California has changed. The state that voted Republican in every presidential election from 1968 to 1988 is now one of the most Democratic in the nation, having rejected Donald Trump by a larger margin than any state except Hawaii. Perhaps an even better illustration of the state’s political-cultural shift, however, is the passage last November of Proposition 58, which reinstates bilingual education. Thanks to the “English for the Children” ballot initiative in 1998, the default pedagogy in California’s public schools had been English immersion, with bilingual programs requiring a parental waiver. Now Prop 58 will allow schools to implement dual-immersion or other types of bilingual instruction unless parents opt out.

Prop 58 is intended to help children “not only learn English as soon as possible, but retain their native language and learn another language, which we know today—in this world, this economy—we have to speak multiple languages in order to excel,” sponsoring Sen. Ricardo Lara told constituents, in Spanish, at a promotional event last fall. His view, that a bilingual populace is a source of strength rather than of division, is increasingly popular in political and academic circles. As we will see, however, advocates are overstating the benefits of bilingualism and ignoring its costs—especially the danger of cultural fragmentation.

It is first important to understand the prevalence of Spanish in the southwestern United States. The percentage of California school-age children who speak Spanish at home rose from 25 percent in 1990 to 35 percent in 2015. The state now has 22 counties where at least one-third of school-age children speak Spanish, including 46 percent of children in Los Angeles, 60 percent in Monterey, and 71 percent in Imperial County. The same figure rises to over 80 percent in parts of southern Texas and Arizona. Even Kansas and Nebraska—hardly traditional immigrant destinations—now have counties where over half the school-age children speak Spanish at home.

Spanish-speaking does decline with acculturation. As a recent Pew Center report points out, the percentage of U.S.-born Hispanics who speak Spanish at home has decreased from 67 percent in 1980 to 59 percent today. But with the U.S. Hispanic population growing, and the size and variety of Spanish-language media growing with it, the number of Spanish speakers in the U.S. has risen dramatically—from 11 million in 1980 to 40 million in 2015.

For many years, Americans worried Spanish speakers may not learn English. Although concern persists over their degree of literacy, the vast majority of U.S-born Hispanics report speaking English well. So the concern here is not about English, but Spanish: what fraction of Hispanic Americans will remain bilingual?

Unlike other foreign languages brought to the U.S., Spanish has had unusual staying power. In a 2002 study, sociologist Richard Alba and his coauthors found that nearly all the grandchildren of European and Asian immigrants speak English exclusively, but over one-third of Mexican Americans still speak Spanish in the third generation and beyond. A 2011 Pew Center poll found that 30 percent of third-generation Hispanics are at least as proficient in Spanish as they are in English, while 47 percent can carry on a Spanish conversation “very well” or “pretty well.” The same survey indicated that Hispanic Americans are nearly unanimous (95 percent to 4 percent) in believing it is important that “future generations of Hispanics living in the U.S. be able to speak Spanish.”

All of this is good news, say bilingual advocates. They offer two main arguments in favor of encouraging immigrants and their descendants to retain their ancestral language (usually Spanish), while also learning English. The first is that bilingualism confers general cognitive benefits on individuals. “A large body of research has demonstrated the cognitive, economic, and long-term academic benefits of multilingualism and multiliteracy,” reads the text of California’s new law.

For decades, many psychologists did believe that speaking two languages improves “executive function,” meaning cognitive skills associated with focus, memory, and reasoning. The theory was that since bilinguals must mentally inhibit one language while speaking the other, they develop keen focus and control over cognitive tasks. But the link between executive function and bilingualism is one of many findings that have recently fallen victim to the “replication crisis.” In a profession that rewards novelty, academics and the journals that publish them often take even the slightest hint of a positive finding and run with it, downplaying or ignoring null results. Since researchers have become increasingly interested in large-scale replications in recent years, they have had trouble verifying some of the most well-known results in social psychology.

The alleged link between executive function and bilingualism is no exception. Writing in the academic journal Cortex in 2016, psychologist Kenneth Paap recalled how he and his colleagues initially “had strong expectations that we would replicate a strong advantage” for bilinguals on executive tasks. But they couldn’t. “Three studies, three additional tasks, and 273 participants later we reconsidered the hypothesis and … what changed our mind was simply the weight of the evidence.” Many psychologists are no longer convinced that bilingualism improves executive function at all. Given the state of the evidence, there is no clear case for encouraging bilingualism in the U.S. on cognitive grounds alone.

