The Science of Language

I recently stumbled upon the following sentence (here, for the curious):

After Windows 7 comes out in October, will Microsoft somehow force us XP users to stop using it?

I am able to figure out from context that the "it" at the end of this sentence is supposed to refer to "(Windows) XP", but no matter how many times I read the sentence, I stumble on that intention. The problem, as I see it, is that there's no noun phrase in this sentence that refers solely to "XP", and the pronoun "it" must be coreferent with a noun phrase. I explain a little more below the fold.

First, here's a relatively detailed syntactic tree for the sentence, with the intended-but-ungrammatical coreference between "it" and "XP" (and the nodes above these terminal items) indicated with identical subscripts. (Click on the image for a larger, more readable version.)

As you can see from the labels in the tree, the noun phrases (NPs) in this sentence are "Windows 7″, "October", "Microsoft", and "us XP users" — as well as the "it" in question, of course. Others with more syntax bona fides than me may disagree with my classification of "us" in "us XP users" as a determiner, but I'm fairly certain that we can all agree that "XP users" is a compound noun, consisting of two nouns, "XP" and "users". There's no noun phrase corresponding to "XP", and thus the only thing that the "it" can grammatically corefer with is "Windows 7″ (the most obvious reading that I inevitably stumble to), "October" (infelicitous but still grammatical), and "Microsoft" (more felicitous). (Of course, the "it" can't corefer with "us XP users" because the former is third person singular and the latter is first person plural.)

The mere fact that it is ungrammatical for the "it" to corefer with "XP" in this sentence should be enough to show that the "XP" part of the sentence is not a noun phrase, but we can also apply further, fairly standard constituency tests to "XP" to convince ourselves. Let's start with the grammatical clause "Microsoft will force us XP users to switch to Windows 7″ and apply some movement tests. (Note that the last of these is technically a substitution + movement test.) These tests clearly show that movement of the compound-internal noun "XP" alone is ungrammatical, but movement of the entire noun phrase "us XP users" is fine.

Constituency movement tests
("*" indicates ungrammaticality; "[__]" indicates the source of movement)

• Fronting:
  *XP, Microsoft will force us [__] users to switch to Windows 7.
  (cf. Us XP users, Microsoft will force [__] to switch to Windows 7.)
• Clefting:
  *It is XP that Microsoft will force us [__] users to switch to Windows 7.
  (cf. It is us XP users that Microsoft will force [__] to switch to Windows 7.)
• Pseudo-clefting:
  *XP is what Microsoft will force us [__] users to switch to Windows 7.
  (cf. Us XP users are who Microsoft will force [__] to switch to Windows 7.)
• Passivization:
  *XP will be forced us [__] users to switch to Windows 7 by Microsoft.
  (cf. Us XP users will be forced [__] to switch to Windows 7 by Microsoft.)
• Wh-questioning:
  *What will Microsoft force us [__] users to switch to Windows 7?
  (cf. Who will Microsoft force [__] to switch to Windows 7?)

But all this now raises a question: why do the ungrammatical sentences here sound so much worse than the original example that I'm claiming is also ungrammatical? Not being an expert on such matters, I'll merely point to some possible avenues for investigating such questions. On the one hand, it could be that the original example sounds a bit better than the ungrammatical examples above because there are at least two grammatical and felicitous ways to interpret the original example; that is, the intended interpretation could be strictly ungrammatical but the simple availability of the other options makes it sound relatively OK. Alternatively, it may be that violation of the rules for pronominal coreference is less consequential than violation of movement rules. (See here for another type of example of this type.) Both of these possibilities might also explain why someone wrote the original example, but it's of course also possible that the writer's correctness conditions simply differ from mine (and yours, if you agree with my judgment of the example).

A nice example of the way singular they works was overlooked (like health care, the economy, and everything else in the past week of "racial politics") during the brouhaha over President Obama's press conference remarks about the arrest in Cambridge, Massachusetts of Professor Henry Louis Gates. Obama said:

. . . the Cambridge police acted stupidly in arresting somebody when there was already proof that they were in their own home.

Why would he use they and their, when the antecedent, somebody, is syntactically singular, and we actually know that the somebody he is talking about in this case was Professor Henry Louis Gates, who is male? Why did he not say proof that he was in his own home?

(By the way, when I write they in this post, in bold italics, I mean the word described by the dictionary entry for they — the word that has the inflected forms they, them, their, theirs, themselves, and occasionally themself. Likewise, by he I mean the item having the inflected forms he, him, his, and himself.)

