as possible was the best treatment. During the period, roughly 1920-30, Claude Regaudof Paris argued that the differential effect of x-rays on cancer and normaltissues could be best obtained by giving the treatment slowly. For example, healing was very much better when skin cancer was treated over a period of a week than in one day. Originally, they used a radium applicator, strapped, say to the arm. Whenthe x-ray machine was introduced in about 1920, they fractionated the treatment - one brief exposure per day for seven days. I began training in 1949, treatment schedules had gradually been improved and become moreorless standardized, for example, treating over the course of five weeks in the case of many types of cancers. The need for relatively large numbers of patients in order to gauge the effectiveness of treatment is readily appreciated when you consider the following hypothetical example. Supposea clinic treats four hundred patients a year, that there are five principle types of cancer, and each maybeclassified in four stages. On the average, there would be twenty patients in each specific subgroup. If you wished to change the treatment, you could haveten in the new group and ten in the standard one. If you expected survival to be improved from five out of ten (fifty percent) to seven out of ten (seventy percent), it would be difficult to establish. Obviously, much larger numbers of patients and goodstatistical analysis would be required. [6. Shields Warren] BERGE: Next question was, you said that you had been brought to Harvard by Shields Warren, and I was wondering if you could talk a little bit about him. KOHN: Shields Warren waschief Pathologist at the Deaconess and Professor of Pathology at the Harvard Medical School. He becameinterested 24