Department News

Remembering Herman Chernoff (1923–2026)

Herman "Chairman" Chernoff

We lost Herman Chernoff on July 6th. He was 103. His remarkable contributions to our field included work on large sample theory, experimental design, sequential analysis, presenting statistical data in visual form, and statistical decision-making. His enthusiasm for mentoring was almost equally well known. He proposed the creation of an annual New England Statistics Symposium to support young researchers and, in honor of his contribution to the symposium and profession in general, the New England Statistical Society established the Chernoff Excellence in Statistics Award in 2019.

His career began in 1943 when he earned a BS in mathematics with a minor in physics from City College, New York. For a year and a half afterwards, he worked as a physicist with the US Navy, building and fixing electronics. According to John Bather in “A Conversation with Herman Chernoff” (Statist. Sci. 1996), Chernoff’s use of statistical ideas in the Navy led him to pursue a master’s and PhD in applied math at Brown University, where he was supervised by Abraham Wald. Chernoff subsequently held faculty positions at the University of Illinois (1949–52), Stanford (1952–74), MIT (1974–85), and Harvard (1985–1997).

He worked briefly as a research instructor at the Cowles Commission for Research in Economics at the University of Chicago, then taught at the University of Illinois at Urbana, rising to associate professor before a visiting appointment brought him to Stanford in 1951, where he was hired permanently in 1952. He stayed for the next 22 years, rising to full professor in 1956 and later chairing the Statistics Department.

It was at Stanford in 1952 that Chernoff introduced what probabilists now call the Chernoff bound. The idea proved far more durable than its original application in hypothesis testing: generations later, it’s become a standard tool in computer science and electrical engineering, used to guarantee that the error rate of an algorithm stays acceptably small. Chernoff kept returning to the mathematics of decisions made under uncertainty for the rest of his career. In 1959 he distilled his broader approach to statistical reasoning into Elementary Decision Theory, co-written with Lincoln Moses. He went on to work out what became known as the Chernoff–Savage theorem in nonparametric statistics and to describe a probability distribution that also carries his name.

Looking back on that body of work decades later, Chernoff pushed back on how colleagues categorized him: “People regard me as a theoretical statistician, but I’ve decided in recent years that I’m really an applied statistician,” he said in a 100th-birthday interview with Harvard professor Xiao-Li Meng. “My theoretical insights have relied upon my work in thinking about applied problems.” An example of Chernoff taking inspiration from applied problems was the creation of Faces, a data visualization tool he developed to help researchers analyze multivariate data, while at Stanford in 1973. These displays let a viewer perceive at a glance embedded patterns that columns of figures could not easily convey, on the premise that the human eye reads a face faster than it reads a table. Printed in a grid, dozens of faces at a time, the technique let an analyst scan a spreadsheet’s worth of cases the way a person scans a crowd. Applications extended well beyond statistics, into political science, geology, and marketing, wherever researchers needed to compare many data points at once.

Chernoff left Stanford for MIT in 1974 to become professor of applied mathematics, then joined Harvard’s Department of Statistics in 1985 until retiring in 1997. In recognition of his many significant contributions, he was elected fellow of the Institute of Mathematical Statistics, where he was elected president in 1967, and the American Statistical Association, receiving their Samuel S. Wilks Memorial Award in 1987. Chernoff was a fellow of the American Academy of Arts and Sciences and the National Academy of Sciences. He received honorary doctorates from Ohio State University, the Technion in Israel, Rome’s La Sapienza, and the University of Athens, and in 2012 he was named one of the inaugural fellows of the American Mathematical Society.

His name persists in the vocabulary of two very different fields. The Chernoff bound remains a standard citation in randomized algorithms and machine learning, a piece of 1952 mathematics still doing daily work in software most of its users have never heard of. Chernoff Faces are still taught in data visualization courses and reproduced in textbooks as an early, stubbornly memorable argument that a chart should be built for the eye that reads it, and Elementary Decision Theory remains in print, still assigned to students learning to reason formally about uncertain choices, more than 60 years after he wrote it.

Sources: