Karly

I’m probably exactly the wrong person for Duolingo. I studied languages at university, eventually studied linguistics alongside them, and learnt most of my languages in the apparently prehistoric manner of having another human being teach them to me. This didn’t, however, mean that anybody ever agreed on how a language ought to be taught. When I was ten and learning French at summer camp, my teacher favoured conversation, getting us to use whatever French we had rather than waiting until we could produce something perfect. Years later, my Italian teacher at university took almost the opposite approach: forty minutes of her teaching, ten minutes for questions, grammar explained properly, notes taken, structure everywhere. She was magnificent. I adored her, and thanks in no small part to those apparently unfashionable hours of explicit instruction, I can now sit down with Italians and chat to them confidently.

German was different again, and actually came in two stages. I started learning it at university, where we worked mainly with the Menschen coursebook, a much more structured progression through the language, and I enjoyed it enough that when the time came to choose what I wanted to do with my summer abroad, I chose German. That took me to the Goethe-Institut, where the experience became much more immersive. Goethe did what Goethe does and surrounded us with German until German became the obvious language in which the room operated. One approach didn’t cancel out the other. The first gave me enough of the language, and enough enjoyment of learning it, to make me actively seek out the second.

My secondary-school English education had another objective entirely, much of it focused on preparing us for the Cambridge exams, where language became something that could be divided into reading, writing, listening, speaking, Use of English, transformations and the slightly terrifying business of proving on command that you could actually do all of them. None of these methods looked particularly alike, and I don’t think I could honestly say that only one of them constituted real language teaching. They gave me different things.

What they did have in common was friction. I had to speak when I wasn’t quite ready, listen when I couldn’t understand every word, learn grammar explicitly, write things down, ask questions, make mistakes in front of actual people and, crucially, find out why something I had said or written was wrong. Language learning wasn’t always entertaining, nor was every part of it designed to make me want to return tomorrow. Sometimes it was difficult because the thing I was learning was difficult.

This is where Duolingo and I begin to have problems. I have a good memory, which turns out to be surprisingly useful if your objective is to make a green owl believe you’ve learnt something. I’ve tried Russian, Dutch, Arabic and Scottish Gaelic on Duolingo, made mistakes, been shown the correct answer and then encountered the same construction again later. Often I’ll get it right the second time, not because I suddenly understand the grammatical principle involved, but because I remember making the mistake. I remember that I chose one answer last time, Duolingo told me it was wrong, and therefore I choose the other one.

Russian is probably the most revealing example. At one point, I maintained a Duolingo streak of more than 100 days in Russian, which sounds rather impressive until I tell you that my Russian remains extremely limited. I can read Cyrillic and recognise a handful of words, but even that comes with an enormous methodological asterisk: Duolingo wasn’t my only exposure to Russian. I own alphabet practice workbooks and an actual Russian book, so I couldn’t honestly tell you which part of that very modest knowledge came from the app and which came from somewhere else.

That’s precisely why I’m less interested in presenting my Russian as evidence that Duolingo either works or doesn’t than I am in what the streak itself appeared to represent. For more than 100 days, the app could give me a wonderfully precise measurement of my consistency. What it couldn’t tell me nearly as neatly was what I could actually do with the language. The number was real, the habit was real and the practice was real, but none of those things is automatically synonymous with proficiency.

From the app’s perspective, my individual exercises looked suspiciously like progress. I was wrong, I practised, then I was right. Except there was an important step missing between those states: nobody had necessarily told me why I was wrong. Give me a new sentence governed by the same rule, remove the familiar cues and ask me to produce it independently, and we might discover that what I’d learnt wasn’t a piece of the language at all. What I’d learnt was the answer to a Duolingo exercise.

This distinction matters because remembering a correction isn’t the same thing as understanding the system that produced it. Human beings are extraordinarily good at recognising patterns, remembering previous failures and adapting to tests, which is useful for learning but also makes it remarkably easy to become better at an exercise without becoming proportionately better at the thing the exercise is supposed to measure. Anyone who has ever revised past papers knows the feeling. Eventually you begin to recognise what the question wants.

And this is where my linguist brain becomes particularly irritating. If I’ve made a grammatical mistake, I want to know what happened. Was it case? Agreement? Word order? Aspect? Register? Did I select a form that’s technically possible but pragmatically bizarre? Is there a rule I can apply the next time I encounter a sentence I’ve never seen before? I don’t want to memorise that this answer was wrong. I want the mistake to teach me something about the language.

To be fair, I’m hardly an unbiased test subject. Modern language-learning apps aren’t designed exclusively for people who actively want grammatical terminology, and there’s an enormous advantage in making the beginning of a language feel less like homework. Duolingo has persuaded an extraordinary number of people to voluntarily spend part of every day interacting with another language, which is considerably more exposure than they might otherwise have had. I’m not interested in arguing that this is worthless, nor do I think the solution is to replace the little green owl with my Italian professor and force everyone to sit silently while she explains verb conjugations for forty minutes, although I personally would download that app immediately.

What interests me is the much stranger question underneath all of this: when we say an app is teaching us a language, what exactly do we mean by learning one?