---
slug: the-forgetting-curve
passed: 2026-11-08
by: claude-gifted-feynman-wvjfa4, one agent on oversight/claims-pass.md (22:29 to 23:05 UTC, while the page was still changing; it audited the state of 22:57), its ten findings each re-read against the source by the author before fixing
checked: 176
confirmed: 152
wrong: 10
unverifiable: 0
observed: 14
stale: 0
---
## Claims
- WRONG | [ebb-29] In 1879 Hermann Ebbinghaus, then 29, began learning rows of meaningless syllables | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | Age 29 is supported (born 24 Jan 1850; M&D: "who was 29 during his experiments in 1879-1880"). But "began learning rows" in 1879 is contradicted by Ebbinghaus himself, memory3.htm Sec. 16: "During a long time preliminary experiments of a similar nature had preceded the definite tests of the first period", and Sec. 13 (metronome "at first", then "very soon the ticking of a watch"); M&D: "The experiment itself was preceded by a period in which he tried out a variety of materials and methods." The tests behind the curve began in 1879-80; his syllable learning did not.
- CONFIRMED | [ebb-1879] In 1879 Hermann Ebbinghaus | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 27: "The investigations in question fell in the year 1879-80 and comprised 163 double tests." Supports the date of the tests (see ebb-29 for "began").
- CONFIRMED | [watch] to the tick of a watch | https://psychclassics.yorku.ca/Ebbinghaus/memory3.htm | memory3.htm Sec. 13 rule 2.
- CONFIRMED | [only-subject] He was the experimenter and the only subject. | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D: "he used only himself as a subject".
- CONFIRMED | [twice] A row counted as learned when he could recite it twice without a slip | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 27; also memory3.htm Sec. 16 "two successive faultless reproductions" for the first period.
- CONFIRMED | [relearn] It was how long it took him to learn it again. | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 27.
- CONFIRMED | [saving-of-work] He called it the saving of work: the share of the first learning he no longer had to do. | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 28: D "the saving of work", Q its relation to the first learning time as a per cent.
- CONFIRMED | [three-points] After 19 minutes he saved 58.2 per cent of it, after a day 33.7, after a month 21.1. | https://psychclassics.yorku.ca/Ebbinghaus/table23.jpg | table23.jpg viewed: 0.33 h 58.2; 24 h 33.7; 31x24 21.1. Readings file matches the image row for row.
- CONFIRMED | [nineteen] After 19 minutes | https://psychclassics.yorku.ca/Ebbinghaus/table15.jpg | table15.jpg viewed: "I. 19 minutes. 12 tests." (record "where" names the 63/525-min notes; the line is the 19-min heading on the same image).
- CONFIRMED | [q-forgotten-ideas] Of ideas that are almost entirely forgotten he asks how their course can be traced / quote: if the almost entirely forgotten ideas return no more to consciousness | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 26, exact.
- CONFIRMED | [bq-27] his answer is to relearn them and time it / quote: After a definite time, the hidden but yet existent dispositions laid down by the learning of a syllable-series may be strengthened by a further memorisation of the series, and thereby the remaining fragments may be united again to a whole. The work necessary for this compared with that necessary when such dispositions and fragments are absent gives a measure for what has been lost as well as for what remains. | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 27, exact; attribution Section 27 correct.
- CONFIRMED | [come-apart] a row you cannot recite a syllable of can still come back faster than a new one | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Discussion: "certain inaccessible representations, seemingly forgotten, can still be relearned faster".
- OBSERVED | [typed-163] all 163 tests of it, which the Artificial Wasteland typed in and added up again | research/the-forgetting-curve/data/ebbinghaus-1885-forgetting-tests.csv | record: CSV header (research/the-forgetting-curve/data/ebbinghaus-1885-forgetting-tests.csv); recreated: I read all 163 rows of table15-22.jpg against the CSV (all match) and reran check-transcription-vs-1885.mjs in scratch: "680 of 680 cells read the same". Note RUNS.md: one cell (1day,B,6) was first typed from arithmetic as 883, since corrected to the printed 863.
- CONFIRMED | [browser-only] Your rows stay in this browser and nowhere else. | public/strata/the-forgetting-curve/app.mjs | data: app.mjs stores rows in localStorage only; its only network call is GET ./data.json; /media/home-banner.js and plain-button.js contain no fetch/sendBeacon/XHR/storage access.
- CONFIRMED | [five-seconds] A wrong letter, or five seconds without one, is a hesitation | public/strata/the-forgetting-curve/app.mjs | data: app.mjs armStall() setTimeout(... 5000); wrong letter -> reciteState onCourse false -> hesitate().
- CONFIRMED | [shortest] Ebbinghaus's shortest interval came from learning eight rows and then going back to the first | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 27.
- CONFIRMED | [twelve-45] each of his points is a mean of 12 to 45 tests of six or eight rows | https://psychclassics.yorku.ca/Ebbinghaus/table15.jpg | n per point 12,16,12,26,26,26,45 (images + CSV). Record nit: the excerpt "VII. 31 days. 45 tests" is on table21.jpg, not table15.jpg, which the record names as source.
- CONFIRMED | [q-150] He read aloud at / quote: 150 strokes per minute | https://psychclassics.yorku.ca/Ebbinghaus/memory3.htm | memory3.htm Sec. 13 rule 2.
- CONFIRMED | [q-escapement] kept by a watch / quote: The mechanism of escapement of most watches swings 300 times per minute. | https://psychclassics.yorku.ca/Ebbinghaus/memory3.htm | memory3.htm Sec. 13 rule 2.
- CONFIRMED | [q-hesitation] He passed freely between reading and trying / quote: upon hesitation the rest of the series was to be read through to the end before beginning it again. | https://psychclassics.yorku.ca/Ebbinghaus/memory3.htm | memory3.htm Sec. 13 rule 1.
- CONFIRMED | [q-two-errorless] In the forgetting tests he learned until / quote: two errorless recitations | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 27.
- CONFIRMED | [md-once] Murre and Dros, replicating him in 2015, learned to once. | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Methods: "We chose to also learn to once correct."
- CONFIRMED | [he-timed] He timed himself. | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 28: L, WL in seconds; memory3.htm Sec. 15.
- CONFIRMED | [q-2300] syllables from the consonants and the eleven vowels and diphthongs of German / quote: about 2,300 | https://psychclassics.yorku.ca/Ebbinghaus/memory3.htm | memory3.htm Sec. 11; footnote lists the 11 vowels/diphthongs.
- CONFIRMED | [syllables-717] not entries in the CMU Pronouncing Dictionary and not on a short list of slang | research/the-forgetting-curve/make-syllables.mjs | data; recreated make-syllables.mjs in scratch: "1350 formed, 717 kept", byte-identical to data/syllables.json and to data.json syllables; positive control cat/dog/bed are in the dictionary.
- CONFIRMED | [syllables-counts] such spellings | research/the-forgetting-curve/data/syllables.json | data; 18 x 5 x 15 = 1350, 717 kept (recreated).
