PAIRING · Part II · The Campfire Contract
The numbers and sources
Chapter 10: The Factor That Fell
The narrated edition tells this chapter with a handful of figures. Here is everything behind it: each figure with its study, its research document and its row in the log, the case for and the case against, where the sources disagree, and what is still open.
- 58 open
- 24 to check first
Numbers to carry
The figures the chapter is built on, each with its research document and its row in the log.
- Woolley's c explained 43 per cent of variance across tasks and correlated at −0.41 with the variance in speaking turns, in 192 groups (Woolley et al. 2010, Science, per OR-01). On the second look, 22 to 35 per cent extracted, an eigenvalue of 1.05, and r = 0.26 (CI 0.10 to 0.40) with criterion tasks across 857 groups (Credé and Howardson 2017, Journal of Applied Psychology, per OR-01). Rows C10-03, C10-04, C10-06, C10-14, C10-15, C10-16.
- Men said almost 70 per cent more words than women in mixed four-person audio teams, 1,368 people, 342 teams (Hardt, Mayer and Rincke 2024, Management Science, per OR-01). Row C10-21.
- Tournament entry: 73 against 35 per cent in the original 80 people (Niederle and Vesterlund 2007, QJE), 52 against 34 across 53 papers (Dariel et al. 2017), and a 13-point unadjusted gap across 110 studies and 409 effect sizes (Markowsky and Beblo 2022), all per OR-03. Rows C10-22, C10-24, C10-25.
- Opt-in: 72.5 per cent of men and 46.7 per cent of women applied. Opt-out: 76.4 and 75.3. About 7,000 job seekers (Flory, Leibbrandt and List 2015, Management Science, per OR-03). Row C10-38.
- Testosterone and competition: 1,000 men, 11 mg intranasal, nine hypotheses, no effect (Bird et al. 2024, PNAS). Testosterone and risk, 52 studies, 94 measurements, CI −0.043 to 0.039 (Dekkers et al. 2022, Neuroscience & Biobehavioral Reviews), both per OR-03. Rows C10-31, C10-33.
Every figure
One row for each figure or finding in the data edition's chapter, whether or not the narrated edition kept it. Layer says whether it belongs to physiology, behaviour or culture. Status is how the finding stands in the research: robust, contested, open or fallen. It is not yet the log's verdict.
| Figure | What it measures | Study (authors, year, venue, N) | Document | Row | Layer | Status |
|---|---|---|---|---|---|---|
| University room, blinds down, three strangers, four kinds of task | The opening scene: the Woolley task battery, visual puzzles, brainstorming, moral judgement, resource negotiation, drawn from the McGrath task circumplex. Place of the experiments not in any document | Woolley, Chabris, Pentland, Hashmi and Malone 2010, Science 330 | OR-01 | C10-01 | culture | open, place not named |
| 30 to 50 per cent | Share of the variance in a battery of mental tests accounted for by general intelligence g, one of the most replicated findings in psychology | No primary named, attributed to the OR-01 summary. The narrated edition gives "one of the most replicated findings" without the percentage | OR-01 | C10-02 | behaviour | open, no primary |
| 40 groups of three, 152 groups of three to five, 192 in all | Sample of the two Woolley studies | Woolley et al. 2010, Science | OR-01 | C10-03 | behaviour | robust |
| 43 per cent | Share of the variance in group performance across tasks explained by a single factor c | Woolley et al. 2010, Science, 192 groups | OR-01 | C10-04 | behaviour | contested, fell on the second look |
| Only moderate | How well the average and the maximum individual intelligence at a table predicted c | Woolley et al. 2010 | OR-01, PS-05 | C10-05 | behaviour | robust |
| Eyes-test average, r = −0.41 with variance in speaking turns, proportion of women mediated by the eyes test | The three correlates of c | Woolley et al. 2010. The −0.41 and the mediation statement to be checked against the supplementary material | OR-01, PS-05 | C10-06 | behaviour | robust as behaviour, the mediation contested |
| Co-authors Chabris, Pentland, Hashmi and Thomas W. Malone | Authorship of the Science paper. First names of the first three not in OR-01 | Woolley et al. 2010 | OR-01 | C10-07 | culture | open, first names |
| A strip of photograph, a pair of eyes, a choice of words | The Reading the Mind in the Eyes Test, developed by Baron-Cohen and colleagues | Baron-Cohen et al. 2001, Journal of Child Psychology and Psychiatry, not named in OR-01 | OR-01 | C10-08 | behaviour | open, primary to confirm |
