---
slug: shapes-of-stories
passed: 2026-10-11
by: claude-vibrant-lamport-79znpp, two independent Claude checkers on oversight/claims-pass.md (source first, every claim the page makes), an adjudicator who re-read each finding at its source and a skeptic who tried to make the page right; every HIGH and every lone WRONG re-read by the instance at the source before fixing; fixes made by a fixer agent and read against the sources by a separate fix-checker
checked: 71
confirmed: 36
wrong: 20
unverifiable: 3
observed: 12
stale: 0
---
# Claims pass: The Shapes of Stories

Reader-side check (live page 200, matching the repository): the page named only repository paths, and the repository is private; https://artwaste.land/checks/research/shapes-of-stories/build_arcs.py was served (200, 11,363 bytes) but linked from nowhere, and verify.py and svd_check.py were 404. In an empty directory, `curl -sSL -O https://artwaste.land/checks/research/shapes-of-stories/build_arcs.py && python3 build_arcs.py` fetched labMT-en-v2 (10,187 words, 3,647 after the lens) and all 29 Gutenberg texts and printed the page's shapes and r values, then exited 1 with FileNotFoundError at its final write (after 4m41s in checker A's run). With the output directory made it exits 0: 28 of 29 books are byte-identical to the shipped arcs.json, and Moby-Dick differs by at most 0.0004 because Gutenberg updated its text on 2026-10-09 (checker B: 29 of 29 at the printed precision). svd_check.py regenerated svd_report.json byte for byte. The repository's verify.py printed "0/0 books reproduce exactly; 0 mismatch(es)." and exited 0 without the gitignored text cache, also with every arc falsified, and gave 28/29, exit 1, with the texts fetched that day. The self-reading probe does not apply, because verify.py never opens index.html. Checker B changed three prose figures and then emptied index.html, and the output did not change, so no prose figure and nothing in svd_report.json was checked.

Of 71 claims, 36 were confirmed, 4 WRONG, 16 MINOR (counted as wrong), 3 unverifiable and 12 observed. Of the 21 merged findings, 19 were fixed. Three were WRONG: F1, the SVD centred across books rather than each arc on its own mean, so mode 1 is now 22% at r = 0.96, not 36% at r = 0.54; F2, the Palm Sunday quotation, now verbatim; F3, Lewis Mitchell was at Adelaide, not Vermont. The page dates these three in a "Corrected 2026-10-11" line. Sixteen were MINOR (F4 to F19), among them the script a stranger could not run, verify.py's vacuous pass, "sixty years later", "computed live", "1,327 novels", the contested thesis date and three of the shape cards. The fourth WRONG line ("a script you can run") is answered by F4, which the director classed MINOR. F20 (the 5.3–6.2 band) and F21 (Vonnegut's three names) were declined as refuted, and no finding is held by a note. After the fixes, verify.py exits 1 when any text is uncached and also checks svd_report.json against the page's verdict; with all 29 texts cached it printed 29/29 and exited 0. arcs.json was rebuilt for the Moby-Dick update, which the page dates in an "Updated 2026-10-11" note.

## Claims
- CONFIRMED | Kurt Vonnegut said every story has a shape you can draw on graph paper | Palm Sunday (1981), archive.org scans palmsundayautobi0000vonn and palmsundayautobi00vonn_0 read via https://openlibrary.org/search/inside.json (both checkers) | The fundamental idea is that stories have shapes which can be drawn on graph paper, and that the shape of a given society's stories is at least as interesting as the shape of its pots or spearheads.; 'Every' is slightly stronger than the source's generic 'stories'. My own re-query of this passage failed (search-engine error), so this rests on both checkers' identical reads.
- CONFIRMED | a society's stories are as worth mapping as its pots and spearheads | Palm Sunday (1981), same scans via https://openlibrary.org/search/inside.json (both checkers) | the shape of a given society's stories is at least as interesting as the shape of its pots or spearheads
- MINOR | Sixty years later a computer read 1,327 novels | https://chronicle.uchicago.edu/070510/vonnegut.shtml; https://chrisblattman.com/?p=5301; https://api.crossref.org/works/10.1140/epjds/s13688-016-0093-1 | "was a graduate student in Anthropology from 1945 to 1947" / "at the University of Iowa in 1965" / issued 2016-11-04; F7. Measured from 1947 it is 69 years; from 1965 it is 51.
- MINOR | A 2016 study read 1,327 novels | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 §3 | We obtain a collection of 1,327 books that are mostly, but not all, fictional stories by using metadata from Project Gutenberg to construct a rough filter.; F9.
- CONFIRMED | found he was mostly right: six shapes cover most of them | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 §3 and §3.1; https://arxiv.org/pdf/1606.07772v3 | We find broad support for the following six emotional arcs ... We observe ... Vonnegut's 'Man in a hole' (mode 2, positive); 'Mostly right' is the page's own gloss; the paper does not grade Vonnegut, but it names his Man in a hole and Cinderella among the six.
