============================================================================== The plan of the third study, on the hero’s journey, written before any text was read, with the notes added afterwards. Printed as it was written, except that one sentence is left out; the place is marked in square brackets. ============================================================================== # The hero's journey as a candidate universal: plan fixed before any reading Written 2026-10-12 from 02:26 UTC, before any text had been read with the new form, before any count of the journey's motifs in the catalogue, and before any figure had been worked out from the earlier readings for this purpose. Everything below was decided in advance. Anything settled later is listed at the end under "Settled after the data were in", which is empty when this file is first saved. ## What is being asked [A sentence naming the person who asked for this study, and quoting the request, is left out here.] Part two (part2/PLAN.md, the page "What Stories Share") tested twenty candidates for something all stories share and kept seven. The hero's journey is the best-known candidate of all and was not among them. The claim. Joseph Campbell, The Hero with a Thousand Faces (1949): "A hero ventures forth from the world of common day into a region of supernatural wonder: fabulous forces are there encountered and a decisive victory is won: the hero comes back from this mysterious adventure with the power to bestow boons on his fellow man." He calls separation, initiation and return "the nuclear unit of the monomyth", divides it into seventeen stages and says it is the one story told everywhere. Christopher Vogler turned it into twelve stages for screenwriters (a memo at Disney in 1985, the book The Writer's Journey in 1992). Star Wars (1977) was written with Campbell's book in hand. The difficulty. The stages are loosely worded, and Campbell allows any of them to be missing, doubled, fused or meant figuratively, so a willing reader can find the journey in almost anything. A test therefore has to say in advance what counts as plainly there and what counts only by stretching, has to be run on texts that are not stories, and has to be run on films that are known to have been built on the scheme. ## What was already known when this was written - Part two's results, including how often and in what order the first reader recorded eighteen plot elements (among them leaving home, returning home, a helper, a test, a confrontation, a reward), and that an inner change of the protagonist was recorded in 25% of stories. No figure had been worked out for any combination of those elements. - The motif catalogue's overall counts (part two, test J). The motifs named below were chosen from the catalogue's titles only (part3/motif/titles.tsv); their counts had not been looked at, except that "Tasks of the in-laws" was already known to be one of the two most widespread motifs (37% of societies). - Nothing had been read with the new form. ## Texts 1. The 711 complete texts of part two, cut into the same eight scenes: 280 folk tales (the second sample of 240 and the 40 first-study tales read again), 66 old written stories, 108 films around the world, 99 Hollywood films (TRIPOD), 38 literary short stories, 40 dreams, 40 stretches of diary, 40 encyclopedia lives. 591 stories and 120 records. 2. The 40 scrambled tales of part two, with the same scene orders. 3. Hollywood by era, a new sample. CMU Movie Summary Corpus, films credited to the United States alone, with a release year, summaries of 400 to 1,500 words, no leftover markup, able to be cut in eight, not in TRIPOD, not in part two's sample and not one of the films of 4. Three eras: 1930 to 1976 (before Star Wars), 1977 to 1992 (after Star Wars, up to Vogler's book), 1993 to 2012. In each era 30 films at random from each of three bands of summary length (400 to 649, 650 to 999, 1,000 to 1,500 words), so that the eras do not differ in how long their summaries are: 270 films. numpy default_rng, seed [20261012, 1]. 