| 42 studies, 4,290 adults, g = 0.177 or d = 0.47, "32 per cent of men above the average woman" | Female advantage on the eyes test. OR-01 gives both effect sizes for the same analysis, chapter 4 found only 0.177 in the PS-02 PDF, and the 32 per cent is derived from the 0.47. The data edition carries 0.177 and withholds the percentage. The narrated edition says "a great many men sit above the average woman" | Kirkland et al. 2013, North American Journal of Psychology 15(1) | OR-01, PS-02 via chapter 4 | C10-09 | behaviour | open, effect size discrepancy |
| 312,739 people, 57 countries, four datasets | On-average female advantage on the eyes test across the lifespan and almost all languages tested | Greenberg et al. 2022, PNAS | OR-01 | C10-10 | behaviour | robust |
| Eyes-test scores predicted c equally well in online text-based groups | Whether the test measures face-reading or something more general about reading people | Engel, Woolley et al. 2014, PLOS One, author list to confirm | OR-01 | C10-11 | behaviour | robust, one study |
| 22 studies, 1,356 groups, 5,279 individuals, average variance extracted 43.6 per cent | Meta-analysis from Woolley's research cluster finding the structure of c intact, the proponents' reply | Riedl, Kim, Gupta, Malone and Woolley 2021, PNAS 118(21), year and venue not in OR-01's text | OR-01 | C10-12 | behaviour | contested, year and venue open |
| Six published samples re-examined, 2017 | The second look at the collective intelligence construct. OR-01 dates it 2017, the chapter's plan 2019 | Credé and Howardson 2017, Journal of Applied Psychology 102(10) | OR-01 | C10-13 | behaviour | robust, date open |
| 22 to 35 per cent against a 50 per cent floor | Average variance extracted by a general factor, Fornell and Larcker formula | Credé and Howardson 2017 | OR-01 | C10-14 | behaviour | robust |
| Eigenvalue 1.05, 17 per cent of variance, threshold about 1.38 | Principal axis factoring for a single dominant factor | Credé and Howardson 2017. The 1.38 threshold to confirm, parallel analysis | OR-01 | C10-15 | behaviour | robust, threshold open |
| r = 0.26, CI 0.10 to 0.40, 857 groups, eight samples | How well c predicted a separate criterion task | Credé and Howardson 2017, cited by OR-01 to an ACU research-bank document rather than the paper | OR-01 | C10-16 | behaviour | robust, source open |
| Low-effort responding and nested data | The two artefact explanations for the covariance between tasks, and the caution against abandoning individual selection | Credé and Howardson 2017 | OR-01 | C10-17 | behaviour | robust |
| "Generally accepted" | Standing of the behavioural finding that social sensitivity and equal turn-taking predict group performance, as against c itself | OR-01's own synthesis, no primary for the consensus claim | OR-01 | C10-18 | behaviour | open, no primary |
| About 550 business students, 45 teams of twelve, a real company for a term, assigned sex ratios, profit rising to 50 per cent women then flattening, no mediator among six measured | Field evidence that balanced teams outsell and outearn male-dominated ones | Hoogendoorn, Oosterbeek and van Praag 2013, Management Science 59(7). First names and place not in OR-01 | OR-01 | C10-19 | behaviour | robust, no mechanism |
| Three-women teams outperformed by all other combinations, bottom decile 0.47 SD below the mean, top decile 0.09 below, MBA optimum two men and one woman, less aggressive pricing, less R&D, more social sustainability | Business simulation in teams of three, undergraduates and MBA students | Apesteguia, Azmat and Iriberri 2012, Management Science 58(1). First names not in documents | OR-01, PS-05 | C10-20 | behaviour | robust |
| 1,368 people, 342 four-person teams, audio-only, all-male teams more topic words and better performance, men almost 70 per cent more words than women in mixed teams, high-skilled women speak less than low-skilled men, skill proxied by GPA | Who does the talking in mixed teams without faces | Hardt, Mayer and Rincke 2024, Management Science. First names not in document | OR-01 | C10-21 | behaviour | robust |
| 40 men and 40 women, sets of four two-digit numbers, three minutes, no performance difference, 73 per cent of men and 35 per cent of women chose the tournament | The original tournament-entry experiment | Niederle and Vesterlund 2007, QJE 122(3). Affiliations not in documents | OR-03, PS-05 | C10-22 | behaviour | robust |
