Psychologists in the civil rights struggle: validity as reform, 1954–67

Though the origins of psychometric testing can be traced back to anthropometry and eugenics, it was in the First World War that testing exploded, when the ‘Army Alpha’ and ‘Army Beta’ aptitude tests were used to determine job classifications and leadership potential during mass military recruitment. Between the wars, tests were increasingly seen as guarantors of fairness and became widely used in industry. The post-war GI Bill and expansion of public education saw the rise of the SAT, or Scholastic Aptitude Test, as well as interest in using tests to assess personality.Footnote ²⁰

However, there was a broad spectrum of opinion on how and what to test. Fallout from the Supreme Court’s unanimous ruling against state-mandated school segregation in Brown, often remembered for Kenneth and Mamie Clark’s doll test, revealed the divide within the field. Segregationists revived arguments linking race and intelligence and invoked the spectre of ‘race suicide’ through miscegenation, while liberal integrationists were drawn to the measurement of personalities in their efforts to uncover the origins of race prejudice.Footnote ²¹ While this divide was clearly political, it reflected epistemological growing pains. With so many different tests in circulation based on different theories, how could one know what was being measured?

To figure this out, the APA did as most professional organizations do: it put together a committee. A panel of education researchers synthesized ideas about how to assess tests, eschewing a broader conversation about disciplinary ethics in favour of a set of technical guidelines for validation, one that tacitly articulated a view of what was scientific about psychology.Footnote ²² Lingering in the background were concerns about biased measurement (for example, a baseline determined by testing only white men) and that widely used tests embodied ‘cultural bias’, selecting for background or education rather than intelligence itself. The technical report, published as a supplement in 1954, outlined information standards for test users, describing three extant kinds of validity – content, concurrent and predictive – and proposing a fourth – construct validity – a measure of the internal consistency of a theory or concept.Footnote ²³ Psychologists were confident that construct validity might one day guide them toward tests that were truly ‘culture-free’, but critics of testing emboldened by the Black freedom struggle remained focused on the practical effects of aptitude tests as barriers to jobs and education. So it became predictive validity, the relationship between test scores and measured outcomes, that most directly addressed the political issues around testing.

Things came to a head after the Illinois Fair Employment Committee pursued a complaint against the Motorola Corporation from Leon Myart, a Black army veteran who had responded to a job advertisement for a television set analyser at a plant west of Chicago in 1963. Although Myart remembered passing the ability test he was given, he filed a complaint after hearing nothing for two weeks that led to a hearing when the company refused the committee’s request to hire him. At the hearing, Motorola was unable to produce Myart’s test, claiming that they destroyed all tests after recording the results on an IBM card, and brought only a witness testifying that he had scored below the cut-off. The hearing examiner found this explanation wanting and ruled against Motorola in February of 1964. Although Myart never claimed that the test itself was flawed – after all, he had passed – in his decision the examiner opined that the test ‘in the light of today’s knowledge, is obsolete’, with a norm ‘derived from standardization on advantaged groups’, and insensitive to the effects of ‘background factors’ on scores. This hit a nerve. A cunning Motorola executive devised a smear campaign charging that the committee had imposed an unlegislated concept of preferential racial hiring at a lower standard.Footnote ²⁴

The timing could hardly have been worse. Word of the decision got out as the 1964 Civil Rights Act was on the Senate floor amid a filibuster by southern Democrats. Title VII, the Act’s provision barring employment discrimination, was a bugbear for Republicans, who leveraged the obstruction to broker a series of compromises to weaken the bill. In response to the Motorola case, Republican John Tower introduced an amendment specifying that it was ‘not unlawful for an employer to give and to act upon the results of any professionally developed ability test provided that such test … is not designed, intended or used to discriminate’.Footnote ²⁵ The tension around discriminatory testing was so pronounced that one lawyer’s willingness to weigh in on social science as the examiner in a public hearing, with the help of some ill-intended publicity, put psychometrics into sweeping civil rights legislation.Footnote ²⁶ Though intended to protect tests, Tower’s amendment was a blessing in disguise for lawyers seeking to test the language of the Act.

