A National Experiment Reveals Where a Growth Mindset Improves Achievement
Nature, 2019 · paper · Psychology
Evaluates a short online growth mindset intervention in a nationally representative sample of US secondary students. It improved grades among lower-achieving students and increased enrolment in advanced mathematics, with effects where peer norms supported the message.
Checked: PDF from nature.com, fetched 5 Oct 2026. Page numbers are those of that PDF.
8 findings
Substance "unstandardized B = −0.07 (95% confidence interval = 0.02, 0.13), s.e. = 0.03, z = −2.76": the interval is printed without minus signs and excludes the negative estimate; Extended Data Table 2 (Model 1a) gives B = -0.074, bounds -0.13 and -0.02. Should read −0.13, −0.02.
- p. 3 Moderation by school achievement level, second paragraph
unstandardized B = −0.07 (95% con-fidence interval = 0.02, 0.13)
- p. 14 Extended Data Table 2, Model 1a, Treatment x School achievement level
-0.074 -0.13 -0.02
- p. 3 Moderation by school achievement level
unstandardized B = −0.07 (95% confidence interval = 0.02, 0.13), s.e. = 0.03, z = −2.76
- p. 14 Extended Data Table 2, Model 1a
-0.074 -0.13 -0.02 2.76 .006
Judge's note: Page 3 image: B and z carry minus signs, the two interval bounds do not, so this is not an extraction artefact. Page 14 image: Model 1a Treatment × School achievement level row reads -0.074, -0.13, -0.02, 2.76, .006. The midpoint of the printed positive interval is +0.075, the mirror of the estimate, so the signs were dropped in the text.
Second check: another model agreed · checked by hand
Report a problem with this finding
Substance Extended Data Table 2, Model 3b, Treatment x High achievement: "-0.132 -0.22 0.04 2.80 .005" - the upper bound is printed as positive 0.04; it should be -0.04.
- p. 14 Extended Data Table 2, Math/Science GPA, Model 3b, Treatment x High achievement
-0.132 -0.22 0.04 2.80 .005
Judge's note: On the page 14 image the upper-bound cell reads 0.04 with no minus sign, while the cells around it (e.g. Model 4b: -0.133, -0.22, -0.04, 2.82, .005) show their minus signs clearly, so this is not an extraction artefact. An interval of -0.22 to 0.04 would include zero, which contradicts z = 2.80, P = .005, and is not symmetric about -0.132. B and z imply an upper bound of -0.040. The Core GPA counterpart (Model 3a: -0.119, -0.20, -0.04) has the same pattern.
Second check: another model agreed
Report a problem with this finding
Reference or label Fig. 2 legend: "the boxes correspond to the middle 75% of the distribution (the interquartile range)"; the interquartile range is the middle 50% (25th to 75th percentile), so the sentence contradicts itself.
- p. 4 Fig. 2 legend, panels c, d
the boxes correspond to the middle 75% of the distribution (the interquartile range)
Judge's note: Crop of the Fig. 2 legend on page 4 confirms the wording. Whatever the boxes were drawn to, 'middle 75%' and 'interquartile range' cannot both be right; the likely slip is 75% for 50% (or for '75th percentile'). The plotted data are unaffected; the label is wrong.
Second check: another model agreed
Report a problem with this finding
Reference or label Extended Data Table 1, Students overall, Math GPA: B = 0.063 with 95% CI printed as "[0.03, 0.01]"; the upper bound is below both the lower bound and the estimate. Under the t statistic the table itself prints (t(46) = 3.556), the upper bound would be about 0.10.
- p. 13 Extended Data Table 1, Students overall, Math GPA row
[0.03, 0.01]
- p. 13 Extended Data Table 1, Students overall, Math GPA
0.063 [0.03, 0.01] 3.556 .001 11539
- p. 13 Extended Data Table 1, Students overall, Math GPA
0.063 [0.03, 0.01]
Judge's note: Read the table on the page 13 image and the enlarged crop of the lower rows. The cell reads [0.03, 0.01]. The standard error implied by B and t is 0.0177, which with the t(46) critical value of 2.013 gives an interval of about [0.027, 0.099], that is [0.03, 0.10]. The same calculation reproduces the neighbouring Science GPA row (0.072, t = 4.811, printed [0.04, 0.10]), so the method is sound and the Math GPA upper bound is a misprint. Nothing in the footnote or text explains it.
