The science behind Hippo
The hippocampus — the brain's memory centre — is where long-term memories are formed. It's why we called this Hippo. Memory is what makes learning permanent, which is why understanding how it works changes what good teaching practice actually looks like.
What is learning, exactly?
A useful working definition: learning is a change in long-term memory. If a student understood something in class but can't retrieve it a week later, the encounter with the material hasn't yet become learning. This shifts the question that matters — from “did students understand the lesson?” to “will they remember it next month?” Those are genuinely different questions with different implications for how we design what happens after a lesson ends.
Your brain forgets. Fast.
In 1885, Hermann Ebbinghaus discovered that without active review, we forget most of what we learn — quickly and predictably. Retention drops to around 40% within 24 hours and to as little as 20% after two weeks. The brain isn't broken; it's efficient. It prunes information it hasn't needed recently to make room for what it does need.
What Ebbinghaus also found is that each successful retrieval resets and flattens the forgetting curve. The brain updates its prediction about whether something is worth keeping — and repeated retrieval is the signal that it is. A 2008 meta-analysis by Cepeda and colleagues confirmed that the timing of reviews matters as much as the fact of reviewing: the optimal spacing depends on how long you need to retain the information, which is why personalised algorithms outperform fixed schedules.
Review at the right moment
Spaced repetition schedules each card based on how well you know it. Cards you struggle with return sooner. Cards you know well wait longer. The result is that you review everything just before you'd forget it — maximising what you retain for the time you put in. Each review doesn't just restore a fading memory; it strengthens it, so the intervals between reviews grow over time as memories consolidate.
Testing yourself beats re-reading
Actively retrieving information from memory produces more durable learning than passively reviewing the same material. Roediger and Karpicke (2006) showed that students who tested themselves retained significantly more after a week than students who spent the same time re-studying — even though the re-studiers felt better prepared.
Robert Bjork's work on desirable difficulty adds a counterintuitive point: the conditions that feel hardest tend to produce the most durable learning. Re-reading is fluent and familiar, which makes it feel productive. Retrieval is effortful, which can feel like failure. The struggle is actually the mechanism — and Hippo is built around it.
Better than Ebbinghaus
Hippo uses FSRS — the Free Spaced Repetition Scheduler — a modern algorithm trained on hundreds of millions of real study sessions. It's more accurate than the SM-2 algorithm used by Anki and early Quizlet, because it constructs a personal forgetting curve for each student-card pair rather than applying average rates across everyone.
A student who knows something well won't see it for months. One who keeps struggling will see it again soon. The system runs continuously to maintain around 90% retention — calibrated enough not to waste review time on material that's already well-consolidated.
Not just memorisation
Pooja Agarwal's 2019 research established that the ceiling of what retrieval practice builds is set by what the questions ask for. Recall-only practice strengthens recall. Questions requiring application, discrimination, and evaluation strengthen deeper understanding as well.
Hippo's five card types address different levels of learning:
Basic and Cloze — foundational recall of facts, vocabulary, and definitions.
Multiple Choice — discrimination between plausible alternatives, revealing where understanding diverges from recognition.
Prediction — students guess before seeing the answer. Even wrong predictions improve subsequent retention through the pretesting effect (Richland et al., 2009). The learning happens in the feedback, which is why prediction cards carry richer explanations.
Error Identification — a statement containing a deliberate mistake. Finding and explaining the error requires genuine understanding, not just recognition. Particularly effective for misconception repair.
Hippo also mixes cards from different decks in a single study session — a technique called interleaving. Studying one topic at a time feels more productive, but research consistently shows the opposite: switching between topics forces the brain to discriminate between concepts and strengthens long-term retention. Kornell and Bjork (2008) demonstrated that interleaved practice produced better learning even when learners themselves believed blocked practice was more effective.
Learning from each other
Hippo's Live Mode supports Peer Instruction — a classroom technique developed by Eric Mazur at Harvard. Students vote individually on a question, discuss with peers who answered differently, then vote again. Across thirty years of replication studies, peer discussion has been shown to significantly improve conceptual understanding, because explaining your reasoning to someone who disagrees forces a depth of processing that passive listening doesn't reach.
Respecting your time
Hippo has no streaks, no guilt mechanics, no notifications designed to manufacture anxiety about falling behind. A well-designed spaced repetition schedule means less time reviewing over time, not more. The goal is deeper, more persistent learning — with less effort spent getting there.
Research
Agarwal, P. K. (2019). Retrieval practice & Bloom’s taxonomy. Journal of Educational Psychology, 111(2), 189–209.
Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way. In Psychology and the Real World. Worth Publishers.
Cepeda, N. J., et al. (2008). Spacing effects in learning. Psychological Science, 19(11), 1095–1102.
Ebbinghaus, H. (1885). Über das Gedächtnis. Duncker & Humblot.
Ingebrigtsen, R. (2025). Teacher-created spaced repetition materials and student exam outcomes. (In press.)
Kornell, N., & Bjork, R. A. (2008). Learning concepts and categories: Is spacing the “enemy of induction”? Psychological Science, 19(6), 585–592.
Mazur, E. (1997). Peer Instruction: A User’s Manual. Prentice Hall.
Pan, S. C., et al. (2022). User-generated digital flashcards yield better learning than premade flashcards. Journal of Applied Research in Memory and Cognition, 12(4), 574–588.
Richland, L. E., Kornell, N., & Kao, L. S. (2009). The pretesting effect. Journal of Experimental Psychology: Applied, 15(3), 243–257.
Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning. Psychological Science, 17(3), 249–255.
Slamecka, N. J., & Graf, P. (1978). The generation effect. Journal of Experimental Psychology: Human Learning and Memory, 4(6), 592–604.
Ye, S., Kim, S., & Mancera, L. (2022). FSRS algorithm. Proceedings of the 28th ACM SIGKDD Conference.