Spaced Repetition Research 2026: The Evidence for Language Learning

The research on spaced repetition is older and deeper than most articles admit - and also more nuanced. Here's the honest 2026 summary, including where the evidence is rock-solid and where it gets shakier.

Every other article on spaced repetition cites "the research" without actually naming anything. That's convenient if you're writing marketing copy. It's less useful if you want to know whether the method has real evidence behind it.

Here's a 2026 summary of what the research actually says - the studies that matter, the findings that have replicated, and the areas where the confident claims you see online are running ahead of the data.

The starting point: Ebbinghaus and the forgetting curve

The underlying observation goes back to the 1880s. Hermann Ebbinghaus spent years memorizing lists of nonsense syllables - deliberately meaningless to eliminate prior knowledge - and testing himself at intervals ranging from minutes to months. What he documented is now called the forgetting curve: information loss follows a predictable shape, with most of the decay happening quickly at first and then slowing down.

The other thing Ebbinghaus documented, which gets less attention, is that each successful review flattens the curve. The second time you recall something, the decay is slower. The third time, slower still. This observation is the entire theoretical foundation of spaced repetition - if you can time reviews to catch material just before it would be forgotten, each review does more work than it otherwise would.

Ebbinghaus's methods were crude by modern standards and his sample size was one (himself). But the core findings have held up under more than a century of replication with proper experimental design.

The spacing effect: Cepeda et al. (2008)

The most cited modern paper on spaced vs. massed practice is a 2008 meta-analysis by Cepeda and colleagues. They pulled together results from dozens of individual studies comparing cramming to spaced review, and the finding was unambiguous: spaced practice improved long-term retention by roughly 64% over massed practice when tested after a delay.

The size of the effect depends on how long after study you test. At very short delays (minutes to hours), crammers and spaced learners perform similarly - sometimes crammers do slightly better. At delays of days to weeks, the gap widens significantly. At delays of months, the gap is large enough to be practically meaningful, not just statistically significant.

The implication for language learners is clear. If you're being tested tomorrow, cramming works. If you need vocabulary that sticks for a year, spacing is the only evidence-based approach. Most language learning falls into the second category.

The testing effect: Roediger and Karpicke (2006)

Half of why spaced repetition works is spacing. The other half is retrieval. A well-known 2006 study by Roediger and Karpicke compared students who studied material by rereading against students who studied it by taking practice tests. Same amount of time, same material.

Immediately after studying, the rereaders felt more confident and actually scored slightly higher on a quick test. A week later, the testers scored substantially better. The act of retrieving a memory strengthens it more than passively reading the same information again - even though retrieval feels harder in the moment and leaves you less confident.

This is why flashcards work better than vocabulary lists, even if you spend the same amount of time on both. Every flashcard flip is a retrieval attempt. Every read through a list is passive encoding that feels productive but doesn't build durable memory nearly as efficiently.

The effect replicates so consistently that it's now considered one of the more settled findings in learning science. It also explains why fill-in-the-blank cards generally outperform straight translation cards - they require slightly more effortful retrieval, and effortful retrieval is exactly what produces the benefit.

Second language vocabulary research specifically

The studies above cover general memory. Research focused specifically on second language vocabulary is narrower but consistent. SRS-based study builds larger working vocabularies faster than wordlist memorization, particularly when cards include context rather than isolated word-translation pairs.

A frequently cited finding: cards with example sentences outperform bare translation cards by a meaningful margin on delayed retention tests. The context appears to do two things. It anchors the word to a scenario your brain can reconstruct during retrieval, and it trains a slightly different kind of memory - knowing how a word behaves in use, not just what it means in isolation.

Research on collocations (words that tend to appear together) has found similar effects. Learners who study vocabulary in natural phrases retain and use it better than learners who study individual words, even controlling for total study time. This is part of why modern SRS decks for language learning typically include example sentences as a matter of default rather than an optional extra.

Where the research gets thinner

Not every claim you'll see online is backed by equally strong evidence. A few areas where the research is more limited than the confident summaries suggest.

The optimal review interval - the specific spacing algorithm - is less settled than it looks. Different scheduling algorithms (SM-2, FSRS, Leitner variants) produce roughly similar results in practice, and no single algorithm has been shown to decisively outperform the others across all learners. The differences matter less than getting the general spacing right and keeping cards coming back before you fully forget them.

The ideal number of new cards per day is based on practical experience rather than controlled studies. The common advice - 10–15 new cards per day for beginners, adjust by tolerance - is a reasonable default, but it's not a research finding, it's a heuristic that seems to work well for most people.

The transfer from SRS-learned vocabulary to real-time speech and listening is the area with the biggest gap between confident claims and hard evidence. It clearly does transfer to some extent. The magnitude, and the factors that predict it, are still open questions. The honest answer is that SRS builds vocabulary reliably, but turning that vocabulary into fluent output depends on other factors that the research hasn't fully pinned down.

What has replicated and what hasn't

Worth being specific about replication, since learning science has had its share of replication failures over the past decade.

The spacing effect itself has replicated many times across many populations. You can treat it as settled. The testing effect has also replicated consistently and is considered robust. The superiority of SRS over wordlist memorization for vocabulary retention has replicated across multiple studies, though the size of the effect varies by design.

Some older claims haven't held up as well. The idea that there's a single magic spacing ratio (e.g., always review at X% of the previous interval) was never well supported, and modern algorithms don't use it. Claims about "learning styles" that sometimes get attached to flashcards in popular writing - visual learners, auditory learners - have largely failed to replicate and should be ignored.

What the 2026 picture looks like in practice

Putting the research together, the current evidence-based view of spaced repetition for language learning looks like this. The core mechanism - spacing reviews to align with the forgetting curve, combined with active retrieval - is one of the best-supported findings in learning science. Learners who use SRS with reasonable habits retain vocabulary substantially better than those who cram or reread.

The method is particularly strong for building the first 1,000–3,000 words in a new language, which is where memorization load is highest and immersion alone struggles to cover enough ground fast enough. It remains useful afterward for maintenance and for acquiring rare vocabulary that doesn't appear often enough in real content to stick from exposure alone.

Where the research is less settled is at the edges: exactly how to pair SRS with immersion, how much the specific algorithm matters, and how directly vocabulary knowledge transfers into fluent output. These gaps matter less than they sound - the optimal SRS-to-immersion ratio article covers the practical framework most successful learners end up using, and it works without needing the research questions to be fully resolved first.

What the research cannot tell you

One last honest note. The research can tell you that spaced repetition works on average, across many learners, when used with reasonable habits. It cannot tell you that it will work for you, on your schedule, with the habits you actually maintain.

Every successful implementation comes down to the same thing: showing up daily, reviewing before adding new cards, grading yourself honestly, and pairing flashcards with real exposure to the language. Research supports the method. Whether you build those habits is still on you.

MindDory is designed to remove some of that friction - sensible defaults, built-in example sentences and audio, short mobile-first sessions so reviews fit into real life. But the core trade is unchanged: the research has done its job in showing the method works. The rest is consistency. For the day-to-day mechanics of building it into a routine, the spaced repetition best practices guide is the place to go next.