4. Films in which the journey is known to be, as a check that the form can find it (CMU summaries): (a) nine that were built on the scheme or are commonly taught as examples of it, named by me before any reading: The Wizard of Oz (1939), Star Wars (1977), The Lion King (1994), The Matrix (1999), Harry Potter and the Philosopher's Stone (2001), Spirited Away (2001), Finding Nemo (2003), Kung Fu Panda (2008), Avatar (2009); (b) three that are not adventure stories and that Vogler's book nonetheless reads as journeys or uses as examples: Titanic (1997), Pulp Fiction (1994), The Full Monty (1997). All texts are mixed at random into packs (at most 9 texts and 11,000 words; the two versions of a scrambled tale never in one pack). A reader is not told what kind a text is. The first reader is a fresh copy of a Claude model for every pack, the same model as in part two. A second reader, a different model, reads the same 96 texts it read in part two (12 of each of eight kinds). ## The reading form (part3/read/READER_INSTRUCTIONS_X.md and _Y.md) The form does not name the hero's journey, Campbell or Vogler, and tells the reader to use the definitions given and no scheme it may know. Per text: `known` (0, 1, 2), the protagonist, `sex` (male, female, mixed, unstated), `age` (child, youth, adult, old, unstated), then - `journey`, 0 to 4, decided before anything else: the reader's own overall impression of how far the text, taken whole, is the story of someone who goes out from their own world into another, wins something there through danger and comes back with it. - `leaves_how`: willing, sent, driven, none. `other_world`: none, ordinary, wild, supernatural. `boon_kind`: spouse, wealth, rank, life, knowledge, kin, other, none. - Twenty-seven items, each coded 2 (the text plainly tells it, as described), 1 (it could be taken for this only by stretching the description, reading it figuratively, counting a faint or partial case, or filling in what the text does not say) or 0 (not in the text), with the scene (1 to 8) where it first happens and, when not 0, up to ten words saying what in the text it is. In doubt between two codes the reader takes the lower. The items, defined by what happens and not by where it stands (full wording in the instructions): `ordinary_life`, `call`, `refusal`, `mentor_aid`, `crossing` (leaves the familiar place for one that is unfamiliar or dangerous), `swallowed`, `trials` (two or more, one after another), `approach`, `ordeal` (a single supreme crisis), `union` (with a partner of more than ordinary power or rank who belongs to the other place), `temptation`, `father`, `apotheosis`, `boon`, `refuses_return`, `pursued_return`, `rescued`, `return` (back to the place or the people they set out from), `new_home` (ends settled in a new place or among new people instead of going back), `final_test` (a last crisis after the prize is won), `benefit_to_others`, `two_worlds`, `freedom`; and four items that are no part of the journey, in part two's wording, for comparison with part two's readings: `trick`, `rule_broken`, `punishment`, `revelation`. The items are listed in a random order that is not the order of the scheme, and in two different random orders (forms X and Y); packs take the two forms in turn. ## Definitions "Plain" means coded 2. "Loose" means coded 1 or 2. Every share below is given both ways; the plain reading is the test and the gap between the two is the measure of stretch. - A departure: `crossing`. - The round trip (Campbell's nuclear unit): `crossing`, and `trials` or `ordeal`, and `boon`, and `return`, with the scene of the crossing no later than that of the boon and the boon no later than the return. - There and back, or there to stay: the same with `return` or `new_home` as the last leg. - Campbell's sentence in full: the round trip, and `other_world` is supernatural, and `benefit_to_others`. - Vogler's twelve: `ordinary_life`, `call`, `refusal`, `mentor_aid`, `crossing`, `trials`, `approach`, `ordeal`, `boon`, `pursued_return`, `final_test`, `return`. V is how many a text has. - Campbell's seventeen: `call`, `refusal`, `mentor_aid`, `crossing`, `swallowed`, `trials`, `union`, `temptation`, `father`, `apotheosis`, `boon`, `refuses_return`, `pursued_return`, `rescued`, `return`, `two_worlds`, `freedom`. C is how many a text has. - Most of the journey: V of 9 or more (C of 13 or more). All of it: V of 12 (C of 17). Words, as in part two: nearly always is 95% or more of the folk tales; usually is 80% or more and at least 60% in every region (the seven regions of part two, counted in the second sample). Three more words are fixed here for what falls below: in most (50% or more), in a minority (20% to under 50%), in few (under 20%). A mark of stories is defined as in part two's test C: at least 80% of the folk tales and at least 60% of every kind of story, and a share among the 120 records at least 20 points lower than among the 591 stories, with a 95% bootstrap interval for the difference that excludes zero. Intervals for shares are Wilson intervals; bootstrap and shuffles use 2,000 draws, seed 20261012. The hero's journey is called a universal of stories only if the plain round trip is usual in folk tales and is found in at least 60% of every kind of story. Otherwise the page says how common it is, where, and which of its parts are common. ## Tests - L1. How common is it? The departure, the round