Second check: another model agreed
Report a problem with this finding
Reference or label Extended Data Table 2: the block containing Model 5b is headed "Core GPA Outcome: Treatment × Self-reported mindset norm", identical to the header above Model 5a; by the table's a/b convention it should read Math/Science GPA Outcome.
- p. 14 Extended Data Table 2, section header above Model 5b
Core GPA Outcome: Treatment × Self-reported mindset norm
Judge's note: Page 14 image: the last two blocks carry the same header 'Core GPA Outcome: Treatment × Self-reported mindset norm', the first over Model 5a (0.055, -0.11, 0.22, 0.64, .524) and the second over Model 5b (0.087, -0.11, 0.28, 0.87, .384). In every other block of the table the 'a' models sit under a Core GPA header and the 'b' models under a Math/Science GPA header. Neither the footnote nor the Methods (which only says self-reported mindset norms did not moderate) describes a second core-GPA specification, so the second header is a copied label.
Second check: another model agreed
Report a problem with this finding
Reference or label Extended Data Table 2, Models 5a and 5b (self-reported mindset norm): the italic note reads "A positive coefficient means larger effects for higher-challenge-seeking schools", although the moderator in these blocks is the self-reported mindset norm, not challenge seeking.
- p. 14 Extended Data Table 2, italic note under the self-reported mindset norm headers (Models 5a, 5b)
A positive coefficient means larger effects for higher-challenge-seeking schools
Only the authors can say what is right here.
Judge's note: Page 14 image: both self-reported mindset norm blocks carry the same italic note as the behavioural challenge-seeking norm blocks above them. Methods (page 11, 'Norms of self-reported mindset of the schools') define this moderator as the school's average fixed mindset self-report before random assignment, a different variable from the make-a-math-worksheet challenge-seeking norm, and one where higher values mean a more fixed mindset. The note therefore does not say what a positive coefficient means for these rows. The authors need to supply the correct direction (it depends on how the norm was coded), hence a query. This is a separate defect from the duplicated outcome header in C033.
Second check: another model agreed
Report a problem with this finding
Wording Methods state the study's policies "require analysts to round all sample sizes to the nearest 10, so this was done here", but Extended Data Table 1 prints unrounded n values (5651, 6315, 5834, 6096, 11351, 12486, 11539, 11585, 12045, 9899), so the table gives 6315 and 12486 where the text and footnote give "n = 6,320" and "n = 12,490".
- p. 7 Methods, Participants
require analysts to round all sample sizes to the nearest 10, so this was done here
- p. 13 Extended Data Table 1, n column
6315
- p. 13 Extended Data Table 1, n column
12486
- p. 13 Extended Data Table 1, Core course GPA rows
0.101 [0.04, 0.16] 3.508 .001 6315
Only the authors can say what is right here.
Judge's note: Read the Methods crop (page 7), the Extended Data Table 1 page image and footnote (page 13), and the results text crop (page 3). The text values are the table values rounded to the nearest 10 (6315 to 6,320; 12486 to 12,490; 5651 to 5,650), so the two sets of numbers are not in conflict as estimates. The defect is that the table breaks the rounding policy the Methods say was applied. The subgroup sum 6315 + 6170 = 12485 against 12486 is within what rounding of 6,170 allows and is not a separate error. No statement in the article exempts the table from the policy.
Second check: another model agreed
Report a problem with this finding
Wording Extended Data Table 3, panel (c), Supportive norm, High (Top 25%): "0.10 [0.04,0.17]" with "z = 0.002, P = .002" - the z statistic repeats the P value; the estimate and interval imply z of about 3.0-3.1.
