#### by John Rolph

I was a statistics graduate student at Berkeley in the mid-1960s. It was my good fortune that David Blackwell taught the inference course to the first-year students in my cohort. I was so taken with him and his teaching that I subsequently enrolled in virtually every course he taught. They embraced a dizzying array of topics, ranging from information theory to game theory to seminars on dynamic programming, bandit problems, search, and related topics. In his teaching as in his research, his interests and knowledge were broad.

David Blackwell was a teacher without parallel. I was particularly impressed by how he could make truly deep concepts transparent — even to the beginner. Indeed, the depth was such that understanding sometimes blossomed mysteriously and gradually. I recall going over class notes after class and only then beginning to understand how deep and subtle some of the concepts he had presented were. To make sure we understood the ideas, a group that included Steve Stigler, Bruce Hoadley, and me made a practice of assigning two to take notes and one to listen intently to his lectures. We would meet afterward, reconstructing the lectures to make sure we all actually understood the concepts and results he covered.

Unlike most faculty members who specialized in one or perhaps two areas, Blackwell was a man of remarkably diverse interests; thus his students worked on a wide variety of topics. While I was there, he supervised dissertations in information theory, dynamic programming, game theory, Bayesian inference, search theory, probability theory, stochastic processes, and stopping rules. He moved from student to student and hence topic to topic with astonishing flexibility and focus.

To say he was a quick study is a remarkable understatement. When you met with him, he would look over what you had done, ask a few probing questions, suggest an approach or two to your current problem, and send you on your way. These sessions typically ran ten to twenty minutes. Indeed, when I finally brought him a draft of what I hoped was my completed dissertation, he read it, asked a few penetrating questions, listened to my responses, then told me I was finished — all in a half hour! This modus operandi obviously worked splendidly — he supervised sixty-five doctoral students during his career.

Although David Blackwell had strong beliefs and points of view, he was not an avid proselytizer. Indeed, it was only gradually that I came to understand and appreciate that David Blackwell was the lone Bayesian in the Berkeley Statistics Department. Interestingly, the content of the first-year inference course he taught us was very similar to the course Erich Lehmann customarily gave — Bayes estimates only came up as a convenient mathematical concept, not as a philosophy of inference. That Blackwell was not a vociferous proponent of his belief in the superiority of the Bayesian approach was typical of his nonconfrontational way of interacting with his colleagues.

David Blackwell was a man we shall all remember with great respect and affection. He was a man of modest demeanor who would solve seemingly intractable problems with mathematical rigor, elegance, and transparency. The impact of his research was both substantial and broad. And he treated his students as equals, both helping them and challenging them to be successful in their research. He was a person of extraordinary character and ability; it was a privilege to know him and to learn from him.