The answer to the question of why Obama did not use he is that he knows intuitively that is not how things work in contemporary Standard English. Obama (like any native speaker) would certainly use he if the antecedent were a name: he would say The Cambridge police acted stupidly in arresting Professor Gates when there was already proof that he was in his own home. (The version with they would be grammatical but with a different meaning: The Cambridge police acted stupidly in arresting Professor Gates when there was already proof that they were in their own home, and that could only have the meaning — absurd in the present context — that the police officers in question shared a home and were provably in it at the time of the arrest.) But antecedents like somebody are different.

Obama was trying to make a general claim about the stupidity of arresting some person x when there was already proof that x was in x's own home. The x in this paraphrase is intended as what a logician would call a bound variable. The issue at hand is which pronoun to use when expressing the same content in English. Now, Obama wasn't intending to limit himself to the claim that arresting Professor Gates was stupid. Doubtless he would think that arresting Harvard president Drew Faust in her own home, if she got snippy after she had shown her driver's license, would also be stupid — unless she had clearly committed an arrestable crime. And in contemporary Standard English, with antecedents like somebody or everyone or any citizen, people typically use the pronoun they for "bound variable" meanings in this sort of syntactic situation.

Strunk and White baldly assert that this is an error. They simply say don't use they with syntactically singular antecedents like somebody. They don't give a reason; and it is pretty clear they didn't know anything much about the literary evidence that they has been grammatical and normal with singular antecedents for six or seven centuries. Strunk and White are just wrong about Standard English syntax, here as nearly everywhere else where they deal with grammar in their book The Elements of Style.

Of course, you have a perfect right to hold the opinion that they with a singular antecedent seems distasteful or ugly to you. In that case I would advise you not to use it. But don't call it a grammatical error, because it clearly isn't one, and never has been. Don't say that it betokens a breakdown in our ability to tell singular from plural, because it doesn't.

And don't allege that it generally introduces ambiguity, because it doesn't. There is (as usual wherever pronouns are found) an ambiguity in what Obama said: it would be linguistically possible to read they and their as referring to the police — or, for that matter, to some group of otherwise unidentified third parties such as the Spice Girls. But to pick up on either of those grammatical possibilities would be going for a crazy interpretation when a sensible one was available. Nobody listening to the president misunderstood him in this way, and none of the journalists writing about it (as far as I know) even mentioned either possibility. Everyone understood his they and their as corresponding to bound variables.

Obama is a fluent and excellent speaker of Standard English, and his grammatical ear (if not his political ear!) was spot-on perfect on this occasion. Singular they was the right pronoun to use in the context. If you talk about arresting a man when he's in his own home, you're talking about arrests of males; if you talk about arresting a woman when she's in her own home, you're talking about arrests of females; if you talk about arresting a man or woman when he or she is in his or her own home, you're talking like a badly written statute or contract. Obama intuitively understood how to avoid all three of those undesired outcomes.

[Language Log reader Michael Straight tells me he once heard a clip on NPR of George W. Bush talking about the need for a father to take care of "his or her children." That would be how someone might put it if they felt anxious about committing a sexist blunder but didn't have the good sense to use singular they. —GKP]

Jonathan Coulton, "Soft Rocked By Me", 11/21/2008:

(The relevant part of the song starts about 1:00 in, or use this link, since time offsets don't work in YouTube embedding. But Coulton's pre-song explanation is also part of the package: "… ladies like a sensitive man — a little bit — but you don't want to go too far…")

Lyrics and download available here:

I thought we’d start our evening off with a glass of cold rose
Then we can sit here on the couch just holding hands
And I’ll say that you’re beautiful
You might say I’m moving too fast
Baby I don’t mind
Because I know in time

You will be soft rocked by me
Though it may take some time I know eventually
You will be soft rocked by me
I use the passive voice to show how gentle I’ll be
When I soft rock you
You will know it’s true
That you’ve never been soft rocked until you’ve been soft rocked by me

I’ll listen to the things you say about the way you feel
I’ll smile an understanding smile when your boyfriend calls
And you’ll go, but you’ll think of me
And one day you’ll knock on my door
Because you want to be
Soft rocked by me

For earlier discussions of the tendency to associate "passive voice" with inadequate masculinity, see "Passive aggression", 7/18/2006; "How long have we been avoiding the passive, and why?", 7/22/2006; "When men were men, and verbs were passive", 8/4/2006.