- MINOR | each computed live from its own words / The tool below does the same thing, live / arcs computed live | public/strata/shapes-of-stories/index.html:706 at 2e60bcb5dd; research/shapes-of-stories/build_arcs.py:252 at ac96b58d24 | fetch('arcs.json').then(function(r){return r.json();}).then(function(d){ DATA=d; ... / OUT.write_text(json.dumps(payload, separators=(",", ":"))); F8. The arcs are precomputed offline (2026-07-13); the browser only draws them.
- OBSERVED | the real emotional arcs of 29 famous books | research/shapes-of-stories/build_arcs.py, served at https://artwaste.land/checks/research/shapes-of-stories/build_arcs.py; recreated: python3 -I build_arcs.py in an empty dir (checker A run2.log; checker B stranger run) | wrote .../arcs.json (33917 bytes, 29 books)
- MINOR | 29 famous novels | research/shapes-of-stories/build_arcs.py:47-77 at ac96b58d24 | (1513, "Romeo and Juliet", "Shakespeare"), (1524, "Hamlet", "Shakespeare"), (1661, "The Adventures of Sherlock Holmes", "Doyle"), (2591, "Grimms' Fairy Tales", "Grimm"); F10.
- MINOR | In the late 1940s, as a graduate student in anthropology at the University of Chicago, Vonnegut proposed a thesis: stories have shapes, and you can draw them. The department rejected it. | https://chronicle.uchicago.edu/070510/vonnegut.shtml; https://www.chicagomag.com/city-life/march-2012/kurt-vonnegut-in-chicago-some-footnotes/; https://chrisblattman.com/?p=5301 | Chronicle: "Vonnegut left Chicago after his Master's thesis, 'The Fluctuations Between Good and Evil in Simple Tasks,' was rejected" / Chicago mag: "In September 1947, with most of his class work finished, Vonnegut left the U. of C. without completing his thesis. ... Twenty years later ... he resumed work on his thesis, now entitled 'Fluctuations Between Good and Ill Fortune in Simple Tales.' The U. of C. rejected it"; F6 (PARTLY). The date is contested: the University's 2007 obituary supports 1947, while biographer Shields dates the story-shapes thesis to 1965.
- CONFIRMED | a graduate student in anthropology at the University of Chicago | https://chronicle.uchicago.edu/070510/vonnegut.shtml | was a graduate student in Anthropology from 1945 to 1947, after serving in World War II
- WRONG | “My prettiest contribution to the culture” was his master's thesis in anthropology, “which was rejected because it was so simple and looked like too much fun.” | Palm Sunday scans palmsundayautobi00vonn_0 and palmsundayautobi0000vonn via https://openlibrary.org/search/inside.json | WHAT has been my prettiest contribution to my culture? I would say it was a master's thesis ... anthropology which was rejected by the University of Chicago a long time ago. It was rejected because; F2. The source says 'my culture', and the second quotation splices two sentences without an ellipsis.
- CONFIRMED | One must not, he noted, be too playful. | Palm Sunday (1981), same scans via https://openlibrary.org/search/inside.json (both checkers) | looked like too much fun. One must not be too playful. The thesis has vanished; My own re-query failed (search-engine error); this rests on both checkers' identical reads of two scans.
- CONFIRMED | He never quite got over how much fun the idea was. | Palm Sunday page reproduced as Fig. S1 (p. S2) of https://arxiv.org/pdf/1606.07772v3 (both checkers) | I was thrilled to discover that years ago, and I am just as thrilled today.
- CONFIRMED | Kurt Vonnegut, Palm Sunday (1981); A Man Without a Country (2005) | https://openlibrary.org/search.json?title=palm+sunday&author=vonnegut (checker A); EPJ ref 18 | "first_publish_year":1981 ... "title":"Palm Sunday" / Vonnegut K (1981) Palm sunday. RosettaBooks LLC, New York
- CONFIRMED | A vertical one running from ill fortune at the bottom to good fortune at the top (he called it the G–I axis), and a horizontal one from the beginning of the story to its end | https://www.laphamsquarterly.org/node/4306 (A Man Without a Country, 2005) | This is the G–I axis: good fortune–ill fortune. Death and terrible poverty, sickness down here—great prosperity, wonderful health up there ... This is the B–E axis. B for beginning, E for entropy.; Vonnegut's E is 'entropy'; the page does not claim it stands for 'end'.
- CONFIRMED | Here are three shapes he drew himself — these three names are genuinely his | A Man Without a Country, list of illustrations, via https://openlibrary.org/search/inside.json; https://www.laphamsquarterly.org/node/4306 | Funniest Joke in the World 22 Man in Hole 26 Boy Meets Girl 28 Cinderella 30 Kafka 32 Hamlet / "The story is 'Man in Hole' ... Another is called 'Boy Meets Girl'"; He also graphed Kafka's story and Hamlet (see F21, REFUTED).
- CONFIRMED | “Somebody gets into trouble, gets out of it again. It is not accidental that the line ends up higher than where it began. This is encouraging to readers.” | https://www.laphamsquarterly.org/node/4306 | It's: somebody gets into trouble, gets out of it again [ draws line A ]. It is not accidental that the line ends up higher than where it began. This is encouraging to readers.; Verbatim, apart from 'It's:' and the stage direction.
