I’ve been reading about the which vs. that dispute.
My summary of the rule as it’s frequently taught in usage guides, and as I learned it growing up:
Use which to head relative clauses that do not restrict the meaning of the sentence—that is, those that are set apart by commas. My car, which I’m trying to sell, is blue. In this sentence, the speaker has one car, which is incidentally about to be sold.
Use that to head relative clauses that do restrict the meaning of the sentence—that is, those not set apart by by commas. My computer that’s a little old has been acting up. In this sentence, the speaker presumably has more than one computer, one of which is a little old.
This rule does not apply when the relative pronoun is preceded by a preposition:
…a story, the movie version of which he hadn’t seen.
*…a story, the movie version of that he hadn’t seen.
Or when it’s preceded by another instance of the word that:
That which does not kill us makes us stronger.
*That that does not kill us makes us stronger.
Many guides admit room for people to do otherwise. And some guides dispute the rule entirely. As always with these things, it’s often taught to students without such nuance, and they can come away thinking that A date which will live in infamy is not just bad style, but ungrammatical.
It is neither. Here’s my summary of the rule that actually exists in English:
Who and whom are available as the head of any relative clause when describing a person. Which is available as the head of any relative clause when describing an object. That is available as the head of some relative clauses that are usually called “restrictive,” and it can describe people or objects. It is not available if it’s preceded by a preposition or another instance of the word that. It is more common than which as the head of relative clauses, when either word is permitted. It is less common than who as the head of relative clauses, when either word is permitted.
The penultimate sentence of my description wasn’t true until usage guides started promoting the prescribed rule.
Now for some disjointed thoughts.
Bryan Garner’s Modern English Usage explains that, though the Fowler brothers are often credited with inventing the rule, they did not. But he gives misleading commentary on the oldest supporting discussion that he can find.
Before quoting, he notes that the author in question and another author “engage in the very misusage they talk about (both noted by a bracketed n.b.)”.
Some critics have lately objected to the use of that, as a relative, conceiving which to be in all cases “the preferable word” [Hugh Blair, Lectures on Rhetoric]. But this is certainly a hasty and erroneous opinion. We have in English three relatives, that, who, and which; and, in their respective and appropriate use, we possess an advantage . . . peculiar to our language, and which [n.b.] I hope we shall not be tempted to relinquish. [In the examples that Odell then gives, the thats are used restrictively, the whiches nonrestrictively.]
James Odell, An Essay on the Elements, Accents, and Prosody of the English Language 49 (1806)
I bolded Garner’s insertions for emphasis. All italics are in the original, as they are in every other blockquote in this post.
James Odell was counterarguing other academics, including Hugh Blair, who said:
Which is a much more definite word than that, being never employed in any other way than as a relative; whereas that is a word of many senses; sometimes a demonstrative pronoun, often a conjunction. In some cases we are indeed obliged to use that for a relative, in order to avoid the ungraceful repetition of which in the same sentence. But when we are laid under no necessity of this kind, which is always the preferable word…
Odell responds as quoted above. He goes on to discuss the restrictive vs. non-restrictive question a little bit:
“All that we can do,” implies the utmost extent of our ability. “All which we can do” is rather an assertion that our ability is, at least, equal to the task. The latter is therefore, if used in any other sense, at best an ambiguous phrase; whereas the former is capable but of one meaning: for the word, we must remember in these phrases, is not that, but thut1. In like manner “the best writer that I am acquainted with,” can only mean the best in the circle of my acquaintance; the best among those whose works I am acquainted with; but, “the best writer with whom I am acquainted,” might be construed to imply not only that he was an excellent writer, but that he was also one with whom I had a personal acquaintance.2
But he doesn’t say these considerations take precedence in all instances of the choice between the words. He says they do not:
The choice between that and which is often, with respect to propriety, a matter of indifference; and then, doubtless, we should be governed by regard to euphony of composition. On this account it is often necessary to use that rather than which, in order to avoid some of those harsh collisions of sound, which, at the best, are too frequent in our language.
