Are IQ Tests Legal in Hiring? What the Rules Actually Say

Ask around an HR department whether you can give job applicants an intelligence test and someone will tell you it is illegal in the United States. It is one of the most durable pieces of workplace folklore there is, it is repeated in good faith, and it is wrong. What the law did was attach a condition to the practice. The condition is demanding enough that many employers gave up rather than meet it, and over fifty years the giving-up got remembered as a ban.

This is an explanation of where the rule came from and what it says, not legal advice. Employment law is jurisdiction-specific and fact-specific, and anyone actually designing a selection process needs a lawyer rather than an article.

What Griggs actually decided

Griggs v. Duke Power Co. was decided by the United States Supreme Court in 1971. Duke Power had required, for transfer into its better-paid departments, a high school diploma and passing scores on two general aptitude tests. Neither requirement had been shown to relate to the work in those departments, and both screened out Black applicants at a substantially higher rate.

The Court held that Title VII of the Civil Rights Act of 1964 reaches practices that are neutral on their face and neutral in intent, if they operate to exclude a protected group and cannot be shown to be related to job performance. Good faith is not a defence: what matters is the consequence of the practice, not the motive behind it. That doctrine is what employment lawyers mean by disparate impact.

The opinion was careful about testing specifically. It said in terms that nothing in the Act forbids the use of testing or measuring procedures — what the Act forbids is giving them controlling force when they have not been shown to measure the person for the job rather than the person in the abstract. That sentence is the whole difference between a condition and a ban, and it is the sentence that dropped out of the folklore.

There is a detail worth noting for anyone who follows this section. The two instruments Duke Power used were the Wonderlic Personnel Test and the Bennett Mechanical Comprehension Test. The Wonderlic is not a historical curiosity — it is still administered, and it is still routinely converted into an IQ-style figure by people who should not be converting it. We went through what a Wonderlic score can and cannot be translated into separately.

The rules that grew out of it

Four years later, Albemarle Paper Co. v. Moody (1975) pressed on what "shown to be job-related" requires, and found the employer's validation evidence in that case inadequate. In 1978 the federal enforcement agencies issued the Uniform Guidelines on Employee Selection Procedures, codified at 29 CFR Part 1607, which remain the operational reference. The Guidelines set out two things worth knowing even if you never read them:

• A rule of thumb for spotting adverse impact — the four-fifths rule. If one group's selection rate is less than four-fifths of the highest group's rate, that is generally treated as evidence of adverse impact worth investigating. It is a screening heuristic, not a legal threshold, and the Guidelines say so.
• Three recognised ways to show a selection procedure is valid: criterion-related validity, evidence that scores actually predict job performance; content validity, evidence that the test samples the job's real tasks and knowledge; and construct validity, evidence that it measures a trait the job demonstrably needs.

Adverse impact is not itself unlawful. It shifts the burden. Once it is shown, the employer has to produce the validation evidence, and if the employer does, a challenger can still argue that an equally effective procedure with less impact was available. That is three steps, and every step is evidentiary. It is a serious amount of work, and it is why a small employer usually concludes that an off-the-shelf ability test is not worth the paperwork — a commercial decision, not a legal prohibition.

Why employers kept using them anyway

It would be a tidier story if employers had simply abandoned cognitive testing, and many did. Plenty did not, and the reason is that general mental ability remains one of the most heavily studied predictors of job performance in the industrial and organisational psychology literature. Recent re-analyses have corrected the older validity estimates downward, and the corrected figures still leave it among the more useful signals available — typically ahead of the unstructured interview, which most organisations run enthusiastically and have never validated at all.

The position that leaves employers in is worth naming. The selection method they could defend with evidence is often the one their lawyers are nervous about, and the method they use without a second thought is the one they could not defend if anybody asked. None of which is an argument that a test score should decide a hire: prediction that holds on average across thousands of people is entirely compatible with wide variation for any one person, and a process resting on a single number discards everything a work sample or a structured interview would have added. We set out what the score-to-performance relationship actually looks like, and how much spread sits around it, in what a workplace test score predicts.

The thing that is genuinely banned

There is one practice around cognitive testing that American law prohibits outright, and it is not the testing. Section 106 of the Civil Rights Act of 1991 made it unlawful, in connection with employment-related tests, to adjust scores, use different cutoffs, or otherwise alter results on the basis of race, colour, religion, sex or national origin. The practice it was aimed at is usually called race-norming: reporting an applicant's standing within their own demographic group rather than against everybody.

The irony is worth sitting with. The provision most people half-remember as "IQ tests are illegal" is in fact the provision that requires everyone to be scored against the same table. Whatever you think of that as policy, it points in the opposite direction to the myth.

Outside the United States

The American framing does not travel. In the United Kingdom, the Equality Act 2010 handles the same problem through indirect discrimination: a provision applied to everyone that puts people sharing a protected characteristic at a particular disadvantage is unlawful unless the employer can show it is a proportionate means of achieving a legitimate aim. The vocabulary is different; the underlying demand — justify it against the job — is recognisably the same.

Across the European Union the pressure has shifted onto the automation around the test rather than the test itself. The General Data Protection Regulation constrains decisions taken solely by automated processing where they produce legal or similarly significant effects, and the EU AI Act treats systems used in recruitment and selection as high-risk, with obligations that phase in over several years. Our piece on what the AI Act means for hiring assessments follows that timetable. A scored aptitude test that a human then interprets sits in a different place from an automated ranking that nobody reviews, and that distinction is doing most of the work.

What this means if you are the one being tested

• A cognitive test in a hiring process is, in most jurisdictions, lawful. If it screens out a protected group at a materially different rate, the employer needs evidence it is about the job.
• Ask what the test is and what the score is used for. A reputable process can tell you the instrument, whether the score is a cutoff or one input among several, and whether it is combined with anything else.
• A score is not a verdict. Cognitive ability tests predict job performance on average across many people, with wide spread around that average for any individual — we set out what the relationship actually looks like in what a workplace test score predicts.
• Practice effects are real and they are not cheating. Second scores on the same or similar tests run higher for reasons that have nothing to do with ability, which is a measurement problem for the employer rather than a moral one for you.
• If you need an accommodation, ask early. Timed tests are the common flashpoint and the request is routine.

If a number has already been handed to you, the first useful move is arithmetic rather than argument: establish which instrument produced it and what standard deviation that instrument uses, then convert it. Our percentile calculator turns a score into a rank once you supply the scale, and the score converter moves a figure between the common scales so that two results from two employers can be compared at all. If you want a baseline of your own to reason from, our assessment reports the scale, the percentile and the confidence range together — which is the format an employer-supplied figure should arrive in too.

The useful correction is small. The rule was never "you may not test". It was "you may not use a test you cannot connect to the job", which is a defensible thing to ask of anyone making a hiring decision, and which sounds a great deal less dramatic than a ban. Fifty years of retelling turned a condition into a prohibition, and the prohibition is now quoted more often than the case. That is worth remembering the next time a workplace claim arrives with no citation attached.

Common questions