give learners tasks with a real outcome, such as planning a trip, solving a puzzle or spotting the differences between two pictures, in which language is the tool. Tasks generate interaction and output naturally.
The limits: many experimental studies take place in a laboratory or a controlled classroom, focusing on one or two specific grammatical forms; in real classrooms with more variables the effect is usually smaller.
Counting how much your child actually speaks in a day of study
One direct way to apply these hypotheses is to estimate how much time your child really gets to speak. The table below is a hypothetical example to show how to work it out; real figures depend on each programme.
| Format | Length | Class size | Estimated speaking time for one student |
|---|---|---|---|
| Large class, mostly teacher-led | 3 hours | 20 students | A few minutes, if called on to answer a few times |
| Small group class with pair activities | 3 hours | 6 students | About 30 to 45 minutes, if half the time is pair work |
| One-to-one tutoring | 2 hours | 1 student | Up to 50–60 minutes, if the teacher knows how to give the floor |
| Task-based afternoon activity, mixed-nationality group | 2 hours | 8 students | About 15 to 30 minutes, depending on the role in the group |
| Dinner with a host family or roommates who speak the target language | 1 hour | 3–4 people | About 10 to 20 minutes |
This calculation shows why the class format and what happens outside class matter more than the total hours printed in the programme. Two courses that both say "30 hours a week" can differ several times over in how much your child actually speaks.
Putting it into preparation and into the trip
Before the trip, parents can help their child prepare for interaction: practise the basic phrases for negotiating meaning, such as asking someone to repeat, to speak more slowly, asking what a word means and checking that they have understood. These phrases are small but they are the key to keeping a conversation going instead of falling silent when your child does not understand. Research on communication strategies shows that learners who use them keep conversations going longer.
When looking at a programme, ask about the activity formats: are there task-based activities, presentations, retelling, small debates, and how do teachers correct mistakes? A lesson where students mostly sit, listen and copy, even with a native-speaking teacher, produces little pushed output.
During the trip, encourage your child to talk about what they did that day, with teachers, friends, or in voice messages to the family in the target language. Every retelling is a piece of output. When your child tells you about an "embarrassing" mistake like ordering the wrong dish, that is often the most valuable moment of negotiation of meaning that day, and it should be welcomed as such.
Common misconceptions about speaking practice
"Just speak a lot and it will come right." Speaking a lot builds fluency, but without feedback and without moments of attention to form, errors can settle in, a phenomenon Selinker called fossilisation. Speaking a lot needs to go hand in hand with clear enough feedback.
"You have to be good before you speak." According to Swain, it is precisely trying to speak while still lacking something that helps learners see what they lack. Waiting until you are "good enough" skips a way of learning, and often makes speaking anxiety worse.
"Only speaking with native speakers is useful." Native speakers provide standard input, but research on interaction between learners shows that students still negotiate meaning, correct each other and try new structures when they work on tasks together. International classmates at the same level are a valuable source of speaking practice.
"The teacher must correct every mistake." Correcting every mistake breaks the flow of communication and makes students reluctant to speak. Research on feedback suggests focusing on repeated errors and errors that affect meaning, and letting students correct themselves where possible.
Frequently asked questions
If my child speaks little in class, does that mean they are not progressing?
Not necessarily. Many learners go through a period of mostly listening before they speak much, and their listening comprehension still improves. But if it goes on too long, your child will miss out on the benefits of output. Create chances to speak in small, low-pressure groups before expecting your child to speak in front of the whole class.
Is talking with other learners useful?
Yes. Research on learner-to-learner interaction shows that learners still negotiate meaning and correct each other, even though the input is less standard than a native speaker's. The worry about picking up classmates' mistakes is not well supported by research, especially when a teacher is there to give feedback.
Are native-speaking teachers better than Vietnamese teachers for speaking practice?
What matters more is how the class is organised: do students get to speak a lot, is the feedback clear? Native-speaking teachers provide natural input; good Vietnamese teachers understand the errors that come from Vietnamese. Both can teach speaking well or badly.
Should my child be banned from using Vietnamese entirely on a language trip?
Rules about using the target language in shared areas increase output. But an absolute ban easily creates stress, especially for young children travelling far for the first time. Many programmes allow the mother tongue when something important needs to be discussed or when calling home, and encourage the target language the rest of the time.
Can speaking practice with AI replace speaking with people?
Chat tools can generate output and feedback and are useful for extra practice. But negotiating meaning with a real person adds a social element: natural reactions, emotion, and a real need to be understood. Research on the long-term effectiveness of speaking practice with machines is still new and limited.
Is task-based learning suitable for low-level students?
Yes, if the tasks are chosen to match their level, such as matching pictures, finding objects or putting together a simple schedule. For low levels, teachers usually prepare the vocabulary and structures needed beforehand, let students do the task, and then look back together at the language they used.