US Multilingual Percentages Don't Mean What You Think

Almost every article you read about American bilingualism quotes the same number: roughly 1 in 4 Americans can hold a conversation in a second language. That number is real. It is also from March 2001. It comes from a single Gallup telephone poll of 1,024 adults, conducted on landlines, before the iPhone existed. Gallup has not repeated the question in the 25 years since. The whole modern discourse about US multilingualism rests on a pre-9/11 phone survey that nobody bothered to replicate.

The other number in circulation is better sourced but worse understood. Per the Census Bureau's five-year American Community Survey (ACS) release covering 2018 to 2022, 78.3% of people age 5 and older spoke only English at home. Flip it around and you get a 21.7% bilingual number. That stat shows up in everything from think-tank reports to airline route announcements. It is closer to correct than the Gallup figure. It is also measuring something different from what most people think.

Worth pausing on what “bilingual” even means before we keep counting. In the research literature it's a spectrum: receptive (understand but can't produce), conversational (Gallup's bar: can hold a conversation), functional (read, write, work in the language), and balanced (equally comfortable in both). Every survey in this article uses a different cutoff, which is most of why the numbers fight. I'll flag which one is which as we go.

Two ideas are worth carrying through the rest of this. The Bilingual Barbell: the US isn't a bell curve from monolingual to multilingual, it's two weights, a heavy one on the immigrant side, a light one on the education side, and a long monolingual bar between them. The Instrument Mismatch: the headline American bilingual number was built in 1975 to trigger bilingual-ballot obligations under the Voting Rights Act, not to measure how many Americans speak two languages. When journalists line it up against Europe's self-reported 59%, they're comparing an administrative compliance tool to an opinion poll. Neither was designed for the job.

The Stat Everyone Quotes Was Measured on a Landline

The Gallup 2001 poll is worth understanding on its own terms, because its survival is the cleanest example of how a number becomes a fact by repetition. Gallup asked 1,024 adults a single question: “Can you hold a conversation in a language other than English, be it fairly or very well?” 26% said yes. The question has not appeared in a published Gallup national survey since. Twenty-five years have passed. The foreign-born share of the US population has grown from about 11% to over 15%. The Census non-English-at-home count has gone from roughly 47 million in 2000 to 67.8 million by 2019. The discourse hasn't caught up because the canonical poll hasn't been redone.

YouGov ran the closest public modern analogue in 2013 and got 75% of Americans reporting no second language at all, which flips to 25% with one. That 25% tracks the Gallup figure, but the sample size was again small and the methodology was online. Nothing at the scale and rigor of the original 2001 Gallup has been published. So when someone tells you “a quarter of Americans are bilingual,” they're citing a 2001 landline poll, a 2013 online poll, or a Census table that's measuring something else entirely.

The headline stat on American bilingualism is a pre-iPhone phone poll. The fact that it's still cited tells you more about the discourse than about the country.

The Census Measures Bilingualism by Accident

The ACS language questions are a remarkable piece of administrative plumbing, and almost nobody reads the methodology. There are three questions, and they cascade. First: does this person speak a language other than English at home? If yes, what language? If yes, how well does this person speak English (very well, well, not well, not at all)? That's it. The Census explains the design on its own site: “The current question design supports the 1975 amendment to the Voting Rights Act,” which requires bilingual ballots in jurisdictions where a minority language population exceeds a statutory threshold. The instrument exists to trigger ballot obligations. It was not built to count bilingual Americans.

The Bureau is honest about this in print. A 2022 Census article on languages in the US states explicitly: “In this article we refer to foreign-language speakers as those who report speaking a language other than English at home, not necessarily all those who can speak that language.” In other words: a native-born Texan who took four years of Spanish, works as a bilingual HR rep, and speaks English at home with her kids does not count as bilingual in the ACS. A kid who grew up in a Korean-speaking household but is more comfortable in English at work does count, because his answer about home is yes.