- CONFIRMED | [eight-rows] He learned eight rows at a sitting (six at the 11 o'clock sitting) | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 27 (38 double tests 11-12 A.M. of six series).
- CONFIRMED | [days-alike] and kept his days alike | https://psychclassics.yorku.ca/Ebbinghaus/memory3.htm | memory3.htm Sec. 13 rule 7.
- CONFIRMED | [q-def] the 85 seconds being his allowance for the two recitations at the end | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 28.
- CONFIRMED | [q-def-formula] how long the first learning of the sitting's rows took, how long the relearning, the difference, and the saving as a percentage | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 28 (L, WL, WLk, D, Q).
- CONFIRMED | [times-of-day] The tests ran in 1879 and 1880 at three times of day | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 28; memory3.htm Sec. 16.
- CONFIRMED | [corrections] he scaled the relearning times down, by about 5 and 12 per cent, to make them comparable | https://psychclassics.yorku.ca/Ebbinghaus/table15.jpg | table15.jpg viewed: "must be decreased by about 5 per cent"; WLk/WL in CSV 0.949-0.951.
- CONFIRMED | [corrections-12] by about 5 and 12 per cent | https://psychclassics.yorku.ca/Ebbinghaus/table15.jpg | table16.jpg viewed: "approximately 12 per cent"; WLk/WL 0.880-0.881.
- CONFIRMED | [evening] 6 to 8 in the evening | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 28 "6-8 P. M."
- CONFIRMED | [range-6d] After six days single tests saved anywhere from | public/strata/the-forgetting-curve/data.json | data; recomputed from CSV: six-day min 2.496 (B6), max 40.334 (B1).
- OBSERVED | [typed-680] The two readings agree on every one. | research/the-forgetting-curve/RUNS.md | record: RUNS.md; recreated: node check-transcription-vs-1885.mjs (scratch copy) -> "680 of 680 cells read the same in the 1885 OCR as in the 1913 transcription (801 numbers in the OCR span, lines 4806 to 6574)"; positive control: 3 planted cells -> 3 DIFFER lines, exit 1.
- CONFIRMED | [de-edition] the Internet Archive's machine reading of the German first edition of 1885 | https://archive.org/metadata/berdasgedcht00ebbi/metadata | archive-metadata.json: publisher "Leipzig, Duncker & Humblot", date 1885, contributor The Library of Congress.
- CONFIRMED | [holds-156] of the 163 rows both hold to the printed decimal | public/strata/the-forgetting-curve/data.json | data; recomputed independently: 7 rows fail (63min A16, 1day B6, 2days C1, 6days B3, 6days C3, 31days C2, 31days C5), 156 hold.
- CONFIRMED | [hour-m] The average printed under them, and carried into his summary table, is | https://psychclassics.yorku.ca/Ebbinghaus/table23.jpg | table23.jpg and table15.jpg viewed: 44.2.
- CONFIRMED | [hour-m-block] Its sixteen printed | https://psychclassics.yorku.ca/Ebbinghaus/table15.jpg | table15.jpg viewed: "II. 63 minutes. 16 tests. The learning at the time A, the relearning at the time B." (excerpt updated during the pass).
- CONFIRMED | [md-table3] So is the figure in the table that Murre and Dros's replication of 2015 compares itself with. | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Table 3 (kept XML): 1 hour 0.442. Online NOT FOUND is the pipe-formatted excerpt, not absence.
- CONFIRMED | [six-day-row] Learning 1,293 seconds, relearning 939, difference 354 | research/the-forgetting-curve/data/ebbinghaus-1885-forgetting-tests.csv | data; table20.jpg viewed: 1293 939 354 21.0; 1885 OCR reads 21.
- CONFIRMED | [mis-l-row] The learning time is printed | research/the-forgetting-curve/data/ebbinghaus-1885-forgetting-tests.csv | data; table17.jpg viewed: 863 761 122 14.9; 1885 OCR reads 863.
- CONFIRMED | [allowance-64] we use 64 seconds, his 0.41 seconds a syllable for two recitations of six rows | research/the-forgetting-curve/analyze.mjs | data; 2 x 6 x 13 x 0.41 = 63.96.
- CONFIRMED | [md-1880] Murre and Dros's comparison table, which cites his handwritten report of 1880, gives 0.442 for this interval too | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Table 3 caption cites [8] (Urmanuskript 1880) p. 56 and gives 0.442. Near-miss: see MINOR finding on the 1880 value in M&D own workbook.
- CONFIRMED | [md-1880-hand] which cites his handwritten report of 1880 | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Introduction.
- CONFIRMED | [md-1983] that manuscript was published in 1983 | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Introduction; ref [8] Passau: Passavia.
- CONFIRMED | [q-2-half] Discussing the day-long point, the 1913 English text says the decrease from 9 to 24 hours / quote: would accordingly have been 2 1/2 per cent | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 29 "2 1/2 per cent"; 1913 print p. 77 "2½".
- CONFIRMED | [p77-image] the 1913 English text says the decrease from 9 to 24 hours | https://archive.org/download/memorycontributi00ebbiuoft/page/n90_w900.jpg | leaf-n90.jpg viewed: "accordingly have been 2½ per cent. In the period 24 to 48 hours it would have been 6.1 per cent".
- CONFIRMED | [de-2-1] the German of 1885 says 2,1, which is what the table gives | https://archive.org/download/berdasgedcht00ebbi/berdasgedcht00ebbi_djvu.txt | 1885 OCR line 6753: "Abnahme der Nachwirkung 2,1 °/o".
- CONFIRMED | [both-6-1] both editions give the decrease from 24 to 48 hours as 6.1 per cent | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 29.
- CONFIRMED | [de-6-1] both editions give the decrease from 24 to 48 hours as 6.1 per cent | https://archive.org/download/berdasgedcht00ebbi/berdasgedcht00ebbi_djvu.txt | 1885 OCR line 6754: "von 24 bis zu 48 Stunden 6,1 °/o".
- CONFIRMED | [pe-one] his own probable errors were about one point | https://psychclassics.yorku.ca/Ebbinghaus/table23.jpg | table23.jpg viewed: P.E.m 1, 1, 1, 1.2, 1.4, 1.3, 0.8. (Record excerpt is only the column heading; the values are on the image.)
- CONFIRMED | [formula-form] He wrote it as | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 29; leaf-n90.jpg.
- CONFIRMED | [q-least-squares] with k = 1.84 and c = 1.25 found / quote: with merely approximate estimates, not involving exact calculation by the method of least squares | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 29.
- CONFIRMED | [k-c] with k = 1.84 and c = 1.25 | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 29.
- CONFIRMED | [calc-column] His column of values calculated from the formula is the formula's to the printed decimal, except at 64 minutes | https://psychclassics.yorku.ca/Ebbinghaus/table24.jpg | table24.jpg viewed (all 7 rows); recomputed b(t): 56.975 57.0, 46.773 46.7 (only miss), 34.491, 30.411, 28.058, 24.916, 21.228.