Title VII came along with a body to enforce it, the EEOC, but the commission was deemed a ‘toothless tiger’ from the get-go.Footnote ²⁷ It could receive complaints and attempt to resolve them with employers, but early commissioners invested heavily in research, reporting and policymaking. In collaboration with the APA, the EEOC developed its own Guidelines on Employment Testing. The only explicit mention of bias in the guidelines invoked differential validity: ‘Only a test which has been validated for minorities can be assumed to be free of inadvertent bias.’ A state-level employment commission recommended asking for validation data as the first recourse in handling a complaint, noting, ‘In full recognition of the [technical] difficulties it is nonetheless recommended that situational validity be enunciated as the ultimate criterion of fairness.’Footnote ²⁸ Situational validity was not part of the APA’s recommended typology (predictive, concurrent, content, and construct), but denoted a best practice: companies needed to validate a test for a given job in a given locale, not simply point to a study in the published literature or conducted elsewhere.

Even if the new fair employment measures spelled good business for psychologists, in the late 1960s the APA was deeply divided over the question of psychology’s social responsibility. Following King’s address and the Division of Evaluation and Measurement’s resolution, pressure only continued to build. Gathered to investigate the causes of uprisings in US cities, President Johnson’s National Advisory Commission on Civil Disorders decried aptitude tests, high-school diplomas and criminal background checks in its bestselling Kerner report as ‘unnecessarily high minimum qualifications [which] often [had] the same prejudicial effect’ as outright discrimination.Footnote ²⁹ At the 1968 APA meeting, marked by the tragedy of King’s assassination, a group of more than two hundred Black psychologists formed their own association, citing their parent group’s failure to acknowledge the potential of the Black Power movement, rather than liberal social science, to address the persistence of racism. They asked leadership to endorse the Kerner report and declare a moratorium on test development ‘pending the thorough review and reassessment of the issue on the highly questionable validity of these measures’.Footnote ³⁰ Even as the proposal challenged the neat separation of the technical and the ethical that the APA sought to maintain with its testing recommendations, validity liberalism offered the nascent Association of Black Psychologists an internal language for challenging the norms of their discipline without upending the entire apparatus.

The view from the trenches

What would a ‘culture-free’ test look like? How did psychologists use validation methods to evaluate tests? Moreover, how did they understand the relationship between validity and fairness? The thought and practice of EEOC staff psychologist William H. Enneis and Richard S. Barrett, the industrial psychologist tapped as an expert witness in Griggs, shed some light on these queries.

Enneis was no opponent of testing, but he found the overuse of general intelligence tests for screening entry-level and menial positions like janitors and dishwashers bordering ‘on the absurd’.Footnote ³¹ In a 1967 hearing, Enneis underscored job-relatedness as the EEOC’s gold standard: carefully analysing the requirements of a given job would, in most cases, indicate that there was no need for pencil-and-paper intelligence tests like the widely used Wonderlic Personnel Test, made up of questions that gauged vocabulary and so-called ‘common’ knowledge – the relationship between similar-looking words (e.g. adopt/adept) was one common question type. If a position did seem to warrant the use of a test to screen applicants, Enneis cautioned against merely providing a rational justification for its use. Rather, employers needed to assess how well a test predicted success on the job, a practice followed by fewer than half of companies using them. For Enneis, validity was precisely the ‘relationship between test scores and standards of job performance’. He also introduced the idea of differential validity between minority groups, which raised important questions about how tests predicted differently – not necessarily better or worse – for various groups. He was confident that pursuing this ‘truly amazing’ line of research beyond the ‘topical variables of race, color, or national origin’ would reveal in time that these were not the significant ones.

Although educational psychologists had been grappling with the implications of testing disparities since before Brown, industrial psychologists had largely ignored the issue before the passage of the 1964 Civil Rights Act. Between 1965 and 1967, under the aegis of a Ford Foundation grant, a group of New York University psychologists conducted studies of differential validity on minority groups in five work situations. The resulting technical report was so popular that it was republished the following year as a monograph, in which the authors posited a simple maxim: ‘Validation irrespective of race may lead to discrimination; separate validation by ethnic group will not.’ Though the Ford data were not conclusive, investigations into test design and evaluation exploded. The US Civil Service Commission partnered with Educational Testing Service (ETS), which oversaw the administration of the SAT, to assess the problem of differential prediction in its General Aptitude Test Battery. Whether or not experts believed that tests could be biased, differential validation raised the bar for evaluation.

