The Multilingual Gift | Psychology Today Singapore

In 2001, Angela Friederici invited me to spend a month at the Max Planck Institute for Cognitive Neuroscience in Leipzig. I arrived speaking no German. That month taught me what it felt like to be a foreigner in another country.

A year later, I went back with my family. My children learned through immersion. On the outside, it seemed effortless to me. I learned differently. Slowly. With effort. Humbled by watching my 6-year-old daughter speak fluently in a year. I was happy to be able to order a taxi, buy bread at the local bakery, and order coffee.

Twenty-three years later, with considerable effort and several return visits, my German remains functional but limited. Or so I thought.

Last year, colleagues were preparing a Festschrift honoring Angela’s retirement. They asked if I would contribute a short reflection. My first instinct was to write it in English. A German academic of her caliber writes almost all of her professional work in English. But I wanted to write it in German. For her retirement, I felt it would be more personal. I was also grateful. Angela had opened a door to Germany for me, one that led to research collaborations and to recovering a lost language and heritage, which I have written about elsewhere. Writing to her in German would be a way of honoring my great-grandfather, who emigrated to Mexico more than a century ago.

I couldn’t write it at the register appropriate for honoring such an important neuroscientist in my field. So I used AI.

Here are the first two sentences of what I actually dictated:

Liebe Angela, vor 25 Jahre habe ich eine Einladung von dich bekommen. Ich habe nach Deutschland anreisen und spreche kein Deutsch.

In plain English:

Dear Angela, 25 years ago I got an invitation from you. I had never traveled to Germany and I spoke no German.

Show More

My spoken German is not bad. But dictating it into my phone revealed how inadequate it was for the occasion. When I asked Claude to create a letter for a book she would be given in honor of her retirement using my German, I got back something entirely different:

Als ich vor fünfundzwanzig Jahren eine Einladung von dir erhielt, ahnte ich nicht, wie sehr dieses Schreiben mein Leben verändern würde.

A closer English equivalent might read:

When I received your summons a quarter century ago, I could not have anticipated the degree to which that correspondence would alter the course of my life.

The AI hadn’t written my letter. I had. The AI had translated my German into another register of German, not across languages, but within one. Claude called it a linguistic prosthetic: a tool that extends expressive range while preserving what the speaker actually means to communicate.

From a neuroscientific perspective, this distinction matters. Multilinguals activate overlapping but distinct neural networks for each of their languages—what researchers call different state spaces in the brain. When I switch languages, it’s not just vocabulary or syntax that changes. Each language feels different. The posture of my tongue changes. I tend to restrain my prosody in languages I don’t fully own. To me, each language harnesses a different combination of grammar, meaning, and intonation.

What surprised me was how differently AI interacts with each one.

In Spanish, my mother tongue, AI changes almost nothing. Spanish lets me be intellectually rigorous and emotionally warm simultaneously, not by switching between registers but by integrating them naturally. When I write una soledad profunda (a deep loneliness), I’m using a sophisticated concept that doesn’t create distance. It feels human, immediate, lived. AI has nothing to add because Spanish already does what I need it to do.

In English, a native language and also my primary academic language, AI sometimes introduces rhetorical tics that aren’t quite mine, a slightly different personality that I have to edit back out. English gives me precision through lexical abundance, but it demands a certain distance. AI can amplify that distance if I’m not careful.

I was only fully immersed in Brazilian Portuguese at 20 in São Paulo. The initial impulse to learn it was driven by Brazilian music, which bled into the language itself. I felt the elongated vowels and the softening of the consonants that make it sound more like the smoothed edges of French than the staccato of Mexican Spanish. AI handles Portuguese competently, but it can’t replicate what makes the language alive for me: its embodied musicality.

German is the language where AI functions as something entirely different—not an editor, not a translator, but a prosthetic that extends my reach into a language I inhabit only partially. I have referred to it as a German Alien, another being that talks through me. The neural circuits exist—they can fire automatically in certain situations—but they feel foreign. In writing, AI helps immensely. I generate the semantic framework. AI provides the syntactic architecture. The result is a collaborative production that neither of us could achieve alone.

My experience with AI has added to my thinking about what multilingual competence actually means. Language proficiency tests measure listening, speaking, reading, and writing across everyday and academic registers. However, multilingualism is not a set of on-and-off switches. I’ve come to see this as a form of distributed multilingualism, where linguistic competence exists across human and artificial systems working together. For me, each language partnership looks fundamentally different.

My German holds Leipzig, my children’s fluency, and the discovery that my family’s Mexican Christmas tradition, round fried dough dusted with sugar, was actually German Pfannkuchen, carried across an ocean by my great-grandfather and converted into something neither German nor Mexican but both. AI’s German holds none of that.

Language choice signals symbolic belonging. When I wrote to Angela in German, I wasn’t just conveying information. I was performing a return. I was saying: This language is mine too, even if I need help to write it proficiently. My children recovered German through immersion. I recovered it through AI partnership. Both are valid forms of linguistic repatriation. The difference is in the mechanism, not the legitimacy.

The multilingual mind has long been studied for what it might engender: interference, cognitive costs, the challenge of maintaining multiple systems, and how that might delay cognitive decline. But multilingual minds carry something else, something that AI never will: grounded, embodied, autobiographically anchored experience in each language. What AI reveals, by contrast, is precisely what human multilingualism is: not parallel fluency, but layered, uneven, deeply personal access to different worlds.

The unevenness is not the failure. It is the architecture.