Slow enough to think, fast enough to care: the Goldilocks pace of chat

Every way of practicing a language has a pace, and the pace decides a lot about what you can learn from it.

A novel lets you go as slowly as you like, but it never responds to you. A podcast keeps moving whether you understood the last sentence or not. A conversation responds to you, but it moves at the speed of speech and won't wait while you look something up.

Text chat is unusual: it's interactive like conversation and self-paced like reading. We think that's the single biggest reason it works so well for learners.

Two questions about any kind of practice

Ask two questions about any practice format:

1. Does it respond to you? Does what you say change what happens next?
2. Does it wait for you? Can you take as long as you need to understand and respond?

Most of the classic advice fills three of these four boxes. Read a lot (waits, doesn't respond). Listen a lot (neither). Speak with natives (responds, doesn't wait). The fourth box, practice that both responds and waits, is where text chat lives, and it's the box beginners and intermediate learners need most.

Why "responds to you" matters

Language research has argued for decades that interaction does something passive input can't. When you try to say something and the other person doesn't quite understand, you rephrase, they clarify, and in that back-and-forth you notice the gap between what you said and what you meant to say.¹ Producing language yourself forces you to process grammar in a way that comprehension alone lets you skip.²

It's also why interaction is motivating. Someone asked you something and wants your answer. They might joke, change the subject, or ask a follow-up you didn't expect. That unpredictability is what keeps a chat engaging in a way a textbook dialogue never is. Practice you look forward to is practice you'll actually do.

Why "waits for you" matters

Spoken conversation is fast. The gap between turns averages around 200 milliseconds across languages,³ while planning even a short spoken reply takes over 600 milliseconds.⁴ That means fluent speakers start formulating their answer while the other person is still talking. Learners can't do that yet, so in real-time conversation they fall back on the words and structures they already have, or freeze.

Text chat stretches that window enormously. In a corpus of about 13,600 real Japanese text chats we analyzed, native speakers took a median of 20.8 seconds to reply, about 100 times longer than the spoken gap, and nobody found that strange.⁵

That time isn't wasted. It's where you:

• reread the message and catch the part you missed,
• look up the one word that unlocks the sentence,
• try a structure you're not sure about, and check it,
• notice how your partner phrased something and borrow it.

Researchers studying chat have seen this happen: in one study, messages that had been revised before sending were more linguistically complex than messages sent as first drafts.⁶ The pause is where the learning fits.

The pace also lowers the stakes

The "waits for you" property has an emotional side, too. Many learners know words they freeze up on in person. In one study comparing text chat and voice chat practice, speaking anxiety dropped only in the text-chat group.⁷ When nobody is watching you search for a word, you're more willing to try harder sentences.

Isn't slow practice inefficient?

In raw words per hour, yes, at least compared to speaking. A spoken lesson gets far more words out of you in an hour than a text chat does, and we're upfront about that in our numbers post.

But words per hour isn't the only thing that matters. What matters is how many of those words come with enough attention to change what you know. A fast conversation in which you understood 60% and reused only phrases you already knew is less useful than a slower one where you understood everything, looked up the three words you needed, and tried two new structures.

And slow doesn't mean dull. A good chat moves: new topics, photos, reactions, jokes. It's slow per message and lively per conversation. That's the Goldilocks pace.

Moving up to speed

None of this means you should text forever and never speak. Think of chat as practicing a piece of music slowly before playing it at tempo. You learn the moves when you have time to think, then you speed up.

As your chat replies get faster, more natural, and need fewer lookups, you're building the same vocabulary and structures you'll need in spoken conversation, just at a pace where you can actually learn them.

That's the idea behind MiraChat. Your language partners are always up for a chat, and they never mind if your reply takes a while. Mira sits beside the conversation for the lookups, translations, and corrections that make that pause useful.

Sources

1. Long, M. H. (1996). The role of the linguistic environment in second language acquisition. In W. Ritchie & T. Bhatia (Eds.), Handbook of Second Language Acquisition (pp. 413–468). Academic Press. ↩
2. Swain, M. (1985). Communicative competence: Some roles of comprehensible input and comprehensible output in its development. In S. Gass & C. Madden (Eds.), Input in Second Language Acquisition (pp. 235–253). Newbury House. ↩
3. Stivers, T., et al. (2009). Universals and cultural variation in turn-taking in conversation. PNAS, 106(26), 10587–10592. https://doi.org/10.1073/pnas.0903616106 ↩
4. Levinson, S. C., & Torreira, F. (2015). Timing in turn-taking and its implications for processing models of language. Frontiers in Psychology, 6, 731. https://doi.org/10.3389/fpsyg.2015.00731 ↩
5. Our analysis of RealPersonaChat (Yamashita et al., 2023). See Chat-based language learning, by the numbers. ↩
6. Sauro, S., & Smith, B. (2010). Investigating L2 performance in text chat. Applied Linguistics, 31(4), 554–577. https://doi.org/10.1093/applin/amq007 ↩
7. Satar, H. M., & Özdener, N. (2008). The effects of synchronous CMC on speaking proficiency and anxiety: Text versus voice chat. The Modern Language Journal, 92(4), 595–613. https://doi.org/10.1111/j.1540-4781.2008.00789.x ↩