The second argument for bilingualism is that it would put the U.S. in a stronger economic and political position in the world. “The goal [of Prop 58] is to learn both languages because the globalized world demands communication beyond the national language,” according to a Spanish-language editorial in La Opinión, the second-largest newspaper in Los Angeles.

But any global advantage that bilingualism would confer upon the U.S. is likely small. Many of the most successful trading nations today, such as South Korea and Japan, are linguistically homogeneous. China has become an economic powerhouse largely by manufacturing products for the North American and European markets, yet only a tiny fraction of the Chinese people are fluent in English or another European language. The U.S. itself became the world’s foremost economic and military power at a time when immigration was low and full assimilation was encouraged.

A useful “reality check” on the benefits of bilingualism is to look directly at educational and economic outcomes. On both the math and reading sections of the National Assessment of Educational Progress, 12th-graders who speak English well but also speak another language at home score slightly lower than students who speak only English at home. Furthermore, adults who were born in the U.S. but remain bilingual have a lower median income and are less likely to complete college than U.S.-born adults who speak only English. This is not to say that bilingualism depresses achievement—there are many other factors at work here—but the claim that bilingual Americans are economic dynamos is unfounded.

Learning a second language can be useful, of course. It is a wonderful way for students to learn about a different culture and to reflect on the structure of their own language. In addition, mastering in-demand languages can be financially lucrative for soldiers, diplomats, intelligence analysts, academics, and business travelers. Foreign-language instruction should continue for all of those purposes. However, widespread bilingualism is neither essential for an individual’s cognitive development nor a prerequisite for a nation’s success on the global stage.

And while the nationwide benefits of bilingualism are limited, the dangers are real. Having 40 million Spanish-speakers—geographically concentrated in the Southwest, sharing Latin American roots, and easily linked by modern communication and proximity to their ancestral countries—risks a cultural bifurcation between native English and Spanish speakers in the U.S.

Signs of linguistic division in the U.S. are already evident, starting with the media. In the first week of the 2016 May “sweeps” period, the top 10 television shows for the nation as a whole were: NCIS, The Big Bang Theory, NCIS: New Orleans, Dancing with the Stars, The Good Wife, Blue Bloods, Madam Secretary, The Voice, Empire, and 60 Minutes. In Hispanic households, shows broadcast by Univision and Telemundo occupied every slot of the top 10: Nuestra Belleza, El Hotel de Los Secretos, Un Camino Hacia el Destino, El Señor de Los Cielos, Yago, Caso Cerrado, La Voz Kids, Sueño de Amor, Crónicas de Sábado, and Eva la Trailera.

Non-Hispanics in the U.S. are unlikely to have heard of any of those Univision and Telemundo shows, nor are they generally aware of other Spanish-language media that are massively popular among Hispanic Americans. Consider the example of Jenni Rivera. The singer born and raised in California was a superstar, having sold over 1 million albums in the U.S. alone. Yet when she died tragically in a plane crash, most Americans had never heard of her. The reason? She sang her songs in Spanish, so almost all of her fans were Hispanic. In admitting that his newspaper had never once mentioned Jenni Rivera’s name before her death, a writer for the Washington Post lamented that “it’s possible to live in parallel Americas, with the larger part only dimly aware of the enormous things happening in the other one.”

Separate media lead to separate political messages. We often hear of “dog whistles” and “coded language” that politicians use to appeal to a particular interest group without alienating the general electorate. In the U.S. today, Spanish is the ultimate “code” for speaking to Hispanics about immigration and other issues on which non-Hispanics may have opposing views.

An egregious example occurred last fall, when Arizona Sen. John McCain put two very different immigration positions on his campaign website. The Spanish version of his site touted his work on behalf of “immigration reform that is humane and sensible to the needs of the immigrant community,” including his leadership on the Schumer-Rubio amnesty bill from 2013. The English text on McCain’s site, however, featured tough talk on border enforcement exclusively. No mention of amnesty. No mention of the “needs of the immigrant community.”

A related controversy arose during the Republican presidential primaries, when Ted Cruz implied that Marco Rubio spoke more favorably of amnesty on Spanish-language Univision than he did on English-language media. “First of all, I don’t know how he knows what I said on Univision because he doesn’t speak Spanish,” Rubio replied, unintentionally highlighting the problem with conducting a campaign in two languages.