trip, there-and-back-or-stay and Campbell's sentence in full, in the 591 stories, in each kind, and by region in the folk tales; with the verdict word. The same for `journey` of 3 or 4. - L2. The stages. Each of the 23 items of the journey, share by kind, with its verdict word; V and C (mean, median, distribution); the share of texts with most and with all of the journey. - L3. Stretch. L1 and L2 under the loose reading, and the gap. - L4. Stories against records. The mark-of-stories rule for the round trip, the departure and every item, plain and loose; and how often each is found in dreams, diaries and lives. - L5. Can the form find it? In the nine films of 4(a): the form counts as able to find the journey if their median V is 9 or more and at least seven of the nine have V of 8 or more. Their round trips and the three films of 4(b) are reported one by one. - L6. Two readers. For every item (plain or not), the round trip and `journey` of 3 or 4: agreement and Cohen's kappa between the two readers on the 96 texts; correlation of V. Kappa of 0.6 or more is called good agreement, 0.4 to 0.6 moderate, under 0.4 poor. - L7. The earlier readings. From part two's elements, recorded by readers who were asked nothing about a journey: "a trip in the earlier readings" is `leaves_home` and `returns_home` both recorded, with leaving no later than returning, and `test` or `confrontation` recorded, and `reward` recorded. (`returns_home` there means back home or settled in a new home.) Its share by kind, and its agreement text by text with there-and-back-or-stay and with the round trip of the new readings. The four comparison items against the same elements in part two. - L8. Order and timing. For texts with at least three of Vogler's twelve plainly present in at least two different scenes: the share whose stages stand in the order of the scheme (Kendall's tau above zero), against the same texts with the scenes dealt out at random among the stages each has (as part two's G2). Parts of this order are forced by the definitions (no one returns before leaving), so it is weak evidence and is reported as such. Mean scene of each stage. Where the ordeal falls: the share in scenes 4 and 5 (Vogler puts it at the midpoint) and in scenes 6 to 8; and the share of texts whose ordeal is in the scene where part two's reader put the peak of tension, or next to it. - L9. Scrambled tales. For the 40 tales read as printed and scrambled: agreement in each item between the two readings; V in each; and the order test of L8 worked out in the order in which the scrambled text was presented, beside the as-printed positions put into that order. If the reader follows the text, the scrambled reading shows no more of the scheme's order than the as-printed positions put into the scrambled order do. - L10. The order of the questions. V and the round trip under form X and under form Y, with a 95% bootstrap interval for the difference. - L11. Hollywood by era. Mean V, the round trip and `journey` of 3 or 4 in the three eras. If the journey is a recipe that Hollywood took up after 1977, they rise from era to era; if it is how stories have always gone, they do not. The test is the difference in mean V between 1993 to 2012 and 1930 to 1976, with a 95% bootstrap interval and a permutation p (eras dealt out at random within each length band). Reported beside it: the same by the corpus's genre labels (adventure, fantasy, science fiction, action and family films against the rest) and the share of such films in each era. - L12. Whose journey? Among stories that follow one human protagonist (part two's fields): the share male and female; the round trip and V for each, with a 95% bootstrap interval for the difference; the same by `age`. How protagonists come to leave (`leaves_how`), where they go (`other_world`) and what they win (`boon_kind`), by kind. - L13. The journey among the marks. Of the stories with the round trip, how many carry all seven marks of part two; of the stories with all seven, how many are round trips. Part two's F13 (what makes a text read as a story) fitted again with one more term, the round trip, and separately with V: the term's size and 95% bootstrap interval. - L14. Memory. The round trip and V among the stories the first reader said it did not know well, and among those it did. - L15. The journey's images in the world motif catalogue (2,564 motifs, 958 societies; no reader involved). Ten images that Campbell names and the catalogue records, chosen from titles alone: 1. the threshold that opens and shuts: i22 Symplegades, i23 snapping door; 2. the belly of the whale: k8a, k8aa, k8c, k8d (Jonah: swallowed by a monster, a bird, an animal, a humanlike being); 3. the road of trials: k27n tasks of the