- CONFIRMED | [q-great-danger] One source of error he called / quote: a source of great danger | https://psychclassics.yorku.ca/Ebbinghaus/memory3.htm | memory3.htm Sec. 14.
- CONFIRMED | [q-secret] and it was himself / quote: the secret influence of theories and opinions which are in the process of formation | https://psychclassics.yorku.ca/Ebbinghaus/memory3.htm | memory3.htm Sec. 14.
- CONFIRMED | [bq-29] And he asked for no more than this from the formula / quote: Of course this statement and the formula upon which it rests have here no other value than that of a shorthand statement of the above results which have been found but once and under the circumstances described. | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 29, exact.
- CONFIRMED | [heller] Heller, Mack and Seitz published a replication with two subjects in 1991, in German | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Introduction: "published only in German".
- MINOR | [heller-ref] Replikation der Ebbinghausschen Vergessenskurve mit der Ersparnismethode | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | Page title "Replikation der Ebbinghausschen Vergessenskurve mit der Ersparnismethode" differs from the excerpt it rests on (M&D ref [21]: "Ebbinghaus’schen ... Ersparnis-methode: \"Das Behalten und Vergessen als Function der Zeit\"", Z. Psychol. 199: 3-18); a library record found by search gives "Ebbinghaus'schen". Apostrophe dropped, subtitle and pages omitted.
- CONFIRMED | [heller-two] a replication with two subjects | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D: "The ages of the two subjects in Heller et al. are not given."
- CONFIRMED | [mack-seitz-names] Murre and Dros tabulate its two curves under the names Mack and Seitz | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Table 3 header Ebbinghaus|Mack|Seitz|Dros (kept XML). Online NOT FOUND = pipe formatting.
- CONFIRMED | [dros-22] Jaap Murre and Joeri Dros published one in 2015, with Dros, 22, as the only subject | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Methods.
- CONFIRMED | [dros-70h] learning 70 hours of Dutch-shaped nonsense over 75 days | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D: "about 70 hours" of data collection (learning and relearning).
- CONFIRMED | [dros-dutch] Dutch-shaped nonsense | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Methods.
- CONFIRMED | [dros-75] over 75 days | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D: "a 75 day period"; experimental phase 01-12-2011 to 13-02-2012.
- CONFIRMED | [dros-online] and put the raw times online | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Results: osf.io/6kfrp.
- CONFIRMED | [dros-time] from the times of each list, Dros saved | research/the-forgetting-curve/data/murre-dros-2015-dros-times.csv | data; the 69 lists in the CSV equal M&D Table 2 cell for cell; mean of per-list savings 42.14 (20 min), 8.70 (31 days). M&D Table 2 prints 0.421 and 0.090 (ratio of mean times) - see MINOR.
- CONFIRMED | [beads] beyond saying that Heller and his colleagues / quote: use a chain with wooden, colored beads, much like a rosary to keep count | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Methods line 65, exact.
- OBSERVED | [label-only] For its Mack and Seitz columns that label is all the paper gives | research/the-forgetting-curve/RUNS.md | record: RUNS.md; recreated: grep of the kept M&D text for Heller|Mack|Seitz -> 24 lines (21 31 55 63 65 71 79 461 480 483 484 529 558 580 581 657 665 666 844 856 866 870 872 932; RUNS.md lists 13, the rest are table cells, the reference and passages on fits/sleep); none says whether the Mack and Seitz columns are times or repetitions.
- CONFIRMED | [reps-data] by repetitions he saved | research/the-forgetting-curve/data/murre-dros-2015-repetitions.csv | data; murre-dros-2015-repetitions.csv = M&D Table 1 as printed (20 min 30.77/16.26); (30.77-16.26)/30.77 = 47.16%; make-replication-data.mjs --check ok (scratch).
- CONFIRMED | [q-time-in-s] Murre and Dros's comparison table introduces its numbers as / quote: based on time in s | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Results line 461, exact.
- CONFIRMED | [dros-col] but its Dros column is his repetition count | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Table 3 Dros 0.472 = (30.77-16.26)/30.77 from Table 1 (repetitions), where Table 2 time mean is 0.421; also 31 days 0.041 vs 0.090.
- CONFIRMED | [dros-col-t1] but its Dros column is his repetition count | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Table 1 (kept XML): 20 min 10 / 30.77 / 2.90 / 10 / 16.26 / 2.27.
- CONFIRMED | [night-data] In three of the four curves, savings after a day is as high as after nine hours or higher | public/strata/the-forgetting-curve/data.json | data; Mack 28.51 -> 31.60, Seitz 26.98 -> 27.00, Dros time 27.32 -> 31.24 (recomputed); M&D Table 3 Mack .285/.316, Seitz .270/.270.
- CONFIRMED | [q-too-large] He did not believe it / quote: It would fit in well with the other observations to consider the number 33.7 per cent for the relearning after 24 hours as somewhat too large and to suppose that with a more accurate repetition of the tests it would be 1 to 2 units smaller. | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 29, exact.
- CONFIRMED | [q-not-credible] He weighed the obvious explanation and set it aside / quote: does not become credible even under the plausible assumption that night and sleep, which form a greater part of the 15 hours but a smaller part of the 24, retard considerably the decrease in the after-effect. | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 29, exact.
- CONFIRMED | [later-tests] Tests he made later, in 1883 and 1884 and in the course of other work, with rows of twelve syllables relearned after a day, gave 33.4 | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 30.
- CONFIRMED | [later-334] with rows of twelve syllables relearned after a day, gave 33.4 | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 30.
- CONFIRMED | [later-other-work] in the course of other work | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 30 "in the course of widely different investigations".
- CONFIRMED | [jenkins] In 1924 Jenkins and Dallenbach read his point as an effect of sleep and tested it | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Discussion (secondary for J&D).
- CONFIRMED | [jenkins-ref] In 1924 Jenkins and Dallenbach | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D ref [50]: Jenkins JG, Dallenbach KM (1924) Am J Psychol 35: 605-612.
- CONFIRMED | [trace] Murre and Dros trace research on sleep and memory back to it | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Discussion.
- CONFIRMED | [four-panels] In each of the four curves they find a point at one or two days lying above a smooth fit by a standard error or more | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D: "In all four panels, at least one of these points [day 1, day 2] is above the fitted power function curve at a distance of at least one standard error."
- CONFIRMED | [q-predict] and they keep the verdict open / quote: would predict such a jump, but for this particular type of experiment this remains to be established. | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Discussion, exact.
- OBSERVED | [tally] of them read the numbers as the share of a lesson that is kept or lost | research/the-forgetting-curve/RUNS.md | Filed STALE by the pass: the source record's excerpt for this tally quoted the first run of check-retellings.mjs (20 pages; 18; 14), true until two retellings were dropped at about 22:50 UTC; the page never showed the old count after that, and its figures (18 pages; 16; 13) are the second run, recreated by the pass. Recorded here as OBSERVED against that run; the excerpt now quotes it. record: RUNS.md. The entry quotes the first run, "20 pages; 18 read the figures as a share ...; 14 give figures not his; 0 failing", true of the page until two retellings were dropped (about 22:50-22:54 UTC, 2026-11-08). The page now says 18 / 16 / 13; RUNS.md has that second line and I recreated it: check-retellings.mjs in the scratch mini-repo -> "18 pages; 16 read the figures as a share remembered or forgotten; 13 give figures not his; 0 failing". My own reading of the 18 kept copies agrees with every share/his flag. Fix: the entry's excerpt in assay/sources/the-forgetting-curve.json (built by research/the-forgetting-curve/make-source-record.mjs).
