by Stephen Stigler
David Blackwell’s research work places him in the pantheon of twentieth-century probability, game theory, and statistical inference, but it is as a teacher that I best recall him. To hear Blackwell lecture was to witness a master of the art. He was not charismatic; he spoke slowly and deliberately, and, when not writing on the board, used slight hand gestures with his palms toward the audience, to conjure up a shape or an entire space in our minds. His mastery came from the way he was thinking through the material with us at our speed and making his thoughts our thoughts. With simple gestures he could create an infinite-dimensional space in our minds and let us see with startling clarity how a result could follow — or, more accurately, how it would be absurd that it could fail to follow. It was magical — but it was lasting magic, since the knowledge imparted remained with us.
Even in a classroom the work he presented acquired a new flavor in the process. When he presented a wonderful generalization of von Neumann’s minimax theorem designed for statistical games in function spaces, he named it after the author of an article he cited, but when I consulted that article later I could see that Blackwell had without comment recast it in a new form; the sparkling clarity of that form was a hallmark of this extraordinary mathematician’s mind and style.
For a few years in the 1960s Blackwell taught an extremely elementary Bayesian statistics course to a very large audience of undergraduates. The book he wrote for that class is almost unknown today, but in his hands it was a gem. The insights he brought to a tired old syllabus were a wonderful reward to the students and teaching assistants alike. At the end of the term he invited the large team of TAs to his home at 5 PM and served soft drinks and a large pitcher of martinis. Many of the foreign TAs were new to this libation, and some took to it too enthusiastically, but the gracious host reined them in, and all felt the warmth of his collegial fellowship.
Blackwell was a second-generation beneficiary of an early 1930s grant from the Carnegie Corporation to Harold Hotelling at Columbia University. In 1932 Hotelling had supported Joe Doob, fresh from a Ph.D. in mathematics at Harvard and with no good job prospects, and introduced Doob to probability and statistics. A decade later Doob did the same for Blackwell at the University of Illinois. No Carnegie money was ever better spent.