Later, Enneis clarified that the EEOC was less concerned with the design of tests than with their implementation, and that criticizing tests as inherently biased was no more rational than claiming them to be ‘culture-free’ or inherently fair based on ‘superficial psychometric considerations’. He also spoke to concerns about charges of ‘reverse discrimination’ should an employer be compelled to hire minority applicants with lower test scores. Predictive validity provided a technical fix:

This approach vested confidence in the use of quantitative performance evaluations to normalize test scores over any concern that judgment of a worker’s value could be laden with biases or impacted by a hostile workplace. Validity liberalism was thus marked by a faith that shifting test standards would help unsettle workplace hierarchies, rather than providing cover for them.

Test scores drew far more scrutiny than performance ratings, even though they were yoked together by the concept of predictive validity. One of the lead investigators in the Ford Foundation study was Richard S. Barrett, an industrial psychologist brought in by the NAACP Legal Defense Fund (LDF) to testify in Griggs and ultimately in more than a hundred fair-employment cases. In the course of this work, he prepared a set of correlation diagrams to illustrate what the adverse impact of tests looked like in relation to performance evaluations. Skewed ellipses indicated validity, their non-overlap a gauge of adverse impact. Though Barrett admitted that the ‘definition of fairness is not yet settled’, he labelled a diagram that showed identical correlation between both groups’ ratings and scores ‘fair’.Footnote ³³ Another diagram portrayed the common scenario that testing experts imagined: Black employees whose performance evaluations were roughly equal to those of their white co-workers, but with lower mean test scores. The simple solution was ‘to use a lower passing score’ for Black applicants, which would benefit employers by helping them select better performers.

Validity was necessary but insufficient to achieve fairness; fairness alone could not satisfy President Johnson’s plea in his 1965 commencement address at Howard University for ‘equality as a fact and equality as a result’, not merely as a theory. Nonetheless, validity denoted a body of technical expertise, territory that gave civil rights advocates the upper hand in the battle over a legal definition of fair employment.

Defining discrimination: validity as regulation, 1967–72

One limitation that the early EEOC contended with was its inability to pursue lawsuits, at least until the Equal Employment Opportunity Act of 1972. The early lawsuits that shaped Title VII therefore depended on close collaboration with the Department of Justice and civil rights organizations like the LDF, which had relied heavily on social scientists in the school desegregation cases. According to one staffer, the LDF devoted nearly half of its resources to employment litigation in the late 1960s and maintained a close working relationship with the EEOC, which filed amicus briefs in what even judges admitted were ‘David and Goliath confrontations’.Footnote ³⁴

In 1967, the EEOC received a complaint from janitors at a North Carolina power plant that remained segregated, with all Black employees confined to the labour department and paid less than half the wages of white employees throughout the plant. When the Civil Rights Act went into effect in 1965, the company adopted a formal transfer policy: a high-school diploma and a sufficient score on the Wonderlic as well as a mechanical-comprehension test. Most of the janitors had no diplomas, and the tests themselves seemed to disincentivize transfer requests. With Census data on high-school graduation rates by race and the promise of a psychological expert, the LDF brought the case, named for lead plaintiff, Willie Griggs, to the courts. They argued that even though the tests were ‘professional’, they served as ‘built-in headwinds’ that ‘froze’ employees into the same old patterns as before the Civil Rights Act.

It was an argument in need of an expert. At first, LDF lawyer Robert Belton reached out to Kenneth Clark, who promptly declined and recommended, among others, educational psychometrician Robert L. Thorndike, son of famed behaviourist Edward Thorndike. After the district court trial was rescheduled multiple times, Thorndike failed to materialize, and Belton ended up calling in Barrett, who was juggling multiple positions at the time. Firm in his conviction that the worth of a test should be measured by performance evaluations rather than vague appeals to competitiveness, Barrett performed well on the stand, despite the tense atmosphere. According to Belton, the court sustained an objection when Duke Power’s lawyer asked if, as an alternative to differential validation, ‘minorities should raise their standards, because industry can’t afford to relinquish [theirs] in the competitive world of today’.Footnote ³⁵

