Cult of Personality

Paul's case is that the major personality instruments were built by people without relevant training, on theoretical foundations that were never validated, and that they persist because they are useful commercially and satisfying personally rather than because they work.

Myers-Briggs was constructed by a mother and daughter with no psychological credentials from Jung's untested typology; the Rorschach was a set of inkblots its creator died before validating.

The most-used instrument in corporate life was built by two people with no training in psychology.

Devmark’s reading — not a quotation from the book

She works through test-retest reliability — a substantial fraction of people get a different Myers-Briggs type weeks later — and through the absence of predictive validity for job performance.

The book also covers the Minnesota Multiphasic Personality Inventory and the use of tests in custody cases and criminal justice, where the stakes make the weak validation considerably more serious.

Paul is clear that personality research itself is not the target. The five-factor model has genuine empirical support; it is simply not what the commercial instruments measure.

The book is from 2004 and the industry has grown since, which has made it more relevant rather than less. It is the standard sceptical account.

Personality science is fine. The commercial instruments are not measuring it.

Devmark’s reading — not a quotation from the book

Paul's answer to the obvious question — if the evidence is this weak, why is the industry this large — is that the instruments sell something the evidence is irrelevant to.

For an organisation they provide a defensible-looking record for decisions that were made on other grounds. For an individual they provide a flattering and non-falsifiable account of the self. Neither purpose requires the test to predict anything.

Paul's most damaging chapter traces the most widely used personality instrument to a mother and daughter with no training in psychology, who built it from a reading of Jung and marketed it through a personnel office during the Second World War. The origin is not an argument against it, and it explains a great deal about its construction.

It was assembled by two people with no psychological training, from one reading of Jung.

Devmark’s reading — not a quotation from the book

The core technical objection she reports is that a substantial proportion of people retake such instruments within weeks and receive a different type, which is fatal for a classification system: a measure that does not return the same answer cannot be measuring a stable property.

A related and more fundamental problem is that the underlying traits these instruments measure are distributed continuously rather than in clumps, with the largest number of people clustered around the middle of each scale. Sorting a continuous distribution into binary categories therefore cuts directly through the densest part of it, so two people scoring one point apart receive opposite labels and are told they are different types.

The chapters on inkblot tests and similar projective instruments document decades of courtroom testimony, clinical diagnosis and child custody decisions made on the basis of evidence with very weak demonstrated validity, alongside sustained professional resistance to admitting as much. This is the book's strongest and most serious material, because the consequences were immediate, irreversible and fell on people with no say in the matter.

Paul's explanation for the persistence of these instruments is institutional rather than scientific. They supply a defensible-looking, quantified basis for decisions that would otherwise have to be admitted as somebody's judgement, they are extremely profitable to license, and a large certification and consulting industry depends entirely on them. Accuracy, she argues, is simply not the property being purchased.

Paul does not claim that personality is unmeasurable or that the field is worthless, and it is worth being clear about that because the book is often summarised as though she does. The five-factor model has substantially better test-retest reliability and predictive validity, and it is the accepted research standard. It is also, notably, not the instrument most workplaces have bought. Her target is bad instruments sold with unearned confidence.