in-laws, and separately every motif of the k27 family; 4. the descent to the dead: h12 Orpheus, k27r to visit the world of the dead, k2a hero marooned in the underworld; 5. the supernatural bride: k24 stolen clothes of a supernatural woman, e9i swan-wife; 6. the father in the sky: j57 son of the Sun, k27n3a the task-giver is the Sun, Moon or Thunder, j56 father tests his sons; 7. the dragon fight: k38f the dragon-slayer; 8. the boon: d4a theft of fire, h52 the search of immortality, h6d stolen immortality, k83 the quest for a wonderful remedy; 9. the magic flight: l72 the obstacle flight, l100 transformation flight, l103 obstacle flight of the Atalanta type; 10. the hero restored to life: k146, k147. For each image (a society has it if it has any of its motifs) and each motif: the share of the 958 societies, of each continent's societies, and of the 81 best-recorded societies; whether it is on all six continents in at least a tenth of the societies of each (part two's J2). Among the best-recorded societies: how many of the ten images a society has, and the share with at least one image from each of the three parts (1 and 2; 3 to 8; 9 and 10). ## What is reported whatever happens Every test above appears on the page or in its tables, including those that fail. The second reader's agreement is given for every item. Readers' own reports of protocol breaches are counted and published. A rating file, once written, is final: none is read again, changed or left out. ## Added before the reading began The four sections below were added on 2026-10-12 between 02:45 and 02:50 UTC, after the pilot and before any text of the sample had been read. Nothing above this line was changed. ### The sample as fixed (src/prep_hj.py, out/sample_table.txt) 1,033 texts, 1,243,365 words: the 711 complete texts of part two (240 folk tales of the second sample, the 40 first-study tales as printed, 66 old written stories, 108 films around the world, 99 Hollywood films of TRIPOD, 38 literary short stories, 40 dreams, 40 stretches of diary, 40 lives), the 40 scrambled tales, 270 Hollywood films by era (30 in each of nine cells; the pools they were drawn from held 698, 296 and 54 films for 1930 to 1976, 397, 318 and 54 for 1977 to 1992, and 915, 853 and 150 for 1993 to 2012, by length band), and the twelve films in which the journey is known to be. 124 packs of 5 to 9 texts for the first reader, form X for odd pack numbers and form Y for even; the second reader's 96 texts in 13 packs, form Y for odd pack numbers and form X for even, so that most texts read twice are read once with each form. ### The form, made final (src/make_form.py, src/collect_hj.py) The two forms are the files read/READER_INSTRUCTIONS_X.md and _Y.md as they stood at 02:45 UTC (sha256 beginning 1399bddc9dc257ed and 7c433573b53168f8). A rating file is checked by rule. Two slips are put right and counted: a scene number given with code 0 is set to 0; a code of 1 or 2 with no usable scene number keeps its code and has no place in time, so it counts in every share and is left out of every check on order and timing. A form letter that differs from the pack's is set to the pack's and counted. Any other fault makes the file invalid. An invalid or missing file is replaced by a fresh reader's reading of that text with the same form; the faulty file is kept aside and counted. A pack whose reader stops early is finished by a fresh reader who is given only the texts not yet rated. No rating is ever changed by hand. ### After the pilot (read/pilot) Seven spare texts that are in no sample (five from part two's pilot: a tale as printed, a scrambled tale, a dream, a diary and a life; and two films, The NeverEnding Story and The Karate Kid) were read by each model with each form, four readers in all. The files were valid and no wording was changed: the pilot forms and the final forms differ only in the folder the answers are written to. What was seen: the four readers agreed closely; code 1 was used as meant (a return "to the king's city, not own home" was coded 1); the dream and the diary got almost nothing; of the two films, both often called journeys, The NeverEnding Story got 7 to 11 of the twelve stages plainly from the four readers and all 12 loosely, and The Karate Kid got 7 plainly from all four and 8 to 10 loosely. All 28 files were valid, with no slip to put right, and no reader reported a breach of the instructions. So the plain reading is strict. The bar of L5 was left as it was written. Whatever L5 shows, every share is given both plainly and loosely and the known-journey films' own figures are shown beside the others. Pilot readings are used in no test. ### Details fixed while writing the analysis (src/analyze_hj.py) - An item with