[Hat tip: Lee Sullivan]

Language Log is loading slowly today, due to heavy traffic from a digg.com link. Our normal weekday load is about 500-600 visits per hour (1,000-1,200 page views); apparently our current server saturates at around 2,000-2,200 visits per hour, which is what it's managing this morning. I expect that the crowds will be gone by tomorrow, and probably sooner.

Ben Schott cites some amusing items from the 1891 Ango-American Telegraphic Code ("Twittergraphy", 8/2/2009):

(Click on the image for a larger sample.)

The whole code-book is available from Google Books here. I'm sorry to say that Log is not a code-word, but a coded telegram reading "Language Hat" would be interpreted "Do not leave if you can help":

Just as la- words apparently have to do with the concept of "leaving", and ha- words have to do with "help", so lo- words deal with "loss":

So I feel that we've dodged a bullet here — "Language Log" might well have translated as "Do not leave You are a loser". But it didn't.

Unfortunately, no one (as far as I know) has written a utility for decoding messages written in (or at least interpreted according to) this code. (And this would be more fun to do if the code gave plaintext equivalents for a larger fraction of common English words.)

(A list of other commercial code-books from the golden age of telegraphy can be found here, with pointers to scanned copies.)

For people interested in language, linguistically-interesting bits grow on  pretty much all of the trees in the forest of communicative interaction. In order to get on with life, we let most of the specimens pass without comment. But the first two segments of this week's Car Talk radio show, which I listened to with half an ear while I waited for a computer program to finish running, rose to the threshold of bloggability: the first segment because it offered a nice exchange on what an "accent" is, suitable for use in my new lecture notes for this year's Linguistics 001;  and the second segment bcause it relates to a recent and celebrated British libel case.

The first listener who called in with a problem was Mary from Altanta, and the interaction started like this:

For non-American readers, I'll note that Tom and Ray Magliozzi are actually from Cambridge MA, and exhibit (perhaps sometimes in exaggerated form) the accent characteristic of working-class residents of the Boston area.  And when Mary suggests that they're from New jersey, this is clearly intended as a jocular insult (New Jersey is the Belgium of the United States, just as Belgium is the New Jersey of Europe), which Tom and Ray recognize and respond to with the remark about the "witness protection program".

The second caller was Mike from Vergennes, VT, who called because his 1993 Cadillac Fleetwood was bouncing up and down whenever he turned hard to one side or the other at low speed.  As is often the case, Tom and Ray disagreed about the diagnosis. (I'm not sure which brother is which in this passage, so I'll transcribe them as B1 and B2):

That sets up the disagreement. Now listen to how it develops:

The critical question is, what does bogus mean in this context?

According to Sir David Eady, the presiding judge in the English High Court, when Simon Singh used the word bogus to describe chiropractic treatments for "children with colic, sleeping and feeding problems, frequent ear infections, asthma and prolonged crying" as "bogus", he was asserting that such treatments are dishonest or fraudulent. (See here, here, and here for some background and discussion.)

Could bogus be used in that sense here? I don't think so. Tom and Ray are not above accusing one another jocularly of fraud, but in this case, I believe it's very clear that the word is used to mean "unsupported by evidence". That's why Brother 1 admits that his diagnosis is "bogus", but defends it anyhow because "that's the feeling I'm getting" and "I'm using all my intuition here". In other words, he has a hunch, not an inference from facts to conclusion.

And of course, "unsupported by evidence" is exactly what Simon Singh clearly (in my opinion) meant bogus to mean in his Guardian opinion piece of April 19, 2008, "Beware the spinal trap".

In the end, B2 suggests a more plausible etiology for Mike's problem: a slipping power-steering belt. And n the real world, Mike has long ago figured out what the problem really was (or junked the car) — this was an "encore presentation", i.e. a re-run, originally broadcast in 1997 or so, more than a decade before Simon Singh's article. But even if the problem turned out to be a wheel not turning properly, that diagnosis would still have been bogus, in the sense that it was unsupported by evidence. Even a bogus idea can be right from time to time.

In the past week my new credit card had been sent by courier service to someone called "Pullem"; a student paper had cited a linguist named "Pollum" for work of mine; and a kindly administrator had sent out an email to a large list in Edinburgh congratulating "our own Geoff Pullman" on being elected to the British Academy. Things had not been going well. But now the general quality of life was improving. United Airlines had asked me to switch to a later (and delayed) flight via London Heathrow on my way back from San Francisco, and for this had given me an upgrade to business class. Definitely a mood-changer. No longer the 247th economy-class passenger from the left in the departure lounge: I'm an F.B.A., and I'm sitting in business class sipping free champagne over the Rocky Mountains. Dinner is coming up soon, with a smoked salmon starter and real metal cutlery. Life is sweet. The long, long wait to board is forgotten, and I'm actually mellow. And now the purser was coming down the aisle with a seating plan on a clipboard so he could ask each passenger by name about their menu choices ("Mr. Fortescue, Mrs. Fortescue: can I ask you about your main course preferences tonight?"). He arrived at my seat and checked his clipboard. "Mr… Pullman?"

