Skip to content

John Chambers

Abstract

John Chambers (born 1941) spent 39 years in the statistics research department at Bell Labs and used them to build S, the language that turned data analysis from a set of programs you ran into a language you wrote in. The project started at a meeting of about five people on May 5, 1976, without management approval, and its goal, in his own words, was to turn ideas into software quickly and faithfully. S became the commercial S-PLUS, won the ACM Software System Award in 1998, and was then copied by two New Zealand lecturers as R, which made it free and made it the language of statistics. Chambers gave the prize money to endow an award for statistical software, joined the R Core Team, and spent the rest of his career improving the copy of his own work.

Toronto to Murray Hill

John McKinley Chambers was born in Toronto on April 28, 1941. He took his bachelor’s degree at the University of Toronto in 1963 and a master’s and doctorate in statistics at Harvard, finishing in 1966. He had spent the summer of 1964 as a graduate intern at Bell Labs, and in 1966 he joined the statistics and data analysis research department at Murray Hill for good.

The department was the one John Tukey had shaped, and Tukey’s 1962 paper “The Future of Data Analysis” was its charter: statistics as the practical examination of data, with computation as a central tool, decades before anyone called it data science. Its work was large by the standards of the time, rain-gauge experiments on microwave transmission that produced millions of observations, and the software was Fortran subroutines. Any new analysis meant telling a programmer what to write and waiting. Chambers’s early work was on the infrastructure: a graphics library called GR-Z with Rick Becker in 1976, and a data-management layer he wrote and never published.

May 5, 1976

The idea for S was that the analyst should type directly to the library. On May 5, 1976, about five people, Chambers and Becker among them and none of them managers, met to present some ideas and preliminary code; Bell Labs research did not require permission to start a project, so they did not ask. The design had three requirements: convenience, so that an analysis could be expressed compactly with the system managing the data; completeness, so that anything in the Fortran library could be reached; and extensibility, so that a researcher could add a new method by writing an interface function rather than a program. The syntax was C-like at a time when statistical systems were command-oriented; Unix colleagues had suggested a shell of pipes instead, and were declined. By 1978 a version ran on the Murray Hill Honeywell system under GCOS. It had no name for its first months and was eventually called “S”, with the quotation marks, for the same reasons C was called C. It moved to Unix in 1979 and was licensed outside Bell Labs from 1981. The 1984 book by Becker and Chambers, S: An Interactive Environment for Data Analysis and Graphics, is the brown book.

The new S of 1988, described in the blue book, The New S Language by Becker, Chambers and Allan Wilks, broke compatibility to make everything a function and everything an object, and added the evaluation semantics that R was later built from. The white book, Statistical Models in S (1992), which Chambers edited with Trevor Hastie, added the formula notation, y ~ x + z, and the informal class system now called S3. Version 4 of S, licensed from 1997 and described in Chambers’s 1998 green book Programming with Data, replaced that with a formal object system, S4, whose classes and generic functions are the ones R still carries. Statistical Sciences of Seattle had licensed S in 1988 and sold it as S-PLUS; through the 1990s that was the product statistics departments ran.

Chambers was head of the Advanced Software Department from 1981 to 1983 and of the Statistics and Data Analysis Research Department from 1983 to 1989, then went back to full-time research; in 1997 he was made the first Bell Labs Fellow, cited for pioneering contributions to statistical computing.

The Award and the Copy

In 1998 the ACM gave S its Software System Award, the prize previously given to Unix, TeX and the World Wide Web, with a citation saying it had forever altered how people analyse, visualise and manipulate data. Chambers donated the $10,000 to the American Statistical Association to endow the John M. Chambers Statistical Software Award for work by students and early-career researchers.

By then the copy existed. Ross Ihaka and Robert Gentleman had built R at Auckland from the blue and white books, since S had no formal definition, and released it under the GPL in 1995. Chambers did not fight it. He joined the R Core Team, contributed the S4 object system and the methods package, and wrote two more books, Software for Data Analysis: Programming with R (2008) and Extending R (2016), that treated R as the continuation of the project he had started in 1976. His account of the whole arc, “S, R, and Data Science”, appeared in the ACM’s fourth History of Programming Languages volume in 2020 and argued that R is best understood as a domain-specific language for the domain Bell Labs had been practising under another name since Tukey. S-PLUS, the paid product, had its last release in 2010 and was formally retired by TIBCO in January 2026; the story of that displacement is in R and Statistical Computing.

Chambers retired from Bell Labs in 2005 and has been at Stanford since 2008, as an adjunct professor of statistics and senior advisor to Stanford Data Science.

📚 Sources