The 1-3-7 Revision Rule vs What Spaced Repetition Shows

"Revise after 1 day, 3 days and 7 days" has no traceable academic source, and the real evidence says it is right for a test next week and wrong for an exam months away. I searched for an origin - a study, a researcher, a book - across seven different search engines and found nothing beyond the phrase repeating itself across Indian exam-prep blogs with no citation anywhere in the chain. What does exist is Cepeda, Pashler, Vul, Wixted and Rohrer's 2006 meta-analysis in Psychological Bulletin, pooling 317 experiments and 14,811 participants, and buried in its Table 7 is the exact comparison this rule needs and never cites: at a retention interval of 10 days, a 1-day gap between study sessions won (74% recall). At a retention interval of 168 days, the same 1-day gap was the worst option tested (33%), and a 28-day gap won instead (56%).

That reversal is the whole story. "1-3-7" is a schedule built entirely from gaps under 8 days. For a class test next week, the evidence agrees with it. For a CAT, JEE, NEET, GATE or UPSC exam months out, the same evidence says the rule stops expanding exactly where it needs to keep growing.

Where the rule actually comes from - or doesn't

I looked for a source before writing anything else, because a rule this specific ("1 day, 3 days, 7 days," sometimes extended to "1-3-7-15-30") sounds like it ought to have one. I searched variations of the phrase across Bing, DuckDuckGo, Brave, Mojeek, Marginalia, Yandex and Yahoo. None returned a study, a named researcher, a textbook, or even a single source that cited one. What exists instead is the phrase itself, repeated across coaching blogs and exam-prep listicles as settled fact, each one citing nothing or citing each other.

The nearest thing to a justification I could find in this genre is a vague appeal to "the forgetting curve," meaning Hermann Ebbinghaus's 1885 self-experiment. It is worth being precise about what that study actually was, because it is invoked constantly and read almost never. Ebbinghaus was his own only subject. He memorised lists of nonsense syllables, let himself forget them for varying periods, then measured how much faster he relearned them the second time - the "savings" method. His curve shows retention dropping sharply within the first hour and levelling off by about a day. Nowhere in that design is a revision schedule of any kind, let alone specific gaps of 1, 3 and 7 days. The forgetting curve describes how fast one 19th-century psychologist forgot nonsense syllables. It does not prescribe when to revise anything, and citing it as the basis for "1-3-7" is citing a paper that does not contain the rule attributed to it.

So the honest position is this: I cannot rule out that "1-3-7" started from a real synthesis somewhere I did not find. But after a genuine search, I found no traceable origin, no cited study, and a popular justification that does not survive contact with the source it names. That absence is itself worth stating plainly, because "widely repeated" and "evidence-based" are being treated as synonyms across a large share of Indian exam content, and here they are not the same thing.

What the real evidence says instead

Cepeda et al.'s 2006 meta-analysis is the paper "1-3-7" should be citing if it is trying to invoke spaced repetition at all. Across all retention intervals it found spaced study produced 47.3% correct on a final test against 36.7% for massed study - a real, well-evidenced effect. But the paper's central and much less quoted finding is that the optimal gap between study sessions is not fixed - it scales with how long you need to remember the material for. Their own words: "for most practical purposes the retention interval of interest is several months or years... so the optimal ISI will likely be well in excess of one day."

Table 7 of the meta-analysis is where that claim becomes a number you can actually use, because it is the one place in the paper where the same experimental family varied the gap between study sessions and the length of time before the final test. Here is the row that matters for anyone treating "1-3-7" as universal advice.

Read the second row carefully, because it is doing two things at once. Recall rises from 33% to 56% as the gap grows from 1 day to 28 days - so far, "longer gaps are better" holds up. Then it falls back to 43% at an 84-day gap. There is an interior optimum, and past it, spacing further apart costs you too. A chapter revised only once, 84 days after you learned it, retained about as much as one revised the very next day. Over-spacing is a real, measured failure mode, not just under-spacing.

Now place "1-3-7" against that same 168-day row. Its three reviews land at gaps of 1, 2 and 4 days between each successive pass (day 1, then two days later, then four days after that) - all of them sitting inside the worst-performing third of the curve above. The rule never reaches anywhere near the 28-day gap the data identifies as the peak, and it stops entirely at day 7, leaving the other 161 days of a five-and-a-half-month retention interval with no schedule at all.

Where "1-3-7" is not wrong

I want to be fair to it, because the first row of that table says something the second one doesn't. At a 10-day retention interval, the 1-day gap won outright (74% against 69% for a 7-day gap). If what you actually need is to hold onto a chapter for a class test next week, or a weekly sectional test on Sunday, a same-week revision schedule is exactly what this evidence supports. The direction fully reverses only once the retention interval stretches into months - which is precisely the situation a CAT, JEE, NEET, GATE or UPSC aspirant is in for most of the material they cover, and precisely the situation "1-3-7" is usually being recommended for.

So the honest verdict is not "the rule is baseless nonsense." It is narrower and more useful than that: the rule is shaped correctly for a short-horizon exam and shaped wrong for a long-horizon one, and almost nobody quoting it says which one they mean.

