The plate says a black-white gap of about one standard deviation sits on the same bell curve.
.jpg)
Two overlapping curves, means marked about 85 and 100.
SteelMan: A mean gap of that size is the standard report from U.S. IQ batteries.
Cut: The plate's 85 and 100 are its marks. One curve does not make the gap genetic.
The plate says the black-white gap is about 15 points, one standard deviation.
.jpg)
A ruler marks 15 points between two means.
SteelMan: The APA task force treated a black-white test-score gap as a known finding and left its cause open.
Cut: Fifteen points is the adult full-scale figure often cited. It is not every subtest, and it is not a cause.
The plate says a one-standard-deviation mean gap becomes a several-fold gap in the right tail.
.jpg)
A threshold line cuts two curves, the higher mean supplying more of the tail.
SteelMan: A mean shift does change tail ratios if the distributions are roughly normal.
Cut: The plate's multiple is not calculated. Tail ratios also depend on variance, which the plate does not show.
The plate says the gap is largest on the g factor, not on a single subtest.
.jpg)
A factor arrow is larger than the subtest arrows.
SteelMan: g is the common factor across cognitive tests. The gap is not confined to one item type.
Cut: The plate does not show a factor analysis. A larger g loading is not a genetic finding.
The plate says elementary cognitive tasks, including reaction time, show a smaller gap in the same direction.
.jpg)
A timer sits beside an IQ card.
SteelMan: Choice reaction time correlates modestly with g. Group differences there are smaller and less stable.
Cut: The plate treats a timer as the same object as a full-scale IQ. It is not.
The plate says the tests predict grades and job performance equally across groups, so the gap is not a biased item.
.jpg)
A regression line is the same for two groups.
SteelMan: Predictive validity is the standard test of bias. Equal slopes cut against a simple item-bias account.
Cut: Equal prediction does not explain the mean gap. The plate's lines are not a validity study.
The plate says the black-white gap is larger on more g-loaded tests.
.jpg)
A scatter of subtests rises with g loading.
SteelMan: That rank-order pattern is the Spearman hypothesis, and it has been reported on several batteries.
Cut: The plate's scatter is unlabeled. A g pattern is not a heritability.
The plate says IQ is substantially heritable within groups, and that this does not by itself explain the group gap.
.jpg)
A twin diagram is marked high heritability, then a separate group arrow.
SteelMan: The APA statement said genetic differences contribute substantially to individual differences, and that the cause of group differences was not settled.
Cut: The plate is right to separate the two. It then leans on the within-group figure as if it transferred.
The plate says a high within-group heritability does not logically settle the between-group gap.
.jpg)
A within arrow and a between arrow are drawn as different questions.
SteelMan: The logical point is correct. Heritability is a within-population variance share.
Cut: The plate uses the logical point and still treats the group gap as genetic on other plates. Both cannot be a proof.
The plate says transracial adoption is the clean test and that the gap remained.
.jpg)
Black children in white homes are still below white adoptees.
SteelMan: The Minnesota study found black adoptees about 15 to 17 points below white adoptees at ages 7 and 17, before Flynn correction.
Cut: The 1992 authors did not call the result a genetic proof. Late testing, selective placement, and pre-adoption conditions remain open.
The plate says the gap was larger at 17 than the early headlines implied.
.jpg)
Two ages, 7 and 17, with the gap marked persistent.
SteelMan: The follow-up retested the children at about 17. The black-white adoptee gap remained.
Cut: Flynn-corrected figures published later are larger, not smaller. The plate does not print those corrections.
The plate says the Eyferth study of occupation children is the usual reply and is a selected sample.
.jpg)
A German sample is marked selected.
SteelMan: Eyferth is a small study of children of Allied soldiers and German mothers. It is not a U.S. population sample.
Cut: The plate dismisses it without the result. A selected sample can still cut a universal genetic claim.