- OBSERVED | [kept-each] and kept a copy of each | research/the-forgetting-curve/data/retellings.json | record: retellings.json + sources/retellings/INDEX.md; recreated: all 18 kept files present and each quotation found in its copy (check-retellings.mjs, 0 failing).
- CONFIRMED | [openstax] section 8.3, in the first edition (2014) and the second (2020) / quote: Due to storage decay, an average person will lose 50% of the memorized information after 20 minutes and 70% of the information after 24 hours (Ebbinghaus, 1885/1964). | https://openstax.org/books/psychology-2e/pages/8-3-problems-with-memory | kept 2e copy and live (sources.mjs FOUND); 1e kept copy has the identical sentence.
- CONFIRMED | [openstax-1e] in the first edition (2014) and the second (2020) | https://openstax.org/books/psychology/pages/8-3-problems-with-memory | kept 1e copy: "Publication date: Dec 8, 2014".
- CONFIRMED | [openstax-2e] in the first edition (2014) and the second (2020) | https://openstax.org/books/psychology-2e/pages/8-3-problems-with-memory | kept 2e copy: "Publication date: Apr 22, 2020".
- CONFIRMED | [pt] in April 2025 / quote: His research showed that we forget about 50 percent of what we learn within an hour, 75 percent within a day, and up to 90 percent within a week. | https://www.psychologytoday.com/us/blog/optimizing-success/202504/why-you-forget-so-much-of-what-you-just-learned | kept copy and live (FOUND).
- CONFIRMED | [pt-date] in April 2025 | https://www.psychologytoday.com/us/blog/optimizing-success/202504/why-you-forget-so-much-of-what-you-just-learned | kept copy: "Posted April 24, 2025".
- CONFIRMED | [kohn] a training article by Art Kohn of March 2014 | https://www.learningguild.com/articles/brain-science-the-forgetting-curvethe-dirty-secret-of-corporate-training | kept copy: 50 percent within one hour, 70 percent within 24 hours, 90 percent within a week; no citation; Figure 1 has no values or source (image viewed).
- CONFIRMED | [kohn-date] of March 2014 | https://www.learningguild.com/articles/brain-science-the-forgetting-curvethe-dirty-secret-of-corporate-training | kept copy JSON-LD datePublished 2014-03-13 (online NOT FOUND: the date sits in a <script> block the tool strips).
- OBSERVED | [kohn-noname] which cites nothing and does not name Ebbinghaus | research/the-forgetting-curve/RUNS.md | record: RUNS.md; recreated: grep -c -i ebbinghaus on the kept page -> 0; positive controls: "forgetting" -> 4 in the same file, "ebbinghaus" -> 5 in the Menemsha copy.
- CONFIRMED | [thalheimer] Will Thalheimer, who collected figures like these in 2010, put it plainly / quote: Rules-of-thumb that show people forgetting at some pre-defined rate are just plain false. | https://www.worklearning.com/2010/12/14/how-much-do-people-forget/ | RESTS ONLY ON www.worklearning.com (ruling: do not fetch); kept copies only. kept blog copy (2010-12-14) and kept report PDF both contain the sentence. Rests only on www.worklearning.com (no-fetch host per ruling) - online UNREADABLE.
- CONFIRMED | [col-iv] 55.8 per cent after an hour, 66.3 after a day, 74.6 after six days | https://psychclassics.yorku.ca/Ebbinghaus/table23.jpg | table23.jpg viewed: v 55.8, 66.3, 74.6.
- CONFIRMED | [edutopia] Edutopia in 2017 has it as information / quote: quickly forgotten—roughly 56 percent in one hour, 66 percent after a day, and 75 percent after six days | https://www.edutopia.org/article/why-students-forget-and-what-you-can-do-about-it | kept copy and live (FOUND).
- CONFIRMED | [edutopia-date] Edutopia in 2017 | https://www.edutopia.org/article/why-students-forget-and-what-you-can-do-about-it | kept copy: September 20, 2017.
- CONFIRMED | [col-iv-head] He headed that column / quote: The amount forgotten was thus equivalent to v % of the original in terms of time of learning | https://psychclassics.yorku.ca/Ebbinghaus/table23.jpg | table23.jpg viewed, column IV heading exact.
- CONFIRMED | [eli-exp] and one takes it for his discovery / quote: Once he'd discovered the exponential decline of memory, Ebbinghaus could identify the factors that contribute to it. | https://elearningindustry.com/what-is-the-forgetting-curve-and-how-to-combat-it | kept eLearning Industry (Denny) copy: "Once he’d discovered the exponential decline of memory, Ebbinghaus could identify the factors that contribute to it." (That page gives no formula; R = e^-t/S is on Case Western, Structural Learning, RemNote.)
- CONFIRMED | [some-exp] Some pages give an exponential | https://structural-learning.com/post/ebbinghaus-forgetting-curve | kept Structural Learning copy: "A simple exponential model is often written as R = e -t/S" (it adds "This simplified model differs from Ebbinghaus's original logarithmic fit").
- CONFIRMED | [wiki-after] English Wikipedia sets the exponential out after his own formula | https://en.wikipedia.org/wiki/Forgetting_curve | kept /wiki/ HTML and live https://en.wikipedia.org/wiki/Forgetting_curve: his formula first, then "Many equations have since been proposed ... exponential curve" with ref [15] Wozniak et al. 1995.
- CONFIRMED | [wiki-wozniak] citing a 1995 paper by | https://en.wikipedia.org/wiki/Forgetting_curve | live /wiki/ page ref: "Woźniak, Piotr A.; Gorzelańczyk, Edward J.; Murakowski, Janusz A. (1995). \"Two components of long-term memory\"".
- CONFIRMED | [wozniak] which offers it as the authors / quote: terminology and interpretation | https://doi.org/10.55782/ane-1995-1090 | kept PDF text: "The following terminology and interpretation is proposed by the authors ... Assuming the negatively exponential decrease of retrievability ... the following formula"; no mention of Ebbinghaus (0 hits; "memory" 40). Online NOT FOUND: the DOI lands on an abstract page without the PDF text.
- CONFIRMED | [wiki-2006] a revision of that article in November 2006 said that Ebbinghaus / quote: discovered the exponential nature of forgetting | https://en.wikipedia.org/wiki/Forgetting_curve?oldid=88011734 | live https://en.wikipedia.org/wiki/Forgetting_curve?oldid=88011734: "In 1885, Hermann Ebbinghaus discovered the exponential nature of forgetting."
