Evolution and Arithmetic

Chapter 6: The Monkeys That Didn't

← Another QuestionWriting the Library of Congress →

A Shocker

Every so often, a story shows up in respectable places — the BBC, CNN, wire services picked up by newspapers around the world — announcing that science has finally settled an old question. In the autumn of 2011, the question was the one we have been struggling with, and the headline was blunt: "Virtual monkeys write Shakespeare." For someone who's been thinking about monkeys and typewriters for a long time, you can imagine a headline like that grabbing my attention.

It hadn't happened. But the way it didn't happen is useful anyway, because it's a good example of a claim that sounds like evidence, gets treated like evidence by people who ought to know better, and falls apart the moment you actually think about it, or do the arithmetic, instead of just reading the headline.

Here's what actually took place. A programmer named Jesse Anderson, wanting to try out some cloud-computing tools, set up a few million virtual "monkeys" — small programs running on Amazon's servers — that generated random nine-letter sequences, a few of them meaningful-looking by pure accident, most of them gibberish. Each sequence got checked against the complete text of Shakespeare's works. If that particular nine-letter string appeared anywhere in that text — anywhere at all, in any of the plays, at any position — the matching spot in Shakespeare was marked "found," and the monkeys moved on. Run this long enough, the claim went, and eventually every nine-letter fragment that exists anywhere in Shakespeare will have been generated and checked off. Anderson's monkeys started on August 21. Within five weeks they had checked off 99.99 percent of Shakespeare, and by early October the last play was done: about six and a half weeks in all. The BBC called it "virtual monkeys write Shakespeare." Nobody at the network, apparently, asked the obvious follow-up question: what does "write" actually mean here?

Last revised 2026-11-17

← Another QuestionWriting the Library of Congress →