The plate says scores rose across the century and the black-white gap did not close by the same amount.
.jpg)
Two rising lines, the gap marked smaller but open.
SteelMan: Scores did rise. The black-white gap narrowed on some tests in some decades and not on others.
Cut: The plate's 'did not close' is too flat. NAEP and some IQ series show partial convergence.
The plate says the gap narrowed and then stalled.
.jpg)
A falling gap flattens.
SteelMan: Dickens and Flynn reported narrowing on some instruments through the late twentieth century.
Cut: The plate's stall date is not given. A stall on one test is not a stall on all of them.
The plate says the NAEP black-white gap narrowed after the 1970s and remains.
.jpg)
Two score lines, a remaining gap.
SteelMan: NAEP long-term trend gaps did narrow and remain on reading and math.
Cut: The plate does not print the scale points. NAEP is not an IQ test.
The plate says the SAT black-white gap is stable.
.jpg)
Two SAT means, a fixed gap.
SteelMan: College Board means differ by race, and the gap has not closed.
Cut: SAT takers are not a population sample. The plate's means are not printed.
The plate says the GRE gap matches the IQ gap.
.jpg)
A GRE card sits on the same 15-point ruler.
SteelMan: Graduate exam means differ by group.
Cut: The plate equates a selected exam with a population IQ. The samples are not the same.
The plate says g predicts job performance and the gap reappears in rated performance.
.jpg)
A validity arrow runs from g to a job rating.
SteelMan: Meta-analyses find g predicts training and job performance.
Cut: A rating gap is not an IQ gap. Supervisor ratings have their own bias.
The plate says the gap appears in training scores, not only in paper tests.
.jpg)
A classroom scorecard differs by group.
SteelMan: Training criteria correlate with g. Group differences there are reported.
Cut: The plate's card is unlabeled. Training is not independent of the same schooling.
The plate says children of high-scoring black parents regress toward a lower group mean.
.jpg)
A parent-child diagram regresses to different means.
SteelMan: Regression to the group mean is the prediction if the means differ and the trait is imperfectly transmitted.
Cut: The plate treats that prediction as proof of a genetic mean. The same pattern follows from any stable mean difference.
The plate says the gap shrinks but remains after a socioeconomic control.
.jpg)
Two bars, uncontrolled and controlled, the second shorter.
SteelMan: SES accounts for a share of the gap and not all of it in most regressions.
Cut: The plate's remaining share is not a coefficient. SES is partly downstream of the same scores.
The plate says matching on parental income does not close the gap.
.jpg)
Two income-matched bars still differ.
SteelMan: Income is a weak stand-in for the cognitive environment. Matched-income gaps are reported.
Cut: The plate's match is not a study. Income bands hide wealth and parental education.
The plate says the gap remains inside the same parental-education cell.
.jpg)
A college-parent cell still shows a gap.
SteelMan: Parent education is the stronger SES covariate, and a gap inside that cell is reported.
Cut: The plate does not show the cell size. College-educated parents are not one environment.
The plate says a wealth control still leaves a gap.
.jpg)
A wealth-matched pair still differs.
SteelMan: Wealth is unequally distributed and is a real covariate.
Cut: The plate's match is fictional. Wealth is partly a result of earlier scores.
The plate says lead exposure is offered as the cause and does not date the gap.
.jpg)
A lead stamp is marked not sufficient.
SteelMan: Lead lowers scores and was unequally distributed. The gap predates the late lead era and remains after blood-lead fell.
Cut: The plate does not estimate the lead share. 'Not sufficient' is not 'not real'.
The plate says nutrition is a real covariate and not a full account.
.jpg)
A diet card is marked partial.
SteelMan: Iodine, iron, and early diet affect scores. They do not, by themselves, produce a one-standard-deviation gap in current U.S. samples.
Cut: The plate does not measure a diet share.
The plate says years of school raise scores and do not close the gap.
