Here is a model answer for the essay question set on November 14th.
First, some background. Singular antecedents for they are most commonly quantified noun phrases like nobody or any student. Indefinite noun phrases are less common, but plentifully attested and quite acceptable. Definite noun phrase antecedents are still less common, and sometimes seem distinct unacceptable to people who otherwise accept singular antecedents. Singular proper name antecedents are almost unattested (for a very rare and somewhat peculiar exception attested one Saturday night in Las Vegas see this post). Basically this is a scale of definiteness of reference: as we move closer toward the kinds of noun phrase that uniquely pick out a specific person with the usual sex characteristics, it becomes less and les plausible to use they as a pronoun dependent on it. This insight is due to the excellent doctoral dissertation Singular They by Rachel Lagunoff (Applied Linguistics, UCLA, 1997).
Now, the factor that makes a personal name all but impossible as an antecedent for they is that if we know we are dealing with a specific person, then we know that they are of either the male or the female sex, and it feels highly unnatural not to use the appropriately gendered pronoun. In the rare cases where the speaker knows the name of someone and can use it to refer to that person but does not know the sex of the person and thus cannot use a gendered singular pronoun, it feels better to say he or she than they.
A definite NP that refers directly to a specific person is also implausible as an antecedent for they. But the more it is clear that it does not have a specific person as referent, but rather means "whoever turns out to meet this definite description", the more singular they will be acceptable.
When the sign on the 29 buses says that a passenger "must not speak to the driver or distract their attention", it means "the driver of this bus, whoever it may be at the time in question". And Lothian Buses does employ female bus drivers. There is no way to tell who might be the person picked out by the driver on a particular occasion. The phrase is interpreted rather like whatever person happens to be driving this bus, which has a covert universal quantification. That is what makes they fully acceptable with a singular definite NP as antecedent in that case.
What is entirely different with the king is that it is so obviously referential. The king (when there is one) is a unique individual, and always male. We always know the sex of the king if he exists at all. So singular they is entirely unmotivated in this case. It is used to allow third person singular pronominal reference in cases where the antecedent is interpreted as a quantifier binding a variable, or is indefinite, has some similar interpretation involving unknown or suppressed sex information and thus unassignable or arbitrary morphological gender choice. A definite NP referring to a unique person guaranteed to be male is just about as unable to support an anaphorically dependent they as a personal name would be.
In short, it is not really about a difference between drivers and kings, but about the difference between (i) a phrase referring to the unknown arbitrary person of whatever sex who may be driving a particular bus at some relevant time in the future, and (ii) a phrase referring directly to the unique male person who currently holds the monarchy. One noteworthy point this illustrates is that pronoun choice is not the province of any one component of grammar on its own: morphology, syntax, semantics, and pragmatics are all relevant to the choice between the four 3rd-person pronoun lexemes he, she, it, and they.
P.S. Paul Postal has written to point out that he finds it utterly ungrammatical to say Congress shall not harass the president nor interfere with their control of the armed forces. All that one can say is that (i) he may be more subliminally influenced by prescriptive injunctions than he would be inclined to think he is; (ii) he may have a stronger sensitivity to the hierarchy of referentiality and draw the cutoff line below indefinites and above definites; and (iii) it is a bit early to judge sentences like this in view of the fact that US presidents have been very much like kings thus far (always male, with no females even nominated by major parties for election to the post, and sometimes members of quasi-royal dynasties that contribute more than one male scion to the presidency). Perhaps ten years from now we shall be looking back on a situation where, although the tendency toward family dynasties in American politics has not lessened, the choice of he as the pronoun to be anaphoric to the president does not look quite so obvious any more. On a point like this one, grammatical acceptability is evolving in parallel with social mores and the politico-cultural context. An enormous amount of English syntax is remarkably stable, but this little bit of it is not.
Posted by Geoffrey K. Pullum at November 16, 2007 03:51 AM