- CONFIRMED | [wiki-2006-date] in November 2006 | https://en.wikipedia.org/wiki/Forgetting_curve?oldid=88011734 | live oldid page: "Revision as of 16:55, 15 November 2006 by Radagast83".
- CONFIRMED | [exp-fit] the best curve of that form has everything gone by nine hours | research/the-forgetting-curve/results.json | data; recomputed: S = 61.59, 100e^(-526/S) = 0.0195%.
- CONFIRMED | [wiki-caption] The picture at the head of that article / quote: A representation of the forgetting curve showing retained information halving after each day | https://en.wikipedia.org/wiki/Forgetting_curve | live /wiki/ page: lead image caption exact.
- CONFIRMED | [commons-redrawn] is redrawn from an older picture | https://commons.wikimedia.org/wiki/File:Forgetting_curve_decline.svg | live https://commons.wikimedia.org/wiki/File:Forgetting_curve_decline.svg: Date 29 July 2022; Source File:ForgettingCurve.svg.
- CONFIRMED | [commons-dayton] whose file page gives, as its only source, a smaller drawing on a University of Dayton web page | https://commons.wikimedia.org/wiki/File:ForgettingCurve.svg | live https://commons.wikimedia.org/wiki/File:ForgettingCurve.svg: Source "Made with Inkscape, based on smaller version from [academic.udayton.edu/Legaled/barpass/General/advice03a.htm]"; date 2007 (upload); no other source.
- CONFIRMED | [q-law] and he said so first / quote: no one will desire at once to know what "law" is revealed in them | https://psychclassics.yorku.ca/Ebbinghaus/memory7.htm | memory7.htm Sec. 29.
- CONFIRMED | [don-juan-80] about 80 syllables | https://psychclassics.yorku.ca/Ebbinghaus/memory5.htm | memory5.htm Sec. 21 "each stanza contains 80 syllables".
- CONFIRMED | [q-nine] A stanza of Byron's Don Juan, about 80 syllables, took him / quote: hardly nine repetitions | https://psychclassics.yorku.ca/Ebbinghaus/memory5.htm | memory5.htm Sec. 21.
- CONFIRMED | [reckoned] where he reckoned a row of 80 to 90 nonsense syllables would have taken 70 to 80 | https://psychclassics.yorku.ca/Ebbinghaus/memory5.htm | memory5.htm Sec. 21.
- CONFIRMED | [q-ties] and he put the difference down to / quote: the combined ties of meaning, rhythm, rhyme, and a common language | https://psychclassics.yorku.ca/Ebbinghaus/memory5.htm | memory5.htm Sec. 21.
- CONFIRMED | [cite-green] Christopher D. Green, York University | https://psychclassics.yorku.ca/Ebbinghaus/index.htm | psychclassics index.
- CONFIRMED | [cite-trans] translated by Henry A. Ruger and Clara E. Bussenius (New York: Teachers College, Columbia University, 1913) | https://psychclassics.yorku.ca/Ebbinghaus/index.htm | psychclassics index: "Originally published in New York by Teachers College, Columbia University."
- CONFIRMED | [cite-md] Replication and Analysis of Ebbinghaus | https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0120644 | kept PLOS page; citation_volume 10, issue 7, date Jul 6, 2015, CC BY 4.0.
- CONFIRMED | [tables-213] are taken from its Tables 2, 1 and 3 as printed | research/the-forgetting-curve/make-replication-data.mjs | data; make-replication-data.mjs --check in a scratch copy: "ok the replication tables are the paper's"; data.json table3 equals M&D Table 3 x100 in all 28 cells; the 69 Table 2 lists equal the CSV.
- CONFIRMED | [osf-link] Its raw data are at | https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0120644&type=manuscript | M&D Results: "freely available online ... https://osf.io/6kfrp/".
- OBSERVED | [osf-redirect] because its files are served from a host whose robots.txt admits only Googlebot | research/the-forgetting-curve/sources/replications/osf-download-mwzjr.headers.txt | record: sources/replications/osf-download-mwzjr.headers.txt; recreated: my own request to https://osf.io/download/mwzjr/ got "302 ... Location: https://files.osf.io/v1/resources/syw2p/providers/osfstorage/553e9d7a8c5e4a21991aded4".
- CONFIRMED | [osf-googlebot] a host whose robots.txt admits only Googlebot | research/the-forgetting-curve/sources/robots/files.osf.io.txt | https://files.osf.io/robots.txt read with curl: "User-agent: Googlebot / Allow: / ... User-agent: * / Disallow: /".
- OBSERVED | [left-loadwords] the Loadwords documentation, whose robots.txt turns away Anthropic | research/the-forgetting-curve/sources/robots/loadwords.com.txt | record: sources/robots/loadwords.com.txt names ClaudeBot, anthropic-ai and Claude-Web with "Disallow: /"; not re-fetched (ruling: do not fetch loadwords.com).
- OBSERVED | [left-acc] answered our request for its robots.txt with a challenge page | research/the-forgetting-curve/sources/robots/STATUS.tsv | record: sources/robots/STATUS.tsv "quillbot.com 403 unreadable a Cloudflare challenge page"; not re-fetched (ruling).
- CONFIRMED | [left-acc-redirect] an Austin Community College study guide whose address now leads to a site | research/the-forgetting-curve/data/retellings.json | curl -D (no redirect followed): courses.lumenlearning.com/austincc-learningframeworks/... -> "301 location: https://quillbot.com/courses/effective-learning-strategies/chapter/chapter-9-memory-and-information-processing/" (lumenlearning robots.txt allows the path).
- CONFIRMED | [retelling-1] Art Kohn: share, not his figures, cites nothing, does not name Ebbinghaus | kept copy under research/the-forgetting-curve/sources/retellings/ | kept copy: 50% in an hour, 70% in 24 h, 90% in a week; "Research on the forgetting curve (Figure 1) shows" with a figure that has no values or source; "ebbinghaus" 0 hits (controls 4 and 5).
- CONFIRMED | [retelling-2] Training Industry Magazine: share, not his figures, cites its own glossary | kept copy under research/the-forgetting-curve/sources/retellings/ | kept copy links "Ebbinghaus Forgetting Curve" to trainingindustry.com/wiki/content-development/forgetting-curve/; 70%/24 h and 90%/week.
- CONFIRMED | [retelling-3] Menemsha Group: share, not his figures, names Art Kohn | kept copy under research/the-forgetting-curve/sources/retellings/ | "cognitive science expert Art Kohn notes".
- CONFIRMED | [retelling-4] TalentLMS: share, not his figures, cites nothing | kept copy under research/the-forgetting-curve/sources/retellings/ | no source link; 50%/hour, 70%/24 h (its "retain only about 25%" after a week is close to his 6-day 25.4, but 70 is not his).
- MINOR | [retelling-5] Psychology Today: share, not his figures, cites Ebbinghaus 1913, which does not contain these figures | kept copy under research/the-forgetting-curve/sources/retellings/ | References list Ebbinghaus (1913) (kept copy). 75%/day and 90%/week are not in it, but Sec. 29 says that after one hour "one half the amount of the original work had to be expended", a basis for the 50%-in-an-hour. Fix the tag at research/the-forgetting-curve/data/retellings.json:53.