.jpg)
A school arrow raises both lines.
SteelMan: Schooling is one of the few environmental factors the APA statement called certainly important. Extra years raise means.
Cut: The plate's parallel lines are not a school study. Quality, not just years, is the open variable.
The plate says stereotype threat moves a score in a lab and does not account for the gap.
.jpg)
A lab arrow is small beside a population gap.
SteelMan: Threat effects are real in some experiments and small or absent in others. They are not the population gap.
Cut: The plate treats a failed full account as a failed effect. Those are different claims.
The plate says anxiety is a covariate and not the gap.
.jpg)
An anxiety stamp sits on a small arrow.
SteelMan: Anxiety correlates with test scores. It does not equal a one-standard-deviation mean shift.
Cut: The plate does not show the partial coefficient.
The plate says the gap is not larger on the most culturally loaded items.
.jpg)
A culture-loaded item and a matrix item show the same gap.
SteelMan: That pattern is reported and cuts against a simple item-content account.
Cut: The plate's two items are cartoons. Item reviews are the evidence, and they are not shown.
The plate says the gap appears on Raven's, a nonverbal test.
.jpg)
A matrix puzzle is marked same gap.
SteelMan: Raven gaps are reported. The test is less verbal, not culture-free.
Cut: The plate equates nonverbal with culture-free. Raven has its own schooling correlation.
The plate says a gap appears on working-memory tasks.
.jpg)
A digit-span card differs by group.
SteelMan: Working-memory differences are reported and are smaller than the full-scale gap.
Cut: The plate's card has no numbers.
The plate says average brain volume differs by group and correlates with g.
.jpg)
A volume figure sits beside an IQ figure.
SteelMan: MRI studies find a modest correlation of brain volume with g, and some group averages differ.
Cut: The plate treats a volume difference as the explanation. The correlation is too small, and the group volume literature is contested.
The plate says within-person reaction-time variance, not just the mean, differs.
.jpg)
A variance bar is higher for one group.
SteelMan: Variability measures have been reported. They are not a standard population result.
Cut: The plate's bar is unlabeled.
The plate says earlier adoption should close the gap if the cause is postnatal, and that it did not.
.jpg)
An early-adoption bar still differs.
SteelMan: Age at adoption is the right test. The Minnesota sample was not adopted at birth in every case.
Cut: The plate's early bar is not the study's age split.
The plate says mixed-ancestry means fall between the two group means.
.jpg)
A middle bar sits between two others.
SteelMan: Intermediate means are reported in some samples and not in others.
Cut: The plate's middle bar is a prediction, not a table. Ancestry is not a parentage checkbox.
The plate says genome-wide ancestry should correlate with the score if the gap is genetic.
.jpg)
An ancestry axis is drawn against a score.
SteelMan: That correlation is the current genetic test. Published estimates disagree in sign and size.
Cut: The plate draws the prediction as a result. It is not.
The plate says polygenic scores predict individual scores and do not yet explain the group gap.
.jpg)
A score arrow works within a group and is marked open between groups.
SteelMan: Within-family prediction is the cleaner genetic result. Cross-population transfer is weaker.
Cut: The plate's open arrow is fair. A later study can close it. The plate cannot.
The plate says environment is real and is not the whole gap.
.jpg)
An environment arrow is marked partial.
SteelMan: The APA statement said environmental factors contribute substantially and are poorly specified.
Cut: The plate uses 'partial' as if the remainder were genetic. The remainder is unexplained.
The plate says shared family environment fades by adulthood.
.jpg)
A family arrow shrinks from childhood to adulthood.
SteelMan: Adoption and twin studies find shared-environment influence on IQ declines with age.
Cut: The plate treats that fade as a group-gap result. It is a within-family result.
The plate says the remaining environment is nonshared and hard to name.
.jpg)
A residual arrow is marked unnamed.
SteelMan: Nonshared environment is the residual in twin models. It includes measurement error.