| Low-ability men entered too often and lost, high-ability women entered too rarely, gap driven by male overconfidence and a residual preference | The mechanism Niederle and Vesterlund gave | Niederle and Vesterlund 2007 | PS-05, OR-03 | C10-23 | behaviour | contested, the residual preference challenged by Gillen |
| 53 papers, 52 per cent of men and 34 per cent of women entering | Pooled tournament entry | Dariel et al. 2017, venue not in OR-03, cited there to a Kotsadam working paper | OR-03 | C10-24 | behaviour | robust, venue open |
| 110 studies, 409 effect sizes, 13 percentage points unadjusted, largest for maths tasks with students, smaller for verbal tasks, other ages, team competition | Meta-analysis of the competition-entry gap and its moderators | Markowsky and Beblo 2022, Journal of Economic Behavior & Organization 198. First names not in document | OR-03 | C10-25 | behaviour | robust |
| 1,480 Dutch secondary students aged 15, four tracks, up to 23 per cent of the sex gap in track choice | Laboratory competitiveness carried into school choice, controlling for ability, risk and confidence | Buser, Niederle and Oosterbeek 2014, QJE 129(3) | OR-03, PS-05 | C10-26 | behaviour | robust |
| After a loss women reduce effort regardless of prize, men only when the prize is large, about 11 per cent of variation in women's effort and 4 per cent in men's, about half the sex gap in repeated competition | How the sexes carry a defeat into the next round | Buser 2016, Management Science 62(12), title and venue to confirm, cited in PS-05 to an Oxford discussion paper | PS-05 | C10-27 | behaviour | robust, venue open |
| 80 to 90 per cent overlap, a random woman more competitive or more risk-taking than a random man 40 to 45 times in a hundred | Overlap of male and female distributions of economic preferences | Nelson 2015 and 2016, venues not in OR-03, cited to a Texas Law Review article | OR-03 | C10-28 | behaviour | robust, venue open |
| Measurement error 30 to 40 per cent of the variance in elicited risk and confidence, Caltech Cohort Study with duplicate elicitations as instruments, sex gap in entry entirely explained by risk attitudes and overconfidence | The measurement-error account of the entry gap | Gillen, Snowberg and Yariv 2019, Econometrica 87(4) | OR-03 | C10-29 | behaviour | contested |
| Null result | Pure replication of Niederle and Vesterlund | Price 2020, Economics Bulletin 40(2), first name not given | OR-03 | C10-30 | behaviour | open |
| 1,000 men, 11 mg intranasal testosterone or placebo, nine primary hypotheses, 80 per cent power for d = 0.23, no treatment effect on any outcome | Pre-registered trial of testosterone on competition, risk and social preference | Bird et al. 2024 or 2025, PNAS. OR-03 works cited item 1 carries a 2025 DOI | OR-03 | C10-31 | physiology | robust, year open |
| 200 postmenopausal women, testosterone or placebo, no effect on risk aversion, altruism or competitiveness | Earlier dosing trial in women | Zethraeus et al. 2009, PNAS 106(16) | OR-03 | C10-32 | physiology | robust |
| 52 studies, 94 measurements, CI −0.0431 to 0.0389, point estimate stripped | Testosterone and risk-taking correlation, near zero | Dekkers et al. 2022, Neuroscience & Biobehavioral Reviews. OR-03 works cited item 2 carries a 2026 DOI | OR-03 | C10-33 | physiology | robust, year open |
| 30 papers, 33 studies, 49 effect sizes, N = 8,538, testosterone × cortisol interaction r = −0.061, p = 0.026, significant but negligible | Meta-analysis of the dual-hormone hypothesis. The narrated edition calls it fallen, said once, so chapter 11 can refer to it as fallen | Dekkers, van Rentergem et al. 2019, Hormones and Behavior, per OR-02. PS-05 attributes the same counts to Geniole et al. 2016, Psychoneuroendocrinology | OR-02, PS-05, chapter 4 rows | C10-34 | physiology | fallen, attribution open |
| Maasai 50 per cent of men and 26 per cent of women compete, Khasi 54 per cent of women and 39 per cent of men, ball tossing | The sign reversal of the competition gap between a patriarchal and a matrilineal society | Gneezy, Leonard and List 2009, Econometrica 77(5) | OR-03, PS-05 | C10-35 | culture | robust |
| Mosuo men more generous than Mosuo women in a dictator game, Mosuo girls more risk-taking than boys at first and more risk-averse by age eleven after schooling with the Han majority | Matrilineal Mosuo and patriarchal Yi of Yunnan, and the fading of the reversal | Gong and Yang 2015, Experimental Economics, and a children's study to identify | OR-03 | C10-36 | culture | robust, children's study open |