- p. 15 Extended Data Table 3, panel (c), Supportive norm, High (Top 25%)
z = 0.002, P = .002
- p. 4 Subgroup effect sizes
schools (P = 0.002; Extended Data Table 3)
- p. 4 Subgroup effect sizes
(P = 0.002; Extended Data Table 3)
Judge's note: Page 15 image: the cell reads CATE 0.10 [0.04,0.17], 'z = 0.002, P = .002'. A z of 0.002 would give P close to 1 and an interval straddling zero. The interval gives s.e. about 0.033 and z about 3.0, which matches P = .002 (z about 3.09). The main text on page 4 cites 'P = 0.002; Extended Data Table 3' for this subgroup, so the P value is the intended one and the z entry is wrong. All other cells in the panel have z values consistent with their P values (e.g. the cell below: z = 0.04, P = .97).
Second check: another model agreed
Report a problem with this finding
23 minor (typos and the like; most checked by the judge only)
Wording Competing interests: "authors D.S.Y, C.S.D., G.W., A.L.D., D.P., and C.H."; "C.H." could be Chris S. Hulleman (C.S.H.) or Cintia P. Hinojosa, "G.W." is "G.M.W." elsewhere, "D.S.Y" lacks its full stop, and Cintia P. Hinojosa has no entry in the author contributions.
- p. 9 Competing interests
D.S.Y, C.S.D., G.W., A.L.D., D.P., and C.H.
- p. 1 Author list
Cintia P. Hinojosa
- p. 9 Competing interests
authors D.S.Y, C.S.D., G.W., A.L.D., D.P., and C.H.
Only the authors can say what is right here.
Judge's note: Read the competing-interests and contributions crops (page 9), the tail of the contributions on page 8 and the byline (page 1), and searched the text for "C.P.H.", which does not occur. The contributions name 24 sets of initials for 25 authors: Hinojosa is absent, and S.Y.H. (see C021) presumably stands for Yang. The ambiguous "C.H." and the missing contribution are for the authors to resolve; the punctuation and "G.W." points are cosmetic.
Second check: another model agreed
Report a problem with this finding
Wording "The model estimates that 5.3% (95% confidence interval = −1.7, −9.0), s.e. = 1.8, t = 2.95, P = 0.005 of 1.5 million students ... would be prevented from being 'off track'": a positive estimate and positive t are paired with a negative interval, listed with the bound nearer zero first, that does not contain 5.3.
- p. 3 Average effects on core course GPAs, second paragraph
5.3% (95% confidence interval = −1.7, −9.0)
- p. 3 Average effects on core course GPAs, second paragraph
5.3% (95% confidence interval = −1.7, −9.0), s.e. = 1.8, t = 2.95
Judge's note: Page 3 image confirms the minus signs on both bounds and none on 5.3 or t. Magnitudes are consistent (5.3 ± 2.013 × 1.8 gives about 1.7 to 8.9), so only the sign convention is mixed: the interval is on the scale of the change in the poor-performance rate (a reduction, as in Extended Data Table 3b where the D/F CATEs are negative) while the estimate is worded as a share prevented. No reader would take away a wrong magnitude or direction, hence minor. The 46% to 41% and 0.05/0.46 = 11% figures are consistent with rounding.
Second check: another model agreed
Report a problem with this finding
Wording Text: "unstandardized B = −0.38 (95% confidence interval = −0.31, −0.46)" versus Extended Data Table 1, lower-achieving students, Fixed mindset: "-0.381 [-0.456, -0.301]"; -0.301 rounds to -0.30, not -0.31.
- p. 3 Average effects on mindset
B = −0.38 (95% confidence interval = −0.31, −0.46)
- p. 13 Extended Data Table 1, Lower-achieving students, Fixed mindset
[-0.456, -0.301]
- p. 3 Average effects on mindset
(95% confidence interval = −0.31, −0.46)
Only the authors can say what is right here.
Judge's note: Page 3 and page 13 images confirm both printed values. The table interval is not symmetric about -0.381: the lower bound is 0.075 below the estimate, and 0.075 above it is -0.306, which would round to the text's -0.31. So the table's -0.301 is the more likely misprint (for -0.306), not the text. A 0.005 to 0.01 difference with no effect on any conclusion. The text also lists the bounds upper first, which is style only.