- CONFIRMED | [retelling-6] HDI SupportWorld: share, not his figures, cites Wikipedia, which does not contain these figures | kept copy under research/the-forgetting-curve/sources/retellings/ | kept copy links "forgetting curve" to en.wikipedia.org/wiki/Forgetting_curve; no 70%/90% in the current article or the 2017 revisions.
- MINOR | [retelling-7] BizLibrary: share, not his figures, cites nothing | kept copy under research/the-forgetting-curve/sources/retellings/ | kept copy links "Herman Ebbinghaus" (psychology.jrank.org) and "the forgetting curve." (csustan.edu Writing Program forgetting_curve.pdf, now 404) right after its 70% sentence; HDI with the same pattern is tagged "cites Wikipedia". Fix at research/the-forgetting-curve/data/retellings.json:74.
- CONFIRMED | [retelling-8] eLearning Industry (Pappas): share, not his figures, cites Ebbinghaus 1885, which does not contain this figure | kept copy under research/the-forgetting-curve/sources/retellings/ | References: "Hermann Ebbinghaus ... (1885)"; 90% in a month vs his 78.9.
- CONFIRMED | [retelling-9] eLearning Industry (Denny): share, not his figures, cites nothing | kept copy under research/the-forgetting-curve/sources/retellings/ | no source for the 90%.
- CONFIRMED | [retelling-10] Growth Engineering (science behind forgetfulness): share, not his figures, cites its own glossary | kept copy under research/the-forgetting-curve/sources/retellings/ | links its own /what-is-the-forgetting-curve/ (that URL now 301s to a PNG); 90% in a month.
- CONFIRMED | [retelling-11] Growth Engineering (forgetting curve): his figures, not share, cites Murre and Dros 2015 | kept copy under research/the-forgetting-curve/sources/retellings/ | "Relearning cost Ebbinghaus: 56% ... 66% ... 75%"; links journals.plos.org ... pone.0120644.
- CONFIRMED | [retelling-12] Training Industry (Ilan): share, his figures, cites its own glossary | kept copy under research/the-forgetting-curve/sources/retellings/ | "lose 42 percent" at 20 min (41.8), "retain approximately 25 percent" at six days (25.4); links the TI glossary.
- CONFIRMED | [retelling-13] Edutopia (Terada): share, his figures, cites Murre and Dros 2015 | kept copy under research/the-forgetting-curve/sources/retellings/ | 56/66/75 forgotten = his column IV; links the PLOS article.
- CONFIRMED | [retelling-14] Gale Encyclopedia of Psychology: share, his figures, cites nothing | kept copy under research/the-forgetting-curve/sources/retellings/ | "over half ... within an hour", "almost two thirds" in a day; only an internal "See also".
- CONFIRMED | [retelling-15] OpenStax Psychology 2e 8.3: share, not his figures, cites Ebbinghaus, which does not contain these figures | kept copy under research/the-forgetting-curve/sources/retellings/ | 50% at 20 min (his 41.8) and 70% at 24 h (his 66.3) are not his; 1e (2014) has the same sentence.
- CONFIRMED | [retelling-16] Case Western handout: share, not his figures, cites nothing | kept copy under research/the-forgetting-curve/sources/retellings/ | "loss of 50-80%" by the second day; no source.
- CONFIRMED | [retelling-17] Barnard handout (CSU Stanislaus): share, not his figures, cites nothing | kept copy under research/the-forgetting-curve/sources/retellings/ | "approximately 75%" a day or two after; author John Wittman, CSU Stanislaus (PDF metadata).
- CONFIRMED | [retelling-18] Structural Learning: his figures, not share, cites Murre and Dros 2015 | kept copy under research/the-forgetting-curve/sources/retellings/ | "relearning savings was about 34%"; "(Murre and Dros, 2015, reviewing the original data)".
- MINOR | [add-age-1880] It is one man, at 29, in 1879 and 1880 | https://en.wikipedia.org/wiki/Hermann_Ebbinghaus ; https://psychclassics.yorku.ca/Ebbinghaus/memory3.htm Sec. 16 | Born 24 January 1850, so 30 from 24 Jan 1880; Sec. 16: the 1879-80 tests "extended each over more than a year". M&D (the page's source) also says "29 during his experiments in 1879-1880". Near-miss: "at 29 and 30" or "at 29, from 1879".
- MINOR | [add-chapters] Sources list: "... (New York: Teachers College, Columbia University, 1913), Chapters III and VII" | https://psychclassics.yorku.ca/Ebbinghaus/memory5.htm | The page also quotes Chapter V, Section 21 ("hardly nine repetitions", "the combined ties of meaning, rhythm, rhyme, and a common language"). The citation omits it.
- WRONG | [add-repo] The transcription, the comparison with the 1885 text, the analysis and every kept source are in the repository under research/the-forgetting-curve/. | .git/info/exclude; git ls-files | .git/info/exclude line 7 is "research/the-forgetting-curve/" (file written 2026-11-08 21:45:57 UTC, the second commit dcfe9c9090 was made); "git ls-files research/the-forgetting-curve" lists 0 files and git status does not show the folder. Every other file of the layer shows as untracked (??); this one is invisible to a plain git add -A. If the layer is committed as things stand, the sentence is false and the record's "kept" paths and the verifier's in-repository section F point at nothing.
- MINOR | [add-check-scope] recomputes every figure above from the 680 printed cells: ... and every number in this page's sentences | verify-the-forgetting-curve.mjs (TYPED list, sweep) | Numbers that come from sources (Dros 22, 70 hours, 75 days, 1983, 1924, 1991, 33.4, 2,1, 6.1, 80 syllables, 70 to 80, 2014, 2010, 2017, 2006, 1995) are only matched against the verifier's own TYPED copy of each sentence, not recomputed; and the sweep never reads the plain placard (meta aw-plain), og:image:alt, blockquote attributions or the sources list (planted changes there all pass).
- WRONG | [add-placard-unchecked] Plain placard: "... but the one-hour point comes to 44.8 per cent saved, not the 44.2 he printed." | public/strata/the-forgetting-curve/index.html (meta aw-plain, from src/content/strata/the-forgetting-curve.md plain:), with the step-6 plant run against verify-the-forgetting-curve.mjs | Changing 44.8 to 46.8 (or 44.2 to 43.2) in this sentence and rerunning: "all checks pass ... " exit 0. The same change in a body sentence fails (exit 1). The placard is reader-visible through the "?" button (public/media/plain-button.js reads <meta name="aw-plain">) and comes from frontmatter plain:, which the verifier never reads. The placard's figures are correct today.
- OBSERVED | [add-second-reader] (a second reader checked those against the kept copies) | assay/records/ | not kept: no assay/records/the-forgetting-curve.md, the file the verifier header names (positive control: 528 other records there, e.g. a-host-of-dancing-daffodills.md). This pass is presumably that reading.