Cut: The plate equates an unnamed residual with a small environment. Those are not the same.
The plate says the Flynn rise is partly nutrition and schooling, and the group gap did not track it one for one.
.jpg)
A rising national line and a slower gap line.
SteelMan: Both claims are argued in the Flynn literature.
Cut: The plate does not show the two series on one scale.
The plate says early programs raise scores and the gain fades.
.jpg)
A preschool arrow rises, then falls.
SteelMan: Fade-out of IQ gains is the usual Head Start finding. Schooling gains sometimes remain.
Cut: The plate treats fade-out as a closed case against environment. A fade is not a zero effect.
The plate says the intensive preschool trials did not close a one-standard-deviation gap.
.jpg)
A trial arrow is shorter than the gap.
SteelMan: Abecedarian raised early scores. Adult IQ effects are smaller than the original gap.
Cut: The plate's arrow is not the trial. A smaller effect is not a null effect.
The plate says the gap appears in AFQT military samples.
.jpg)
An AFQT card differs by group.
SteelMan: The NLSY AFQT gap is one of the most cited adult figures.
Cut: The plate does not print the means. AFQT is a vocational battery, not a clinical IQ test.
The plate says high-school GPA gaps are smaller than IQ gaps and do not erase them.
.jpg)
A GPA gap is shorter than a test gap.
SteelMan: Grades and tests are different instruments. Grades include effort and course choice.
Cut: The plate's lengths are props.
The plate says the test gap reappears as a college-grade gap after admission.
.jpg)
A grade card differs inside one campus.
SteelMan: Mismatch research argues that admission gaps reappear in grades and completion.
Cut: The plate does not show a campus table. Major choice changes the grade.
The plate says professional licensing exams show the same gap.
.jpg)
A bar-pass card differs by group.
SteelMan: Pass rates differ by race on several licensing exams.
Cut: The plate's card is unlabeled. A pass-rate gap is not a 15-point IQ gap.
The plate says medical-board gaps match the admission gap.
.jpg)
A board score sits on the same ruler.
SteelMan: Board scores differ by group.
Cut: The plate equates a selected profession with a population. The samples differ.
The plate says the IQ gap reappears as an occupational-status gap.
.jpg)
A status ladder differs by group.
SteelMan: The APA statement said a correlation of scores with occupational status remains after education and family background are controlled.
Cut: The plate treats that correlation as the group gap. Status also tracks discrimination, networks, and field choice.
The plate says low IQ predicts crime within groups, so a mean gap predicts a crime gap.
.jpg)
An IQ arrow points at a crime rate.
SteelMan: IQ predicts offending within groups. A mean shift changes the expected rate if nothing else differs.
Cut: The plate skips age, sex, and neighborhood. A prediction is not the crime table.
The plate says a group mean is not a judgment of a named person.
.jpg)
A mean card is crossed out over one figure.
SteelMan: The logical point is correct. Overlap is large.
Cut: The plate uses the point and still draws the tail as a group verdict on other plates.
The plate says the distributions overlap, so most individuals are not in the exclusive tail.
.jpg)
Two curves share most of their area.
SteelMan: A one-standard-deviation shift leaves most people inside the shared range.
Cut: The plate's shaded share is not calculated.
The plate says the black distribution is sometimes drawn wider, and the evidence is mixed.
.jpg)
Two widths, one marked uncertain.
SteelMan: Variance differences are reported in some batteries and not in others.
Cut: The plate's wider curve is a prop.
The plate says a male variance difference is accepted and a group mean difference is not.
.jpg)
A variance diagram is marked publishable. A mean diagram is marked not.
SteelMan: Male variance on some tests is a reported finding. The publication difference is a norm, not a result.
Cut: The plate equates a variance claim with a mean claim. They are different measurements.
The plate says the Ashkenazi mean is above the European mean, so group means are not a taboo only in one direction.