So it doesn’t mean anything to catch him out using which restrictively. He never said that you shouldn’t; he said that euphony is “often” the governing factor.
Garner says that in all of Odell’s examples, which is used non-restrictively and that restrictively. But in most of Odell’s examples, he is talking about euphony/melody. Which needs to be stressed when that does not, so swapping that out for which can ruin poetry.
It seems Garner wants to claim Odell as an early proponent of the rule, but I don’t think the text supports that.
From what I can find, the first time a commentator explicitly suggested the modern rule was in 1851. Goold Brown published a book called The Grammar of English Grammars. Here is what he said about which vs. that, also commenting on Blair’s 1793 suggestion that we not use that as a relative:
The relative that, though usually reckoned equivalent to who or which, evidently differs from both, in being more generally, and perhaps more appropriately, taken in the restrictive sense. It ought therefore, for distinction’s sake, to be preferred to who or which, whenever an antecedent not otherwise limited, is to be restricted by the relative clause; as “Men that grasp after riches, are never satisfied.”—”I love wisdom that is gay and civilized.”—Art of thinking, p. 34. This phraseology leaves not the limitation of the meaning to depend solely upon the absence of a pause after the antecedent; because the relative that is seldom, if ever, used by good writers in any other than a restrictive sense. Again: “A man of a polite imagination is let into a great many pleasures that the vulgar are not capable of receiving.”—Addison, Spect. No 411. Here, too, according to my notion, that is obviously preferable to which; though a great many critic, very widely known, has taken some pains to establish a different opinion. The “many pleasures” here spoken of, are no otherwise defined, than as being such as “the vulgar are not capable of receiving.” The writer did not mean to deny that the vulgar are capable of receiving a great many pleasures; but certainly, if that were changed to which, this would be the meaning conveyed, unless the reader were very careful to avoid a pause where he would be apt to make one. I therefore prefer Addison’s expression to that which Dr. Blair would substitute.
Brown’s advice is much more limiting than Odell’s. Ergo, unlike Odell, it is funny to catch Brown out not following his own advice:
…thinking it practicable to effect some improvement upon the manuals which explain our own [language]…
…giving thanks for the signal mercy which has upborne me to the long-continued effort.
The nature and design of this treatise are perhaps sufficiently developed in connexion with the various topics which are successively treated of in the Introduction.
The best thing about the above three examples is that I didn’t pull them from random passages throughout this gargantuan text. They are all from the first page. No restrictive that clause appears on the same page. On the second page, there are 4 restrictive whiches and no restrictive thats. He didn’t follow his own advice at all!
The Fowler brothers are probably most responsible for popularizing the rule. In The King’s English from 1908:
2. 'Who' or 'which' should not be used in defining clauses except when custom, euphony, or convenience is decidedly against the use of 'that'.
H.W. Fowler advocated the rule again in 1922. From the first edition of A Dictionary of Modern English Usage:
The two kinds of relative clause, to one of which that and to the other of which which is appropriate, are the defining and the non-defining; and if writers would agree to regard that as the defining relative pronoun, and which as the non-defining, there would be much gain both in lucidity and in ease. Some there are who follow this principle now; but it would be idle to pretend that it is the practice either of most or of the best writers.
The Evans siblings—Bergen and Cornelia—responded in their 1957 book A Dictionary of Contemporary American Usage:
What is not the practice of most, or of the best, is not part of our common language.
Great quip against those who pretend that Fowler’s suggestion is incontrovertible. (Not so against Fowler himself, who was not claiming the rule to be part of our common language—only that he wanted it to be.)
The same year his book was published, Bergen Evans began hosting a tv show called The Last Word which contained grammarians arguing about various usages. Did you know that there was a tv show in which an English professor hosted arguments about grammar and word usage? Unfortunately, I can’t find any footage.