| 1,943 Norwegians, parents from 59 countries, gap smaller for ancestries from more gender-equal countries, from both sides | Second-generation evidence that the gap tracks the parents' culture | Hauge and Kotsadam 2023, Economic Journal, volume to confirm, cited in OR-03 to a 2020 working paper | OR-03 | C10-37 | culture | robust |
| About 7,000 job seekers, opt-in 72.5 per cent of men and 46.7 per cent of women, opt-out 76.4 and 75.3, gap "25 per cent" in OR-03, 25.8 points by arithmetic, "nearly 26" in the data edition | Field experiment on competitive pay as an opt-in or an opt-out default | Flory, Leibbrandt and List 2015, Management Science 61(1). The opt-out figures cited in OR-03 to a PMC paper that may be a different, later paper. First names and place not in document | OR-03 | C10-38 | culture | robust, which paper carries the opt-out figures open |
| A quota guaranteeing equal representation among winners increased female entry, drew in able women, did not lower the performance threshold | Affirmative action in a tournament | Niederle, Segal and Vesterlund 2013, Management Science 59(1) | OR-03 | C10-39 | culture | robust |
| Penalty for initiating salary negotiation "nearly tripled" for women, figures stripped, women less inclined to negotiate with a male evaluator, mediated by anticipated social cost | Four experiments on the social cost of asking | Bowles, Babcock and Lai 2007, Organizational Behavior and Human Decision Processes 103(1) | OR-03 | C10-40 | culture | robust, penalty sizes open |
| Over 4,000 online adults and over 10,000 school-aged youth, self-evaluation gap 13 points, 24 per cent of the mean, performance controlled, persists when private, present from sixth grade, employers hired and paid women less when shown self-evaluations | The self-promotion gap | Exley and Kessler 2022, QJE 137(3) | OR-03 | C10-41 | behaviour | robust |
| A 12-percentage-point gap in ambitious career advice eliminated by informing advisers that women self-promote less | Whether telling the reader of a self-rating about the gap closes it | OR-03 attributes it to Exley and Kessler 2022 but cites an EconStor discussion paper and an RFBerlin paper, likely a separate paper. Primary to identify. Dropped from the narrated edition | OR-03 | C10-42 | culture | open, attribution |
| Women volunteer for non-promotable tasks 50 per cent more often, baseline about 20 percentage points higher, managers of both sexes ask women more, women accept more | Who does the task that helps the group and promotes no one | Babcock, Recalde, Vesterlund and Weingart 2017, American Economic Review 107(3) | OR-03 | C10-43 | behaviour | robust |
| 1993 amendment, one third of village council chief posts by lot, 265 councils in West Bengal and Rajasthan, more drinking water and (West Bengal) roads, less education, women constituents twice as likely to raise concerns | Randomised reservation of council leadership for women. Dropped from the narrated edition, to chapter 14 | Chattopadhyay and Duflo 2004, Econometrica 72(5). First names not in SO-07. SO-01 and SO-07 disagree on roads | SO-07, the SO digest | C10-44 | culture | robust, roads open |
| 40 per cent women on boards, 2003 | Norway's board quota. Year, enforcement dates and scope to confirm | Primary legal source not named | SO-07, OR-01 | C10-45 | culture | open |
| 12.4 per cent decline in Tobin's Q, worse operating performance, younger and less experienced boards | The headline that the quota cut firm value | Ahern and Dittmar 2012, QJE 127(1). First names not in documents | SO-07, OR-01 | C10-46 | culture | fallen |
| Event dates wrong, no correction for cross-correlation of returns, corrected effect indistinguishable from zero, supply of qualified women sufficient | The re-analysis that overturned Ahern and Dittmar | Eckbo, Nygaard and Thorburn 2022, ECGI Working Paper per OR-01, Management Science 68(6) per SO-07. First names not in documents | OR-01, SO-07 | C10-47 | culture | robust, venue open |
| 140 studies, positive with accounting returns, near zero with market performance | Meta-analysis of female board representation. Dropped from the narrated edition | Post and Byron 2015, Academy of Management Journal 58(5) | OR-01, SO-07 | C10-48 | behaviour | robust |
| No evidence that female board representation affects equity risk | Boards and risk. Dropped from the narrated edition | Sila, Gonzalez and Hagendorff 2016, Journal of Corporate Finance 36 | SO-07 | C10-49 | behaviour | robust |