Political systems obviously do not function well when the message one constituency hears is unintelligible to another constituency. How often politicians deliver these mixed messages is impossible to know, but there are plenty of opportunities. Mitt Romney ran over 3,000 ads on Spanish television in 2012, and Barack Obama ran a remarkable 13,000, totaling about $23 million in spending between the two of them. In fact, it is so common for candidates of both parties to run Spanish-language campaigns that it was news when Donald Trump chose not to do so.

Outside the national spotlight, linguistic divisions have subtle but important effects on community life. Researchers have found that “social capital”—meaning the bonds of trust and reciprocity that people forge within their communities—plays a crucial role in life satisfaction. Political scientist Robert Putnam explains: “Where levels of social capital are higher, children grow up healthier, safer, and better educated; people live longer, happier lives; and democracy and the economy work better.”

Linguistic diversity reduces social capital. Two Danish academics, Cong Wang and Bodo Steiner, published a 2015 paper in Economic Record showing that more national languages correlate with lower levels of social capital across a sample of 68 nations. Specifically, multilingualism leads to “lower levels of social trust, fewer memberships in social organizations, and deteriorated social norms and structures.” These relationships survived the authors’ rigorous tests for causality.

The impact of linguistic diversity on social capital is felt within nations as well as across. Data from Putnam’s Social Capital Community Benchmark Survey show that American communities with larger proportions of foreign-language speakers suffer from less neighborly trust, even after controlling for a host of socioeconomic variables. Putnam is famous for finding that ethno-cultural diversity causes people to trust everyone less, including members of their own group. He described people in diverse communities as turtles retreating into their shells, disengaged from civic life. Returning to the example of bilingual California, Putnam’s data show that Los Angeles and San Francisco are among the U.S. cities with the least social capital.

Are there any reasons for optimism? One view is that the U.S. will eventually embrace bilingualism, learn to value ethno-cultural diversity, and develop the kind of “bridging” social capital (meaning cross-group ties) that Putnam envisions. This scenario requires the U.S. to be a very exceptional nation. For example, despite a reputation for accommodating diversity, Canada has failed to prevent bilingualism from dividing its society along ethno-cultural lines. Many of Canada’s Québécois can speak English, but their shared French language helps maintain a culture and identity distinct from the rest of Canada. The resulting tensions have left Quebec flirting with secession for decades. Western Europe has experienced similar conflicts involving Basque-speakers in Spain and Dutch-speaking Flemings in Belgium. These examples are not perfectly analogous to the U.S. situation, of course, but they illustrate how intractable linguistic differences can be, even for Western nations with progressive values.

If our goal is to foster a more culturally unified, high-trust nation where social institutions are effective and reliable, then an initiative such as California’s Prop 58 is a step in the wrong direction. Encouraging bilingualism risks further fragmentation of a country that, as the 2016 election illustrated, already suffers from internal divisions that will be difficult to heal. It is not clear, however, that promoting English immersion is sufficient in an age of mass immigration. Even if immersion were the norm at school and in the workplace, the children of immigrants would continue to learn their ancestral language at home, and there will always be a supply of foreign-language media as long as there is a demand. To promote assimilation, perhaps we also must slow down the arrival of newcomers who speak foreign languages.

Optimists say that restricting immigration is not necessary. They argue that European immigrants brought many different languages to the U.S. in the early 20th century, but the relentless power of assimilation converted their descendants to English-only. Therefore, the optimists say, assimilation will not only wring Spanish from the children and grandchildren of today’s immigrants, but it will also enable the U.S. to absorb an unremitting flow of immigrants without any foreign language ever establishing a permanent foothold.

Such faith in the power of assimilation may be misplaced. For one thing, the non-English media and social institutions present in the U.S. in the early 20th century did not simply fade away with assimilation’s quiet tide. It was World War I and attendant rumors of disloyalty that jumpstarted the process. What followed was a period of low immigration from the mid-1920s to the mid-1960s that gave assimilation a crucial boost. Furthermore, much has changed in 100 years about both American society and the immigrants who are entering it. In contrast to their European forerunners, immigrants to the U.S. after 1965 had a dominant language (Spanish) and came mainly from countries in close proximity to the U.S. The ease of travel and communication between nations better preserves ties to the old country these days, and the ideology of multiculturalism in the U.S. now encourages immigrants to retain their old identities.

In short, there is no law of nature that says all groups will always assimilate in America. Times change, and so does the ability of our society to manage cultural and linguistic differences. As the source of legal immigration to the U.S. now shifts more toward Asia and Africa, we must carefully weigh the consequences of even greater linguistic diversity on the horizon.

Jason Richwine is a public-policy analyst in Washington, DC.