"I'm past the age where you can learn this properly" is the most common reason adults give up on speaking.
It sounds scientific, and there *is* research showing age matters. But most people only ever hear the first half of that research. Add the second half and the conclusion flips.
First, the number that sounds like a sentence
In 2018, Hartshorne, Tenenbaum and Pinker published an unusually large study: data from 669,498 English speakers, modelled to separate current age, age of first exposure, and years of experience.
Their finding: grammar-learning ability holds up until roughly 17.4 years old, then declines.
That number travelled widely, usually under headlines along the lines of "after 17 it's too late."
Second, that finding is contested
In 2022, van der Slik and colleagues reanalysed the same data in *Language Learning* and reached a different conclusion: there is no sharp cliff at 17.4, but a gradual, continuous decline, with large differences between types of learner. The sharply defined critical age may have been an artefact of the analysis.
In other words, no door slams shut on a particular birthday. The change is gradual, and people vary enormously.
Third — and this is the part that gets skipped
Those studies measured grammar judgement: whether a sentence is well-formed.
They did not measure whether you can explain something in a meeting. They did not measure how much effort a listener spends understanding you.
And in pronunciation research specifically, the field made a clear shift over the past two decades:
The goal of speaking is not to sound native. It's to be understood.
Those two things have completely different age sensitivities:
- Accentedness — how native you sound — does get genuinely harder with age. That part is real. Be sceptical of anyone promising a native accent at any age.
- Intelligibility and comprehensibility — whether people catch your words, and how hard they have to work — improve measurably in adult learners who train.
The thing that gets harder is the thing you don't need. Plenty of people with obvious foreign accents work, present and teach in English with no friction at all. An accent isn't a problem. Not being understood is.
What actually decides whether you're understood
One more finding deserves its own line:
Errors of stress, rhythm and intonation hurt intelligibility more than errors on individual sounds.
Saying *th* as *s* usually doesn't stop anyone understanding you. But giving every word equal weight, failing to reduce the small words, putting the stress in the wrong place — that genuinely loses your listener.
Which lines up exactly with the 44-study shadowing review: the layer where shadowing has the strongest evidence is rhythm, stress and intonation — the same layer that decides whether you're understood.
Chain the three together:
1. What gets harder with age is mainly *sounding native*. 2. What actually determines communication is *being understood*. 3. What determines that is mostly prosody — and prosody responds to training, at any age.
Adult disadvantages and advantages
Disadvantages, honestly:
- A native-like accent is very unlikely.
- Your time is fragmented; you don't get the uninterrupted years a child gets.
Advantages, usually ignored:
- You already know what you want to say. A child learning a language is also learning how the world works. You aren't.
- You can think about your own thinking. You can diagnose your errors and target them. A child can only soak.
- Your goal is specific. You know whether you need meetings, interviews or ordering food, so you can train just that.
Over the time frame adults actually work with — months to a couple of years — these usually matter more than any notion of talent.
So what should the goal be
- Change the target from "no accent" to "understood without effort." The first is mostly unattainable; the second is very attainable, and it's the one you actually wanted.
- Train rhythm, not phonemes. On hearing a sentence, ask which word carries the beat before asking how any sound is made.
- Accept the accent. Clearly-spoken accented English is the global norm, not a defect.
- Don't use age as an excuse, and don't believe anyone selling a native accent at 40. Both are dishonest.
MirrorSay works on the rhythm layer: you say a sentence, it shows the native version and reads it aloud, you shadow it, and it compares your rhythm word by word. It doesn't promise to remove your accent — the research says that's hard, and we won't market what we can't do. It works on making you easier to understand, which by the evidence above is both trainable and the thing that matters.
Keep reading: Does Shadowing Work? · Connected Speech · You Understand English but Can't Speak It
References
- Hartshorne, J. K., Tenenbaum, J. B., & Pinker, S. (2018). A critical period for second language acquisition: Evidence from 2/3 million English speakers. *Cognition*, 177, 263–277.
- van der Slik, F., Schepens, J., Bongaerts, T., & van Hout, R. (2022). Critical Period Claim Revisited: Reanalysis of Hartshorne, Tenenbaum, and Pinker (2018) Suggests Steady Decline and Learner-Type Differences. *Language Learning*, 72(1).
- On intelligibility (rather than native-likeness) as the goal of pronunciation teaching, and on prosodic errors outweighing segmental errors for intelligibility, see the reviews by Levis and colleagues.
*MirrorSay is a free AI speaking-practice app: say a sentence, see how a native would say it, then shadow it. No sign-up, and your voice stays on your phone.*