- MINOR | An ordinary person finds something wonderful, loses it, and gets it back forever. | https://www.laphamsquarterly.org/node/4306; Palm Sunday via https://openlibrary.org/search/inside.json | somebody, an ordinary person, on a day like any other day, comes across something perfectly wonderful: 'Oh boy, this is my lucky day!' … 'Shit!' … [ drawing line back up again ]. And gets back up again.; F17. 'Forever' is not in the source.
- MINOR | Incremental gifts lift her step by step; midnight drops her to the floor; then off-the-chart bliss. | Palm Sunday via https://openlibrary.org/search/inside.json; https://arxiv.org/pdf/1606.07772v3 intro | AMWAC: "Her fairy godmother shows up [ draws incremental rise ] ... so she drops down. Does she drop down to the same level? Hell, no. ... she becomes off-scale happy" / Palm Sunday: "the stroke of midnight at the ball. Cinderella is in rags again. All the presents have been repossessed"; F16. The steps and the bliss are confirmed. 'To the floor', drawn below her start (.20 vs .28), contradicts AMWAC.
- MINOR | Vonnegut was struck that this curve is the shape of the New Testament. | Palm Sunday via https://openlibrary.org/search/inside.json; arXiv:1606.07772v3 intro | stroke of midnight looked exactly like the unique creation myth in the Old Testament. And then I saw that / "Vonnegut finds a remarkable similarity between Cinderella and the origin story of Christianity in the Old Testament"; F16. The comparison is to the Christian story as a whole: Old Testament fall, then redemption.
- WRONG | In 2016, five researchers at the University of Vermont did exactly that. | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 abstract; https://arxiv.org/pdf/1606.07772v3 §II.B; https://api.crossref.org/works/10.1140/epjds/s13688-016-0093-1 | 1 University of Vermont, 85 South Prospect St, Burlington, VT 05405, USA. 2 University of Adelaide, Adelaide, SA 5005, Australia.; F3. Lewis Mitchell was at Adelaide.
- CONFIRMED | In 2016 ... They took 1,327 works of English fiction from Project Gutenberg | EPJ PDF abstract; arXiv:1606.07772v3 §II.B; https://api.crossref.org/works/10.1140/epjds/s13688-016-0093-1 | by classifying the emotional arcs for a filtered subset of 1,327 stories from Project Gutenberg's fiction collection / in an attempt to obtain a set of books that represent English works of fiction / issued 2016-11-04; The paper adds 'mostly, but not all, fictional stories'. The phrasing follows the paper's own abstract (B rated MINOR; resolved CONFIRMED).
- CONFIRMED | ran a sliding window across each one, scoring every word for happiness against a lexicon of 10,000+ words | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 Fig. 1 caption; Dodds et al. 2011, https://www.ebi.ac.uk/europepmc/webservices/rest/PMC3233600/fullTextXML | fixed window size set at Nw = 10,000 words / We merged the top 5,000 words from each source, resulting in a composite set of 10,222 unique words.
- UNVERIFIABLE | rated by thousands of people | Dodds et al. 2011, https://www.ebi.ac.uk/europepmc/webservices/rest/PMC3233600/fullTextXML (searched workers, raters, participants, people, individuals, distinct; positive control '50 independent' found) | obtaining 50 independent evaluations per word. We broke the overall assignment into 100 smaller tasks of rating approximately 100 randomly assigned words at a time.; F12 (UNSUPPORTED). No count of distinct raters exists in the source.
- MINOR | scored word-by-word with the same labMT lexicon and the same “lens” that drops neutral words | https://raw.githubusercontent.com/andyreagan/core-stories/master/README.md; https://hedonometer.org/api/v1/words/?format=json&wordlist__title=labMT-en-v2&limit=1 | Applies the labMT stopper: `mask = /score - 5/ >= 1.0` (drops neutral words; 3731 of 10222 retained at default threshold). / 'title': 'labMT-en-v2', 'date': '2020-03-28'; F13. The lens is the same. The lexicon is the revised v2, not the paper's labMT 1.0; the page corrects this at :291.
- CONFIRMED | Each was downloaded from Project Gutenberg, stripped of its front and back matter | research/shapes-of-stories/build_arcs.py:111-124 at ac96b58d24; checker A's reader-side run | GUT_START = ... START OF (?:THE/THIS) PROJECT GUTENBERG ... body = body[:m2.start()]
- OBSERVED | Tom Sawyer and Heart of Darkness live near the floor; Little Women floats at the top | public/strata/shapes-of-stories/arcs.json; recreated: mean of each arc in shipped arcs.json (both checkers) | lowest: Tom Sawyer 5.7125, Heart of Darkness 5.7204 ... highest: Little Women 6.0760
- CONFIRMED | human language carries a measurable positivity bias — one of the same team's other findings ... Across 10 languages and 24 corpora, the words people use most skew happy | https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=DOI:10.1073/pnas.1411678112 (abstract) | Using human evaluation of 100,000 words spread across 24 corpora in 10 languages diverse in origin and culture, we present evidence of a deep imprint of human sociality in language, observing that (i) the words of natural human language possess a universal positivity bias
- OBSERVED | every arc here floats in the 5.3–6.2 band | recreated: python3 -I adjudicator/svd2.py public/strata/shapes-of-stories/arcs.json | arc min 5.3061 max 6.2281; F20 REFUTED: at the one decimal the page states, the range is 5.3–6.2.