Second check: another model agreed
Report a problem with this finding
Wording Main text: "earning GPAs at or below the school-specific median"; Methods: "(those below the within-school median)"; Extended Data Fig. 1 legend: "(below the school-specific median)". The definition of the lower-achieving group is stated with and without students exactly at the median.
- p. 3 Defining the primary outcome and student subgroup
earning GPAs at or below the school-specific median
- p. 7 Methods, Participants
(those below the within-school median)
- p. 10 Extended Data Fig. 1 legend
(below the school-specific median)
- p. 10 Extended Data Fig. 1 legend
below the school-specific median
Judge's note: Page 3 image and the page 7 and page 10 crops confirm the three wordings. The main-text sentence itself uses 'at or below' for prior GPA and 'below' for the imputation variables, so the parentheticals in Methods and the legend read as shorthand for the fuller main-text definition. A minor wording inconsistency; the analysis plan (not in the files) holds the exact rule.
Second check: another model agreed
Report a problem with this finding
Wording Extended Data Table 1, Students overall, Fixed mindset: "-0.420 [-0.425, -0.337] -18.686 <.001 11351". The printed interval and the printed t statistic do not correspond under the t reference the table reports (46 df, per Methods): B = -0.420 with t = -18.686 would go with an interval of about [-0.465, -0.375], whereas the printed interval is centred on -0.381, the estimate of the lower-achieving row. We cannot tell which number is intended.
- p. 13 Extended Data Table 1, Students overall, Fixed mindset row
-0.420
- p. 13 Extended Data Table 1, Students overall, Fixed mindset row
[-0.425, -0.337]
- p. 13 Extended Data Table 1, Students overall, Fixed mindset
-0.420 [-0.425, -0.337] -18.686 <.001 11351
- p. 13 Extended Data Table 1, Students overall, Fixed mindset
-0.420 [-0.425, -0.337]
Only the authors can say what is right here.
Judge's note: Read the row on the page 13 image: B -0.420, CI [-0.425, -0.337], t(46) -18.686. The interval's midpoint is -0.381 and its half-width 0.044. B/t gives s.e. 0.0225 and a half-width of about 0.045, so the width fits the t statistic but the interval is shifted by about 0.04 from B. Every other row in the table is symmetric about B within rounding (e.g. lower-achieving -0.381 [-0.456, -0.301]). The main text gives only the lower-achieving value (B = -0.38), so nothing in the article says whether B or the bounds are wrong; authors must say.
Second check: another model agreed
Report a problem with this finding
Wording Extended Data Table 3, panel (b), Supportive norm, Low and Medium Combined: "-.08 [-.13,-.02]" with "z = 3.78, P <.001". Under the z reference the table itself reports, the printed interval and the printed z statistic do not correspond: the interval is wider than z = 3.78 allows, which would be about [-.12, -.04]. We cannot tell which is intended.
- p. 15 Extended Data Table 3, panel (b), Supportive norm, Low and Medium Combined
-.08 [-.13,-.02]
- p. 15 Extended Data Table 3, panel (b), Supportive norm, Low and Medium Combined, test statistics
z = 3.78, P <.001
Only the authors can say what is right here.
Judge's note: Page 15 image: the cell reads -.08 [-.13,-.02], z = 3.78, P <.001. The printed interval gives s.e. 0.028 and z about 2.85; even taking the most favourable rounding of all three numbers (estimate -.085, bounds -.125 and -.025) z reaches only about 3.3. From B and z the s.e. is 0.021 and the interval about [-.12, -.04]. The adjacent Medium column prints -.08 [-.13, -.04] with z = 3.74, and in panels (a) and (c) the Combined and Medium supportive-norm cells have near-identical intervals, so the upper bound is probably -.04 (the -.02 matches the unsupportive row directly below). Nothing in the table footnote explains a different interval method for this cell. Which number is wrong is for the authors to confirm.
