Why Are Some Countries Bilingual? Borders, Conquests and Deals

The routes are historical and the route shapes the politics that follow:

• •Borders drawn around multiple existing communities, as in Switzerland, where German, French, Italian and Romansh speaking cantons joined a confederation that never claimed to be one nation
• •Conquest or annexation, where a population speaking one language comes under a state speaking another, as with French in Canada after 1763 and Welsh, Irish and Scots Gaelic under English rule
• •Colonial imposition, leaving a European language as the language of administration alongside many indigenous ones, which describes most of Africa and South Asia
• •Migration, producing large communities of speakers without any territorial claim, as with Spanish in the United States
• •Deliberate revival or elevation, as with Hebrew in Israel, which was rebuilt as a spoken vernacular from a liturgical language within a few generations
• •Indigenous recognition after a period of suppression, as with Maori in New Zealand, recognised as official in 1987 after decades in which children were punished for speaking it

The central policy choice is between two models. Under the territorial principle, each region has one official language and services are provided in it, so a citizen moving between regions changes language; this is the Swiss and Belgian arrangement and the basis of most Canadian provincial policy. Under the personal principle, a citizen carries a right to use their language anywhere in the country, which is how Canadian federal services operate and which is far more expensive. The territorial model produces stability and protects a minority language within its area by making it dominant there, and it produces hard borders and disputes about exactly where the line runs, which is why Belgium's language frontier was fixed by law in 1962 and why the municipalities along it have their own contested arrangements. The personal model is more generous to individuals and, applied alone, tends to favour the larger language, since a minority speaker who can be served in the majority language eventually is.

The comparison is instructive because the arrangements look similar on paper. Switzerland's cantons predate the federal state and hold substantial powers over education and culture, the language communities do not map onto the religious or political divisions, and no single group can dominate since German speakers, though a clear majority, are divided among cantons and use a dialect distinct from standard German that limits their identification with Germany. Belgium's Dutch and French communities have overlapping grievances stretching back to a nineteenth-century state in which French was the language of government and advancement despite Dutch speakers being the majority, and the subsequent economic reversal, with Flanders prospering as Wallonia's heavy industry declined, layered resentment on both sides. Belgium's response has been to federalise progressively, transferring powers to language communities and regions to the point where the country has multiple parliaments and no national political parties, and government formation has repeatedly taken over a year.

The research on individual bilingualism has been through a correction worth knowing about. A body of work from the 1960s onward reported a bilingual advantage in executive function, meaning better attention control and task switching from the constant management of two active languages, and it was widely reported. Subsequent analysis found that studies showing an advantage were far more likely to be published than those that did not, and large well-controlled studies have often failed to find the effect, so the current position is that any cognitive advantage is small, inconsistent and possibly absent. What is well established is different and more practical: bilingual children pass language milestones on a normal schedule when both languages are counted together, mixing languages within a sentence is a sign of competence rather than confusion, there is no evidence that early bilingualism causes delay, and the evidence on dementia onset being delayed in lifelong bilinguals is suggestive and not settled. The social effects are clearer than the cognitive ones, since access to education, employment and public services in a language a person actually speaks is a substantial determinant of outcomes.

Multilingual settlements fail in identifiable ways. A language loses ground when it is excluded from the domains that carry status and income, chiefly schooling, administration and employment, which is what happened to Irish despite its constitutional position and to numerous minority languages across Europe. Explicit suppression, including punishing children for speaking a language at school, was policy in many countries within living memory and its effects persist for generations. Demographic change alters the balance regardless of policy. And symbolic recognition without resources achieves little, since a language that is official in law and absent from broadcasting, publishing and the workplace continues to decline. The measures that have demonstrably reversed decline, in Catalonia, Wales, Quebec and to an extent among Maori, share features: immersion education, a guaranteed presence in media, a requirement for the language in public employment, and a large enough population of young speakers to use it with each other, which is the part that cannot be legislated.

Countries become multilingual through federation of existing communities, conquest, colonial administration, migration or deliberate revival, and the route determines how contested the arrangement is. The central policy choice is between territorial rules, where each region has one language, and personal rights carried anywhere, which is more generous and tends to favour the larger language. Switzerland's arrangement holds because cantons predate the state and no group can dominate, while Belgium's has federalised repeatedly under pressure. Claims of a cognitive bilingual advantage have largely not replicated.