- CONFIRMED | three different methods — a matrix decomposition (SVD), hierarchical clustering, and a self-organizing neural map | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 §1 and §3 | matrix decomposition by singular value decomposition (SVD), supervised learning by agglomerative (hierarchical) clustering with Ward's method, and unsupervised learning by a self-organizing map (SOM, a type of neural network)
- CONFIRMED | The paper's exact and carefully hedged title: the emotional arcs of stories are dominated by six basic shapes | https://api.crossref.org/works/10.1140/epjds/s13688-016-0093-1 | ['The emotional arcs of stories are dominated by six basic shapes']
- CONFIRMED | They come as three arcs and their mirror images (Rags to riches, Tragedy “riches to rags”, Man in a hole, Icarus, Cinderella, Oedipus) | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 §3 and §3.1 | • 'Rags to riches' (rise). • 'Tragedy', or 'Riches to rags' (fall). • 'Man in a hole' (fall-rise). • 'Icarus' (rise-fall). • 'Cinderella' (rise-fall-rise). • 'Oedipus' (fall-rise-fall). ... so we need to consider both the modes and their negation
- MINOR | "Man in a hole":"fall, then rise — ends higher than it began" | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 §3; research/shapes-of-stories/build_arcs.py:189 at ac96b58d24 | 'Man in a hole' (fall-rise). / "Man in a hole": _tpl(lambda x: -math.sin(math.pi * x)); F18. The template drawn on the card ends where it began.
- OBSERVED | some books (here, Crime and Punishment and The Wizard of Oz) match nothing cleanly | research/shapes-of-stories/build_arcs.py; recreated: build_arcs.py reader-side run (checker A run1.log) | Crime and Punishment 209022 w win=10000 weak-fit r=+0.15 ... The Wizard of Oz 40000 w win= 4000 weak-fit r=+0.17
- CONFIRMED | Gregor Samsa wakes as an insect and slowly dies ... the final pages turn to the family's relief and the sister blooming into a young woman | https://www.gutenberg.org/cache/epub/5200/pg5200.txt (both checkers) | he found himself transformed in his bed into a horrible vermin ... their daughter was blossoming into a well built and beautiful young lady.; This translation says 'vermin', with insect anatomy.
- OBSERVED | Kafka's novella scores as a rise ... Emma reads as a fall, Hamlet as a rise-fall-rise | research/shapes-of-stories/build_arcs.py; recreated: build_arcs.py reader-side run (checker A run1.log) | The Metamorphosis 22377 w ... Rags to riches r=+0.58 / Hamlet ... Cinderella r=+0.58 / Emma ... Tragedy r=+0.73
- CONFIRMED | an emotional arc “does not give us direct information about the plot.” | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 p.2 | not give us direct information about the plot or the intended meaning of the story
- CONFIRMED | Two independent inputs, both fetched live: the labMT-en-v2 happiness lexicon from the study authors' own hedonometer.org, and the plain-text novels from Project Gutenberg | https://hedonometer.org/about.html (checker A); research/shapes-of-stories/build_arcs.py, recreated: build_arcs.py run | Hedonometer.org is based on the research of Peter Dodds and Chris Danforth and their team in the Computational Story Lab, including visualization by Andy Reagan / labMT: 10187 words, 3647 after lens [4.0,6.0] removed
- CONFIRMED | The paper's core move is a singular-value decomposition of the stacked, mean-centred arcs | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 §3.1; https://arxiv.org/pdf/1606.07772v3 appendix | The modes are from mean-centered emotional arcs, such that the first SVD mode need not extract the average from the labMT scores nor the positivity bias present in language / for each emotional arc we subtract the mean
- WRONG | We ran the identical decomposition on our 29 books (svd_check.py) ... mode 1 — the overall rise-vs-fall axis — reproduces cleanly and is the single biggest way these stories differ | https://raw.githubusercontent.com/andyreagan/core-stories/master/src/PCA_SVD/PCA-SVD-006.py lines 80-87; public/strata/shapes-of-stories/arcs.json, recreated: python3 -I adjudicator/svd2.py arcs.json | # big_matrix_mean0 = big_matrix-big_matrix.mean(axis=0) / # but I should do it individually... / big_matrix_mean0 = data-np.tile(data.mean(axis=1),(200,1)).transpose() / column-centred mode 1 var 0.360 linear +0.54 flat-energy 0.96; row-centred mode 1 var 0.219 linear -0.96; F1. svd_check.py centres across books; the paper centres each arc.
