While cleaning up some paperwork today, I came across something I printed from a web site in October of 2000, entitled "Grammar and Choice of Words" by Eugene A Burdick.
Reading through it, I realized why I'd saved it: it contained one of the very few accurate explanations of using "that" and "which" that also included how to properly reference remote and near antecedents.
I tried to open the web site, but it no longer existed. I then searched for Judge Burdick by name and found that he had died less than a month after I printed his page.
Here, then, is his explanation of using "that" and "which". I will make a point of notifying his relatives about this posting. A liberty I've taken with the original is to add some quotes to clarify when he is speaking of "that" vs. using "that" in a sentence. I have also corrected what appear to be errors, not in grammar but in missing words. I don't doubt Judge Burdick was writing quickly in between many other tasks. Any errors in copying his work are strictly my own.
Many writers have not found a workable distinction between "that" and "which". We all know that a pronoun takes the place of a previously mentioned noun, so as not to repeat the noun. The noun replaced by the pronoun is the pronoun's antecedent. The pronoun substitutes for the previously mentioned noun. We say the pronoun modifies the noun. We all know that "that", when used as a pronoun, modifies an animal or a thing. Example: I wrote a poem that expressed my sentiments. In that example, "that" is a pronoun modifying "poem." Rarely is "that" correctly used to modify a human, except when referring to a group such as a team. Example: I threw the ball to the boy who was on the mound. In that example, "who was on the mound" is a restrictive clause. It identifies the individual to whom I threw the ball. On the other hand: It was the Boy Blues team that won. In the first example, "that expressed my sentiments" is also a restrictive clause, because it indicates [what] the poem intended, namely that the poem expressed my sentiments. In each of those examples, the pronoun is obvious. No problem. Let us complicate the examples a little.
Example: The poem, which was a sonnet, won first place in the contest. In that example, the clause "which was a sonnet" was gratuitously injected into the sentence. It is a nonrestrictive clause because it is merely parenthetical to the main thought. The sentence reads correctly without it. It may convey information, but the information does not otherwise restrict the sentence. In a nonrestrictive clause, which is always set off by commas, the pronoun that introduces it is always "which", in speaking of non-human antecedents, such as animals, plants, and mental concepts. Let us complicate the examples further.
Let us construct a sentence that has two possible antecedents of the pronoun in a restrictive clause. Example: I saw a rat in the house that was very old. Without a rule to guide us, it is impossible [to] tell whether the writer intended the pronoun "that" to modify the rat or the house. As noted above, if there is only a single antecedent, we are all comfortable with the use of "that". But, we are confused when our pronoun, considering the grammar used, can modify either, but not both, of the antecedents. We solve the dilemma by providing another use for "which". If we intend to refer to the house, the nearest of the two antecedents, as being very old, we use the pronoun "that", just as if the house were the only antecedent. If we intend to refer to the rat, the remote antecedent, as being very old, we use the pronoun "which". The reader reasons: If the writer intended the pronoun to modify the house, the writer would have used the pronoun "that". Because the writer used, instead, the pronoun "which", the writer must have intended to modify the remote antecedent in the above example. The National Conference of Commissioners on Uniform State Laws has a neat solution. The Conference expresses the rule thus:
Rule 9 c). Use "which" to introduce a restrictive clause that is intended to modify the remote antecedent, rather than the nearest of two possible antecedents.
The conference has used this differentiation for many years. I have yet to hear that a reader was confused by this rule. I have also noticed that the New York Times has applied the rule, perhaps unwittingly, to a number of its news items.