| Higher share of women on central bank boards associated with higher interest rates for the same expected inflation, forward-looking Taylor rule, selection or signalling | The hawks at the central bank. Kept in the narrated edition's In practice only | Masciandaro, Profeta and Romelli 2020, European Journal of Political Economy 63. First names not in document | SO-07 | C10-50 | behaviour | robust |
| Trained minority (yellow) and majority (blue), minority steers when alone, adding 5 to 10 uninformed fish restores majority control | Golden shiners and the uninformed majority | Couzin et al. 2011, Science 334(6062). First name not in document | SO-07 | C10-51 | behaviour | robust |
| Shared decisions less extreme, lower consensus costs than despotic ones | Why a group might evolve to share its decisions. Dropped from the narrated edition | Conradt and Roper 2003, Nature 421(6919) | SO-07 | C10-52 | behaviour | robust |
| 150 studies, men take more risk in 14 of 16 categories, financial risk d about 0.43 | Sex and risk-taking. In the data edition's case against only, dropped from the narrated edition | Byrnes, Miller and Schafer 1999, Psychological Bulletin 125(3). The 0.43 from chapter 4's PDF reading, stripped in PS-05 | PS-05, chapter 4 rows | C10-53 | behaviour | robust |
| 97,678 working-age adults, women's participation fell 25.5 per cent under competition outside STEM, gap absent among STEM adults, authors read it as selection | Field evidence on the competition gap and selection into careers. In the data edition's case against only, dropped from the narrated edition | Saccardo, Pietrasz and Gneezy 2022, NBER WP 30062 or published version. N from chapter 4's PDF reading | PS-05, chapter 4 rows | C10-54 | behaviour | contested, selection or context |
| 10,888 participants, male negotiation advantage vanishes when negotiating on behalf of others | The second jaw of the data edition's case against. Not among this chapter's documents. Dropped from the narrated edition | Mazei et al. 2015, Psychological Bulletin 141(1), to confirm from SY-01 and the paper | concept note §11 citing SY-01 | C10-55 | behaviour | open, not in this chapter's documents |
| Solo women in majority-male groups less talkative and less influential than women in balanced groups | The research on tokens, the narrated edition's first practice | Bear and Woolley 2011, Interdisciplinary Science Reviews 36(2), venue to confirm | OR-01 | C10-56 | behaviour | robust, venue open |
| Chapter 11 treats the dual-hormone hypothesis, chapter 14 owns Chattopadhyay and Duflo and the state, chapter 4 carried Niederle, Gneezy, Byrnes and Saccardo, chapter 3 owns hormones, chapter 5 left the paradox open | Cross-references between chapters | concept note §6 and the drafted chapters 03, 04, 05. To cross-check once chapters 11 and 14 are drafted | none | C10-57 | culture | open |
The case for and the case against
The case the chapter makes
- A group's performance is not the average of its members: the cleverest person and the average cleverness at a table only moderately predicted how the table did (Woolley et al. 2010, OR-01, PS-05, C10-05).
- What predicted it was the reading of faces, the evenness of turns and, mostly through the first, the share of women (Woolley et al. 2010, OR-01, C10-06). The eyes-test advantage is real on average and widely overlapping (Kirkland 2013, Greenberg 2022, OR-01, C10-09, C10-10).
- The factor c did not survive its second look: a third of the variance at most, an eigenvalue under the line, a weak link to the next task, two cheap artefact explanations (Credé and Howardson 2017, OR-01, C10-14 to C10-17). The headline "more women, smarter teams" fell with it.
- The behavioural finding underneath survived and is generally accepted: tables where a few dominate the talk do worse, tables whose people read each other do better (OR-01, C10-18). Neither belongs to a sex: the turn can be written into the rules, the eyes test is a score many men beat the average woman at.
- Field evidence is messier than either headline: balanced teams outperformed male-dominated ones with no mechanism found (Hoogendoorn 2013, OR-01, C10-19), all-women teams lost a game that paid for aggression (Apesteguia 2012, OR-01, PS-05, C10-20), and in audio-only rooms the even turn went missing, a finding about a medium (Hardt 2024, OR-01, C10-21).