Second check: another model agreed
Report a problem with this finding
Typo "as described in the pre-analysis plan found in the Supplementary Information section 13 and at https://osf.io35": the in-text URL is the bare domain, whereas the plan is at https://osf.io/tn6g4 (Fig. 1 legend, Data availability, reference 35).
- p. 3 Defining the primary outcome and student subgroup
and at https://osf.io35
- p. 2 Fig. 1 legend
https://osf.io/tn6g4
Judge's note: Page 3 image shows 'https://osf.io' followed by superscript 35; the Fig. 1 legend crop and reference 35 give https://osf.io/tn6g4. The superscript leads the reader to the full address, so nothing is lost; the in-text URL is simply incomplete and inconsistent with the other three mentions.
Second check: none (judge only)
Report a problem with this finding
Typo Text gives "z = −2.76" for the intervention × school achievement interaction; Extended Data Table 2 Model 1a prints "-0.074 -0.13 -0.02 2.76 .006", a positive z for a negative coefficient, as it does for every negative coefficient in the table.
- p. 3 Moderation by school achievement level
z = −2.76
- p. 14 Extended Data Table 2, Model 1a
-0.074 -0.13 -0.02 2.76
- p. 3 Extended Data Table 2, Model 1a
2.76
Judge's note: Page 14 image: all z values are unsigned (2.76, 3.05, 0.87, 2.89, 2.91, 3.71, 3.80, 2.80, 2.82 for negative B). The table evidently reports absolute z values without saying so. Extended Data Table 3 is internally mixed: panel (a) prints z = -0.13 for a negative CATE while panel (b) prints positive z for negative CATEs. The magnitude agrees with the text and no conclusion is affected; a presentation inconsistency only.
Second check: none (judge only)
Report a problem with this finding
Typo "the conditional average treatment effect (CATEs) on core GPAs within low- and medium-achieving schools (combined) was 0.14 grade points ... and 0.18 grade points": singular 'effect' is glossed with the plural abbreviation 'CATEs'.
- p. 4 Subgroup effect sizes
the conditional average treatment effect (CATEs) on core GPAs
Judge's note: Crop of page 4 confirms the text. Number disagreement between the noun and its abbreviation (two values follow, so 'effects (CATEs) ... were' was probably intended). Cosmetic.
Second check: none (judge only)
Report a problem with this finding
Typo Course-enrolment tests "t = 3.18, P = 0.001", "(t = 2.37, P = 0.018)" and "(t = 2.00, P = 0.045)" carry P values that match a standard normal reference, not the t distribution described in Methods ("For the t distribution, the degrees of freedom is 46"); with 46 degrees of freedom t = 2.00 gives P of about 0.051.
- p. 5 Advanced mathematics course enrolment, continuation paragraph
(t = 2.00, P = 0.045)
- p. 5 Advanced mathematics course enrolment, continuation paragraph
(t = 2.37, P = 0.018)
- p. 5 Advanced mathematics course enrolment, continuation paragraph
t = 3.18, P = 0.001
- p. 8 Methods, Model for average treatment effects
For the t distribution, the degrees of freedom is 46
Judge's note: Page 5 image confirms the three t/P pairs; the page 8 crop confirms the stated 46 degrees of freedom for average-treatment-effect models. The two-sided 5% critical value of t with 46 df is about 2.013, so t = 2.00 cannot give P = 0.045 on that reference; the printed P values are what a normal distribution gives (0.0455, 0.018, 0.0015). The GPA results on page 3 (t = 3.51, P = 0.001; t = 2.95, P = 0.005) are consistent with 46 df. Methods (Course enrolment to advanced mathematics, page 8) says the analysis used the full sample in 41 schools but states no different model or reference distribution; fewer clusters would give fewer degrees of freedom and larger P. Tail probabilities cannot be written in the plain-arithmetic field; the one expression given is the gap between the statistic and the critical value. The practical effect is confined to one secondary subgroup result crossing 0.05. The supplementary methods would settle which reference distribution was used.