Now compare Europe. The Special Eurobarometer 540, published May 2024 from fieldwork in September and October 2023, asks respondents whether they can “have a conversation in a language other than their mother tongue.” No domain restriction. Not just at home, not just at work. Self-report, no test, no threshold beyond “a conversation.” The headline result: 59% of EU citizens say yes, up three points from 2012. 28% say they speak two or more foreign languages. 11% claim three.

• ACS asks about language use at home only. Eurobarometer asks about language ability in any setting.
• ACS measures English proficiency, not the other language's. Eurobarometer asks about the foreign language at a “conversation” threshold.
• ACS is a compliance instrument for bilingual ballots. Eurobarometer is an attitudinal opinion poll with no behavioral verification.
• Neither was designed to measure bilingualism. Both get cited as if they were.

Put it in one sentence: the Census undercounts educated bilinguals who don't speak their second language at home, and Eurobarometer overcounts Europeans who rated themselves generously with no proficiency test behind the answer. The 37-point gap between 22% and 59% is real in direction, but a meaningful chunk of it is pure definition.

America Is a Barbell, Not a Bell Curve

Here's the framework worth carrying. American multilingualism is a barbell with three distinct populations, and the aggregate “22% bilingual” number averages them into something nobody actually looks like. Precise slicing is hard because no single national survey asks the right question, but the shape of the distribution is clear.

1. Immigration bilinguals (the heavy weight). The foreign-born plus a large chunk of the second generation, very roughly 18-20% of the population. Per the Migration Policy Institute, only 17% of immigrants speak only English at home. Almost all of the ACS 67.8 million non-English-at-home speakers live here. It's the Spanish speaker in Houston, the Tagalog speaker in Daly City, the Mandarin speaker in Flushing.
2. Education bilinguals (the light weight). Native-born Americans who actually acquired a second language through schooling and kept it to a usable level. No national survey measures this cleanly, and self-reported fluency from high school language classes is notoriously inflated. What we have is direction: Modern Language Association (MLA) data shows total language enrollments in US higher education fell 16.6% between 2016 and 2021, the steepest drop on record, against a 29.3% decline from the 2009 peak. German alone fell 33.6%. The education pipeline for native-born bilinguals has been shrinking for a decade, whatever the current absolute size is.
3. The monolingual middle (the bar).Native-born English-dominant Americans who never acquired a second language to a usable level. Roughly three-quarters of the country sits here. This is the population the Gallup 2001 poll was really measuring against, and it's the part every European is describing when they say “Americans don't speak languages.” It almost never shows up in discussions of US multilingualism because the ACS doesn't ask them anything.

The averaging problem is real. If you take the 22% headline and treat it as “22% of Americans are conversationally bilingual,” you're conflating a Salvadoran immigrant who speaks Spanish at home and struggles with English, a second-generation Vietnamese American who speaks English at work and Vietnamese with grandma, and a Harvard grad who took four years of French in high school and can order coffee in Paris. Those three populations need three different numbers.

Europe Averages 59%. Luxembourg Hits 98%. The World Average Doesn't Exist.

The other favorite headline, “more than half the world is bilingual,” has an even shakier provenance than the US one. The claim traces to linguist François Grosjean, who wrote in his 2010 book Bilingual: Life and Reality that “more than half the world's population is bilingual.” He didn't cite a source. On his own site he repeats the claim without a primary citation. It appears to be an educated estimate by one scholar that got laundered through repetition into a UNESCO-flavored fact. There is no study, no census series, no cross-national aggregation that produces it.

What we do have are country-level measurements, and the direction of the pattern is clear even if the decimals aren't apples-to-apples.

Two things jump out. First, Canada outside Quebec posts a 9.5% English-French bilingual rate, and that number has been declining for 20 years. That's a narrower measure than the US 22% (it ignores Punjabi, Mandarin, Tagalog, and every other home language), but the direction is the opposite of what the “Americans are uniquely bad at languages” frame predicts. Second, the spread within Europe is enormous. Hungary at 13% is closer to the US than to Luxembourg, and Luxembourg's 98% is a product of a small country with three official languages (Luxembourgish, French, German), not a pan-European virtue.

