Spaced Retrieval and the Enrolled Agent (EA) Exam

Two candidates start preparing for the same EA Part 1 sitting in mid-summer. Both buy the same review course. Both have the same number of study hours available.

Candidate A reads each chapter twice, highlights the important parts, and works the end-of-chapter quiz once. The night before the exam, she re-reads the key chapters.

Candidate B never re-reads. After a single first pass through each chapter, she drills herself on the material at expanding intervals. One day later. Three days later. A week later. Three weeks later.

If both walk into the PSI testing center on the same day, the cognitive science predicts Candidate B walks out with a substantially higher score. That is not folk wisdom. It is the most replicated finding in the experimental study of human memory, and it has been sitting on the shelf for more than a century while exam candidates keep doing what Candidate A does.

This week's note is on what the evidence actually says, why it matters specifically for an exam shaped like the Special Enrollment Examination, and what to do about it.

Hermann Ebbinghaus published the first quantitative study of memory in 1885. He memorized lists of nonsense syllables, then tested himself at intervals from 20 minutes to 31 days, and plotted what happened. Within a day, he had lost roughly half of what he had learned. Most of the loss happened in the first few hours.

That graph, the famous forgetting curve, has been challenged, refined, and replicated for 140 years. The cleanest direct replication, Murre and Dros in PLOS One (2015), redid the experiment carefully and confirmed Ebbinghaus's basic shape, with one notable wrinkle: the curve flattens slightly between 24 hours and 31 days, consistent with sleep-driven consolidation. The big picture stands. Without intervention, what you study today is mostly gone in a week.

For the EA exam, that is the core problem. You are being tested on hundreds of dollar thresholds, phase-outs, depreciation lives, due dates, and code section concepts. A study schedule that does not actively counter the forgetting curve is a study schedule that loses ground every day.

The intervention that fights the forgetting curve is called distributed practice, or spacing. Instead of studying the same material massed together (Monday: read Chapter 4 three times), you space the encounters out (Monday, Thursday, the following Wednesday).

Cepeda, Pashler, Vul, Wixted, and Rohrer published the definitive meta-analysis in Psychological Bulletin in 2006. They pooled 839 separate measurements from 317 experiments across 184 articles. The conclusion was unambiguous. Spacing produces better retention than massing. The effect is large, robust across age groups, and shows up across every type of verbal material researchers have tested.

This is not one of those psychology effects that has been embarrassed by the replication crisis. The spacing effect has been confirmed since the late 1800s, in lab settings, in classrooms, with children, with adults, with motor skills, with vocabulary, with concepts. If you wanted to bet on a finding from cognitive psychology, this is the one to bet on.

The 2006 meta-analysis told you spacing works. The 2008 follow-up told you how much spacing is right.

Cepeda and colleagues taught more than 1,350 participants a set of facts, then brought them back for a second study session at gaps ranging from minutes to 3.5 months, then tested them at delays of up to a year. They mapped what they called a "temporal ridgeline" of optimal performance: the best gap between sessions depends on how long you need to remember.

The rough rule that fell out of the data: the optimal gap is about 10 to 20 percent of the retention interval.

Apply that to your EA timeline. If you sit on July 1 and study a topic on May 9, the gap to your test is roughly eight weeks. The 10-to-20-percent rule says you want to revisit that material at gaps of about one to two weeks. Restudying it the same day adds little. Saving it for the comprehensive review at the end is past the optimum and probably loses ground.

Spacing alone is half the picture. The other half is what you do during each spaced encounter.

Karpicke and Roediger published a paper in Science in 2008 with a striking design. Students learned vocabulary items to a criterion of one correct production each. They were then split into four conditions: keep studying and testing, keep studying but stop testing, keep testing but stop studying, or stop both. A week later, they were given a final test.

Repeated studying after initial mastery added almost nothing. Repeated testing produced a large gain. The effect is so robust it has its own name now, the testing effect, and it has been replicated hundreds of times.

The mechanism matters. The act of pulling a fact out of memory is what strengthens the memory trace. Reading the answer does not. Recognizing the answer in a multiple-choice list barely does. Cued recall and free recall do the work.

For the EA exam, this means the study activity that feels productive (reading and re-reading the textbook, watching videos at 1.5x) is mostly inert. The activity that feels uncomfortable (closing the book and trying to write down the §179 limit from memory) is what actually moves the needle.

In 2013, John Dunlosky and four colleagues published a systematic review in Psychological Science in the Public Interest evaluating ten common study techniques. They rated each as low, moderate, or high utility based on the breadth and quality of the supporting evidence.

Of the ten techniques, exactly two received a "high utility" rating: practice testing and distributed practice. Re-reading was rated low. Highlighting was rated low. Summarization was rated low. The keyword mnemonic was rated low.

That is the cleanest single citation in the field for the claim that spaced retrieval is the best-documented study technique we have. The other techniques are not useless, but the evidence for them is much weaker, and several are weaker than students assume.

Three operational moves drop out of the evidence.

First, never re-read passively. If you find yourself reading the same paragraph twice, close the book and try to write down what it said. Look back only after you have tried.

Second, build expanding intervals into your schedule. A topic studied today should be reviewed tomorrow, then in three or four days, then in about a week, then in two to three weeks. The intervals stretch as the memory consolidates.

Third, use questions, not summaries. Every chapter should generate a stack of self-quiz items you actually drill yourself on.

One tool: Anki. Free on desktop, free on Android, paid on iOS. Open-source. Used by medical students, language learners, and anyone who has taken long-term retention seriously over the last twenty years. Its scheduling algorithm handles the expanding-interval math for you, so you stop worrying about gaps and just answer cards when they come up.

The catch is that Anki gives you no content. You build your own cards as you work through your review course, or you import a deck someone else built. Shared EA decks exist on AnkiWeb, but quality varies and most are out of date on current dollar thresholds. Plan to build cards yourself. The act of writing the card is part of the encoding step.

Three rules are non-negotiable, whatever the tool.

Cards must require recall, not recognition. "What is the §179 expensing limit for 2025?" is a recall prompt. The multiple-choice version of the same question is a recognition prompt and does perhaps a third of the work.

One concept per card. If a card asks four things at once, you cannot tell which part you missed, and the algorithm cannot schedule the missed piece correctly.

Build cards as you go, not at the end. The card you write while reading Chapter 4 is the encoding event. Postponing card-building to a dedicated "review phase" reinvents cramming with extra steps.

Two qualifications, because the cognitive science is not a magic wand.

Spaced retrieval is a tool for declarative knowledge. The EA exam has plenty of that, especially in Part 1 and Part 3. Part 2's complex business returns also reward worked-problem practice with full scenarios, where you reason through a partnership basis adjustment or a corporate distribution. Spaced retrieval supports that work. It does not replace it.

And spaced retrieval works best on material you have understood once. If you never grasped how §1245 recapture differs from §1250 recapture, no amount of card drilling will install the concept. Encoding has to happen first. The schedule keeps it from leaking out.

What do we know about pass rates, and where does that information come from?