Though sympathetic to the case’s merits, judges were loath to endorse the LDF’s argument that the ‘use of tests and educational requirements [was] but one example of the new breed of racial discrimination’ when Title VII barred only tests adopted with the deliberate intent to discriminate.Footnote ³⁶ Nonetheless, the Supreme Court eventually agreed to hear Griggs, the first hearing on the testing issue. In anticipation, the EEOC issued new guidelines specifying that a test was not ‘professional’ merely because it ‘was prepared by an individual or organization claiming expertise in test preparation’. The guidelines even included a new section explicitly endorsing differential validity ‘where technically feasible’. Still, Griggs seemed like a long shot. Few Duke Power employees had attempted the tests, and the company was willing to subsidize education for the remaining plaintiffs to get their high-school diplomas. An EEOC lawyer even drafted a letter to the LDF suggesting that Griggs presented ‘a most unappealing situation for finding tests unlawful’, urging that they wait on a more favourable case in the works with better data on test score disparities. However, he had a change of heart and never sent the letter.Footnote ³⁷

In a unanimous 1971 decision, the Court interpreted Title VII as applying to ‘the consequences of employment practices, not simply the motive’. In other words, it recognized that a racially disparate impact of an otherwise ‘neutral’ policy was unacceptable. Just as importantly, the decision enshrined the EEOC as the legal authority on the fair use of tests and employment requirements: ‘Diplomas and tests are useful servants, but Congress has mandated the commonsense proposition that they are not to become masters of reality.’Footnote ³⁸

It is difficult to overstate the importance of Griggs and disparate-impact theory for civil rights advocates. The decision was celebrated by many as giving sanction to affirmative-action policies of compensatory hiring to prevent lawsuits. As discussed above, former EEOC chief of conciliations Alfred Blumrosen joked that the decision rendered Title VII akin to ‘a full-employment act for industrial psychologists’.Footnote ³⁹ But psychologists did not all see it that way. The narrative quickly got away from aptitude tests as lawyers interested in legal remedies for what Charles V. Hamilton and Kwame Ture (Stokely Carmichael) termed institutional racism appeared within reach.Footnote ⁴⁰ Companies and governments began ditching tests – the Wonderlic in particular – to avoid liability. Meanwhile, Black Power activists pushed the city of San Francisco to declare a moratorium on student testing.Footnote ⁴¹ Many perceived these developments as a threat to psychologists’ livelihoods.

‘Rational’ validation

The Griggs litigation began in February of 1968, weeks before President Johnson ended his campaign for re-election. With a liberal Democrat having overseen the passage of the 1964 Act and the original EEOC appointments, the election of Richard Nixon signalled changes to come in federal civil rights policy. Under John W. Macy Jr, a liberal administrator who nonetheless oversaw the Lavender Scare persecution of gay federal employees, the Civil Service Commission (CSC) pursued validation studies with ETS.Footnote ⁴² Macy had cooperated with the EEOC, but, after he resigned in 1969, his successor took things in a different direction. In the aftermath of Griggs, state agencies ceased using the Wonderlic. As the government’s largest test administrator, the CSC recognized that it needed to ensure that its Federal Service Entrance Examination (FSEE) was up to par and announced an overhaul later that year. A subsequent lawsuit revealed that more than revising the test to comply with the EEOC guidelines, the CSC had developed its own standards of evaluation.

Eight Black college graduates who had been hired into the Department of Housing and Urban Development through an internship programme sought permanent positions through the FSEE. About half of them passed, while others failed and were terminated or resigned. The group filed an injunction against the CSC, claiming that the FSEE had a disparate impact on Black applicants. A DC District Court judge threw out the case, citing the plaintiff’s lack of statistical data on the racial breakdown of pass rates, something the CSC did not systematically record. In its affidavit, the CSC took a further step, detailing its own draft guidelines for appraisal procedures. Deviating from the EEOC guidelines, the CSC proposal described a principle of ‘rational’ validation as ‘a demonstrable rational relationship to important job-related performance objectives’, which it contrasted with ‘statistical validation’ procedures it deemed infeasible and limited. The judge found this perfectly suitable even though no psychologist had endorsed ‘rational’ validity.

