You spent three weeks on Fixed Income. You ground through duration calculations, convexity adjustments, yield curve shapes, credit spreads. By the end of week three, you felt solid. You moved on to Derivatives.
That was six weeks ago. Today, if you sat down with a Fixed Income question set you'd never seen before, how much would you get right?
Less than you think. Possibly much less. And the scale of the problem is sitting in a dataset that has been public since 1885.
What Ebbinghaus actually found
Hermann Ebbinghaus spent years memorising nonsense syllables — random consonant-vowel-consonant strings with no prior associations — and testing his own recall at intervals. What he discovered became the most replicated finding in cognitive psychology: the forgetting curve.
The curve is not gradual. It is steep, then asymptotic. Within 20 minutes of learning something, roughly 42% is gone. Within an hour, about 56% is gone. After 24 hours, approximately two-thirds has been lost — Ebbinghaus recorded approximately 67% forgetting at the 24-hour mark. Within six days, roughly three-quarters of the original material has faded. The rate of forgetting then slows, but by the time the curve flattens, most of the material has already been lost.
Ebbinghaus also measured what he called the "savings" when relearning material. Material that had been learned and forgotten was much faster to relearn than completely new material — proof that memory traces persist below the threshold of active recall, and that distributed practice leaves deeper traces than massed study.
What this means for a 300-hour study plan
A typical CFA Level 1 candidate studies across four to six months. Ethics in month one, Quant in month two, FRA across months two and three, and so on. By the time they sit the exam in month five or six, their Ethics material is five months old.
Without systematic review, the forgetting curve predicts that this material is largely gone. Bahrick, Bahrick & Wittlinger (1975) conducted one of the most striking studies on long-term forgetting: they tested people on their high school classmates' names and faces at intervals from weeks after graduation to 50 years later. The decay patterns closely followed Ebbinghaus's curve, with meaningful retention only in subjects who had frequent exposure during the years after graduation — the equivalent of active retrieval.
For CFA and FRM material — abstract financial concepts without daily social reinforcement — the curve is likely steeper than Bahrick's interpersonal data suggests.
Studying without reviewing is like filling a bathtub with the drain open. The question is never whether forgetting is happening — only how fast.
The compounding problem
The forgetting curve doesn't just affect individual topics — it compounds across a study plan. While you're studying FRA in month two, you're simultaneously forgetting Ethics from month one. While you're studying Fixed Income in month three, you're forgetting both Ethics and FRA. By exam month, you have a stack of individually-studied topics, each at a different position on its forgetting curve.
The common response is a "review week" at the end of the study plan: intensive re-reading before the exam. This is massed study — exactly the format that produces the shallowest encoding and the steepest subsequent forgetting. A review week produces a temporary retention boost that peaks around exam day and then drops steeply. The deeper problem is that re-reading is not retrieval.
Cepeda et al. (2008) showed that the benefit of a review session is directly proportional to the retrieval effort required — the more you struggle to pull something back before being reminded of it, the stronger the subsequent memory trace. Re-reading produces almost no retrieval effort. Testing yourself produces a great deal of it.
The recognition trap
There is a specific cognitive illusion that makes the compounding problem worse: the gap between recognition and recall. They feel identical from the inside. They produce radically different results on exam day.
Recognition is a low-threshold process. It asks only: "Have I seen this before?" When you re-read a page of Fixed Income notes six weeks after first studying them, the content feels familiar. Terms surface. Relationships between concepts seem obvious. The material "comes back." That sense of fluency is genuine — but it is recognition, not recall. Your brain is identifying previously encountered information, not reconstructing it from scratch.
Recall asks a harder question: "Can I produce this from nothing?" Under exam conditions, you are given a question, not a passage to recognise. You need to retrieve the definition of modified duration, apply it to a bond's price sensitivity, and do this under time pressure with no contextual cues from a textbook page in front of you. Recognition gives you no meaningful preparation for that task.
Roediger & Karpicke (2006) demonstrated this gap with unusual precision. In their study, one group of students re-read a passage after initial study. A second group took a retrieval test immediately after initial study instead of re-reading. Both groups then took a delayed recall test one week later. The re-reading group felt more confident going into the delayed test — recognition had given them a false sense of preparation. But the retrieval-test group scored significantly higher on actual recall.
The implication for CFA and FRM candidates is direct: if you can recognise the term "duration" when you see it in your notes, that does not mean you can define it cleanly, distinguish it from modified duration and effective duration, or apply it correctly under exam pressure. Re-reading builds recognition. Only retrieval practice builds recall. And exams test recall.
Why the curve feels gentler than it is
The most dangerous feature of the forgetting curve is that it's invisible from the inside. When you re-read your Fixed Income notes after six weeks, they feel familiar. The material comes back quickly. You feel like you knew it all along and just needed a reminder.
This is the recognition-recall gap. Recognition — identifying something as previously encountered — is a much lower-threshold process than recall. You can recognise a concept when prompted with it and be completely unable to produce it unprompted under exam conditions. The feeling of "it's coming back" during re-reading is recognition, not recall recovery. Under exam conditions, you need recall.
