Quick thought experiment. You're scrolling through language-app TikToks, half-watching a "polyglot" claim he picked up Mandarin in 3 months, and you think: sounds too good to be true, but what does the actual research say?
Good news: a lot. Researchers in linguistics and cognitive science have been quietly running meta-analyses — studies of studies, basically — that pool decades of experiments on how humans actually go about learning language. The findings are way more interesting (and more useful) than the average TikTok would have you believe. Some of them might even change how you spend your next thirty minutes of practice.
We've got 9 recent and fascinating metareviews, packed into a coffee break sized article.
1. Does Watching TV Help You Learn a Language? (Yes, With the Right Settings)
Good news for your Netflix queue: a 2025 meta-analysis of captioned video as a vocabulary-learning tool found a solid medium effect on word learning (g ≈ 0.56). An earlier synthesis on captioned video and L2 listening reached similar conclusions. Captions, as a rule, beat no captions.
But here's the wrinkle that doesn't make it into the explainer videos: subtitle type matters. Several meta-analyses suggest bilingual subtitles (your native language plus the target language together on screen) tend to outperform native-language-only subtitles, which tend to outperform target-language-only subtitles for raw vocabulary gains — though L2-on-L2 captions also help listening comprehension and reading. (L2 = Second Language) The takeaway isn't "captions = magic." It's: the medium is doing real work, but your settings are doing real work too.
For the Friends fans: yes, this is also why Joey's strategy in The One Where Joey Speaks French — pure mimicry, zero structure, no feedback loop — produces "blu blu blu blu blu" instead of French. Phoebe was right to give up.
2. Is It Too Late to Learn a Second Language? (You're Probably Not As Old As You Think)
The most stubborn myth in language learning is the idea that there's some magical childhood window — usually pegged at five, or maybe ten — and after that, you're cooked. Forever an awkward foreigner, mispronouncing things at restaurants.
In 2018, MIT and Harvard ran one of the largest language-learning studies ever conducted: an internet quiz that pulled grammar judgments from roughly two-thirds of a million English speakers worldwide. The conclusion? The cliff in foreign language ability shows up much later than the bedtime-story version of the "critical period" suggests — closer to age 17 or 18, not 5. Before that, learners are roughly equally capable. After it, learning continues; it just slows.
Translation: if you're 43 and panicking about being "too old" for Italian, you are not too old for Italian. You are, however, possibly procrastinating on Italian.
3. Spaced Repetition Is the Most Evidence-Backed Language Learning Method
Here's a number worth tattooing somewhere visible: in a 2022 meta-analysis pooling 48 spaced-practice experiments, learners who spaced out their practice landed at an effect size of g = 1.51, while learners who crammed it all together — same total time — landed at g = 0.97.
In meta-analysis terms, that gap is enormous. In normal-human terms: same hours, much more language stuck.
The mechanism, by the way, is not mysterious — it's the spacing effect, one of the most replicated findings in cognitive science, period. Twenty minutes today plus twenty minutes Thursday plus twenty minutes next Tuesday will absolutely demolish a single ninety-minute Sunday session. Your brain treats forgetting-and-re-remembering as a signal that something is worth keeping. Crammed material gets flagged as "we already have this, no need to file it." Then it gets quietly thrown out.
If you take exactly one thing from this post, take this one.
4. Does Explicit Grammar Instruction Work? Why Pure Immersion Leaves Gains on the Table
The bumper-sticker version of language pedagogy is "just immerse yourself, like a child!" It's romantic. It's also, according to the meta-analyses, not quite right.
Across three big meta-analyses spanning 25 years — Norris and Ortega in 2000, Spada and Tomita in 2010, and Goo and colleagues' 2015 update — explicit instruction (someone actually telling you the rule, or directing your attention to a specific pattern) consistently outperformed implicit instruction (figuring it out from exposure alone) on most outcome measures.
There are caveats. Critics point out that the tests in these studies tend to measure explicit knowledge, which favors explicit teaching by design. Implicit exposure probably builds different, slower-blooming skills — the kind that show up when you're tired and ordering wine and stop having brain-cycles for rules.
But the practical implication is clear: the "no rules, just vibes" approach to a new language is leaving gains on the table. A short, clear explanation of why a verb conjugates the way it does isn't a betrayal of authentic learning. It's a shortcut your brain quietly appreciates.
5. Language Learning Motivation vs. Consistency: What the Data Actually Shows
There's a popular framework in the motivation literature called the L2 Motivational Self System — the idea that the more vividly you can imagine yourself as a fluent speaker, the more likely you are to put in the work. Vision boards for verbs, basically.
A 2018 meta-analysis aggregating data from over 32,000 language learners found that this "ideal L2 self" really does correlate with how much intended effort people report. The catch? When you look at actual achievement — not "how hard I'm planning to try," but "how good I got" — the correlations get noticeably thinner.
The implication is one of those mildly uncomfortable truths the data keeps surfacing across fields: visualization is great for starting. It is not, by itself, a substitute for showing up. The system that gets you back to the lesson on a tired Wednesday matters more than the daydream that got you to download an app on a fired-up Sunday.
