Let's start by giving streaks their due, because the case against them is usually made by people who have never shipped anything and it's unconvincing.
The hardest problem in language learning is not method. It's attrition. Most people who start a language stop within a few weeks, and they don't stop because they picked the wrong app — they stop because nothing in their day reminds them and nothing in their head punishes them for skipping. A streak fixes exactly that. It converts a vague intention into a small, visible, daily object you don't want to break. That is a real piece of engineering and it has kept millions of people studying who would otherwise have quit in February.
The problem isn't that streaks work. It's what they work on.
A streak measures attendance
A streak counts one thing: did you open the app today. It cannot see whether you learned anything, whether the session was three minutes of tapping words you already knew, whether you could produce a single sentence of what you covered, or whether you completed the lesson at 11:58pm on autopilot to protect the number.
In any system, the metric you display becomes the metric people optimize. Show someone a streak and they will optimize for not breaking it — which, on a tired Tuesday, means finding the shortest possible action that counts. The app is satisfied. The learner is satisfied. No learning occurred. Do that often enough and the streak stops being a record of study and becomes a record of streak maintenance.
Ask what your last session actually contained
Not "did I study yesterday" — what did I do. If you can't name a word you learned, a pattern you noticed, or something you understood that you wouldn't have understood a month ago, the session was attendance. One in a row is fine. Thirty in a row is a plateau with a number on top of it.
What the outcome data shows
The most useful evidence here is Duolingo's own. They funded a study in which researchers gave standardized ACTFL reading and listening tests to 225 people who had completed the beginner Spanish and French courses using only the app.
Median study time was about 112 hours. Spanish learners averaged Intermediate Low in reading — a real, verified A2-ish result that many critics of app learning would not have predicted. But listening averaged just short of Novice High, with two-thirds of Spanish listening scores still at the Novice level. And speaking and writing were not assessed at all.
Read it straight: a hundred hours of gamified study reliably produces decent beginner reading, weak listening, and an unmeasured ability to actually talk. That's a fair description of what a lot of long-streak learners quietly discover on their first trip.
Why the motivation curve bends
There's a second-order problem that shows up later. Psychologists have spent decades on what happens when you attach an external reward to something a person was doing for its own sake — the overjustification effect. The short version is that the reward can crowd out the original motivation: the activity becomes about the points, and when the points stop mattering, so does the activity.
You can watch this happen in the wild. Someone loses a 400-day streak to a flight delay and simply never comes back. Not because they stopped wanting Spanish, but because the thing they had been protecting was the streak, and the streak is gone. If the number were a side effect of caring about Spanish, losing it would be annoying. When it's load-bearing, losing it is terminal.
The comfort ceiling
Gamified exercises are designed to be completable. Completable means calibrated slightly below your level, because failure feels bad and bad feelings cost retention. That's the right call for a beginner and the wrong one for everyone after. Learning happens at the edge of what you can do — the sentence you nearly understand, the reply you have to construct rather than recognize. A system tuned to keep you succeeding is, by construction, tuned to keep you comfortable.
Recognition isn't production
Tap-the-right-tile exercises test whether you can pick the answer out of a lineup. Speaking requires you to generate it from nothing, under time pressure, while also planning the next clause. These are different skills with different difficulty curves, and months of the first builds surprisingly little of the second.
Four things worth tracking instead
None of this means turn the streak off. It means don't let it be the only number you look at. Better ones:
- Minutes, not days. A streak treats 90 seconds and 40 minutes identically. Total time is crude but honest, and it maps directly onto the hour counts that actually predict proficiency.
- Words you can retrieve, not words you've seen. The count that matters is what survives a review a month later without a prompt.
- Unknown words per page. Pick something you want to read and count what stops you. Watching that number fall from fifteen a page to three is the clearest progress signal in language learning, and no app displays it for you.
- Minutes of output. Time spent speaking or writing, where you had to produce rather than recognize. For most app learners this number is zero, and moving it to ten minutes a week changes more than doubling everything else.
Set a floor, not a maximum
Keep the streak, but define the minimum session as something that has to cost you: five retrieval-based reviews you actually had to think about, or one paragraph read, or two sentences said out loud. The streak still does its job — showing up — and it can no longer be satisfied by tapping. And when you do break it, restart the same day. The habit is the asset; the number was only ever a receipt.
Progress measured in words, not days
Lernlang tracks the thing a streak can't: how many words you can actually retrieve, and when each one is due before you lose it. Reviews are scheduled at the point of useful difficulty rather than tuned to keep you succeeding, grammar is checked against what you can already do so you skip what you know, and the stories are written at your level so reading stays just over your head instead of comfortably under it.
The honest position
Gamification solved the biggest problem in consumer language learning: getting people to come back tomorrow. That's not a small thing, and the people who sneer at it usually have the luxury of already being motivated.
But a streak is a measure of showing up, and showing up is the entry fee, not the outcome. If you've got a long one, keep it — then go and find out what it's actually bought you. Read a page of something real and count what stops you. Say three sentences out loud to a person. The number you get back is the one worth improving.