Language Log

What I currently know about which and that

Following on my postings about that, which, Bellow, and Safire and about the Committee for the Certification of Good Writing, people have been writing me about versions of the That Rule in the style sheets of institutions that should really know better.  Roger Shuy, in particular, wrote, somewhat sheepishly, about his sheeplike compliance for one of these institutions.  In reply, I summarized for him what I currently know, and think, about the That Rule, and now I'm passing a version of this summary on to you.

Two of my correspondents have dealt with large, famous, venerable, and transatlantic university presses that publish extensively in linguistics (actually, I've dealt with these presses too, but I didn't have their style sheets to hand).  In exhibit #1, which came to me on 5/3/05, the style sheet says sternly:

This one comes to me from someone who has done copyediting for the U.S. office of one of these large, famous, etc. presses.  I'm keeping my correspondent's name confidential, just in case there's a chance of more copyediting gigs in the future.

In exhibit #2, dated 5/8/05, the safely nameable Roger Shuy passes on a style sheet from the other of these transatlantic academic behemoths, a style sheet in which we are (merely) advised:

This is just to show that the That Rule lives on even in the genteel ivied walls of academia, not just in PSAT prep manuals and the like.

So, what do I currently know about the justification for the That Rule?

1. The usual claim (now, although this was not Fowler's claim in 1926) is that the That Rule avoids ambiguity between the two sorts of relative clauses -- or, in some muddier formulations, that it avoids "unclarity".
2. Taken at face value, this claim is just silly.  The guidelines in exhibits #1 and #2 recommend marking the distinction in TWO ways -- by punctuation, and by the choice of that (vs. which).  But the punctuation ought to be sufficient, and it's required in other cases (see below) where the that/which choice is not involved.  In any case, there should be no chance of ambiguity, or any other sort of unclarity, so long as the relative clause is properly punctuated.
3. So, apparently, the ambiguity justification really is cogent only if people don't punctuate correctly.  Well, plenty of writers are not especially competent at punctuation; in particular, commas with restrictive which AND that are both pretty common (not in material from "good writers", but in, say, student writing).  So what the That Rule does is try to fix the punctuation problem by having the punctuation follow automatically from the choice of relative marker.  It seems to me it would make more sense to get at the punctuation directly; in fact, as I argue below, it's necessary to get at the punctuation directly.
4. One baleful consequence of trying to use the that/which choice to fix a punctuation problem is that students are likely to learn the alternative rule, the Relative Punctuation Rule: use a comma with which, no comma with that.  Now, since virtually everybody uses some restrictive which in speaking and writing, this alternative rule actually INDUCES incorrect punctuation with restrictive which.

   (It's one thing to promulgate a rule.  How your formulation will be understood by your audience is quite another matter.  The universe of "rules" of grammar and usage is peppered with unintended consequences.)

   The underlying problem is that almost every English speaker's Sprachgefühl occasionally calls for restrictive which, and then the Relative Punctuation Rule can lead people to mispunctuate.  (For the brief story on restrictive which, see the MWDEU entry for that; the studies mentioned there, and now a number of others, show that edited prose contains quite a lot of restrictive which.)

   However, there are at least two cases where the That Rule isn't sufficient to fix the punctuation of relative clauses.

5. The first case involves relative markers in combination with prepositions.  Contrast:
   
    The only case of which I have direct knowledge occurred in 1972.
    The only case, of which I have direct knowledge, occurred in 1972.
   

   Here, the punctuation alone marks the restrictive/nonrestrictive distinction, since (as is well known) that is unacceptable in combination with a preposition:
   

   *The only case of that I have direct knowledge...
   *The only case, of that I have direct knowledge,...
   
6. If you insist that restrictive relatives must ALWAYS have that rather than which, then the only option open to you is to strand the preposition:
   
   The only case that I have direct knowledge of...
   
   Unfortunately, the same manuals that insist on restrictive that also recommend against preposition stranding.

   The That Rule could, of course, be altered to exclude this case, by saying something like:

   That Rule 2: Use that for restrictive relatives, unless the marker is the object in a prepositional phrase; otherwise, use which.

   I assume that something like That Rule 2 is what the manuals have in mind, though how a student is supposed to figure that out, I have no idea.

7. The second case involves relative who(m).  If there is ambiguity or unclarity with relative which, the same ambiguity or unclarity exists with relative who(m).  The following contrast is made entirely by punctuation:
   

   The only cheater who(m) I know is Kim.
   The only cheater, who(m) I know, is Kim.
   

8. You might expect the manuals to insist on restrictive that here as well as in the earlier cases; it would, after all, fix any problem of unclarity.  And restrictive that is available:
   

   The only cheater that I know is Kim.
   
   Unfortunately, the same manuals that insist on restrictive that in non-human relatives disfavor it in human relatives.

   The That Rule could, of course, be altered to explicitly apply only to non-human relatives:

   That Rule 3: Use that for non-human restrictive relatives, unless the marker is the object in a prepositional phrase; otherwise, use the appropriate wh-word.

   Again, I assume that something like this is what the manuals have in mind, though, again, I don't know how students are supposed to figure that out.

   I'd guess, by the way, that the standard That Rule actually ENCOURAGES people to use that in human relatives (against the intentions of the manual writers).  More unintended consequences.

9. So far I've stated the rule as, literally, a prescription: advice about which words to use where.  Viewed this way, the rule is baroque.  But, actually, the manuals (in this case and most others) provide prescriptions only secondarily.  Their primary purpose is to issue PROSCRIPTIONS; the prescriptions are fixes for the bad stuff.

   The proscription is one that's pretty easy to check for mechanically:

   Which Hunting Rule: Look for a noun immediately followed, without a comma, by which; the which probably should be replaced by that.

   (There are a few cases where this sequence doesn't involve a relative clause, as in "I asked the man which (one) he wanted", so you can't be COMPLETELY mechanical about it.)  The near-mechanical character of the rule in this form undoubtedly has contributed much to its popularity on tests of "grammatical competence", its endurance in style sheets, and its incorporation into "grammar checkers".  An actual STYLISTIC CHOICE would involve judgment.

   Note that the Which Hunting Rule works well only when the writing you apply it to is generally punctuated correctly.  Which is almost always the case for the material that the venerable transatlantic presses' style sheets are applied to.  That is, this fairly mechanical rule can be used only to alter the writing of people who are in fact making a stylistic choice between which and that; the effect of the rule is to limit this choice.  That's why I object to it.

   As for those who are less competent writers than I am, I think their problem in this domain is mostly one of punctuation, and that's where they need help.  Trying to package the punctuation together with the choice of relative marker is a really bad idea.

zwicky at-sign csli period stanford period edu