Lawrence Plotkin, a psychologist working with former EEOC research head Phyllis A. Wallace at Kenneth Clark’s Metropolitan Applied Research Center (MARC) in Harlem, sounded the alarm. He argued that rational validity most closely resembled the lay notions of ‘face validity’ that the APA standards were designed to counteract – whether a test was reasonably related to the job, without rigorous evaluation. He called this ‘regression to unscientific’ standards ‘an attempt to undercut the Griggs decision’.Footnote ⁴³ EEOC chairman William H. Brown III chastised the CSC head for deviating from the federal coordinating council, writing, ‘I fear that almost any test, examining procedure, standard, or employment practice could be squeezed to fit into its loose definition and proposed application’. In a memo subtitled ‘Once more to the barricades’, Wallace made the case to Clark that MARC was the organization best situated to coordinate a defence against what Plotkin called ‘a potent weapon for (perpetuating) racial discrimination’.Footnote ⁴⁴ In her previous position, Wallace had sought to use the tools of social science to create a computer system for detecting workplace discrimination. This project was sensationalized in a Wall Street Journal profile of second EEOC chair Stephen N. Shulman, who quipped that despite its technological prowess, the agency was ‘fighting an elephant with a fly gun’.Footnote ⁴⁵ With some distance from government service, Wallace – the first Black woman to receive an economics PhD from Yale and, later, to earn tenure at MIT’s Sloan School of Business – had come to realize the truth this offhand remark expressed.Footnote ⁴⁶

The Civil Service Commission’s pivot showed how pliable validity was. Judges might be willing to enjoin the use of tests developed by private industry, but undercutting a federal agency was a bridge too far. Even as the Equal Employment Act of 1972 expanded the jurisdiction of the EEOC to federal, state and local government employees, the first move had been made in what was becoming a struggle over the meaning of fairness. In the years following Griggs, validity as regulation became contorted through litigation, with psychologists on both sides of the stand.

Bias backlash: validity unbound, 1972–91

From the national level to the local, US politics during the 1970s was shaped by backlash against civil rights policies as the compact between interest groups that had secured their passage fractured. Battle lines were drawn over education: white opposition to bussing and redistricting was bolstered by an emboldened hereditarian defence of racial inequality that had been brewing since attempts to measure the effects of educational equality during the years of President Johnson’s Great Society.Footnote ⁴⁷ Some psychologists, like Richard S. Barrett, saw their careers bolstered by employment lawsuits, while others feared that the regulatory climate had compromised their scientific autonomy.

For liberal personnel psychologists, diminished reliance on testing meant that employers would have an easier time hiring better-quality employees, but members of the public readily conflated test scores and performance. For example, in a 1975 letter, American Jewish Committee lobbyist Hy Bookbinder chastised EEOC chairman John H. Powell Jr for his comments on a television programme minimizing the threat of ‘reverse discrimination’ against whites, asking him to reaffirm his commitment to the idea that employment goals for minorities would never require ‘the hiring of a person who is less likely to do well on the job (“less qualified”) over a person more likely to do well on the job (“better qualified”), under valid selection processes’.Footnote ⁴⁸ Validity was the very measure of how well a test predicted such success on the job but sceptics of civil rights policy weaponized the language as a blanket defence of aptitude testing.

With parallel debates over merit in education and new popular understandings of psychology on the table, the next phase in the legal battle over fair testing was shaped by a third factor: the redefinition of minority disfranchisement as a crime problem during the Nixon years. Political scientists and historians have shown how policies intended to supplement a more holistic Great Society became the leading edge of a new punitive agenda.Footnote ⁴⁹ Chief among these was the Law Enforcement Assistance Act which, among other things, provided steady funding to improve police training and recruitment – legislation like this, in combination with the civil rights Acts, seemed to promise full employment for industrial psychologists. But because blue-collar professionals employed by municipalities were not initially covered by Title VII, when a lawsuit about police testing did emerge, lawyers needed to make an analogy between the statute and the US Constitution’s Due Process clause.