.jpg)
A higher mean is marked about one half to one standard deviation up.
SteelMan: A higher Ashkenazi verbal mean is a reported finding.
Cut: The plate's size is its own. A higher mean elsewhere does not identify the cause of the black-white gap.
The plate says East Asian means are higher on visuospatial tests.
.jpg)
A visuospatial bar is higher.
SteelMan: That pattern is reported on some batteries and is smaller than the plate implies.
Cut: The plate's bar is unlabeled. National samples and American samples disagree in size.
The plate says sub-Saharan means are lower than African American means.
.jpg)
A lower bar is marked Africa.
SteelMan: Some compilations report lower means. Schooling, nutrition, and test conditions differ sharply.
Cut: The plate treats a compilation as a population IQ. Test conditions are not held constant.
The plate says African and Caribbean immigrants are selected and their children do not match the selection story.
.jpg)
A selected parent bar and a lower child bar.
SteelMan: Selection is real. Second-generation means are the test of a pure-selection account.
Cut: The plate's child bar is not a study.
The plate says a within-group color gradient is offered as genetic and is confounded by status.
.jpg)
A skin-tone axis is marked confounded.
SteelMan: Color correlates with status and with measured ancestry in some samples.
Cut: The plate does not separate the two. A gradient is not an ancestry coefficient.
The plate says a caste account predicts a gap that should close when caste weakens, and that the test-score gap did not close in step.
.jpg)
A caste arrow shrinks. A score gap does not.
SteelMan: Legal caste changed faster than the test gap. That cuts a pure caste story.
Cut: The plate's arrows are not dated. Caste is not only a statute.
The plate says discrimination is real and does not, by itself, produce a g-loaded gap.
.jpg)
A discrimination stamp is marked not sufficient.
SteelMan: Hiring and schooling discrimination are documented. A g pattern is a separate fact.
Cut: The plate treats 'not sufficient' as 'not real'.
The plate says teacher expectation is too small to be the gap.
.jpg)
A small expectation arrow sits beside the gap.
SteelMan: Expectation effects exist and are smaller than one standard deviation in the better studies.
Cut: The plate's arrow is not a study.
The plate says incentive studies do not close the gap.
.jpg)
A paid-test bar still differs.
SteelMan: Incentives raise scores. They do not erase group means in the studies usually cited.
Cut: The plate does not name the study.
The plate says the gap remains among native English speakers.
.jpg)
A native-speaker filter still shows a gap.
SteelMan: Dialect and language are real covariates. They do not account for a nonverbal gap.
Cut: The plate's filter is not a sample.
The plate says the gap appears in every U.S. region.
.jpg)
Four regional pairs all differ.
SteelMan: Regional means differ, and the black-white gap is reported in each census region.
Cut: The plate's four pairs are unlabeled.
The plate says the gap appears in cohorts born after civil rights.
.jpg)
A post-1964 cohort still shows a gap.
SteelMan: Later cohorts narrowed on some tests and not on others.
Cut: The plate's post-1964 line is not NAEP. A remaining gap is not an unchanged gap.
The plate says the AFQT gap remained after the integrated force.
.jpg)
A post-1970s military card still differs.
SteelMan: The volunteer-force AFQT gap is documented.
Cut: The plate does not print the means.
The plate says special-education placement follows the left tail.
.jpg)
A placement bar is higher for the lower mean.
SteelMan: A mean shift predicts more placements below a cutoff if the cutoff is shared.
Cut: Placement also follows school policy. The plate's bar is not a count.
The plate says gifted placement follows the right tail.
.jpg)
A gifted bar is higher for the higher mean.
SteelMan: The same cutoff logic applies.
Cut: Selection into gifted programs is not only a test. The plate ignores referral.
The plate says a shared cutoff, not a different test, produces the placement gap.
.jpg)
One line cuts two curves.
SteelMan: That is the tail arithmetic.
