Is 20% of the day enough for a heritage language?

Nobody can say yet, and this article does not pick a side. King and Mackey, in The Bilingual Edge, report a study in which as little as 20% of total exposure was enough for children to develop a spoken vocabulary, and below that they were very hesitant to speak. Carroll (2017) argues that no single exposure percentage can work as a general rule. Treat 20% as a rough line to check against, not as a pass mark.

What the 20% figure says

King and Mackey describe a 1997 study by Pearson, Fernández, Lewedeg and Oller. It followed 25 Spanish-English bilingual children with different amounts of exposure. Vocabulary in each language grew in proportion to exposure. But with as little as 20 percent of total exposure in one language, children were still able to develop a productive vocabulary in it. Below that, they were very hesitant to speak the language at all.

A caution on strength. The notes behind this site record that King and Mackey's summary was read, and the 1997 paper was not. That makes the figure REPORTED, secondhand, and not something checked at the source.

What the challenge says

Susanne Carroll published "Exposure and input in bilingual development" in Bilingualism: Language and Cognition (online 2015, in print 2017). Its existence was checked against the journal's page. As this site's notes describe it, from the abstract and a secondary description and not the full text, the argument is that treating the amount of exposure as one variable that predicts outcomes is conceptually unsound. Terms like "exposure" and "input" are underspecified, and different parts of a language have different learning requirements, so one percentage cannot be a general law.

The same notes say Carroll does not appear to name Pearson or the 20% figure. So this is a general objection to threshold thinking, not a direct rebuttal of one study. Whether the objection weakens the 20% line, and by how much, is the open question. Both statements stand side by side here: the study exists, and the objection exists.

What the app does with the line

The app keeps 20% as a single constant and shows it with the hedge attached. The Family Language Audit measures your heritage-language share against it, and the plan generator uses it to say what 20% would mean in minutes a day. Both phrase it as one study's debated figure. With three or more languages the audit does not use the 80% line as a verdict, and shows a per-language breakdown instead.

To turn a share into hours, the app needs the child's waking hours. These are the app's own working estimates by age band, not from a cited source (UNVERIFIED). Twenty percent of them:

• Ages 0 to 2: 11 waking hours a day, so 2.2 hours a day, or 15.4 a week.
• Ages 3 to 5: 11.5 waking hours a day, so 2.3 hours a day, or 16.1 a week.
• Ages 6 to 9: 12.5 waking hours a day, so 2.5 hours a day, or 17.5 a week.
• Ages 10 and up: 13.5 waking hours a day, so 2.7 hours a day, or 18.9 a week.

An illustration, not data

ILLUSTRATION built from the app's arithmetic, not data about real families. Take a child aged 4: 80.5 waking hours a week, so the 20% line sits at 16.1 hours. I ran three invented weeks through the audit code:

• Week A: a parent with 10 hours in the heritage language. Share: about 12%. Below the line.
• Week B: the same 10 hours, plus a grandparent with 4 hours in the heritage language. 14 hours, about 17%. Still below the line.
• Week C: the same 14 hours, plus a nanny who switches between languages for 12 hours. The audit counts a switching caregiver at half weight, so 6 of those hours count. 20 counted hours, about 25%. The audit reports it as past the 20% line.

The weeks are invented. What comes from the app is the method: full weight for heritage-only hours, half weight for switching hours, and a division by waking hours. The half weight is the app's own choice, with no source behind it (UNVERIFIED). A different weight would move week C across or away from the line.

How to use a line that is not settled

King and Mackey also give a different and higher figure, for a different purpose. For the early years they suggest a goal of roughly 80 percent of child interaction in the weaker, minority language, because the majority language reaches the child from everywhere else. That is a target for protecting the minority language, read directly in the book (VERIFIED). It is not a correction of the 20% line, and the 20% line is not a correction of it. They answer different questions, and this article does not rank them.

1. Log one real week in the audit, with the caregivers you tend to forget.
2. Read your share against both lines, and read the 20% line as one study's debated figure.
3. If you are below a line, use the audit to see which caregivers and hours add to the count, then decide what to change. A share is a count of hours, not a verdict on your child.

What is still unknown

• Whether 20% suits any particular child. No source read for this site says so.
• Whether Carroll's objection applies to Pearson's specific figure. The notes say it does not name it, and the full text was not read.
• What the 1997 paper itself says. Only King and Mackey's account was read.
• How close the app's waking-hours estimates and half weight are to real life. Both are UNVERIFIED.