A teaching method built on outdated barriers and assumptions was mistaken for the best way to learn.

Every language app in your pocket has inherited a built-in Latin tutorial. Understanding why that happened is a more useful design lesson than anything the apps will teach you.

Article Continues Below

In 1788, Prussia introduced the Abitur, a standardized national examination required for entry into universities and the civil service. To pass it, students had to demonstrate measurable, programmable knowledge. A necessary program to teach the language in large classes, produce consistent results, and do it with one teacher and thirty students. The teachers who were responsible for designing this system had access to the only teaching template they had, which had been used in European schools for two centuries: the established method of teaching Latin.

Latin, in 1788, was a dead language. No one needed to speak. Scholars who read it read Cicero and Virgil, not conducting discussions. The method built around, memorizing the rules of grammar, creating translations, analyzing written texts, showed exactly that fact. Verbal skills were useless. Understanding the script was everything. This method was not designed to produce speakers. It was designed to produce readers of literature in a language no one spoke.

When Prussia applied this template to French and German, living languages ​​spoken by living people, the basis did not change. Book by Johann Valentin Meidinger Practical French Grammarpublished in 1804, published in 37 editions throughout Europe in 1857 [1]. Karl Plotz formalized the method into what became the dominant model for modern language teaching throughout Europe and eventually in the United States, where it became known simply as the Prussian Method. [2]. The host institution trains teachers, who train students to become teachers. The limitation that created the method, how to scale the language with limited resources, is not visible within the method itself. All that was left was to think: language is a body of rules that must be carefully studied and measured. It was a design decision dressed up, over time, as an educational reality.

A realization that had to be eliminated#section 2

There are people in the world who can read and write a language and speak it well. There are children who hold full conversations for years before they learn a single word. There are immigrants who come to the country without knowing anything about its language and leave, after years speaking naturally, not because they studied, but because they live in it. Literacy and fluency are different things produced by completely different methods. The Grammar-Translation method, as it is known, assumes that they are the same thing. That assumption was inherited in a way that was designed for a language that no one had to speak, and it was incorrect when applied to the language that people actually use.

The evidence against it is slowly piling up. In the mid- to late-nineteenth century, reformers including François Gouin in France and Maximilian Berlitz in the United States argued independently that language should be taught as it is, through deep exposure to actual communication with the target language, not through analysis of its rules. Berlitz is building a school-wide network around this system. The reformers were right. They were also ignored by mainstream education systems, because the grammar translation method had one decisive advantage that was not watered down: it could be edited.

In 1982, linguist Stephen Krashen gave the argument its formal articulation in what he called the Monitor Model of second language acquisition. His distinction was precise: language learning, the unconscious process by which children learn their native language and where adults thrive in concentrated environments, is very different from language learning, the careful study of grammatical rules and vocabulary that classrooms bring. [3]. Finding produces fluency. Studying, at its best, produces the ability to pass a test. The evidence supporting this distinction, and the realization that intensive exposure to real native-speaker communication is the way to produce real fluency, has grown since then.

I went to Brazil without a word of Portuguese and came out speaking. I studied French in class for many years and I can’t speak French today. This is not unusual. It is an expected result, and it has been an expected result for a long time as we learn structured language.

The same decision, made in a different way#section 3

Prussian educators faced the question: How to deliver language learning at scale, measure progress, and retain users over time? The answer he got was the same as the building he got in 1788. Duolingo reduced the grammar to one line. Anki formalized the translation work on the blank multiplication card. Babbel has organized the grammar lessons into structured modules. The interfaces were new. The basic assumption, that language is something you learn rather than where you live, was not the case.

This was not a failure of design skill. The products that came out of these decisions are, in many ways, really well built. Duolingo’s maintenance tools are complex. Anki’s space repetition is based on real psychology. They are really good at what they do. The problem is what they actually do: produce measurable interactions with the language agent rather than the conditions that produce the language itself. The streak is measurable. Vocabulary scores are measurable. The moment when the user exits the application and holds a real conversation in another language, which happens in the world, outside the product, and cannot be used.

If the result the user needs is difficult to measure directly, the design process often reaches something that can be measured. The agent becomes the goal. The interface is self-configuring. The gap between what the product delivers and what the user actually needs is widening. This is not a unique pattern of language learning. It’s a pattern that repeats itself in all product categories whenever a design constraint—the need to scale, the need to scale, the need to produce a grade—is built into the system so deeply that it stops being seen as a constraint and begins to be mistaken for the truth of the problem itself.

What happens when the pressure changes# section 4

The obstacle that necessitated the translation of the language in 1788 was both real and logical. Another teacher. Thirty students. Standard test. You can’t rate a conversation on a scale. You can rate translation work. The method was not chosen because it produces smoothness. It was chosen because it produced points.

That limitation is no longer the same. Technology has made it possible to bring immersive, real-time chat to anyone with a smartphone, at a cost that continues to fall. The design problem is no longer how to organize language learning at scale. It is a way to make the conditions of acquiring the original language accessible to people who cannot immigrate or pay for a teacher who speaks the native language.

The products that are now closest to solving the real problem are not the ones that invented the new pedagogy. They are the ones who removed the barrier of access to the elder. Praktika creates AI conversation partners with unique personalities, dialects, and cultural contexts, replicating the clarity of a real native speaker rather than a typical language learning voice. Langua combines the words of a native speaker so that the interaction feels like a real conversation rather than a lecture. The basic method of Rosetta Stone, the association of the image with the target language without translation, is built on the same understanding that Berlitz arrived at in the nineteenth century: language is acquired through deep exposure, not through the analysis of its rules. [4]. A 2025 study found that students who used AI conversation practice tools showed a 75 percent improvement in speaking scores in eight weeks, a result that no amount of flashcard practice ever produced. [5].

None of these products developed a new theory of language acquisition. They translate what is there into something that more people can access.

The question of designing this leaves#section 5

The Grammar-Translation method did not persist because the teachers were wrong about the design, but because the design decision taken under a certain constraint became, during two centuries, inseparable from the thing itself. Obstacle, how do you put the tongue on the scale, you forgot. The method it produced was inherited as if it were an explanation of how language works, passed from Prussia to Europe to America to the App Store, from grammar training to sequencing.

Every time a design team prepares a metric because the actual outcome is difficult to measure, they make a version of the same decision. It is often the right decision given the real constraints. The question you have to ask yourself is whether the constraint that made it necessary is still there, or has it simply disappeared from within the system that produced it in the first place.

Before getting to what can be measured, it’s worth asking what exactly the user needs to do, and what has prevented them from doing it before. Sometimes the answer is a new solution. Usually it’s the old one that was always out of reach.