Turning back to Garner, he does try to pretend that it is the practice of the best writers. But he fumbles in doing so:
You’ll encounter two schools of thought on this point. First are those who don’t care about any distinction between these words as relative pronouns, who think that which is more formal than that, and who point to many historical examples of copious whiches. They say that modern usage is a muddle. Second are those who insist that both words have useful functions that ought to be separated, and who observe the distinction rigorously in their own writing. They view departures from this distinction as “mistakes.”
Before reading any further, you ought to know something more about these two groups: those in the first probably don’t write very well; those in the second just might.
Not to impugn his capabilities generally, but the final sentence in this passage is itself bad writing.
Garner can’t dismiss everyone who flouts the rule. Later he admits, “British writers mostly ignore the distinction,” and the rule wasn’t even popular until the 1920s. So he hedges with probably not.
The problem is that probably not X implies maybe/might X.3 The sentence is logically equivalent to:
“Those in the first might write very well; those in the second just might.”
Huh?
I think he means to say something like, “The ratio of good to bad writers is higher among those who follow the rule,” but he botched it.
Bonus bad writing points for not providing a scintilla of evidence for the assertion I think he intended to make. In his defense, the assertion he actually makes—most people are probably not great writers either way—is self evident.
Which and who have a symmetry: where you can use one, you can use the other in an equivalent construction. As one example:
This gun, with which he carried out many murders…
His accomplice, with whom he carried out many murders…
If alleviating the confusion between restrictive and non-restrictive clauses were the consideration über alles, you’d expect that who in restrictive clauses would get the same censure as which. Why doesn’t it?
Brown’s original suggestion actually does include who:
It ought therefore, for distinction’s sake, to be preferred to who or which, whenever an antecedent not otherwise limited, is to be restricted by the relative clause; as “Men that grasp after riches, are never satisfied.”
Garner’s guide provides numerous quotes from grammarians between Brown and Fowler approving the which/that rule, and some of them also include who in the recommendation.
“He was the best man that could be found for the place.”
Joseph Angus (1868)
Some teachers insist that the relative that should be used, instead of who or which, when the relative clause serves to restrict the meaning of the antecedent, and that who or which should be used, instead of that, when the relative clause adds something to the meaning of the antecedent, or explains it.
Adams Sherman Hill (1888)
The relative who or which may, and theoretically does, introduce a new fact about its antecedent; its office is, therefore, to head a coordinate clause, as may be shown by using its equivalent and he, and it, and they.
John F. Genung (1894)
Even though he includes these quotes, Garner devotes not a single word to explaining why the rule should apply only to which and not to who. To their credit, the Fowler brothers did:
'That', used of persons, has in fact come to look archaic: the only cases in which it is now to be preferred to ‘who' are those mentioned above as particularly requiring 'that' instead of 'which' ; those, namely, in which the antecedent is 'it', or has attached to it a superlative or other word of exclusive meaning…
In 1908, apparently The child that I’m putting through school sounded antiquated. But in the nearly 120 years since, it no longer does. Who in such a construction still dominates, but that is common enough.
There’s an extra reason given in Fowler’s Dictionary of Modern English Usage (1922):
Politeness plays a great part in idiom, and to write The ladies that were present, or The general that most distinguished himself, is perhaps felt to be a sort of slight, depriving them of their humanity as one deprives a man of his gentility by writing him Mr. instead of Esq. At any rate, the necessarily defining that is displaced by the not necessarily defining who where the relative refers to a particular person or persons, but holds its own better when the person is a type or generic… such antecedents as all, no one, a man, ask for nothing better than that.
Neither of the two examples he provides sound unusual or rude to me. Maybe other people feel differently, but I think the 100 years since the publication of this book have weathered away this consideration as well.
So there is now little or no reason for the disparity between which and who in usage guides. But the modern guides like Garner’s, or Fiske’s Dictionary of Unendurable English, don’t seem to have noticed. If they truly thought that differentiating these two types of clauses is so important that it needs more than a couple of commas, they would expand the recommendation to include the word who.
I wonder if the disparity is part of the reason for the folk belief that using that to refer to people is wrong—that it should always be who. That is, since teachers only talk about that vs. which, some people infer that that is like which generally. This Grammarly page, for example, claims that You’re just somebody that I used to know is ungrammatical.