- OBSERVED | mode 1 · 36.0% of variance ... top 3 modes together: 61% ... top 6: 83% | research/shapes-of-stories/svd_check.py and public/strata/shapes-of-stories/svd_report.json at ac96b58d24; recreated: python3 svd_check.py (both checkers; byte-identical to the shipped svd_report.json); adjudicator/svd2.py | top-3 modes explain 61.2% of variance; top-6 83.0% / mode 1: 36.0% var ~ +linear r=+0.54; mode 2 13.6% linear r=0.64; mode 3 11.7% linear r=0.52; These are faithful outputs of svd_check.py, but of the across-book centring, not the paper's (F1). Under the paper's centring: 21.9% / 56.3% / 78.8%.
- CONFIRMED | the tidy “mode 2 = one swing, mode 3 = two swings” ordering of the full study | https://arxiv.org/pdf/1606.07772v3 §III.A | Vonnegut's 'Man in a hole' (mode 2, positive) ... 'Cinderella' (mode 3, positive)
- OBSERVED | does not emerge at 29 books; the swing shapes scatter into weaker, noisier modes | public/strata/shapes-of-stories/arcs.json; recreated: python3 -I adjudicator/svd2.py arcs.json | row-centred: mode 2 single -0.22, mode 3 single -0.43, mode 4 single +0.61; column-centred: modes 2 and 3 nearest linear (0.64, 0.52); Holds under both centrings.
- CONFIRMED | labMT (“language assessment by Mechanical Turk”) ... each rated 1–9 for happiness by many independent raters | Dodds et al. 2011, https://www.ebi.ac.uk/europepmc/webservices/rest/PMC3233600/fullTextXML | We refer to our data set as 'language assessment by Mechanical Turk 1.0', which abbreviates as labMT 1.0 ... rate how a given word made them feel on a nine point integer scale, obtaining 50 independent evaluations per word
- MINOR | 10,187 English words ... (Dodds et al., 2011) | https://hedonometer.org/api/v1/words/?format=json&wordlist__title=labMT-en-v2&limit=1; Dodds 2011, https://www.ebi.ac.uk/europepmc/webservices/rest/PMC3233600/fullTextXML | 'total_count': 10187 ... 'date': '2020-03-28' / we provide our set of 10,222 words; F13. 10,187 is the 2020 v2 count, but it is cited to the 2011 paper (10,222).
- CONFIRMED | We use the authors' revised labMT-en-v2 list from hedonometer.org. Reagan et al. used the earlier labMT | https://hedonometer.org/api/v1/words/?format=json&wordlist__title=labMT-en-v2&limit=1 (wordlist metadata); https://raw.githubusercontent.com/andyreagan/core-stories/master/README.md | 'title': 'labMT-en-v2' / rows: word counts in `labMTsimple/data/LabMT/labMT1.txt` order (10222 words
- CONFIRMED | Following the hedonometer method, we drop every word whose score falls in the neutral band [4.0, 6.0] (Δh = 1) | Dodds et al. 2011, https://www.ebi.ac.uk/europepmc/webservices/rest/PMC3233600/fullTextXML | we exclude words whose average happiness lies within Δh of the neutral score of 5 ... we remove all words lying in a centered band of width 2Δh; The page's band is inclusive at 4.0 and 6.0 while Reagan keeps /h−5/ ≥ 1, a boundary detail only.
- OBSERVED | leaving 3,647 clearly-valenced words (and 10,187 words in the lexicon) | research/shapes-of-stories/build_arcs.py; recreated: python3 -I build_arcs.py in an empty dir (both checkers) | labMT: 10187 words, 3647 after lens [4.0,6.0] removed
- CONFIRMED | Reagan et al. slid a fixed 10,000-word window (their books were 20k–100k words) | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 Fig. 1; https://arxiv.org/pdf/1606.07772v3 §II.B | fixed window size set at Nw = 10,000 words / with total words between 20,000 and 100,000
- OBSERVED | For books over 100k words this is exactly the paper's window ... a 30k-word novella and a 580k-word epic ... the best match (if /r/ ≥ 0.20) is the book's family | research/shapes-of-stories/build_arcs.py:43,131-134,211 at ac96b58d24; public/strata/shapes-of-stories/arcs.json | return min(n, max(WIN_MIN, min(WIN_CAP, round(WIN_FRAC * n)))) / WEAK_FIT = 0.20 / A Christmas Carol words 28850; War and Peace words 583368 win=10000
- CONFIRMED | It performs worse than chance on single sentences; the large window is what rescues it | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 §2.1 | dictionary-based methods for sentiment analysis usually perform worse than random on individual sentences [20, 21], and although this issue can be resolved by using a rolling average; The page drops 'usually'.
- MINOR | across an English, fiction-only ... public-domain corpus | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 §3 | a collection of 1,327 books that are mostly, but not all, fictional stories; F11.
- UNVERIFIABLE | mostly pre-1920s public-domain corpus | not kept: https://arxiv.org/pdf/1606.07772v3 Table S1 gives Gutenberg IDs, downloads and LoC class, but no publication years | the most categorical tragedies we find Lady Susan (946) and Warlord of Kor (17958) ... This World is Taboo (18172); F11. Plausible but unsupported; the named examples include 1960s science fiction.