Cut: The plate assumes the cutoff is the only screen. It is not.
The plate says the curve is a description, not a permit and not a ban.
.jpg)
A curve sits between a stamp that says do not cite and a stamp that says therefore.
SteelMan: A mean describes a recorded distribution. It does not try a person.
Cut: The plate's moral is a reading rule. It is not a new measurement.
The plate says null environmental results are filed as inconclusive and genetic results as unpublishable.
.jpg)
Two folders, different stamps.
SteelMan: The publication norm is real. It is not a finding about the gap.
Cut: The plate equates a filing rule with a result.
The plate says the 1996 APA statement recorded the gap and did not assign its cause.
.jpg)
A statement is marked gap known, cause open.
SteelMan: The task force said genetic differences matter for individual differences and did not settle the group difference.
Cut: The plate's paraphrase is fair. The press release is not the full report.
The plate says The Bell Curve stated the gap and a partial genetic hypothesis, and that the hypothesis is not the data.
.jpg)
A book is split into data and hypothesis.
SteelMan: The data chapters and the genetic chapters are separable.
Cut: The plate does not cite a page. The hypothesis is the book's, not a later proof.
The plate says the 2005 review concluded the gap is partly genetic.
.jpg)
A review is marked 50 to 80 percent genetic.
SteelMan: That range is the authors' conclusion, not a measurement.
Cut: The plate prints the conclusion as a result. It is an inference.
The plate says the environmental reply exists and does not close the gap.
.jpg)
A reply book is marked partial.
SteelMan: Nisbett's review argues for a larger environmental share.
Cut: The plate treats the reply as a failed close. It is a competing account.
The plate says intelligence researchers split on the cause and do not split on the gap.
.jpg)
A survey bar is marked gap agreed, cause split.
SteelMan: Snyderman and Rothman, and later surveys, found that pattern.
Cut: The plate's bars are not a survey. Expert opinion is not a heritability.
The plate says a sibling comparison is the clean genetic test and is rarely shown.
.jpg)
A within-family arrow is marked missing.
SteelMan: Within-family ancestry or polygenic tests are the current design.
Cut: The plate's missing arrow is not a result. Absence of a figure is not a finding.
The plate says adoptee grades followed the test gap.
.jpg)
A grade card differs inside the adoptive sample.
SteelMan: The Minnesota follow-up reported academic outcomes as well as IQ.
Cut: The plate's card is not the table.
The plate says intervention gains are largest before school and fade after.
.jpg)
A preschool peak, then a decline.
SteelMan: That shape is the usual early-program result.
Cut: The plate's peak is unlabeled.
The plate says iodine deficiency lowers scores and is not the U.S. gap.
.jpg)
An iodine stamp is marked elsewhere.
SteelMan: Iodine matters where deficiency is common. It is not the current U.S. black-white gap.
Cut: The plate does not show a deficiency rate.
The plate says breastfeeding raises scores and does not close the gap.
.jpg)
A feeding arrow is small.
SteelMan: Breastfeeding effects on IQ are small in the better studies.
Cut: The plate's arrow is not a trial.
The plate says father absence is more common and does not account for the g gap.
.jpg)
A household arrow is marked partial.
SteelMan: Family structure correlates with scores. It is not a one-standard-deviation account in the regressions usually cited.
Cut: The plate's arrow is not a coefficient.
The plate says hours of television are offered as the cause and are too small.
.jpg)
A screen arrow is short.
SteelMan: Media time correlates weakly with scores.
Cut: The plate's arrow is not a study.
The plate says dialect differences do not explain a matrix gap.
.jpg)
A dialect stamp sits on a verbal item, not on a matrix.
SteelMan: That split is the right test.
Cut: The plate's two items are cartoons.
The plate says examiner race does not close the gap.
.jpg)
A matched-examiner pair still differs.
SteelMan: Examiner effects are small in the studies usually cited.
