In November, 1949, all students in Princeton High School were given a sheet of paper divided by a horizontal line across the middle of the page. The top half bore the heading: "Something that happened recently that made you like a person better." The bottom half bore the heading: "Something that happened recently that made you like a person less." 718 favorable incidents and 669 unfavorable incidents were reported, or 1387 in all, of which 39 had to be discarded because they did not indicate clearly enough what behavior was involved. Very often an incident involved more than one type of behavior and had to be classified in two or more categories. Thus, the 697 favorable incidents that were used yielded 927 specific behaviors, and the 651 unfavorable incidents that were used yielded 965, or a total of 1,892. These behaviors were classified independently by two different people until each had a list of headings under which he could classify all the incidents he had examined. The two lists were than compared, argued about, and finally combined into one. To find out whether the new headings were meaningful and distinct, the investigators then took a fresh sample of 324 incidents and classified them under these new headings independently. They agreed in 91 per cent of their classifications without consulting one another, and they agreed in all their classifications after consultation.