- MINOR | That is not a universal law of all storytelling, and the authors don't claim it is. | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 §3.3 | Each of these top stories are all readily identifiable, yet again demonstrating the universality of these story types.; F11.
- CONFIRMED | The paper notes that Icarus, Oedipus and double “man in a hole” arcs are the most-downloaded ... the authors call downloads “only a rough proxy for success.” | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 §3.4 and §4 | We find 'Icarus' (-SV 2), 'Oedipus' (-SV 3), and two sequential 'Man in a hole' arcs (SV 4), are the three most successful emotional arcs ... Of course, downloads are only a rough proxy for success
- CONFIRMED | Our corpus is 29 hand-picked books | research/shapes-of-stories/build_arcs.py:45-77 at ac96b58d24 | # --- the curated corpus: famous, recognisable, public-domain, varied shapes --- (29 entries)
- CONFIRMED | Reagan, Mitchell, Kiley, Danforth & Dodds, “The emotional arcs of stories are dominated by six basic shapes,” EPJ Data Science 5:31 (2016). doi:10.1140/epjds/s13688-016-0093-1 · arXiv:1606.07772 | https://api.crossref.org/works/10.1140/epjds/s13688-016-0093-1; https://arxiv.org/pdf/1606.07772v3 | ['The emotional arcs of stories are dominated by six basic shapes'] ['Reagan', 'Mitchell', 'Kiley', 'Danforth', 'Dodds'] 5 31 {'date-parts': [[2016, 11, 4]]}
- MINOR | Dodds, Harris, Kloumann, Bliss & Danforth, “Temporal Patterns of Happiness and Information in a Global Social Network,” PLOS ONE 6(12):e26752 (2011) ... (CC-BY) | https://api.crossref.org/works/10.1371/journal.pone.0026752 | ['Temporal Patterns of Happiness and Information in a Global Social Network: Hedonometrics and Twitter'] ['Dodds', 'Harris', 'Kloumann', 'Bliss', 'Danforth'] 6 12 ... 2011 ... creativecommons.org/licenses/by/4.0/; F14. Only the subtitle is missing.
- CONFIRMED | Dodds et al., “Human language reveals a universal positivity bias,” PNAS 112(8):2389–2394 (2015) | https://api.crossref.org/works/10.1073/pnas.1411678112 | ['Human language reveals a universal positivity bias'] 112 8 2389-2394 {'date-parts': [[2015, 2, 9]]}
- CONFIRMED | labMT-en-v2 scores: hedonometer.org. Texts: Project Gutenberg (public domain in the US). | https://hedonometer.org/words/labMT-en-v2/ (both checkers, 200); https://hedonometer.org/api/v1/words/?format=json&wordlist__title=labMT-en-v2&limit=1 | Hedonometer word list: labMT-en-v2 / 'total_count': 10187
- CONFIRMED | A Man Without a Country (2005), “Here is a lesson in creative writing.” The famous “blackboard” lecture is where the drawn shapes appear. | https://www.laphamsquarterly.org/node/4306; A Man Without a Country scans via https://openlibrary.org/search/inside.json (checker A) | 2005 / New York City At the Blackboard Kurt Vonnegut diagrams the shapes of stories. ... I'll draw it on the blackboard behind me so you can follow more easily / "3 Here is a lesson in creative writing 23"
- CONFIRMED | Only Man in Hole, Boy Meets Girl, and Cinderella are names Vonnegut used himself. The tidy set of eight titles that circulates online (“From Bad to Worse,” “Which Way Is Up?,” and so on) comes from Maya Eilam's infographic interpreting him | A Man Without a Country illustration list via https://openlibrary.org/search/inside.json; https://www.aerogrammestudio.com/2013/03/26/the-shapes-of-stories-by-kurt-vonnegut/ (checker A) | Funniest Joke in the World 22 Man in Hole 26 Boy Meets Girl 28 Cinderella 30 Kafka 32 Hamlet / Designer: Maya Eilam Sources: A Man Without a Country and Palm Sunday; F21 REFUTED. He also graphed Kafka and Hamlet, but as stories, not named shapes.
- UNVERIFIABLE | Maya Eilam's 2011 infographic | https://ds106.us/2012/05/31/what-shape-is-your-story/; WebSearch 'Maya Eilam Shapes of Stories Vonnegut infographic 2011'; Wayback (no snapshot) | href="http://mayaeilam.com/2012/01/01/the-shapes-of-stories-a-kurt-vonnegut-infographic/" ; img src="http://mayaeilam.files.wordpress.com/2012/01/maya_eilam_vonnegut.png"; F15 (UNSUPPORTED). The earliest dated record is Eilam's post of 1 January 2012.
- CONFIRMED | The corpus was filtered to English-language fiction, 20k–100k words, with more than 40 downloads, from roughly 50,000 Gutenberg books — leaving 1,327. | https://arxiv.org/pdf/1606.07772v3 §II.B | We apply rough filters to the collection (roughly 50,000 books) in an attempt to obtain a set of books that represent English works of fiction. We start by selecting for only English books, with total words between 20,000 and 100,000, with more than 40 downloads from the Project Gutenberg website, and with Library of Congress Class corresponding to English fiction; The paper's appendix restates the threshold as 'more than 20 downloads'; the page follows the main text.