Washington v. Davis began like any other testing case. Two Black applicants rejected from the DC police academy for failing a written verbal aptitude test – developed by the CSC, no less – sued the District of Columbia with the help of the some of the Griggs team. Since the failure rate for Black applicants was four times that for whites, they made an argument for disparate impact under the Griggs standard. On this basis, the Court of Appeals overruled a lower court finding for the police department. When the department appealed to the Supreme Court in the spring of 1975, psychologists declared their allegiances. The APA’s Division of Industrial-Organizational Psychology filed a brief in favour of the plaintiffs, arguing that the test validly predicted police academy grades, a reasonable proxy for performance on the job. ETS also filed a brief, arguing that the respondents adhered too rigidly to technical validation data. Their view, ‘perhaps paradoxically, [assigned] a less dispositive role to statistical validation evidence’, which could not ‘be substituted for basic analytical judgments about fairness and rationality’. The company responsible for the expansion of testing was arguing that civil rights groups had gone too far in demanding technical studies rather than a substantive discussion about fairness. Such rhetoric offered convenient cover for what was certainly a deregulatory agenda, but the irony resonates nonetheless: in their efforts to court scientific authority, had champions of civil rights lost control of their own narrative?

In a 7–2 decision, the Court held that the Constitution did not recognize disparate impact and instead required proof that the government intended to discriminate. It was a shock to civil rights advocates, who regarded the decision as a major setback, an attempt by the Court to stem the tide of racial-discrimination lawsuits. The majority even admitted as much, acknowledging that a constitutional recognition of disparate impact could invalidate ‘a whole range of tax, welfare, public service, regulatory, and licensing statutes that may be more burdensome to the poor and to the average black than to the more affluent white’. These words cemented Davis’s place among the most infamous decisions in constitutional law of the twentieth century. According to a leading casebook, it set the standard for proving discrimination so high as to ensure that ‘every lawsuit involving constitutional claims of racial discrimination directed at facially race-neutral rules would be conducted as a search for a bigoted decisionmaker’.Footnote ⁵⁰

The constitutional drama of Davis has largely overshadowed a key technical detail. In contrast with perceptions at the time, not to mention the text of the opinions, hardly any of the leading legal commentaries provide so much as a hint that the case was about the regulation of aptitude testing or the niceties of validation.Footnote ⁵¹ The majority’s concluding argument invoked the CSC’s revision to the Federal Personnel Manual encouraging ‘rational validation’ to affirm the district court’s endorsement of training programme grades as a validity criterion.Footnote ⁵² What Plotkin called the CSC’s gambit to undercut the EEOC’s authority over testing proved successful.

The vanishing horizon of reform

But what do the politics of validation teach us about Davis? More bluntly, could the Court have made such a sweeping ruling without testing? The latter question is largely out of bounds for the historian, but it is nonetheless instructive to follow the path trod by legal scholars and consider these testing cases against the backdrop of the Burger Court’s evolution.

The early years of the Burger Court saw it wrestling with the relationship between motive, context and effect in interpretations of the Fourteenth Amendment. This is perhaps best illustrated by Palmer v. Thompson (1971), a case argued on the very same day as Griggs that signalled a short horizon for equal protection as an instrument for structural reform. The city of Jackson, Mississippi opted to close four of its public swimming pools following a 1962 integration mandate, alleging financial difficulties, while leaving the fifth segregated by transferring ownership to its operator, the private YMCA. Palmer raised the question whether the city had a responsibility to ameliorate this racially disproportionate outcome – the Court answered in the negative.Footnote ⁵³ Decisions in the years leading up to Davis were inconsistent on these grounds, but the clear lack of professional consensus around validity and vocal backlash against affirmative action made the case a prime candidate for drawing a bright line between the new civil rights administration and the Constitution.

Another element was at play, one more particular to psychology and subtler, yet no less significant. Where the Court spoke of fair processes in Griggs, Davis was concerned solely with validity, mirroring the EEOC’s own insistence on rigorous measurement of performance and job-relatedness. While it was clear that there were many ways for a test to be unfair or biased, it became clearer still that a positive definition of fairness was elusive, ill-suited to technical approaches. Validity offered a seemingly objective solution that kept fraught matters of justice at arm’s length. However, it turned out to be just as pliable. Testing professionals embraced validation as a practice, the first step in designing fair, context-specific personnel systems. With Griggs, anti-discrimination law invited inquiry into the consequences of such systems. Context and consequences went hand in hand. Davis was effective in sealing off equal-protection doctrine from the world beyond the four corners of the page in part because the question whether tests had any place in a post-civil rights society was off the table. Litigation, backlash, and professional dissensus had narrowed the conversation about test fairness to the technicalities of validation. It was a matter of adjustment.

