How to Almost Learn Italian

Language apps like Duolingo are addictive—but not particularly effective.

Tiago Galo

Late one chilly evening last September, I excused myself from a small group huddled around a campfire to peck at and mumble into my phone.

No way was a camping trip going to make me miss my Italian lesson.

For most of the preceding year, I had religiously attended to my 15-minute-or-so daily encounters with the language-learning app Duolingo. I used it on trains, while walking across town, during previews at the movie theater. I was planning a trip to Rome in the late spring, and I’ve always been of the mind that to properly visit a country, you’ve got to give the language a shot.

But I had another reason for sticking with it: Duolingo is addictive. It pulled me right in, helping me set daily goals and then launching into simple phrases. Sometimes it demanded that I speak an Italian phrase or sentence (which I always did correctly, to hear Duolingo tell it). But more often it asked me to translate Italian phrases and sentences into English, or vice versa, providing multiple-choice responses. No tedious grammar or vocabulary drills—that stuff, apparently, would seep into my consciousness via exposure to increasingly varied, complex, and interesting sentences.

Duolingo praised me constantly: for responding correctly several times in a row, for completing a chunk of the day’s lesson, for learning from my sloppy mistakes. Finishing a lesson was a full-out digital celebration featuring treasure chests with flapping lids. The app kept me apprised of my progress via various point schemes, and used email and phone notifications to nudge me to keep my routine going, even betting me points that I wouldn’t keep my streak up for another week. Sucker! I became rich in worthless points, and cherished them.

I’m not a serious polyglot, but I’ve tackled a handful of languages in just about every way a language can be learned: classroom, tutor, textbook, audio recordings, flash cards, software, and more. Learning languages was always a chore—until Duolingo. I looked forward to my lessons. And I was learning Italian! I walked through my home confidently talking about hiding a knife in my boot, when my master’s thesis was due, and how important it was to pay attention to the will of the people. Okay, so Duolingo’s sentences covered some strange ground. But surely, I figured, that would work to my advantage when I was faced with more mundane language demands as a tourist.

A week before we were to leave for Rome, my wife, Laurie, put me to the test. You’re at the airport outside Rome, she said, and you want to get downtown; how would you ask? I gaped like a fish. Words and phrases swam through my mind, but they didn’t add up to anything useful. Laurie switched to a restaurant scenario: “Do you have a table for four?” “I’d like two glasses of red wine.” I knew I had seen all the pieces in Duolingo’s sentences. But I was utterly unable to recall them and pull them together.

Panicking, I fired up Duolingo and almost instantly saw the problem. The app had made me a master of multiple-choice Italian. Given a bunch of words to choose from, I could correctly assemble impressive communiqués. But without a prompt, I was as speechless in even the most basic situations as any boorish American tourist. And this in spite of 70-plus hours of study.

But I still had a week. I got my hands on a self-study book, a travel phrase book, and a pocket dictionary, and started cramming. A funny thing happened: I started easily picking up what I hadn’t been able to get from Duolingo—grammar, vocabulary, and, most important, an ability to engage in simple conversations in typical situations. It seemed I had been getting something useful from my hours with Duolingo. The app had exposed me to a considerable vocabulary; I needed only minimal drilling with books to remember the words. Learning the verb conjugations was a breeze, too.

In the end, I did pretty well in Rome, engaging in simple, fractured semi-conversation in most of my encounters. Was that how the app was supposed to work?

I recently got in touch with Luis von Ahn, a co-founder and the CEO of Duolingo, to ask whether my experience was typical. I expected some defensiveness from him about my need to use books to get the conversational skills I had hoped to get from Duolingo. But instead he laughed and told me the app had done exactly what it was built to do. “The biggest problem that people trying to learn a language by themselves face is the motivation to stay with it,” he told me. “That’s why we spend a lot of our energy just trying to keep people hooked.”

Duolingo is essentially a product of crowdsourcing; volunteers build much of the teaching content, and the in-app behavior of its 27.5 million active monthly users is continuously analyzed to determine which exercises, sentences, and techniques lead to better adherence and faster learning. The challenge, von Ahn told me, is that the two metrics tend to be at odds: Making the lessons more difficult reliably speeds up learning—but also increases dropout rates. “We prefer to be more on the addictive side than the fast-learning side,” he explained. “If someone drops out, their rate of learning is zero.”

This emphasis on user retention helps explain why Duolingo is by far the most popular language app in the U.S. In other countries, von Ahn notes, learning a language is often crucial to communicating with partners and their families, and for work; learning English, in particular, can be a ticket out of poverty. “In the U.S., about half of our users aren’t even really motivated to learn a language; they just want to pass the time on something besides Candy Crush,” he said.

Joey J. Lee, the director of the Games Research Lab at Columbia University, who did a study of 50 language apps in 2016, told me that he suspects the addictiveness of tools like Duolingo has more to do with business models than with language learning. Where most apps really fall short, he said, is in language “pragmatics.” “That’s the learning that’s based on real-world settings—you’re in a restaurant, in an interview, waiting for a bus,” he explained. “It’s usually lost in apps.”

That sounded right to me. And the claim was echoed by Geoff Stead, the chief product officer for Babbel, a rival language app. “What most helps someone learn a language is when they’re immersed in a situation and they’re struggling to speak,” he told me. “Our approach is to help you have the confidence to speak in those situations, and to get you there as rapidly as possible.”

Judging by the fact that Babbel’s user base is about one-15th the size of Duolingo’s, that struggle is apparently less addictive than multiple-choice sentence translation. (Also, Babbel doesn’t offer a free version of its app, as Duolingo does.) But the approach has real rewards, Stead insists. Most of Babbel’s lessons, he says, are focused on giving users the ability to get by in social settings (meeting people, traveling, ordering food and drinks)—which tends to fire up interest in learning more. “Once we get the ball rolling, we bring in more classic, cognitive learning techniques,” he said, such as more vocabulary and grammar.

My problem, then, is that I’m a pragmatics kind of guy living in a Candy Crush world. But if I had swapped the crowdsourced appeal of Duolingo for the when’s-the-bus-getting-here practicality of Babbel, would I have put enough time into the app to get what I got out of my 70-plus hours with Duolingo?

In the future, I may not have to choose. Duolingo has been rolling out new features—including podcasts, social interaction among users, and character-driven narratives—that aim to raise its language pragmatics as well as its addictiveness. Lee predicts that language apps will eventually also incorporate AI-based chatbots that will engage and guide users through realistic conversations. (Microsoft offers one called Microsoft Learn Chinese, but I tried it, and it seemed buggy.)

But I also take heed of the caution offered by Tom Roeper, a linguistics professor at the University of Massachusetts who studies language acquisition. Roeper told me that apps aren’t likely to soon overcome the two essential advantages of a human teacher: the ability to hold a student’s attention, and to continually tailor a lesson to the individual’s progress, difficulties, and interests. “There are all kinds of contextual factors in language learning,” he said. “It would be hard for an app to take them all into account.”

Then again, teachers aren’t around when you have a little spare time sitting by the campfire.

This article appears in the December 2018 print edition with the headline “How to Almost Learn Italian.”