The perfidy of percentiles

Standard texts in neuropsychology, forensic, and educational psychology recommend the use of percentile rank scores (PRs) in reports and in oral “feedback” on the grounds that percentiles are easily understood by nonpsychologists. This study tested that assumption, testing predictions that errors would be made consistent with misunderstanding PR values as units of equal intervals. Four hypotheses about errors in interpretation were tested using a 12-item task to assess third-year psychology undergraduates' estimates comparing PR scores against the familiar metric of IQ. All predictions of significant asymmetrical and systematic errors of interpretation were supported. Even psychometrically educated subjects grossly misinterpret the meaning of PRs. Commonly recommended graphical display formats may unwittingly enhance these errors. Implications are most significant for forensic neuropsychology. Apparently low PRs representing cognitive performance within the Average range will typically be misinterpreted to mean significant impairment and, thus, may distort compensation and personal injury awards. Conversely, apparently high PRs representing offender recidivism risk within the Average range will be misinterpreted as high risk and impede release.