no place in time is left out of the scene-order condition of the round trip. - One more step is reported between the departure and the round trip, for a figure: "sets out and wins" is `crossing`, and `trials` or `ordeal`, and `boon`, with the crossing no later than the boon. - L5: the films' counts are of plainly present stages, as everywhere. - L8: the scenes in which Vogler's diagram puts each stage, as I read it (Act One is the first quarter, Act Two the middle half, Act Three the last quarter; a stage on an act boundary may fall on either side): ordinary world 1; call 1; refusal 1 or 2; mentor 2; crossing 2 or 3; tests 3 or 4; approach 4; ordeal 4 or 5; reward 5 or 6; road back 6 or 7; last test 7 or 8; return 8. For each stage, the share of its occurrences in those scenes. - L10: the interval is a bootstrap over packs (packs drawn whole), because a pack has one reader. - L11: a film counts as an adventure-kind film if any of its genre labels contains Adventure, Fantasy, Action, Science Fiction or Sci-Fi, or is Family Film, Children's/Family or Children's. The labels of the 270 films were looked at to fix this spelling (40%, 47% and 40% of the three eras carry one of the five plain labels); no reading existed. - L12: sex is taken from the new reading (male or female; mixed and unstated left out); "one human protagonist" from part two's fields. - L15 was run at about 02:35 UTC, before the reading (it needs no reader); the best-recorded societies are the 81 with 200 or more motifs, as in part two. - One more test, written here before any reading. L16, do the stages hang together? Among the 591 stories: (a) the share with all four legs of the round trip, in any order, against the share expected if each leg were dealt out independently of the others within each kind of story (each leg shuffled among the texts of a kind, 2,000 times); (b) the spread of V (its standard deviation, and the share with V of 9 or more) against the same shuffles of the twelve stages; (c) the mean correlation between pairs of the twelve stages, worked out within kind. If the journey is one pattern, its stages come together more often than chance; if they are separate common events, they do not. - A disclosure. When the analysis script was tried on made-up ratings it printed one real figure, because L7 uses part two's readings: "a trip in the earlier readings" is found in 52.5% of the 280 folk tales. This was seen after the plan was saved and before any new reading. The definition was not changed. ## Settled after the data were in (empty when first saved) Written 2026-10-12 at 05:30 UTC, after the analysis and after the page's text had been read against the numbers. Nothing above this line was changed; PLAN_as_fixed_before_reading.md is the file as it stood before the reading. The same list is in settled_notes.txt and on the page under Methods. - All 1,129 rating files passed the checking script and are used exactly as written. One file as first written failed it (a stray extra entry among its items); by the plan's rule it was set aside and its text was read again by a fresh copy of the same reader. - Of the 138 readers, 41 reported something about how they had worked; the reports are counted and listed above. No rating was changed, left out or read again because of a report. One reader rewrote one of its own files at once, before opening the next text, and that file is used as it stands. Two readers each reported one code they had not meant and left the file alone, as the instructions require; those codes are used as written. - The analysis was first run on the real ratings at 04:39 UTC on 12 October 2026 and run again two minutes later, with the same result to the last digit. No figure had been worked out from the real ratings before that; only the checking script, which reports faults and nothing else, had been run on them. A quick run with fewer shuffles, started beside the first, was deleted unread. Each time a check was added afterwards the whole analysis was run again, and every earlier figure came out the same to the last digit. - Every sentence with a number on this page is made by a rule from the analysis file. The rules were drafted and tried on two sets of made-up ratings before the analysis was run, and a copy of them as they stood then is kept with the project. Where the real numbers showed that a rule said the wrong thing or too much, the rule was changed. The notes that follow list each such change and each figure that was added. - The films built on the scheme. The drafted conclusion was that what the form does not find elsewhere "is not there to be plainly read". Only four of the nine films make the round trip as the plan defines it, so that was withdrawn. The section