by

## For Want of a Copy Editor the Sense Was Lost

Yuri Matiyasevitch in 1969

When a Russian mathematician collaborates with a French computer scientist on a paper published by Elsevier in the Netherlands, what language do they choose?

English, of course. Unsuitable it may be, but it’s the unavoidable language of science these days.

And that means Elsevier will need to provide expert editors to assist non-native-speaking authors, right?

Wrong. Elsevier’s two and a half billion dollars of annual revenue (only about a billion of it operating profit) is not squandered on part-time English-speaking copy editors. Not even the titles of papers are checked.

I discovered that recently when I looked up a fascinating paper by Yuri Matiyasevitch and Géraud Sénizergues and noticed that the first error came at the seventh word of the title.

Matiyasevich is famous for completing (at age 23) the proof that Hilbert’s 10th problem (finding an algorithm for determining whether a diophantine equation has an integer solution) is unsolvable: It is impossible for such an algorithm to exist.

Sénizergues is a French computer scientist who in 2002 won the Gödel Prize for proving that a computer program could be written to ascertain the equivalence of two deterministic pushdown automata. (Roughly, that means telling whether two primitive word-checking machines of a certain sort accept exactly the same set of words.)

These two prize-winning computability theorists collaborated on a problem about semi-Thue systems, which are sets of rules for manipulating words to obtain other words. If wuv is a rule, then from the word xwy (where x and y are any sequences of zero or more letters) you can derive the word xuvy. If uvz is another rule of the system, you can then derive xzy. And so on.

A semi-Thue system can have any finite number of rules, so you can imagine that it might get complicated to figure out whether you can get from one word to another using the rules. It turns out that’s an understatement: Figuring out whether you could start with some word w and end up with another word z is actually undecidable: No computer program could guarantee to discover the answer.

Matiyasevitch and Géraud Sénizergues studied semi-Thue systems with very small numbers of rules to see whether they are perhaps more tractable. Suppose there are just three rules, for example. Surely that wouldn’t be quite so hard?

Wrong. Matiyasevitch and Sénizergues show, inter alia, that even with only three rules it is undecidable whether a given semi-Thue system can get you from word w to word z, or whether there is a word x from which you can reach both w and z, or whether there is a maximum limit on the length of the words you can obtain from w.

Their paper was published in 2005 as “Decision problems for semi-Thue systems with a few rules” (Theoretical Computer Science 330: 145–169). The mistake there is using a few instead of few.

Using a few merely sets a lower bound on some number, while using few asserts that the number is really very small. To say there are a few pickpockets downtown is bad news, and unfortunately it’s still true if it understates things and in truth there are whole gangs of them. But to say there are few pickpockets downtown is good news — it says they’re rare.

Matiyasevitch and Sénizergues meant “semi-Thue systems with few rules” — that is, with a very small number of rules. And that isn’t the only minor grammatical slip in the paper. The abstract mentions “3-rules semi-Thue systems,” but you don’t normally put the plural -s on a nominal used to modify another noun in English: We say a 10-foot pole, a seven-year drought, or a 3-rule system. Then four lines into the text of the paper they write about “the development of a general notion of algorithm and Church Thesis.” They mean Church’s Thesis. And we haven’t left the first page yet.

Elsevier undertook to publish this work in English. They owed it to the authors to provide half an hour of a native-speaking editor’s time to check for small usage errors. There are hundreds of English native-speaking editors in the Netherlands; I gave a lecture to a conference audience of about two hundred of them in Utrecht.

Elsevier (now part of the RELX Group) is a wealthy megacorporation making billions of dollars by fleecing the libraries of the world; they make extra money by (almost unbelievably) selling access to papers they have already been paid to provide open access to; yet they are such penny-pinching Scrooges that they won’t even provide the minimal labor that would permit them to do their moral duty to the brilliant academics whose work they publish without paying for it (and whom they sometimes threaten for distributing it to colleagues). But talk to me one-to-one some evening over a beer and I’ll tell you what I really think of those venal, grasping, unscrupulous skinflints.