- The tournament-entry gap is real, replicated smaller, and its composition is open: preference (Niederle and Vesterlund 2007, Dariel 2017, Markowsky and Beblo 2022, OR-03, C10-22 to C10-25) against risk and confidence measured with error (Gillen, Snowberg and Yariv 2019, OR-03, C10-29) and a null replication (Price 2020, C10-30).
- No hormone moves it: a pre-registered dose of testosterone did nothing (Bird 2024, OR-03, C10-31), nor in women (Zethraeus 2009, C10-32), the testosterone-risk correlation sits on zero (Dekkers 2022, C10-33), and the dual-hormone interaction is negligible, so the hypothesis fell (Dekkers 2019, OR-02, PS-05, C10-34).
- Culture reverses the gap: the Khasi against the Maasai (Gneezy, Leonard and List 2009, OR-03, C10-35), the Mosuo reversal erased by schooling (Gong and Yang, C10-36), and the gap among Norwegians tracking the parents' country (Hauge and Kotsadam 2023, C10-37).
- Change the form and the gap goes: an opt-out default closed the application gap to a point (Flory, Leibbrandt and List 2015, OR-03, C10-38), a quota drew in able women without lowering the bar (Niederle, Segal and Vesterlund 2013, C10-39).
- The rest of the architecture carries the same lesson: the penalty for asking (Bowles, Babcock and Lai 2007, C10-40), the self-rating gap and the employers who used it (Exley and Kessler 2022, C10-41), the non-promotable task everyone expected the woman to do (Babcock et al. 2017, C10-43).
- Scaled up, the same shape: the Norway board headline fell on re-analysis (Ahern and Dittmar 2012, Eckbo, Nygaard and Thorburn 2022, OR-01, SO-07, C10-46, C10-47), and a body's decisions depend on how many in it hold no brief (Couzin 2011, SO-07, C10-51).
- The factor is in the room, not in the sexes, and the room is what a team can change (narrator's synthesis).
The case against
- The entry gap survives every design averaged in: 13 points across 110 studies, 52 against 34 across 53 papers (Markowsky and Beblo 2022, Dariel 2017, OR-03, C10-24, C10-25). The measurement critique explains it and does not remove it (Gillen 2019, C10-29).
- The laboratory measure predicted school track at 23 per cent of the sex gap in 1,480 Dutch fifteen-year-olds, before any employer or form had seen them (Buser, Niederle and Oosterbeek 2014, OR-03, PS-05, C10-26).
- The self-rating gap was there in private and in eleven-year-olds (Exley and Kessler 2022, C10-41). Women and men carry a defeat differently into the next round, and no form fixes the next round (Buser 2016, PS-05, C10-27).
- Men take more risk in 14 of 16 domains, financial risk at about 0.43 of a standard deviation (Byrnes, Miller and Schafer 1999, PS-05, C10-53).
- All-women triads scored below every other combination because of the decisions they took (Apesteguia 2012, C10-20). All-male teams beat mixed and all-female teams on a sex-neutral puzzle (Hardt 2024, C10-21). The balanced-team premium flattened at 50 per cent and has no mechanism (Hoogendoorn 2013, C10-19).
- The vanishing gap among people in science and engineering is, on its authors' reading, selection: the competition-averse women never arrive (Saccardo, Pietrasz and Gneezy 2022, PS-05, C10-54).
- The Khasi are one society, the Mosuo reversal was erased by a decade of school, the Norwegian result runs with the paradox chapter 5 left open, and Flory's opt-out equalised who applied and said nothing about who performed or stayed (C10-35 to C10-38).
- Woolley's three correlates were never refuted as behaviour, and the eyes-test advantage now rests on 312,739 people in 57 countries (Greenberg 2022, C10-10, C10-18). On that reading the women at the table carried something real and the c-factor was only a bad name for it.
- The second jaw: the male negotiation advantage vanishes when people negotiate for others, so even the "real on average" difference is a difference in who may ask for what (Mazei et al. 2015, SY-01 via the concept note, C10-55). Not in this chapter's documents.
Where the sources disagree
The data edition's Contested list:
- Whether c exists at all: Woolley 2010 and Riedl 2021 against Credé and Howardson 2017 (OR-01, C10-04, C10-12 to C10-16).
