Abstract: We determined the frequency of antenatal corticosteroid use for mothers with threatened premature deliver in 1985, 1990, 1995, and 2000. We next compared published data to the surveyed recollections of 302 obstetricians who were practicing during these years. Two points emerged. First, published reports reveal that the use of antenatal corticosteroids increased steadily, from 8% in 1985 to 20% in 1990, 52% in 1995, and 75% in 2000 (P < .001). Second, "expert" opinions derived from the recollections of practicing obstetricians consistently overestimated the actual use of antenatal corticosteroids during the year in question - 31% versus 8% for 1985, 56% versus 20% for 1990, 78% versus 52% for 1995, and 92% versus 75% for 2000 (all Ps < .001). The use of antenatal corticosteroids by obstetricians in the past 15 years reveals a phenomenon that is widely recognized elsewhere - retrospective memories are often wrong, and when they are wrong they are not randomly wrong. Rather, recollections are systematically skewed toward an outcome that, in hindsight, is considered desirable (the "Monday morning quarterback" phenomenon). We offer a simple proposal. In determining the "standard of medical care," the legal system should rely on statistical data about doctors' performance rather than the recollections of experts about doctors' performance. The fallible memories of isolated experts are a crude second-best, far inferior to the data that they approximate. Widespread adoption of the view by professional physician organizations would dramatically increase the rationality of expert testimony in medical malpractice tort law.