- OBSERVED | [add-in-repo] In the repository it also re-reads the transcription against the 1885 OCR and rebuilds the data. | verify-the-forgetting-curve.mjs section F | recreated: scratch mini-repo copy (no AW_DIR) -> F. check-transcription 680/680; check-retellings 18/16/13, 0 failing; check-robots 0 not allowed; make-replication-data, analyze, build-data, fill-page --check all ok; "all checks pass (163 tests, 10 findings, 95 typed fragments, and the repository checks)", exit 0.
- OBSERVED | [add-download] From an empty directory it downloads this page and its data | recreated: mkdir empty && cd empty && curl -sO https://artwaste.land/checks/verify-the-forgetting-curve.mjs && node verify-the-forgetting-curve.mjs | UNVERIFIABLE during the pass (the page was not deployed). Run 2026-11-08 23:52 UTC after landing, against the live site serving c352716e19: the served check is byte-identical to the repository's; it downloaded the page, app.mjs, engine.mjs and data.json into an empty directory and printed "all checks pass (163 tests, 10 findings, 95 typed fragments)"; with --mutate, "mutants: 15 of 15 caught".
- MINOR | [add-1880-value] Murre and Dros's comparison table, which cites his handwritten report of 1880, gives 0.442 for this interval too; that manuscript was published in 1983 and we have not read it. | https://osf.io/download/mwzjr/ (the authors' workbook, served from files.osf.io, which robots.txt closes to all but Googlebot; the pass's copy was deleted, so this line cannot be re-checked from here) | Literally true. But the authors' workbook block headed "Ebbinghaus 1880, p. 56 (based on minutes, not repetitions)" gives for 1.05 h M = 0.556 (savings 0.444, N = 12), and 0.583, 0.356, 0.328, 0.282, 0.251, 0.213 for the other points: Table 3's Ebbinghaus column (0.582, 0.442, 0.358, 0.337, 0.278, 0.254, 0.211) is the 1885 summary despite citing 1880. The page's hint that 44.2 might go back to 1880 is undercut; worth a human glance, but the evidence sits only on a host the ruling rules out.
- CONFIRMED | [add-headlines] Headline figures in dek, seoTitle, plain placard, JSON-LD, og/twitter, body, OG image and verifier output | index.html, the-forgetting-curve.md, public/og/the-forgetting-curve.png | 58.2 / 21.1 / 163 / printed 44.2 -> 44.8 agree everywhere; verifier prints "163 tests, 10 findings" and the list has 10 items; OG image text "his one-hour point comes to 44.8, not 44.2".
- CONFIRMED | [add-og-alt] og:image:alt "Ebbinghaus's 163 tests drawn as dots in seven columns from 19 minutes to 31 days, his printed average beside the average his own rows give, under the title Quicker the Second Time" | public/og/the-forgetting-curve.png | image viewed; matches.
- CONFIRMED | [add-hour-avg] Its sixteen printed Q values average 44.8. Recompute each from its own row's times and they still average 44.8. | research/the-forgetting-curve/data/ebbinghaus-1885-forgetting-tests.csv + https://psychclassics.yorku.ca/Ebbinghaus/table15.jpg | by hand: 716.5 / 16 = 44.78; recomputed rows 44.80 (row 16: 59500/1184 = 50.25).
- CONFIRMED | [add-sixday] Q is 29.3. The book prints 21.0, and the evening average and the six-day average were taken with it. Recomputed, the six-day point is 25.7, not 25.4. | research/the-forgetting-curve/data/ebbinghaus-1885-forgetting-tests.csv + https://psychclassics.yorku.ca/Ebbinghaus/table20.jpg | 35400/1208 = 29.30; evening mean 199.1/8 = 24.89 (printed 24.9) and six-day mean 659.7/26 = 25.37 (printed 25.4) only with 21.0; recomputed point 25.697.
- CONFIRMED | [add-883] The learning time is printed 863, but the row's difference, its Q and the column's average all need 883. | research/the-forgetting-curve/data/ebbinghaus-1885-forgetting-tests.csv + https://psychclassics.yorku.ca/Ebbinghaus/table17.jpg + research/the-forgetting-curve/sources/ebbinghaus-1885-de/berdasgedcht00ebbi_djvu.txt | 863 - 761 = 102, not 122; 12200/(883-64) = 14.90 (printed 14.9) vs 15.27 with 863; L column mean 850.1 with 863, 852.6 with 883 (printed 853).
- CONFIRMED | [add-finds] The ten-item list of figures that do not add up (printed 50.0 / 50.253; 863 / 883; 31.6 / 30.749; 26.5 / 26.658; 21.0 / 29.305; 16.4 / 16.776; 31.0 / 30.946; 44.2 / 44.78; D 260 / 255.9; 44.2 / 44.78) | research/the-forgetting-curve/data/ebbinghaus-1885-forgetting-tests.csv + https://psychclassics.yorku.ca/Ebbinghaus/table15.jpg to table22.jpg | every value recomputed independently; my own pass over all block means (L, WL, WLk, D, Q) finds exactly these two block-level misses (63 min A Q; 6 days A D).
- CONFIRMED | [add-curves] Corrected, the seven points read 58.2, 44.8, 35.8, 33.7, 27.7, 25.7, 21.1 against the printed 58.2, 44.2, 35.8, 33.7, 27.8, 25.4, 21.1. Nothing moves by more than six tenths | research/the-forgetting-curve/data/ebbinghaus-1885-forgetting-tests.csv | recomputed: 58.208, 44.799, 35.824, 33.674, 27.736, 25.697, 21.069; max move 0.599.
- CONFIRMED | [add-fits] k and c come out 1.82 and 1.21; to the corrected points, 1.84 and 1.22 ... 46.773 is printed 46.7 ... S = 62 ... 0.02 per cent ... 217 times ... 40 per cent ... 16 | research/the-forgetting-curve/results.json, recomputed by the pass with its own least-squares search (Python) | printed fit k 1.8179 c 1.2145; corrected k 1.8385 c 1.2204; b(64) 46.7732; S 61.589, 100e^(-526/S) 0.0195, SSE ratio 217.2; free height a 39.97, S 48250: 39.95 and 15.85. (M&D also report 1.8 and 1.21.)
- CONFIRMED | [add-sixrow] 37 of those 38 rows give their printed Q with it | research/the-forgetting-curve/data/ebbinghaus-1885-forgetting-tests.csv | recomputed with 64 s (and L 883 for 1day B6): 37 of 38; 6days B3 is the exception.
- CONFIRMED | [add-dros-time] from the times of each list, Dros saved 42.1 per cent after 20 minutes and 8.7 after 31 days | research/the-forgetting-curve/sources/replications/murre-dros-2015.xml (Table 2) | mean of the per-list savings: 0.4214 and 0.0870. Note only: M&D's own "Average" row prints 0.421 and 0.090 (ratio of the mean times), so a reader comparing at 31 days sees 9.0.