When President Carter appointed Eleanor Holmes Norton to chair the EEOC in 1977, her priority was getting all federal agencies to agree on a single set of guidelines after five years of competing ones.Footnote ⁵⁴ Norton was a well-known civil rights advocate who represented women at Newsweek denied bylines for the articles they wrote. In her new position, she saw herself as more a mediator than a firebrand. Norton took a far more conciliatory attitude than prior EEOC chairs toward industry concerns about ‘compliance with complex, technical and expensive validation standards’. The new inter-agency guidelines she helped to develop stated that employers could simply engage in recruitment activities or provide any kind of alternative test to avoid disparate-impact liability. This was a quiet change, but a profound one. In the 1960s, civil rights advocates and activist psychologists put their stock in validation as a tool for enforcing anti-discrimination law. But in the back-and-forth of lawsuits and the political maelstrom of the 1970s, civil rights advocates lost this tool and much of the force of law that came with it.

Wonderlic’s revenge

Just weeks before Davis was handed down, a psychology professor forwarded an advertisement to the chair of the APA’s Committee on Psychological Tests and Assessment from none other than E.F. Wonderlic & Associates, whose namesake exam had been the subject of numerous lawsuits since the late 1960s, most importantly Griggs. The advertisement claimed that the Wonderlic Personnel Test had been ‘approved’ by federal agencies, so long as an ‘ethnic conversion table’ was used.Footnote ⁵⁵ The exchange among psychologists that resulted is telling. While they chuckled at the suggestion that any test could be ‘federally approved’, the bigger issue was the implication that a test need not be validated for a given use. Some were concerned with the implications of the included ethnic conversion table, that corrected score formulas were a relic like the fantasy of the ‘culture-free’ test: ‘not all women are the same, nor are all blacks the same’. Ultimately, more than the fear that companies might try to shirk off legal responsibility by adopting plug-and-chug race corrections, the psychologists in the exchange were ‘concerned about the probability that nonprofessionals who criticize tests, testing and test publishers will generalize such behavior to all professionals involved in these activities’. In other words, it would easily be weaponized by critics of affirmative action.

Perhaps they had the right idea. The APA wrote to the company asking that the advertisement be corrected and received a response from E.F. Wonderlic himself. He claimed that the letter was an honest attempt to signal the company’s compliance, but kept on talking:

Was Wonderlic simply out for revenge? The updated letter remained steadfast in its equation of adverse-impact elimination with ethnic conversion tables. He seemed to insist that no further validation evidence was necessary because a company using quotas and cut-offs had already foregone its ‘psychological responsibilities’ to avoid becoming the target of a lawsuit. Invocations of ‘race-corrected’ scores cynically oversimplified what was mostly a good-faith effort behind differential validation studies: to denaturalize the association between test scores and job performance, and to do so without massively disrupting the testing industry or compromising psychologists’ authority. The tension between ameliorative efforts that took racial categories as pragmatic social classifications and emboldened hereditarian insistence on the immutable biological roots of racial inequality created a political and intellectual environment that allowed colour-blind ideology to flourish. As influential as claims for a genetic basis of intelligence have been in dismantling redistributive policies, psychological politics also played a part in shaping the jurisprudence invoked to do the same.

Conclusion: divergence, retrenchment and the enduring politics of measurement

In the decade following the passage of the 1964 Civil Rights Act, energy once channelled into marches and protests was redirected toward administering the ideals of equal opportunity. This article has considered how validation became a key regulatory resource for fair employment law, and how the politics of knowledge shaped legal doctrine with a lasting impact. Fissures within psychology deepened as courtroom contests ratcheted up expectations of certainty on issues that lacked consensus. The relationship between validity and fairness was a moving target; at times, they were regarded as separate measures, and at others they were conflated. The practical thrust of validity was directed toward situational specificity, which required employers to periodically re-evaluate their personnel systems, but, by the end of the 1980s, the rise of meta-analysis had led psychologists to embrace validity generalization, the idea that cognitive tests predict performance similarly across situations.Footnote ⁵⁶ Scientific understanding thus diverged from what Title VII case law required.

