What most clinicians do when they receive a laboratory report is, of course, to look up the normal range for the tests in question. … Traditionally, a normal range is calculated in such a way that it includes 95% of the results found in a group of normal or healthy persons, and, consequently, there is a 5% risk that a healthy person will present with an abnormal laboratory result. Then, imagine that you do ten tests on a normal person. In that case the risk that at least one of these tests is abnormal is (1 – 0.95¹⁰) which amounts to 0.40 or 40%. If you do twenty-five tests (and that is not uusual in clinical practice), this chance is 72%! As Edmond A. Murphy puts it so aptly, ‘Therefore, a normal person is anyone who has not been sufficiently investigated.’

— Henrik R. Wulff, Stig Andur Pedersen, and Raben Rosenberg, Philosophy of Medicine: An Introduction, 1990, citing Murphy’s The Logic of Medicine, 1976