and its heading, the summary and the paragraph on the bar now say that the form finds the stages, that the round trip as defined is strict even for these films, and that every share of round trips is to be read against four in nine. - Why the round trip is strict. The form records only the first scene in which a thing happens, so a story that tells a return twice, or a prize after the homecoming, fails the rule that the legs come in order. The plan did not foresee that. Its counts and its verdict were left as they were. Added beside them, and marked as added wherever they appear: the round trip without the rule about order and with every loosening (the table under "How far stories get"), and the reason each of the five films is not a round trip. - One of the nine films' texts is damaged: the corpus's summary of Avatar has lost its first half where a citation was stripped out. It was named in the plan and is kept, and the figures are given with and without it. Five of the other eleven named films have stray markup in their summaries that removes none of the plot. The 108 films around the world and the 270 Hollywood films by era were searched for the same marks and none was found. - Hollywood by era. The drafted heading said Hollywood has told more of the journey "since Vogler", and the drafted conclusion that the rise "is there within both kinds" of film. The rise beats chance only weakly, the step comes between the first era and the second, before Vogler's book, and films that are not adventures show no change. The heading, the conclusion and the summary now say so. The intervals for the two kinds of film apart, and the differences by length of summary, were worked out after the results and are marked. - The order of the questions. The drafted sentence was that the order "makes no difference that can be told from chance", and it rested on the count of stages alone. The round trip's interval only just takes in no difference, and the reader's overall impression does differ between the two forms, so the paragraph now gives all three. - Whose journey. The drafted heading, for a gap whose interval reaches zero, was that the round trip is "no more a man's story than a woman's", and the drafted text gave the count of stages for men and for women without the interval the plan asks for. With the intervals in, the count of stages differs clearly across all stories. Because kinds of story differ in whose story they tell, the gaps were then worked out inside each kind and averaged, which the plan does not ask for. That figure is marked as added, and the rule that words the heading was written before it was computed. The summary gained a sentence on it, which also says that for the round trip itself the gap could be chance. - Three things the plan asks for had been left out of the drafted page and were put in: the plan's words and part two's rule applied to the looser count (two parts, the call and the ordeal, then pass as marks of stories); the agreement item by item between the printed and the scrambled readings; and the interval for the difference in the count of stages between men's and women's stories. The median of Campbell's seventeen and the count of stages by age were added for the same reason. - The two readers. A kappa of 0.596 is printed to three decimals, because 0.60 would read as "good" where the plan's cut point makes it "moderate". The two items with poor agreement are now said to be rarities, each found in no more than two of the 96 texts. - Smaller changes that alter no claim: the average order of the stages is described by how many keep their place, not by a list of those that do not; the share of stories whose stages stand exactly in the scheme's order is given beside the share that lean towards it; a sentence that began with a figure was turned round; "who go" became "where they go", "100% of texts" became "every text", and two equal figures became one; the regions whose films have no round trip are counted and not one of them named; and titles and notes are shown with typographic apostrophes and quotation marks. - The plan's opening sketch of the claim has three details written from memory that could not be confirmed from the sources used: that Vogler's memo was written "at Disney in 1985", that Star Wars was "written with Campbell's book in hand", and that stages may be "doubled, fused". The page says only what could be confirmed. - Rules of display that the plan does not mention: a share that is not exactly none or all is never printed as 0% or 100% (it gets one decimal); the reader's own impression is put to the plan's bar, as a check the plan did not list; and the summary's words for how far the stages hang together are the plan's words for chance.