These developments took place amid a broad political realignment that saw old civil rights coalitions fracture and new ones – some militant, others liberal and multiracial – crop up, not to mention a resurgent right wing that championed colour-blindness.Footnote ⁵⁷ Alongside decisions like Regents of the University of California v. Bakke (1978), which championed a new value of ‘diversity’ over the non-discrimination that civil rights leaders had fought for, Davis met its moment. It shut the door that Griggs had seemed to open for a legal approach to the structural inequalities that advocates were most concerned with. In the formulation of legal scholar Alan David Freeman, Davis re-enshrined the ‘perpetrator perspective’ as the dominant frame of anti-discrimination law, one in which racism is treated ‘not as a social phenomenon, but merely as the misguided conduct of particular actors’.Footnote ⁵⁸ No doubt this was helped by Eleanor Holmes Norton’s effort to coordinate test standards by affording greater deference to employers – a trend that has unfortunately become calcified in employment law, making it harder for employees to challenge the use of new technologies in the workplace.Footnote ⁵⁹

My goal has been to show how the politics of measurement both paralleled and propelled changes in the legal order. To many psychologists and civil rights advocates, validity seemed like a lifeline in an uphill legal battle to deliver on the promise of landmark legislation. But in the tumult of adversarial litigation, they unwittingly set themselves up to be challenged on technical grounds – a Trojan Horse for a broader deregulatory agenda. Throughout this erosion of civil rights protections, validity became central to the very methodology of quantitative social science and machine learning, furnishing the very language in which arguments about racial bias in recidivism risk algorithms are conducted, for one prominent example.Footnote ⁶⁰

Historicizing validity in this way yields two further insights. Despite the unmitigated success of validity within methodology, the same issues that civil rights advocates tried to combat in the 1960s have been amplified, even transformed, in the age of artificial intelligence (AI).Footnote ⁶¹ Scholars associated with fairness, accountability and transparency efforts in machine learning and artificial intelligence have called attention to the historicity and contextual specificity attached to definitions of fairness.Footnote ⁶² But the lessons they derive fall short of a proper historical reckoning with how law and science allocate power. Of course, not everyone needs to wade through the thicket of constitutional history and interpretation, hard as it may be to ignore in the current moment. But historians of science interested in metrics and regulation need to re-examine metanarratives about the relationship between law, politics and expertise if we want to intervene in conversations about the politics of knowledge today.Footnote ⁶³ Here, historical analysis might better guide policy. Rather than calling for researchers of machine learning to recognize their own cultural biases, language that hearkens back to the 1950s or earlier, we should push for standards that account for how common-law systems shape evidence and expertise along recognizable lines of inequality. Addressing the shortcomings of validity liberalism requires building popular consensus, not merely winning in court.Footnote ⁶⁴

The episodes recounted here also reveal a largely unexplored connection between the differential-validity research promoted by the EEOC and widely decried ‘race correction’ practices in medicine today.Footnote ⁶⁵ Differential validity was a capacious concept, agnostic as to the underlying reality of the groups compared. But this same flexibility made it ripe for selective interpretation. Psychologists thought they could compel employers to dispense with the practice of deriving test norms for exclusively white, male subjects, but test manufacturers sought shortcuts like publishing racial score conversion tables – not exactly what civil rights advocates had in mind. The history of differential validity suggests that race correction draws its authority not only from entrenched assumptions about biological difference, but also from well-meaning attempts to account for social disparities that reify more than rectify. Perhaps dislodging these practices requires reactivating the more radical critiques of measurement that the brief consensus on validation took off the table.

Finally, the larger project from which this article is drawn cuts toward what we might call a critical history of scientific anti-racism, one in which legal liberalism figures prominently. Social science and statistical evidence became tools of choice for lawyers looking to shape the legal order of the post-civil rights era. Whether out of political conviction or a yen for professional advancement, experts were active collaborators, but they did not always succeed in dismantling injustice even on their own terms. Foundational work in critical race theory – which by no accident emerged from the same moment as the developments I have traced here – argued that the formal equality promised by anti-discrimination law legitimates inequality and conceals racism.Footnote ⁶⁶ Not unlike law, scientific authority has been an unstable ally in the quest for social justice, and we need histories that can hold both to account.