- CONFIRMED | 1,327 books in the original study (stat tile) / Across all 1,327 books | https://epjdatascience.springeropen.com/counter/pdf/10.1140/epjds/s13688-016-0093-1 abstract | by classifying the emotional arcs for a filtered subset of 1,327 stories from Project Gutenberg's fiction collection
- MINOR | every number checked (seoDescription) / the numbers on this page are recomputed from raw text by the scripts | research/shapes-of-stories/verify.py:38-49 at ac96b58d24; checker B's prose-figure probe | if got != exp["arc"] ... if shape != exp["shape"] ... if round(bestc, 3) != exp["shapeCorr"] ... if nwords != exp["words"]; F19. svd_report.json and prose figures are not checked.
- MINOR | This layer's check reads the bytes this page serves and asserts against them. | research/shapes-of-stories/verify.py at ac96b58d24; public/strata/shapes-of-stories/arcs.json; recreated: verify.py in adjudicator/replica with falsified arcs.json and no data/texts/ | 0/0 books reproduce exactly; 0 mismatch(es). exit=0; F5 (MED, BROKEN). Literally true, but the check is vacuous without the gitignored text cache.
- OBSERVED | Measured, not assumed: with every file this layer ships emptied, it goes red. | research/shapes-of-stories/verify.py at ac96b58d24; recreated: emptied arcs.json, svd_report.json and index.html in a replica, then python3 verify.py (both checkers) | json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0); exit=1
- OBSERVED | Last verified 13 July 2026, when it was written. | memory/log.d/2026-07-13T0101Z-shapes-of-stories.md:26-27; commit ac96b58d24 | `verify.py` recomputes all 29 from cached text and diffs the shipped JSON (**29/29 exact, 0 mismatch**); Recreated today with fresh texts: 28/29. Gutenberg updated Moby-Dick (#2701) on 2026-10-09, causing a 4th-decimal drift that leaves no printed figure changed. The dated line stands.
- WRONG | Everything above is recomputed from raw text by a script you can run (research/shapes-of-stories/build_arcs.py) / svd_check.py / research/shapes-of-stories/verify.py | recreated: curl https://artwaste.land/checks/research/shapes-of-stories/{build_arcs,verify,svd_check}.py; checker A's run1.log | build_arcs.py 200 11363; verify.py 404; svd_check.py 404 / FileNotFoundError: [Errno 2] No such file or directory: '.../public/strata/shapes-of-stories/arcs.json' exit=1; F4 (BROKEN). The arcs themselves do reproduce from raw text once the output folder exists.
- OBSERVED | put a real, reproducible measurement under it (relates note) | research/shapes-of-stories/build_arcs.py; public/strata/shapes-of-stories/arcs.json at ac96b58d24; recreated: reader-side build_arcs.py vs shipped arcs.json (both checkers) | A: 28 of 29 books identical, Moby-Dick max arc diff 0.0004 (Gutenberg text updated 2026-10-09); B: 29/29 match at printed precision

## Findings
- [fixed] F1 (WRONG): The page called its SVD run "the identical decomposition", but svd_check.py centred each time point across books where Reagan et al. centre each arc on its own mean, so the printed mode 1 (36.0%, r = 0.54) mostly measured how high each book sits; svd_check.py now centres each arc on its own mean, both copies of svd_report.json were regenerated (mode 1 21.9% at r = +0.96, mode 2 18.7%, mode 3 15.8%; top 3 56%, top 6 79%), and the page now says "We ran the same decomposition ... centring each arc on its own mean as the paper's code does", with the verdict that mode 1 "reproduces cleanly (r = 0.96 with a straight line) and is the single biggest way these stories' shapes differ, though only just (22% of the variance, against 19% for mode 2)".
- [fixed] F2 (WRONG): The Palm Sunday blockquote said "the culture" where Vonnegut wrote "my culture" and joined two sentences without marking the cut; it now gives the passage whole: "What has been my prettiest contribution to my culture? I would say it was a master's thesis in anthropology which was rejected by the University of Chicago a long time ago. It was rejected because it was so simple and looked like too much fun. One must not be too playful."
- [fixed] F3 (WRONG): The page put all five authors of the 2016 paper at the University of Vermont, but the paper's author details place Lewis Mitchell at the University of Adelaide; it now says "five researchers, four at the University of Vermont and one at the University of Adelaide".
- [fixed] F4 (MINOR): The page named only paths in a private repository, the served copy of build_arcs.py was linked from nowhere and crashed at its final write when run in an empty directory, and verify.py and svd_check.py were 404; the check section and the footer now link /checks/research/shapes-of-stories/build_arcs.py ("a script you can download and run ... Python 3, standard library only"), the page and the .md say verify.py and svd_check.py are "in the site's repository only", and build_arcs.py now writes arcs.json beside itself when it is not inside the repository tree.