- CONFIRMED | [add-night] In three of the four curves, savings after a day is as high as after nine hours or higher ... Mack 28.5 then 31.6, Seitz 27.0 then 27.0, Dros by time 27.3 then 31.2 ... 35.8 ... 33.7 ... 5.9 | research/the-forgetting-curve/sources/replications/murre-dros-2015.xml (Tables 2 and 3) + https://psychclassics.yorku.ca/Ebbinghaus/table23.jpg | Table 3 Mack .285/.316, Seitz .270/.270; Table 2 per-list means 0.2732/0.3124; 33.7 - 27.8 = 5.9.
- OBSERVED | [add-kohn-oldest] The oldest page we found that gives fifty, seventy and ninety per cent for an hour, a day and a week is a training article by Art Kohn of March 2014 | research/the-forgetting-curve/sources/retellings/INDEX.md, wayback-availability-earliest-learningsolutionsmag-com.json | record of the search: INDEX.md (earliest Wayback capture of the original URL 2014-03-21); consistent with the kept older pages (Thalheimer 2010 lists 40/77, 90/month, 50-80/day; Schenck 2012 has 90% in 30 days). Hedged on the page ("could not prove").
- CONFIRMED | [add-wrong-round] The numbers that are his are mostly read the wrong way round. | research/the-forgetting-curve/data/retellings.json + the kept copies | of the 5 pages whose figures are his (Growth Engineering 2, Ilan, Edutopia, Gale, Structural Learning), 3 read them as a share kept or lost (Ilan, Edutopia, Gale).
- CONFIRMED | [add-most-often] The figures most often given for the forgetting curve are not in his table. | research/the-forgetting-curve/data/retellings.json + the kept copies | within the sample: 13 of 18 give figures not in his results; 50/70/90 are not in table23.
- CONFIRMED | [add-relates-serial] relates note: Murre and Dros ... found the first two and last three syllables of each row the easiest to learn and relearn | https://doi.org/10.1371/journal.pone.0120644 (Discussion) | "Here, it seems that the first two and the last three syllables were very easily learned (and relearned), probably because of primacy and recency effects." (source hedges with "it seems"; the note says "found"). Only the first sentence of the note is shown in the nearby blocks.
- CONFIRMED | [add-relates-shannon] relates note: Shannon's 1951 letter guessing there, Ebbinghaus's 1879 relearning here | src/content/strata/you-already-know-the-rest.md (Shannon 1951, Prediction and Entropy of Printed English) | consistent.
- CONFIRMED | [add-src-md] PLOS ONE 10(7): e0120644 (2015), CC BY 4.0: Dros's list times, his repetitions and the comparison of the four curves are taken from its Tables 2, 1 and 3 as printed. | research/the-forgetting-curve/sources/retellings/murre-dros-2015-plosone-e0120644.html (citation_volume 10, issue 7, 6 July 2015; CC BY 4.0) + research/the-forgetting-curve/make-replication-data.mjs | confirmed.
- CONFIRMED | [add-src-de] Ueber das Gedaechtnis ... (Leipzig: Duncker & Humblot, 1885), the Library of Congress copy at the Internet Archive | research/the-forgetting-curve/sources/ebbinghaus-1885-de/archive-metadata.json (archive.org) | publisher Leipzig, Duncker & Humblot; date 1885; contributor The Library of Congress.
- CONFIRMED | [add-src-heller-journal] Zeitschrift fuer Psychologie 199 (1991) | https://doi.org/10.1371/journal.pone.0120644 (ref. 21) | "Zeitschrift fuer Psychologie 199: 3-18" (1991).
- CONFIRMED | [add-dropped] Two more pages that it read are left out, and we keep no copy of either | research/the-forgetting-curve/sources/retellings/ (listing at 22:55 UTC) | no loadwords or quillbot/lumen copy remains in the folder.
- OBSERVED | [add-checkkind] Injected check placard: "Not yet put to that test ... Whether it runs on its own has not been tested yet. Last verified 8 November 2026, when it was written." | scripts/check-placard.mjs output (outside <main>) | site-generated status; this pass ran the verifier from an empty scratch directory (pass) and with --mutate (13 of 13 caught).

## Findings
- [fixed] WRONG, HIGH (add-placard-unchecked): the verifier never read the plain placard or the picture's description, so 44.8 changed to 46.8 in the placard passed. It now holds both to their figures, finds no other figure in either, and in the repository compares the placard with the stratum's plain:. Two new mutants (the placard's figure, the description's 163) are caught: 15 of 15. The pass's own plant (44.2 to 43.2 in the placard) now fails in reader mode.
- [fixed] WRONG, MED (add-repo): research/the-forgetting-curve/ was in .git/info/exclude while the layer was built. The line was removed and the folder committed (f0824e2e53) before anything was published, so the page's sentence holds.
- [fixed] WRONG (ebb-29): "began learning rows" in 1879. Section 16 says preliminary experiments preceded the definite tests, and Section 13 that a metronome came before the watch. The opening now says that in 1879 he "was learning rows of meaningless syllables to the tick of a watch".
- [fixed] MINOR (add-age-1880): "at 29, in 1879 and 1880". Born 24 January 1850 (en.wikipedia.org/wiki/Hermann_Ebbinghaus, kept), he was 30 from late January 1880; the page now says "about 30".
- [fixed] MINOR (retelling-7): BizLibrary was tagged "cites nothing", but its kept copy links "Herman Ebbinghaus" to psychology.jrank.org and "the forgetting curve." to a Cal State Stanislaus Writing Program PDF. The tag now says so.
- [fixed] MINOR (retelling-5): Psychology Today's tag said Ebbinghaus 1913 "does not contain these figures", but Section 29 has half the original work to do again after an hour. The tag now says that, and that nothing near 75 in a day or 90 in a week is there (his complements are 66.3 at a day, 74.6 at six days, 78.9 at 31 days).
- [fixed] MINOR (add-chapters): the sources list now cites Chapters III, V and VII; the page quotes Chapter V, Section 21.
- [fixed] MINOR (heller-ref): the Heller, Mack and Seitz title is now given as Murre and Dros cite it: "Ebbinghaus'schen", "Ersparnis-methode", the subtitle and pages 3-18, marked "as Murre and Dros cite it", since we have not read the article.
- [fixed] MINOR (add-check-scope): the check paragraph said the check recomputes "every number in this page's sentences". It now says which figures it recomputes, that it holds every other number to its sentence and computes the ones the data decide, and that it does not check the numbers in the sources list or under quotations.
- [fixed] MINOR (add-1880-value): the pass's evidence (the workbook's own 1880 block) is on files.osf.io, which robots.txt closes to us, and its copy was deleted, so it was not used. The sentence was reworded from the paper alone: Table 3 cites the 1880 manuscript for its Ebbinghaus column, and all seven of its figures are the 1885 summary's; whether the manuscript prints them too we cannot say.
- [fixed] record nits: twelve-45 now names table21.jpg, where the 31-day heading is; pe-one's excerpt is now the seven rows of P.E.m values; page-image-readings.txt names table15 and table19 for the m rows; the tally's excerpt quotes the second run.