Swain's noticing function applies to methods too. The gap between what you think you're doing and what you're actually doing is where immersion plans die. Each claim below gets checked against the literature, and most of them turn out to be half right, which is the annoying part.
half true, mostly irrelevant
Hartshorne's 670,000-person study did find a critical period, but read what it says. Grammar-learning ability stays high until about seventeen or eighteen, and the native-level ceiling really only bites for people aiming at grammar indistinguishable from a native's. Adults in lab studies out-learn children at the start. The honest version of the myth is that adults rarely reach native accent and grammar intuition. The honest rebuttal is that fluent, comfortable, fully functional adult-acquired language is common, and the same study found immersive exposure time, not age, was the biggest controllable predictor of how far people got. You're not competing with a seven-year-old. You're competing with last month's you.
true, with three asterisks
First asterisk: comprehensibility. Native TV aimed at natives is i+10 for a beginner, and Krashen's hypothesis only fires when you understand. Below about ninety percent comprehension it's noise, however many hours you sit through. Second: subtitles. With English subs on, eye-tracking work consistently shows viewers read the English and largely ignore the audio, so you're practising English reading. Target-language subs are the rung that works. Third: production. Canada's immersion students got years of massive input and still couldn't speak accurately. Watching builds comprehension, which is most of the battle but not all of it.
an overcorrection
Krashen argued learned rules never become acquired competence, and the Monitor's three conditions make explicit grammar nearly useless in live conversation. Fair. But even the researchers who argue with him keep a role for it. Ellis's work shows low-salience features, the quiet particles and subtle endings, are hard to notice in the flood unless someone points at them, and a lot of Chinese grammar is exactly that kind of feature. Long's focus on form is precisely brief attention to structure inside real communication. Nation budgets about a quarter of course time for deliberate language-focused learning. Grammar won't build your fluency, but as a noticing aid it speeds up what the input is doing. The failure mode isn't studying grammar. It's studying grammar instead of getting input.
it counts. it isn't the same.
There's no decent evidence for osmosis through fully-ignored audio. What passive immersion genuinely does is build phonological segmentation (hearing where words begin and end, which adults find nearly impossible in a new language at first), tune prosody and rhythm, which is make-or-break for tones, and add sample size during hours you'd otherwise spend in silence. Per minute it's worth a fraction of active immersion, but the minutes are nearly free, so the total is real. The dangerous version of this myth is replacing active hours with passive because passive feels easier.
depends on the language, and the week
The silent period Krashen described is real. Forcing production before the ear can parse the sound stream mostly practises errors and spikes the affective filter, and for tonal languages, speaking before you can reliably hear the four tones risks fossilising bad ones. But "never speak until perfect" is the opposite ditch, because Swain's noticing only happens when you actually attempt output, and tutor sessions become enormously productive from around month three or four. The synthesis: listen first for a few months, then start speaking and never stop, with someone willing to correct you.
it's the SRS layer with a mascot
Translation exercises, isolated sentences, streak gamification. That's deliberate language-focused learning, which is useful and which Nation budgets a quarter of course time for, but it isn't message-focused input and it isn't negotiation of meaning. The gamification exploits streak psychology, which is genuinely good for the consistency problem and genuinely bad if the streak becomes the goal. Use apps as a warm-up and a vocabulary supplement. A 500-day owl streak is not hours of comprehensible input and shouldn't be counted as them.
the country is now a URL
Hartshorne's data says immersive exposure predicts outcomes, and immersion is defined by using the language as a medium, not by passport stamps. Every layer in the At Home build reproduces a mechanism the country provides: native media for input, tutors and exchange apps for interaction and negotiation, phone-OS flips for ambient exposure, interest-driven content for a low affective filter. What moving abroad adds is social pressure and no escape hatch, which is real but reproducible with commitments: booked tutors, exchange partners who expect you, content you actually want to watch. The Canadian kids in 1965 didn't go to France either.
efficient at forgetting
Cepeda's 317-experiment meta-analysis: massed presentation is reliably worse than the same total time spaced out, and the optimal gap scales with your retention target. A word you want for life wants reviews spread over weeks and months, not one intense encounter. Nation adds the second problem: a word needs about twelve varied meetings, and a flashcard list gives you twelve identical ones. Mining sentences from real content spaces the meetings automatically and varies the contexts for free.
The number one killer. Enthusiasm buys a native crime drama in week two, comprehension sits at twenty percent, the affective filter slams shut, and the language becomes associated with feeling stupid. The fix feels wrong: go easier than seems worth your time. Easy content you finish beats hard content you abandon.
Beginner progress is fast and visible. Months six to eighteen feel flat because the remaining vocabulary is mid-frequency, rare enough that you meet each word slowly, and the wins stop being obvious. Everyone hits it. The people who come out the other side are the ones who kept the hours constant through the boring stretch. It's a cost-curve artefact, not a verdict on your talent.
Ten hours a day for a week, then nothing for three. Spacing punishes exactly this pattern. The floor-day exists so the chain never fully breaks, and a plan you can do tired beats a plan you can only do inspired.
Eleven apps installed, four half-built Anki decks, zero episodes watched. Setting up tools feels like progress and is a classic avoidance loop. One tool per layer, installed once, then the next hour goes to actual input. The pinyin chart doesn't teach you Mandarin, and neither does the hundredth video about it.