- [fixed] F5 (MINOR): verify.py skipped every book without a cached Gutenberg text, so in a fresh checkout it printed "0/0 books reproduce exactly; 0 mismatch(es)." and exited 0 even with every arc falsified; it now exits 1 with a FAIL line when no book, or any book, is skipped, and takes --fetch to download missing texts without touching arcs.json (the placard's wording is unchanged).
- [fixed] F6 (MINOR): The lede stated as settled that Vonnegut proposed the thesis in the late 1940s as a Chicago graduate student, a date his biographer disputes; it now says "After the war Vonnegut studied anthropology at the University of Chicago, and the master's thesis he eventually submitted there argued that stories have shapes you can draw. The university rejected it.", followed by both datings (the University's 2007 obituary, rejected before he left Chicago in 1947; Charles Shields, 1965, written while he taught at Iowa).
- [fixed] F7 (MINOR): "Sixty years later" fits no dating of the thesis (69 years from 1947, 51 from 1965); the dek now says "Decades later".
- [fixed] F8 (MINOR): The dek, metadata, body and stat tile said the arcs were computed "live", but the page draws a precomputed arcs.json; "live" is gone from all of them, which now say "computed from its own words", "The tool below shows the result of doing the same thing to 29 books you know" and "famous books, arcs computed from their full text".
- [fixed] F9 (MINOR): The dek and metadata called the paper's corpus "1,327 novels", though the paper calls it "mostly, but not all, fictional stories"; the dek, meta description, JSON-LD and the .md dek and seoDescription now say "1,327 books from Project Gutenberg's fiction collection", following the paper's abstract, and the TV narration line now says "more than a thousand books".
- [fixed] F10 (MINOR): og:description and twitter:description called the 29 "famous novels", though two are plays and two are story collections; both now say "29 famous books", and the check section's "plain-text novels" now reads "plain-text books".
- [fixed] F11 (MINOR): The limits note called the paper's corpus "English, fiction-only, mostly pre-1920s" and said the authors don't claim universality; it now reads "an English-language Project Gutenberg corpus of mostly, but not all, fiction" and "That is not shown to be a law of all storytelling, though the authors do write of 'the universality of these story types.'"
- [fixed] F12 (MINOR): "Rated by thousands of people" had no source, since the labMT paper gives 50 ratings per word and no count of raters; the page now says each word was "rated 50 times by people on Amazon's Mechanical Turk".
- [fixed] F13 (MINOR): The method note gave 10,187 words, the 2020 labMT-en-v2 count, under a citation to Dodds et al. 2011, whose labMT 1.0 has 10,222, and the body called v2 "the same labMT lexicon"; the note now gives "10,222 English words, each rated 1–9 for happiness in 50 independent evaluations (Dodds et al., 2011)" and "labMT-en-v2 list (10,187 words)", and the body says "a revised version of the same labMT lexicon".
- [fixed] F14 (MINOR): The Dodds et al. 2011 reference dropped its subtitle; it now reads "Temporal Patterns of Happiness and Information in a Global Social Network: Hedonometrics and Twitter".
- [fixed] F15 (MINOR): Footnote 1 dated Maya Eilam's infographic to 2011, which no source supports; it now says the infographic was "posted January 2012", the date of her own post.
- [fixed] F16 (MINOR): The Cinderella card, labelled "Vonnegut's own", drew the midnight drop below her starting level and said Vonnegut saw "the shape of the New Testament"; the drop is now drawn at .38 against her .28 start, and the text reads "midnight drops her back into rags; then off-the-scale bliss. Vonnegut was struck that this is the Christian story: midnight looked to him like the Old Testament's creation myth, and the final rise like the redemption early Christianity expected."
- [fixed] F17 (MINOR): The Boy Meets Girl card, labelled "Vonnegut's own", added "forever"; it now reads "An ordinary person, on a day like any other, comes across something perfectly wonderful, loses it, and gets back up again."
- [fixed] F18 (MINOR): The six-shape gallery described the paper's Man in a hole as ending "higher than it began", which is Vonnegut's description and not the paper's, on a template that ends where it began; it now says only "fall, then rise".
- [fixed] F19 (MINOR): The seoDescription's "every number checked" and the footer's "the numbers on this page are recomputed from raw text" overstated a check that compared four fields per book; the seoDescription now says "every arc is recomputed from the raw text and checked against the shipped data", the check section says "Every arc above" and the footer "the arcs on this page", and verify.py now also fails if svd_report.json or the SVD figures in the page's verdict do not match.
- [declined] F20 (MINOR): Checker B reported that two arcs rise above the footnote's 5.3–6.2 band (The Wizard of Oz peaks at 6.228, Little Women at 6.211). Declined: refuted, since 6.228 is 6.2 at the one decimal the page states.
- [declined] F21 (MINOR): Checker B reported that Vonnegut drew and titled five graphs in A Man Without a Country, adding Kafka and Hamlet, not three. Declined: refuted, since Kafka and Hamlet are stories Vonnegut graphed, not shape names, and the footnote's point about Eilam's names holds.