Absurd! I can’t find anybody claiming this in a serious language guide, even the very conservative ones.
Usage guide authors often have a methodology that I find off-putting. Starting with Brown and on to Garner, they note one thing: “There would be less room for confusion over whether a clause is restrictive or non-restrictive if we eliminated which from the restrictive role.” They then dust their hands as if the entire matter has been settled, no other considerations even meriting discussion. Not even worth a thought, apparently, are:
The inelegance of destroying the who/which symmetry.
Whether that is overused in our language.
Prosody, euphony, melody. The fact that which tends to be stressed and that does not.
Formality. Does that sound too informal in some instances? Can which add gravitas?
And as in so many of these prescriptions, there’s one factor that seems to me overwhelmingly more important than others, and that usage guides never address:
Ease with which the proposed rule is instituted.
Are you trying to change a fringe usage that only a small portion of the population employs? Is it something new that hasn’t taken hold in the general populace yet? Or are you trying to change a common usage employed by almost every English speaker for hundreds of years? In this case, it’s the latter.
In such cases, the costs of trying to force such a change are much larger. Do the usage guide authors ever care about the costs? How much hypercorrection and nervous cluelessness they’re going to introduce? How much extra work they’ll add to the practice of writing and editing? Doesn’t adding such confusion to the language increase bad writing as well?
Such costs are typically ignored. Worse yet, the benefits in this case are exaggerated at best. We have a natural experiment on what happens when you do and don’t follow the rule. British writers and editors mostly don’t bother with it. Is their writing markedly worse? I’ve never seen anyone try to argue so. Any such argument would run into a problem anyway: When people train themselves to believe invented rules, previously acceptable writing becomes bad to them. An argument that “this sentence sounds bad” will therefore only work on people who are already convinced of the rule.
Influential usage guides have the ability to turn good writing bad, simply by fiat. It’s such a loss, when they do.
EDIT: Something brought up in Geoffrey Pullum’s The Truth About English Grammar that I had never considered before:
Here’s a nice example of a sentence about reporting wrong-doers in which choosing that instead of which happened to be a disaster for intelligibility:
IEEE has a way to denounce someone that protects the whistle-blower.
This is grammatical but ambiguous: It could be talking about a method for denouncing the protector of a whistle-blower, or it could be talking about a way of denouncing people that provides protection for the whistle-blower. The sentence should have been written like this (violating the Fowler brothers’ pseudo-rule):
IEEE has a way to denounce someone which protects the whistle-blower.
That makes it clear (because of non-human-referring which rather than human-referring who) that the relative clause modifies the method of denouncing. The Institute of Electrical and Electronics Engineers is not offering a way to denounce someone who protects whistle-blowers. You can see that because the relative word which is not compatible with someone. The version with that obeys the Fowlers’ rule, but is clearly the wrong choice: because it doesn’t have a relative pronoun, it can’t use the who/which distinction to make clear whether we’re talking about a person or not.
The people who invented the rule didn’t think through every complication. Do they ever?
It’s funny that prescriptivism is so often associated with social conservatism. Aren’t conservatives supposed to be afraid of making deliberate changes to evolved social orders? Chesterton’s Fence and all that? Trying to perform surgery on language can have unintended consequences too.
1 I believe Odell is pointing out that sometimes “that” is pronounced with a schwa. Consider the sentence, “I believe that that is true.” Personally, I pronounce these two instances of “that” slightly differently.
2 Odell’s point seems a bit more subtle than the ordinary restrictive vs. non-restrictive problem. “All, which we can do” reads to me like an assertion that we are omnipotent. “The best writer, with whom I am acquainted” reads to me like an assertion that they’re not only an “excellent” writer, but the best of all time.
3 They’re not synonymous both ways. “Maybe/might” can mean any probability that isn’t 0% or 100%, depending on context. But you can always map “probably X” to “maybe not X,” and “probably not X” to “maybe X.”