Listen for the missing word exercise


While at Duolingo, I worked with a team with a PM, iOS engineers, and a language expert to create a listening exercise for beginners. I worked end-to-end from problem discovery, to running usability tests, to iterative design, to final implementation. My crowning achievement was introducing rapid prototyping to our process


  • My Role: Product Design
  • Team Size: 5


  • Increase in learners passing more challenging listening Exercises (without a hit to Time spent Learning)
  • Convinced team to purchase team license to protopie


Duolingo's listening activies take off the training wheels a little too soon...

The current line up of listening exercises ramp up the difficulty very quickly. As seen in the digram below, the jump expects a learner to move from listening and tapping to listening and writing.Such a jump not only affects our learners, but also the product. For learners, this increase causes frustration (and eventually temptation to skip these exercises). On the other side, with little variety, the product suffers from not being able to effectively teach an important skill. Combined, these issues lead to a drop in time spent learning and a drop in overall completion within our listening exercises (which is no bueno).​

How might we ensure learners are properly prepared for harder listening activities?

A peak at the science working behind the scenes

Looks simple, right? On the surface it is, but underneath there's a lot to unpack. We built this experience around two key learning principles: sound discrimination and sentence segmentation. Turns out, both are necessary in building up folks' listening abilities early on and setting them up for future success.

Sound Discrimination is the ability to recognizing different sounds (think telling apart your "j" sound from your "a" sound. On the other hand, Sentence Segmentation is the ability to break up sentences into its meaningful parts (nouns, verbs, etc). The final design reflects these two skills, and utilizes them to create an engaging activity where learners have to listen and select the right word that completes the sentence.

A breakdown of the experience

For this exercise, I worked on designing experience. While designing the main goals I had in the back of my head were speed and good friction. Duolingo exercises should be snappy, but for this I knew we needed to introduce a bit of fraction to ensure learners were utilizing both sentence segmentation and sound discrimination. The good bit of friction here is intentionally using waveforms, which causes learners to have to rely on their ears to get this correct.

To get the interaction  perf, I created interactive prototypes in Protopie that the team and I could then test with learners to refine the experience. Having this step in our design process enhanced our ability to apply qualitative insights into our final solution.

Hero Flow

No capes, just a look at the core flow

Looking back at it

Tbh my brain is still doing backflips thinking about all of the science that goes into processing language. Though simple at face value, a lot went into making this experience just right for beginner learners, just starting out learning a language. This project stressed the importance of prototyping an experience not just once, but several times and use actual media wherever possible. Though it may add additional time, the insights are worth it (so note to self, plan ahead!).

Though this experience was a success, I'd love to go back and spend more time stress testing the edge cases for this exercise to get a sense of example words that would not work for this. Also, I'd love to create a vision for the many types of listening exercises we could create to level the difficulty ramp. This was just scratching the surface.

Many thanks to to thanks to my team and all the learners that helped test this thing!

Made in Webflow