Second check: another model agreed
Report a problem with this finding
Typo "interaction, (B = 0.04 (95% confidence interval = 0.00, 0.08), s.e. = 0.02, z = 2.26, P = 0.024, the opposite of what we found for core course GPAs": the parenthesis opened before 'B = 0.04' is never closed.
- p. 5 Advanced mathematics course enrolment, continuation paragraph
interaction, (B = 0.04 (95% confidence interval = 0.00, 0.08)
- p. 5 Advanced mathematics course enrolment in tenth grade
interaction, (B = 0.04 (95% confidence interval = 0.00, 0.08), s.e. = 0.02, z = 2.26, P = 0.024, the opposite
- p. 5 Advanced mathematics paragraph
interaction, (B = 0.04 (95% confidence interval = 0.00, 0.08),
Judge's note: Page 5 image confirms: the inner interval parenthesis closes, the outer one does not, and the sentence runs on after P = 0.024. Cosmetic punctuation error (a closing parenthesis is missing after 'P = 0.024', or the opening one is spurious).
Second check: none (judge only)
Report a problem with this finding
Typo "a high-quality seed (an adaptive belief system conveyed by a compelling intervention) and conductive soil in which that seed can grow": 'conductive' should be 'conducive'.
- p. 5 Discussion, third paragraph
conductive soil in which that seed can grow
Judge's note: Page 5 image confirms 'conductive soil'. The seed-and-soil metaphor calls for 'conducive' (favourable); 'conductive' has no fitting sense here. Wrong word, cosmetic.
Second check: none (judge only)
Report a problem with this finding
Typo Reference 6 prints the title-case book title as "Navigating the Social Challenges of high School to Get an Education"; "high" should be capitalised ("High School").
- p. 5 Reference 6
Navigating the Social Challenges of high School to Get an Education
- p. 5 Reference 6
Social Challenges of high School to Get an Education
Judge's note: Read the crop of the reference list on page 5. Every other main word of the italic title is capitalised and "high" alone is lower-case, so the inconsistency is internal to the entry and needs no outside source.
Second check: none (judge only)
Report a problem with this finding
Typo Methods (GPAs): "Schools provided the grades of each student in each course for the eight and ninth grade"; "eight" should be "eighth".
- p. 7 Methods, Measures, GPAs
for the eight and ninth grade
Judge's note: Read the crop of page 7. The sentence is printed as quoted, with the cardinal "eight" where the ordinal is needed.
Second check: none (judge only)
Report a problem with this finding
Typo Methods: "in the median school 91% students reported that most or all of their peers worked carefully and quietly"; the word "of" is missing ("91% of students").
- p. 7 Methods, Intervention delivery and fidelity
in the median school 91% students reported
- p. 7
in the median school 91% students reported that most
Judge's note: Read the crop of page 7. The text runs "91%" at the end of one line and "students reported" at the start of the next, with no "of".
Second check: none (judge only)
Report a problem with this finding
Typo Methods (Bayesian robustness analysis): "see the Supplemental Information, section 13, p. 12", where every other mention in the article is "Supplementary Information".
- p. 8 Bayesian robustness analysis (Methods)
see the Supplemental Information, section 13, p. 12
Judge's note: Read the crop of page 8 and searched the article text for both forms. "Supplemental Information" occurs once; all other occurrences (about twenty) are "Supplementary Information". It is a terminology inconsistency only, and the pointer itself is not shown to be wrong.
Second check: none (judge only)
Report a problem with this finding
Typo Extended Data Fig. 1 legend: "because this does not involve coding pre- and post-treatment making periods"; "making" should be "marking".
- p. 10 Extended Data Fig. 1 legend
post-treatment making periods
- p. 10 Extended Data Fig. 1 legend
pre- and post-treatment making periods
- p. 10 Extended Data Fig. 1 caption
coding pre- and post-treatment making periods
Judge's note: Read the crop of the legend on page 10. Three lines earlier the same sentence has "revising the post-treatment marking period", so "making periods" is a misspelling.