- Whether the entry gap is a preference or risk plus confidence measured with error: Niederle and Vesterlund 2007 against Gillen, Snowberg and Yariv 2019, with Price 2020's null (OR-03, C10-22, C10-23, C10-29, C10-30).
- The Kirkland effect size, 0.177 or 0.47 (OR-01, C10-09).
- The dual-hormone meta-analysis attribution, Dekkers 2019 or Geniole 2016 (OR-02, PS-05, C10-34).
- Whether Hoogendoorn's premium has a mechanism (OR-01, C10-19).
- Whether Saccardo's vanishing gap is selection or context (PS-05, C10-54).
- Whether the Eckbo correction is the last word on Norway (OR-01, SO-07, C10-47).
- Whether the Mazei advocacy result, not in this chapter's documents, removes the negotiation difference entirely (C10-55).
Report-to-report discrepancies logged in the data edition's working notes:
- Dual-hormone meta-analysis. OR-02 PDF page 2: Dekkers et al. 2019, Hormones and Behavior, 30 papers, 33 studies, 49 effect sizes, N = 8,538, interaction r = −0.061. PS-05 PDF page 2: Geniole et al. 2016, Psychoneuroendocrinology, same 30, 33, 49 and 8,538, effect size stripped in both the text file and the PDF text layer (chapter 4 read r = −0.061, p = 0.026 from the rendered page). PS-05's works cited item 1 links a ResearchGate page carrying the Dekkers title, so PS-05's attribution looks like a mislabel of the same paper. Both editions carry Dekkers 2019, the narrated edition without the name. Chapter 11 should carry the same resolution (C10-34).
- Kirkland 2013: OR-01 gives Hedges' g = 0.177 and Cohen's d = 0.47 for the same meta-analysis. Chapter 4 found only 0.177 in the PS-02 PDF. OR-01's "32 per cent of men above the average woman" derives from 0.47. The data edition carries 0.177 and withholds the percentage, the narrated edition carries neither (C10-09).
- Credé and Howardson: OR-01 dates the paper 2017, the chapter's plan says 2019. OR-01's works cited also lists a separate 2019 meta-analysis, "Clarifying the structure of collective intelligence in teams", author not named. Both editions carry 2017 (C10-13).
- Eckbo, Nygaard and Thorburn: ECGI Working Paper in OR-01, Management Science 68(6), 4112 to 4134, in SO-07 (C10-47).
- Chattopadhyay and Duflo spending: SO-07 says water and roads in West Bengal, the SO digest records that SO-01 disagrees on roads (C10-44).
- Flory's gap: OR-03 calls 72.5 against 46.7 "a massive 25% gender gap", the arithmetic is 25.8 points, both editions say "nearly 26 points" (C10-38).
- Riedl meta-analysis: OR-01's text gives no year or venue, its works cited links a PNAS paper "Quantifying collective intelligence in human groups" (C10-12).
- Figures stripped from the text file and not recoverable from the PDF text layer (pypdf extraction, page rendering unavailable in the data edition's session): the Geniole or Dekkers r in PS-05 (taken from OR-02 and chapter 4), the Bowles, Babcock and Lai penalty sizes (OR-03 p. 5), the Saccardo N (PS-05 p. 5, taken from chapter 4), the Byrnes financial and social d (PS-05 p. 4, taken from chapter 4), Hyde's d thresholds and Nelson's height d (PS-05 p. 3, not used), the Dekkers 2022 point estimate (OR-03 p. 2, CI used instead), Niederle and Vesterlund's N as a symbol (given in words as 80).
Open questions
The data edition's marks for checking, 25, each with its row:
- Place of the Woolley experiments (C10-01).
- Woolley's affiliation (C10-01).
- First names of Woolley's co-authors Chabris, Pentland and Hashmi (C10-07).
- Kirkland 2013 effect size, 0.177 or 0.47 (C10-09).
- OR-01's "32 per cent of men above the average woman", computed from the larger figure (C10-09).
- Riedl meta-analysis year and venue (C10-12).
- Credé and Howardson date, 2017 or 2019 (C10-13).
- Hoogendoorn, Oosterbeek and van Praag first names and place (C10-19).
- Apesteguia, Azmat and Iriberri first names (C10-20).
- Hardt, Mayer and Rincke first names (C10-21).
- Niederle and Vesterlund affiliations (C10-22).