1-3-7 against Karma Yogi's own R0-R7 schedule

This site's own revision-round model is built for the long-horizon case specifically, so it is worth holding the two side by side rather than treating "spaced repetition" as one undifferentiated idea. Karma Yogi's rounds run R1 three days after a mock, then each later round from when the previous round was completed - R2 seven days after R1, R3 fourteen days after R2, R4 twenty-one days after R3, R5 thirty days after R4, R6 forty-five days after R5, R7 sixty days after R6 (the exact schedule implemented in frontend/src/lib/revisionUtils.ts and specified in this site's own PRD). If none of the rounds slip, here is what both schedules look like measured the same way: cumulative days since the original event.

Two things fall out of that table. First, all three schedules are technically "expanding interval" designs - even the plain "1-3-7" mildly expands its own gaps (1, then 2, then 4 days). The real difference is not expanding-versus-fixed, it is how far the schedule keeps expanding before it quits. Second, Karma Yogi's R7 lands at day 180 - close to the 168-day retention interval Cepeda et al. actually tested, and in the same neighbourhood as the 28-day-gap-wins finding once you're revising material for the fourth or fifth time. "1-3-7" and its extended cousin are both out of runway by day 7 or day 30, which is most of a CAT prep season and a small fraction of a JEE or UPSC one.

I have to be honest about the limits of that comparison too, and this site has made the point elsewhere so I will only state it rather than re-argue it: Cepeda et al.'s own Table 8 found expanding schedules are not significantly better than fixed ones (62.0% versus 58.6% across 18 studies, p = .61). So Karma Yogi's shape being closer to the evidence's preferred horizon does not mean its expanding shape is proven superior to a fixed-interval alternative reaching the same eventual gap. Both points are true at once: reaching day 180 is defensible on this evidence, and doing so via expanding rather than fixed intervals is a design choice, not a finding. The fuller treatment of that specific question is in the CAT spacing post and the NEET spacing post, and I am not repeating their tables here.

What this means for revising a mock or a chapter

• Ask what your retention interval actually is before picking a schedule. A weekly class test and a November board exam are not the same problem, and the evidence genuinely points in different directions for each.
• If the exam is under two weeks away, a short fixed schedule is defensible. This is the one place "1-3-7" and the data agree.
• If the exam is months away, plan reviews that keep expanding past day 7. The 28-day-gap result is the single most directly relevant number in this literature for anyone with a multi-month horizon, and it is also the one number in this whole area of study nobody quoting "1-3-7" seems to have read.
• Do not treat 28 days as a universal answer either. It is one measured optimum from one 168-day experiment, not a law. Cepeda et al.'s own conclusion is that they cannot say with certainty what the optimal gap is at any given horizon - only that it grows with the horizon.
• A schedule that runs out of instructions is a real design flaw. Whatever you use, it should still have something to say in month four of a six-month plan.

Where this is weak

• "No traceable source" is a search result, not a proof of non-existence. I searched extensively and found nothing; I cannot certify that no such source exists anywhere, only that I could not find one after a genuine attempt across seven engines.
• Table 7's 168-day row is thin. It is one experimental series (Cepeda et al.'s own Experiment 2a), not an aggregate across dozens of studies the way the paper's topline 47.3%-versus-36.7% figure is. Treat the 28-day figure as the best single data point available for a long horizon, not as a robust average.
• The materials are word lists, not mock papers. Cepeda et al.'s title says "verbal recall tasks" - paired associates and lists. Nothing in it was a DILR set, a Physics numerical, or an NCERT diagram.
• 85% of the meta-analysis's data are young adults in a laboratory, a fact the authors state themselves. The mechanism plausibly transfers to exam revision; the exact percentages do not.
• Neither the folk rule nor this rebuttal has been tested on Indian competitive-exam populations. No randomised trial of revision spacing exists for CAT, JEE, NEET, GATE or UPSC preparation, to my knowledge.
• I have not read Ebbinghaus's 1885 monograph in the original. What I have is a secondary summary of his method and findings, which is enough to say his study contained no 1-3-7 schedule, but not enough to characterise every detail of his century-old work with full confidence.

Sources

• Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354-380. Read in full at eScholarship. All figures above are from its Tables 1, 7 and 8.
• Ebbinghaus's 1885 method (single-subject self-experiment, nonsense syllables, the savings method) is described here from a secondary summary, not from the original 1885 monograph, which I did not read.
• Search for an origin of "1-3-7" conducted across Bing, DuckDuckGo, Brave, Mojeek, Marginalia, Yandex and Yahoo, 2026-09-10. No cited academic source was found on any of them.
• Karma Yogi's own revision-round schedule (R1 at 3 days, then 7/14/21/30/45/60 from the previous round) as implemented in frontend/src/lib/revisionUtils.ts and specified in this site's PRD.

Karma Yogi tracks revision rounds against the actual date each pass happened, so a schedule that assumes no slippage - like every table above - can be checked against what really occurred. Whether 3-7-14-21-30-45-60 beats a fixed alternative reaching the same eventual gap remains, honestly, an open question.