6. Does Study Abroad Improve Language Proficiency? What Research Says About Program Length
A 2024 multilevel meta-analysis on study-abroad programs sorted decades of research by trip length and got a brutally clean result. Long-term programs (six months to a year) produced enormous gains (g = 1.77). Mid-length (one to six months) produced solid ones (g = 0.82). Short-term (a month or less) produced something pretty close to a souvenir keychain (g = 0.36).
This is the part that should redistribute some life decisions. The two-week language vacation is genuinely fun. It is not, statistically, a major proficiency event. (Becky, nobody believes that you went to Cancún for 2 weeks for "your Spanish"). The actual ingredients — time, immersion, daily forced output — do their thing on a calendar measured in months, not weekends.
The good news, hidden in this finding: most of what makes study abroad work (frequent contact with speakers, daily speaking, low-stakes mistakes, exposure across contexts) can be approximated at home. You don't need a plane ticket. You need a routine that mimics the conditions.
7. How Many Times Do You Need to See a Word to Learn It?
A 2019 meta-analysis pooled 26 studies on incidental vocabulary learning and asked a deceptively simple question: how many times does someone need to encounter a word before it sticks? The answer, in effect-size form, was a clear medium-sized boost from repeated encounters (r ≈ 0.34) — and the more controlled variables they layered in (spacing, engagement, visual support), the bigger the effect got.
This is unsexy. Nobody wants to hear "you have to see quotidien about fifteen more times." But this is one of the rare findings where the science and the grandma advice agree completely. Words you encounter once die. Words you encounter many times, in different contexts, with little hits of effortful recall in between, become yours.
And if you had to pick one word to practice in every language 15 more times, it's indubitably 'indubitably' because how else will you let people know you're both a scholar and a gentleman?
How to Learn a Language Based on the Research
| Finding | Study | Key Number |
|---|---|---|
| Captioned video builds vocabulary — subtitle type matters | 2025 meta-analysis of captioned video | Effect size g ≈ 0.56 |
| The "critical period" ends at ~17–18, not childhood | MIT/Harvard 2018, 670,000 participants | Ability holds constant until ~age 17–18 |
| Spaced practice outperforms cramming significantly | 2022 meta-analysis, 48 studies | g = 1.51 (spaced) vs. 0.97 (massed) |
| Explicit grammar instruction outperforms immersion alone | Norris & Ortega 2000, Spada & Tomita 2010, Goo 2015 | Consistent across 25 years of meta-analyses |
| Motivation predicts intended effort, not actual achievement | 2018 meta-analysis, 32,000+ learners | Moderate correlation; visualization helps starting, not sustaining |
| Study abroad payoff scales sharply with program length | 2024 multilevel meta-analysis | g = 1.77 (6–12 mo), 0.82 (1–6 mo), 0.36 (≤1 mo) |
| Repeated word encounters drive incidental vocabulary retention | 2019 meta-analysis, 26 studies | r ≈ 0.34; grows with spacing and engagement |
Stack the deck. The research, taken together, draws a pretty specific picture of what effective language learning looks like:
You space your practice instead of cramming it. You get clear, structured explanations of the patterns — not just exposure. You consume real media with the right kind of subtitles. You see new vocabulary repeatedly, in varied contexts, ideally with little forced retrievals along the way. You convert "I want to be fluent someday" into a system that survives the Wednesday slump. And you stop waiting until you can afford six months in Lisbon.
If you've been learning a language for a while and quietly suspect that your routine doesn't actually do most of those things — well, you're in good company, and you're noticing the right thing. The gap between how the science says we should learn and how the average app or class actually teaches is wide.
Atlas Runa is built around exactly this research — spaced review timed to when you're about to forget, reading pitched to where you actually are so vocabulary stacks through real encounters rather than isolated lists, and a Progress Log that turns a week of practice into a concrete next move.
Sources
- Hartshorne, J. K., Tenenbaum, J. B., & Pinker, S. (2018). A critical period for second language acquisition: Evidence from 2/3 million English speakers. Cognition.
- Kim, S. K., & Webb, S. (2022). The effects of spaced practice on second language learning: A meta-analysis. Language Learning.
- Kurokawa, M. (2025). Incidental vocabulary acquisition through captioned viewing: A meta-analysis. Language Learning.
- Montero Perez, M., Van Den Noortgate, W., & Desmet, P. (2013). Captioned video for L2 listening and vocabulary learning: A meta-analysis. System.
- Goo, J., Granena, G., Yilmaz, Y., & Novella, M. (2015). Norris and Ortega (2000) revisited and updated: Implicit and explicit instruction in L2 learning. In Implicit and Explicit Learning of Languages.
- Al-Hoorie, A. H. (2018). The L2 motivational self system: A meta-analysis. Studies in Second Language Learning and Teaching.
- Tseng, W.-T., Liu, Y.-T., Hsu, Y.-T., & Chu, H.-C. (2024). Revisiting the effectiveness of study abroad language programs: A multi-level meta-analysis. Language Teaching Research.
- Uchihara, T., Webb, S., & Yanagisawa, A. (2019). The effects of repetition on incidental vocabulary learning: A meta-analysis of correlational studies. Language Learning.
- Vos, J. F., Schriefers, H., Nivard, M. G., & Lemhöfer, K. (2018). A meta-analysis and meta-regression of incidental second language word learning from spoken input. Language Learning.