How Copy Editors Are Killing Restrictive "Which"

The distinction between that and which is a favorite among usage writers. I've written about it several times on my blog and for the Copyediting newsletter. It's an interesting usage item for several reasons: first, it is an invention that was first proposed in the early 1800s yet didn't catch on until the 1900s; second, it's primarily, though not exclusively, an American distinction; and third, it has been very successful in print, though I think a good portion of its success is attributable to copy editors.

That is used almost exclusively for restrictive relative clauses, or clauses that limit the reference or define something, as in The car that I drive is black. Traditionally, which has been used for either restrictive or nonrestrictive relative clauses. A nonrestrictive relative clause simply adds supplemental information and is set off by commas, as in My car, which is black, gets good gas mileage. The rule forbids the use of which restrictively, so many people avoid sentences like The car which I drive is black.

In his Modern American Usage, Bryan Garner says that "the Fowler brothers are often credited with 'inventing' this distinction" and cites several examples of earlier proponents of the rule. But this misses the point. The Fowlers may not have invented it, but the fact remains that the rule is an invention that does not reflect the practices of the best writers, which the Fowlers admitted. Although it's hard to get accurate data from corpora, a search in the the Corpus of Historical American English for the strings "[noun] that [verb]" and "[noun] which [verb]" shows that the rule has increased in popularity since the Fowlers' day.

A comparison of the Corpus of Contemporary American English to the British National Corpus shows the disparity between American and British usage. The British use which nine times as often as Americans do. American newspapers are strictest of all in following the rule, with a rate of which one-fourth that of the overall American rate. For whatever reason, the rule has never caught on in Britain, and usage there is similar to American usage in the nineteenth and early twentieth centuries.

It appears that Americans have been following the that/which rule for the better part of a century. But corpus data doesn't tell us who has been following it. In the research for my master's thesis in linguistics, I found that changing restrictive which to that was one of the usage rules that editors enforce most often.

It's true that nonrestrictive clauses almost always use which, but there are many cases in which which is required in restrictive clauses, as in this sentence. Which is always required after prepositions and after the demonstrative pronoun that, and it frequently appears when two relative clauses are conjoined or when the relative clause is separated from its noun phrase. The Cambridge Grammar of the English Language spells out more constructions that favor either that or which, but the point is that usage is much more complicated than a simple restrictive/nonrestrictive distinction.

Looking strictly at cases in which either that or which could be used, I found that relative that appeared 106 times in just under 35,000 words of unedited text. I found 20 instances of restrictive relative which in the unedited text, 12 of which editors replaced with that, for an editing rate of 60 percent. This means that the original 5:1 ratio of that to which jumped to about 14:1 in the edited text. And this was after just one round of editing; most published text gets at least two rounds of editing between author and audience.

Note that editing doesn't completely account for the difference between American and British usage. I suspect in this case that other factors are also to blame: writing instruction, professional style guides, and especially Microsoft Word's over-vigilant grammar checker, which ruthlessly enforces the that/which rule. Though some writers turn it off, many accept its help and change every green-underlined restrictive which to that before a copy editor sees the text.

The change over the last century is an impressive piece of language engineering, and it's an excellent example of language standardization as the elimination of optional variation. But the question remains whether it's worth it. Readers outside the United States get by fine without the that/which rule and are often flummoxed by our insistence on it. We don't insist on a parallel restrictive/nonrestrictive distinction with other relative pronouns and adverbs such as who, where, or when, and it's impossible to make one in some constructions. Even the Fowlers didn't insist on it, and E. B. White admitted in The Elements of Style that "occasionally which seems preferable to that." Garner is less forgiving, however, calling violations of the rule "misusages" and saying that those who don't follow the rule "probably don't write very well."

In a century the rule has gone from a suggestion to an ironbound law. It's clear that usage writers can effect widespread changes to the language, especially when aided by copy editors, but this does not mean they should. I'm open to the idea that language engineering can be beneficial, but the burden of proof should lie with the proponents of the rule to show that it's worth the effort.