Most of the World Speaks Two Languages. It Just Isn't in Europe.

The Europe-vs-America framing is how most American readers encounter multilingualism, and it misses the bigger story. Outside Europe, in places where nobody is running an Eurobarometer and where the ACS would melt down, bilingualism isn't a policy goal. It's furniture.

Afrobarometer's 20-country survey found 70% of Sub-Saharan Africans speak at least two languages, and 29% speak three or more. Indonesia's 2020 long-form census documents 694 regional languages alongside Bahasa Indonesia, which produces near-universal bilingualism for the 275 million people living there. Singapore's official bilingual education policy produces a 62.3% English-plus-Chinese rate among the ethnic Chinese population per the 2020 SingStat release. India's 2011 Census clocked 26% bilingual and 7% trilingual, and state-level variance is dramatic: Goa is 77% bilingual and more than half trilingual. The Philippines has had official bilingual Filipino-English education since the 1987 Constitution.

In all of those places, bilingualism happens for the same reason it happens in immigrant households in the US: one language is useful at home and in the community, another one is useful outside it. Nothing about this is European.

Monolingualism is a rich-English-country luxury, not a human default. The US, UK, Japan, and Australia are the weird ones, and it's because their first language happens to already be the global second language.

What We Actually Lost in 1917

The part of the story that rarely makes it into US multilingualism pieces is that America used to be much more multilingual than it is now, and the drop wasn't an accident. Before World War I, German was the second language of daily American life: hundreds of German-language newspapers, bilingual German-English public schools in more than a dozen states, and a foreign-born population share of 14.8% in 1890 (per Census historical working-paper series). That share collapsed to 4.7% by 1970 after the 1924 National Origins Act closed most immigration streams for four decades.

Inside that period, states banned non-English-language instruction outright. Nebraska's 1919 Siman Act outlawed teaching in any language other than English to students below high school. Similar laws passed in roughly 20 states across the same window. The Supreme Court struck Nebraska's law down in Meyer v. Nebraska (1923), but the damage was done: an entire generation of schools stopped teaching in German, Polish, Italian, and Czech. Within about 10 years the US was reengineered from one of the more multilingual developed countries into one of the least. That reengineering is what produced the monolingual middle we measure today.

Look at what's happening now. Dual Language Immersion (DLI) programs, where native-born American kids take core subjects in two languages through public schools, reached over 3,600 programs in 44 states by 2021, roughly triple the count a decade earlier. Pew's 2023 Latino language survey found that 49% of third-generation Latino parents still speak some Spanish to their children, stickier than the old three-generation assimilation model predicts. And Duolingo reported crossing 100 million monthly active users in 2024, and the US is its largest single market by revenue.

The monolingual middle isn't static. DLI pulls native-born kids into the bilingual weight from one side. Heritage-language retention is running longer than the old models expected on the other. The barbell is flattening, unevenly and mostly outside the ACS's field of view.

The Take

Every load-bearing number in this conversation is broken in its own way. The 25% Gallup figure is a 2001 landline poll nobody updated. The 22% Census figure is a Voting Rights Act compliance instrument that asks about home use, not conversational ability, and the Bureau is upfront about that. The 59% Eurobarometer figure is a self-report with no test and no threshold beyond “can hold a conversation.” The “half the world is bilingual” line traces to one scholar's 2010 book with no citation.

The real story isn't that Americans can't learn languages. It's that from roughly 1917 through 1970, US policy and demographics conspired to strip multilingualism out of the country. From roughly 1970 to today, immigration and immersion education have been putting it back. The canonical “Americans are bad at languages” frame is a snapshot of a system mid-reversal. What I'd watch from here: DLI program growth through the next American Councils canvass, third-generation heritage-language retention in the next Pew survey, and whether anyone finally updates the Gallup question. The interesting part of the American language story isn't the current 22%. It's the direction of travel, and the fact that the instrument we use to track it was built to do a different job.