Cut: The plate's pair is not a study.
The plate says coaching raises a score and does not close the gap.
.jpg)
A coaching arrow is shorter than the gap.
SteelMan: Coaching effects are real and smaller than one standard deviation.
Cut: The plate's arrow is not a coaching trial.
The plate says the gap remains inside selective schools.
.jpg)
A campus pair still differs.
SteelMan: Admitted students still differ on the admission test, so a remaining gap is expected.
Cut: The plate treats the campus as an environment held constant. The admission scores were not.
The plate says standardized military training does not close the AFQT gap.
.jpg)
A training arrow raises both lines.
SteelMan: Shared training is a useful control. It does not erase prior differences.
Cut: The plate's lines are not a military table.
The plate says birth weight differs and does not account for the later gap.
.jpg)
A birth-weight arrow is marked partial.
SteelMan: Birth weight correlates with later scores and differs by group.
Cut: The plate does not show the partial coefficient.
The plate says maternal age is a covariate and not the gap.
.jpg)
An age stamp is marked partial.
SteelMan: Age at birth correlates with child scores.
Cut: The plate's stamp is not a regression.
The plate says neighborhood controls shrink the gap and do not close it.
.jpg)
A neighborhood-matched pair still differs.
SteelMan: Neighborhood is a real covariate. Matching on it is partly matching on prior scores.
Cut: The plate's pair is fictional.
The plate says a within-family ancestry correlation, or a closed gap after a named intervention, would decide the cause.
.jpg)
Two tests are drawn, neither checked.
SteelMan: Those are the decisive designs.
Cut: The plate leaves them blank and still writes as if the cause were known.
The plate says a 15-point shift implies a large ratio above 130.
.jpg)
A 130 line cuts two curves.
SteelMan: The arithmetic is right if both distributions are normal with equal variance.
Cut: The plate does not state the variance assumption. Real tails are thinner than a normal curve.
The plate says the gap is about one standard deviation, not a law of fifteen points.
.jpg)
A ruler is marked about 15.
SteelMan: The adult figure is near that on several batteries and lower on others.
Cut: The plate's 'about' is the right word. Other plates in the set drop it.
The plate says the gap is smaller on clerical speed tests.
.jpg)
A speed bar is shorter than a reasoning bar.
SteelMan: That pattern is reported and fits a g gradient.
Cut: The plate's bars are unlabeled.
The plate says forward digit span shows a smaller gap than backward span.
.jpg)
Two span bars, forward shorter.
SteelMan: The pattern is reported.
Cut: The plate's bars have no numbers.
The plate says the black-white gap is not confined to verbal items.
.jpg)
A verbal bar and a spatial bar both differ.
SteelMan: Both differences are reported. The verbal gap is often larger.
Cut: The plate's equal bars overstate the match.
The plate says the Hispanic mean falls between the black and white means.
.jpg)
A middle bar.
SteelMan: That order is the usual NAEP and IQ report.
Cut: The plate's bar is unlabeled. Hispanic samples differ by national origin.
The plate says Native American means are closer to the white mean on some tests and not on others.
.jpg)
A split bar.
SteelMan: The pattern is reported and varies by test and nation.
Cut: The plate's split is not a table.
The plate says the gap is a finding and the cause is not assigned.
.jpg)
A known stamp and an open stamp.
SteelMan: That split matches the APA statement.
Cut: The set's earlier plates assign the cause the frame plate leaves open.
The plate says the gap is treated as unprintable rather than as a table.
.jpg)
A table is stamped do not cite.
SteelMan: A publication norm is not a measurement.
Cut: The plate equates silence with a false table. Silence is not a result.
The plate says the curve describes a distribution and does not license a judgment of a person.
.jpg)
A curve sits under a crossed-out verdict.
SteelMan: Overlap makes a person-level verdict invalid.
Cut: The plate's rule is sound. It does not identify the cause of the mean.