Roediger & Karpicke (2006) demonstrated this directly: students who re-read a passage scored significantly lower on recall tests one week later than students who had taken a retrieval test immediately after initial study — even though the rereading group felt more confident going into the delayed test.
What actually fights the curve
The solution is not more initial study time. It is review at the right intervals, using retrieval rather than re-reading.
Spaced repetition algorithms — particularly FSRS 4.5, which models both memory stability and retrievability — are specifically designed to schedule review at the point of maximum benefit: just as material is beginning to fade, before it has faded completely. Reviewing too early wastes time. Reviewing too late wastes the material.
For a CFA or FRM candidate with 300+ hours of material to retain across six months, manually scheduling this is impossible. The volume of concepts across 10–14 topic areas, at varying individual decay rates, produces thousands of review scheduling decisions. This is precisely what adaptive spaced repetition handles — and why candidates who use it consistently outperform those who don't.
How FSRS works in practice
Understanding why FSRS outperforms older algorithms — and simple re-reading schedules — requires understanding what it actually models. FSRS tracks four interconnected parameters for every concept you study: memory stability (S), retrievability (R), difficulty (D), and lapses.
Memory stability represents how long a memory will persist at a given level of retrievability. A concept with high stability can go weeks between reviews without falling below a usable threshold. A concept with low stability needs to be revisited within days or it drops below the point where recall is reliable. Stability increases each time you successfully retrieve a concept — and the increase is larger when you retrieve it just as it was about to fade, rather than when it was still fresh.
Retrievability is the probability that you can successfully recall a concept right now. It decays exponentially between reviews, following a curve that is mathematically identical to the forgetting curve Ebbinghaus described in 1885. FSRS makes this decay visible and computable: it knows, based on your history with a concept, roughly what your probability of recall is at any point in time.
The scheduling decision FSRS makes is straightforward in principle: schedule the next review when retrievability falls to approximately 90%. At that point, you still have a high probability of recalling the concept — so the review session is likely to succeed — but the memory is under enough pressure that successfully retrieving it produces a meaningful boost to stability. This is the optimal window. Too early, and you're reviewing material you'd have remembered anyway, wasting review time. Too late, and you've already lost the concept and must relearn it from scratch.
Unlike earlier spaced repetition algorithms such as SM-2, FSRS recalibrates its model after every single review based on your actual response time and accuracy. It does not apply a fixed schedule to everyone. It learns your specific forgetting patterns for each concept and adjusts continuously.
For a CFA or FRM candidate, this per-concept tracking produces a meaningful practical advantage. Ethics concepts — which often involve nuanced judgment calls and verbal reasoning — may degrade at a different rate than Fixed Income arbitrage mechanics, which require tighter procedural recall. FSRS tracks each concept separately across all your topic areas and knows that Ethics degrades at a different rate than Fixed Income arbitrage for you specifically, based on your actual review history. The practical result is a personalised schedule: you might review Ethics concepts every eight days while Fixed Income concepts require review every three days — because your data shows they decay at different rates. No manual intervention required. The algorithm watches the curve for each concept and tells you exactly what to review today.
What this means for your study schedule
The evidence across Ebbinghaus, Bahrick, Roediger, Karpicke, Cepeda, and the FSRS literature converges on a single conclusion: the passive study loop that most candidates use is empirically broken. Read, highlight, re-read, panic — this sequence produces recognition, not recall, and it offers no defence against the forgetting curve. The review week at the end of the study plan, however intensive, cannot compensate for months of unreviewed material decaying to near-zero retrievability.
The correct loop is structurally different at every step: study the material, attempt immediate recall before looking anything up, then schedule spaced reviews at the intervals the data says are optimal, and arrive at the exam having already reviewed every concept at the right time.
The practical protocol follows directly from this. After each study session, do not re-read your notes. Close them and write down everything you remember from the session — concepts, definitions, relationships, worked examples. The struggle to retrieve material without the notes in front of you is not a sign that you haven't learned it well enough. That struggle is the learning. The cognitive effort of attempted retrieval, even when partially unsuccessful, encodes the material more durably than any amount of passive re-reading. For more on how to build this into a full study method, see the spaced repetition study method guide.
For weekly review, the same principle applies at a larger scale. Do not block out time to review everything. Review only what FSRS tells you is due — the concepts sitting at approximately 90% retrievability that are about to fade below the recall threshold. Reviewing concepts that are still at 99% retrievability wastes time and provides almost no stability benefit. The algorithm's scheduling exists precisely to prevent both over-reviewing (inefficient) and under-reviewing (forgetting).
The exam week insight that follows from all of this is counterintuitive but well-supported: the goal is not to cram the night before the exam. Candidates who arrive at exam week with a healthy spaced repetition schedule behind them do not need to panic-review everything. Their retrieval probability across all topics is already high — maintained by months of optimally timed reviews. Exam week becomes consolidation: working through practice questions under timed conditions, identifying any residual weak spots, and sleeping enough for memory consolidation to complete. The candidates who struggle in exam week are the ones who neglected ongoing review and are now attempting to compress months of spaced repetition into seventy-two hours. The forgetting curve makes that mathematically unlikely to succeed.
Start the review schedule early. Keep it consistent. Trust the intervals. The curve works against every candidate equally — the only variable is whether you have a system designed to fight it.