Second check: none (judge only)
Report a problem with this finding
Typo Extended Data Fig. 2 legend: "Mindset treatment effects on for core course GPAs (a) and mathematics/science GPAs (b)"; there is a stray preposition ("on for").
- p. 11 Extended Data Fig. 2 legend
Mindset treatment effects on for
- p. 11 Extended Data Fig. 2 legend
Mindset treatment effects on for core course GPAs
- p. 11 Extended Data Fig. 2 legend
Mindset treatment effects on for core course GPAs (a)
Judge's note: Read the legend on the page 11 image and its crop. The words "on for" are printed as quoted.
Second check: none (judge only)
Report a problem with this finding
Typo Extended Data Fig. 3 legend: "The area between thevertical lines is the interquartile range"; a space is missing between "the" and "vertical".
- p. 12 Extended Data Fig. 3 legend
The area between thevertical lines
- p. 12 Extended Data Fig. 3 legend
(since it is a nonlinear model).Recall BCF
- p. 12 Extended Data Fig. 3 legend title
as the pre-registered linear mixed-effects model
Judge's note: Read the legend on the page 12 image. "thevertical" is typeset as one word, so it is not an extraction artefact. The other two defects reviewers attached to this cluster are judged separately: the missing full stop after the title in C027 and "model).Recall" in C028.
Second check: none (judge only)
Report a problem with this finding
Typo Extended Data Fig. 3 legend title: "...moderation by norms as the pre-registered linear mixed-effects model The BCF analysis uses..."; the bold title has no closing full stop.
- p. 12 Extended Data Fig. 3 legend, title
moderation by norms as the pre-registered linear mixed-effects model
- p. 12 Extended Data Fig. 3 title
pre-registered linear mixed-effects model The BCF analysis
Judge's note: Read the legend on the page 12 image. The bold title ends at "model" with no full stop before the legend text starts. The Extended Data Fig. 2 title on page 11 ends with a full stop ("heterogeneity across schools."), so this is not house style.
Second check: none (judge only)
Report a problem with this finding
Typo Extended Data Fig. 3 legend: "(since it is a nonlinear model).Recall BCF is designed to shrink"; there is no space after the full stop.
- p. 12 Extended Data Fig. 3 legend
(since it is a nonlinear model).Recall BCF
Judge's note: Read the legend on the page 12 image. "model).Recall" is typeset with no space. It is a different place in the legend from "thevertical" (C026), so it is kept as its own item.
Second check: none (judge only)
Report a problem with this finding
Typo Extended Data Fig. 3 legend: "LOESS smoothing curves for the trend in norms of the estimated treatment effects"; there is a double space before "norms".
- p. 12 Extended Data Fig. 3 legend
the trend in norms of the estimated treatment effects
Judge's note: Read the legend on the page 12 image. The gap before "norms" is visibly wider than the other word spaces in a ragged-right legend, and the text layer has two spaces. The same phrasing without the extra space appears earlier in the legend ("the linear trend of the estimated treatment effects in norms"), so no word is evidently missing. It is a spacing blemish only.
Second check: none (judge only)
Report a problem with this finding
Typo Extended Data Table 3, panel (b), Unsupportive norm, High (Top 25%): "z = 1.23, P = 2.19" - a P value cannot exceed 1; it should be P = .219.
- p. 15 Extended Data Table 3, panel (b), Unsupportive norm, High (Top 25%)
z = 1.23, P = 2.19
- p. 15 Extended Data Table 3b, Low and Medium (not shown in extraction), High column, unsupportive norm
z = 1.23, P = .2.19
Judge's note: Page 15 image: the cell reads 'z = 1.23, P = 2.19' (at this resolution a stray leading point, i.e. '.2.19', cannot be excluded; either way the value is malformed). The CATE in the cell is .04 [-.02, .09], which includes zero, and a two-tailed P for z = 1.23 is about .219, so the fix is evident: a misplaced or missing decimal point. Rated cosmetic because no reader could take 2.19 as a real P value. The normal-tail value is a standard table lookup, not computable with the four operations, so the arithmetic below only checks that z fits the interval.
Second check: none (judge only)