- Dual-hormone meta-analysis attribution, Dekkers 2019 or Geniole 2016 (C10-34).
- Flory, Leibbrandt and List first names and place (C10-38).
- Bowles, Babcock and Lai penalty sizes, stripped in text and PDF (C10-40).
- Chattopadhyay and Duflo first names (C10-44).
- Roads in West Bengal, SO-01 against SO-07 (C10-44).
- Norway quota date and scope (C10-45).
- Ahern and Dittmar first names (C10-46).
- Eckbo, Nygaard and Thorburn first names and venue (C10-47).
- Masciandaro, Profeta and Romelli first names (C10-50).
- Couzin first name (C10-51).
- Mazei via the concept note and SY-01, not among this chapter's documents (C10-55).
- Exley and Kessler career-advice attribution, likely a separate paper (C10-42).
- The 1.38 eigenvalue threshold and whether it is from a parallel analysis (C10-15).
- Whether the 857 groups and the confidence interval are Credé and Howardson's own, since OR-01 cites an ACU research-bank document rather than the paper (C10-16).
Further open points from the verification rows:
- No primary source for g's 30 to 50 per cent (C10-02) or for OR-01's "generally accepted" (C10-18).
- Dariel 2017 venue (C10-24), Buser 2016 title and venue (C10-27), Nelson venues (C10-28), Price first name (C10-30), Bird year, 2024 or 2025 (C10-31), Dekkers 2022 year against a 2026 DOI (C10-33), Gong and Yang children's study (C10-36), Hauge and Kotsadam volume (C10-37), which paper carries Flory's opt-out figures (C10-38), Bear and Woolley venue (C10-56), Engel author list (C10-11), Baron-Cohen primary for the eyes test (C10-08).
- The −0.41 and the mediation statement against Woolley's supplementary material (C10-06).
- Cross-references to chapters 11 and 14, to be checked once those are drafted (C10-57).
- No affiliation or place is given for any researcher in the chapter, because none of the four documents gives one. The style sheet's apposition rule is unmet throughout and needs a pass once the primary papers are in hand.
The data edition's open questions:
- Whether "The room decides" should lose the boardroom entirely to chapter 11 and keep only the fish. The narrated edition keeps one paragraph on the Norway fall and the fish, and hands Chattopadhyay and Duflo to chapter 14. If chapter 11 also tells Norway, cut the paragraph here.
- Whether the Mazei paragraph in the case against should stand without the primary paper or SY-01 in this chapter's library. The narrated edition drops it until the primary is in hand.
- Whether the one-paragraph hormone section is short enough for chapter 3's and chapter 11's ownership. It is one paragraph in both editions, and the narrated edition settles the dual-hormone hypothesis as fallen so that chapter 11 can refer to it.
- The eyes test is filed under behaviour in both editions, against OR-01's placement under physiology, because it is a score on a test with a large overlap.
- Chapter 4's working notes sent Buser, Niederle and Oosterbeek's 23 per cent to chapter 15. It lands here in both editions, in the case against. Chapter 15 should not repeat it.
The documents
- OR-01 Mixed-sex teams and collective intelligence: Woolley, Credé and Howardson, Riedl, Engel, Hoogendoorn, Apesteguia, Hardt, Kirkland, Greenberg, Bear and Woolley, Lau and Murnighan, Adams and Ferreira, Ahern and Dittmar, Eckbo, Post and Byron.
- OR-03 Competition, negotiation and promotion systems: Niederle and Vesterlund, Dariel, Markowsky and Beblo, Buser, Gillen, Price, Bird, Zethraeus, Dekkers 2022, Knight, Gneezy, Gong and Yang, Hauge and Kotsadam, Flory, Niederle Segal and Vesterlund, Bowles, Exley and Kessler, Babcock 2017, Nelson.
- PS-05 Competition, status, risk and hierarchy: the dual-hormone meta-analysis as Geniole 2016, Byrnes, Buser 2016, Saccardo, Woolley and Apesteguia restated, Hyde 2005.
- SO-07 Mixed vs single-sex decision bodies and collective risk: Eisenegger, Couzin, Conradt and Roper, Chattopadhyay and Duflo, Ahern and Dittmar, Matsa and Miller, Eckbo, Sila, Post and Byron, Masciandaro, the Capital Jury Project.
- OR-02 Leadership, hormones and sex (consulted for the dual-hormone attribution only): Dekkers 2019.