The Forgetting Curve: Why You Forget 70% of New Words Within a Day
In 1885, the German psychologist Hermann Ebbinghaus conducted an experiment on himself that has shaped cognitive science ever since. He memorised lists of nonsense syllables and then tested his own retention at intervals. The result was what we now call the forgetting curve: a steep, predictable decline in memory retention that sees roughly half of new information forgotten within an hour and around 70 percent lost within 24 hours, absent any review.
Ebbinghaus also discovered the antidote, which he called the spacing effect: reviewing material at increasing intervals dramatically slowed the rate of forgetting. Each successful review extended the time before the next one was needed — so the total number of reviews required to achieve long-term retention was far lower than if you simply reviewed everything at fixed short intervals.
For language learners, the forgetting curve explains a familiar frustration. You study 30 new vocabulary words in a session, feel confident you know them, and return three days later to find that most of them have effectively vanished. The problem is not failure of intelligence or effort. It is that human memory without systematic reinforcement discards information it does not repeatedly encounter.
What SM-2 Actually Does (Without the Scary Maths)
SM-2 is a spaced repetition scheduling algorithm developed by Piotr Wozniak and first published in 1987. It is the foundation underlying Anki, the most widely used flashcard tool in serious language-learning circles, and it powers the review systems in many structured language applications.
The algorithm's core job is to decide when you should next see a given piece of information. It does this by tracking two things: how long it has been since you last reviewed an item, and how well you remembered it when you did. A word you recall easily gets scheduled for review further in the future. A word you struggle with gets scheduled for review very soon. Over time, easy items move to intervals of weeks and months; difficult items stay in frequent rotation until they stabilise.
The elegance of SM-2 is that it personalises your review schedule automatically. Rather than reviewing everything at the same interval, the algorithm concentrates your limited review time on the items that are actually on the verge of being forgotten. This makes review sessions dramatically more efficient than any fixed-interval or random approach.
Ease Factor, Intervals, and Grade Scores: A Practical Walkthrough
SM-2 manages each item through three variables: the ease factor, the interval, and the grade score you give after each review.
The ease factor represents how easy a particular item is for you personally, not in absolute terms, but relative to your response history. It starts at a default value (typically 2.5) and adjusts upward when you answer correctly and downward when you struggle. An item you have found consistently easy ends up with a high ease factor and gets pushed further into the future with each review. An item that keeps tripping you up has a depressed ease factor and will keep appearing more frequently.
The interval is the number of days until the item reappears. For a new item you answer correctly, the initial interval is short: one day, then perhaps four or six days after the next successful review. From there, each successful review multiplies the interval by the ease factor, so the gaps grow rapidly: 6 days, 15 days, 38 days, and so on.
The grade score is your self-assessment after each review: did you remember the item immediately and confidently, did you recall it but with effort, or did you fail to recall it entirely? Your grade directly determines whether the interval continues to extend or resets toward zero.
How Spaced Repetition Differs from Random Flashcard Review
The difference between SM-2 scheduling and shuffling a deck of flashcards at random is not merely organisational; it is fundamental to what you achieve with a given number of review minutes.
Random flashcard review treats all items equally regardless of your memory state for each. You spend as much time reviewing words you have known for months as words you first saw yesterday. You review items when they are already overlearned and therefore not on the verge of being lost. The cognitive work is genuine but the timing is inefficient.
SM-2 solves this by surfacing each item at the point of maximum return — the moment just before it would likely slip below the threshold of reliable recall. This is called the optimal retrieval practice hypothesis: retrieval is most beneficial for long-term consolidation when it occurs when the item is still accessible but requires some effort to retrieve. Too easy and the review adds little; too late (after forgetting) and you are starting over.
The practical consequence is that learners using well-implemented spaced repetition can maintain a vocabulary of several thousand words with as little as 15 to 20 minutes of daily review, where equivalent maintenance through random review would require substantially longer sessions.
Integrating Spaced Repetition With Skill-Tree Courses
One of the most effective study architectures combines a forward-facing skill-tree course with a backward-facing spaced repetition review system. The skill tree introduces new vocabulary, grammar patterns, and sentence structures in a pedagogically sequenced order. The spaced repetition system ensures that what you learned in last week's lessons is still accessible when you need it this week.
This is the model that well-designed language platforms use natively, and it is why the reviewed English-to-French course with built-in spaced review structures its lessons to feed directly into a review queue, rather than treating new-content study and maintenance review as separate unconnected activities.
Without the integration, learners typically find that their new-content study races ahead of their retention baseline: they can do the latest lessons but have forgotten large portions of what they covered three months ago. With it, the skill tree and the review queue work together: the tree expands the learner's total vocabulary and grammar range, and the review system consolidates it.
Common Mistakes That Break Your Review System
Several predictable errors undermine spaced repetition practice. The most common is review debt accumulation: missing review sessions allows due items to pile up until the queue feels overwhelming, at which point many learners either quit the system or rush through reviews without genuine recall effort. Both outcomes destroy the algorithm's data quality.
The second mistake is over-adding new items while under-servicing the review queue. Every new item you add creates future review obligations. Adding 50 new words per day when your daily review capacity is already strained means your ease factors will degrade, your intervals will compress, and the system will essentially stop functioning as designed.
A third mistake is passive reviewing, glancing at an item and deciding you "sort of know it" without actually performing the retrieval. SM-2 works because retrieval practice itself strengthens memory consolidation. Skipping the actual recall step and just confirming what you can see defeats the mechanism entirely.
Building a Sustainable Daily Review Habit
The most important design principle for a spaced repetition practice is that it must occur every day, or very close to it. Unlike new-content study, which can be batched on longer sessions at weekends, review is time-sensitive because items are due on specific days. Skipping review sessions shifts items past their optimal retrieval window and degrades the accuracy of the algorithm's scheduling.
Fifteen minutes of faithful daily review consistently outperforms ninety-minute cramming sessions twice a week. The habit formation challenge is therefore simpler than for most study activities: keep the session short, anchor it to a fixed daily trigger (morning coffee, commute, bedtime), and treat it as non-negotiable maintenance rather than optional supplementary study. Over six to twelve months, a learner who maintains this practice will have a retention baseline that most other learners (even those who study more hours overall) simply cannot match.