I snapped open my seat belt buckle in a half second and sprang up, grabbing him by his scrawny throat. Propelling him back down the aisle I slammed him into a bulkhead and held him there with his heels eight inches off the floor.

"What. Is. My. Name?" I demanded, in my James Earl Jones voice.

"P… Pu…", he choked out weakly, his clipboard now flapping like the wing of an injured bird. But it was too late; he had displeased me. With a twist of my arm I snapped his neck and hurled his twitching corpse against the door of the starboard side toilet.

No I didn't. I just sighed and told him that the rosemary chicken with fresh spring vegetables would be fine. But sometimes…

From Iceland, via Thor Lawrence, a Zits cartoon (from a free daily newspaper) with Eskimo snow words in it:

Says Lawrence:

I have no idea what the original text said (as far as I can see the original is dated 2004), but loosely translated it goes:

Pane 1. Son: When something has become a fixed part of society, language develops around it.
Pane 2. Son: Did you know that the Eskimo have 49 words for snow?
Pane 3. Father: Really?
Pane 4. Father: What does that signify? Son: I could find 50 expletives for when one's Internet connection drops!

Snow words and expletives!

[Addendum. It seems that we've been here before, back in the early days of snowclonology. Barry Ross points out that Mark Liberman posted the Zits cartoon back on 1/14/04, and referred to that posting a couple of days ago. I somehow missed the 2004 posting in my search through the Language Log archives. But it's nice to have another version of the text.]

Another opinion piece for our passive voice file: Marie Murray, "The passive voice is the penultimate weapon of denial", The Irish Times, 7/31/2009:

The passive voice is especially useful where apologies are required: personal apologies for what people have done personally. Because instead of having to say, “I’m sorry”, the passive voice allows a culprit to say “It is regrettable”. Instead of saying “I made a mistake” the abstract term “mistakes happened” can be evoked.

On the basis of the examples provided, it seems that Ms. Murray, who is "the director of student counselling services in UCD", subscribes to the now-dominant view that passive voice means "unclear about agency, and especially about blame". It's interesting to see that this new meaning has become international, and that the traditional interpretation of grammatical terminology has apparently not survived even in Ireland, where the government subcontracts primary and secondary education to the Catholic Church.

The essay doesn't identify the ultimate weapon of denial, suggesting that Murray is also among those who have adopted the new meaning of penultimate as "especially or intensely ultimate". She strengthens this impression by asserting that "of all the duplicitous linguistic devices designed to deny civil rights to citizens, the passive voice is supreme".

Ironically, Murray herself uses the passive voice frequently, in entirely appropriate ways, e.g. the bold-face verb groups in the quote below:

It is time to speak out against the passive voice. It is time to insist that it is not used when people are asked questions about their personal responsibility or the responsibility of the institutions they represent.

Or this passage:

Repetition of unreality is a powerful denial of reality. Repetition weakens resistance, dismantles resolve and assumes a veracity against which there is little defence. The most courageous and loquacious may try to challenge it, but they are defeated by the impenetrability of the passive voice.

Actually, that last sentence might be better in the active voice:

…but the impenetrability of the passive voice defeats them.

If so, however, it's not because the passive version leaves any doubt about who's to blame. Either way, it's clear that the writer believes that the passive voice is at fault. And that equality of clarity underlines the fact that she's wrong.

As I reported here earlier this week, I used my most recent Word Routes column on the Visual Thesaurus to debunk a widely circulated myth about Walter Cronkite: that in Sweden (or Holland) news anchors are known as "Cronkiters" (or "Kronkiters"). I got the opportunity to talk about this linguistic legend on the NPR show "On the Media," airing this weekend and available online here. I also address the shaky claim that the word "anchorman" was coined for Cronkite in 1952, the topic of a recent piece I wrote for Slate. I'm starting to worry that I'm going to get a reputation as some sort of nitpicking Cronkite-hater, but I'd like to think Uncle Walter would appreciate my fact-checking of his mythos.

In my original round of research, the two earliest examples I found for the "Cronkiter" story (claiming its use in Sweden) were in Gary Paul Gates' Air Time: The Inside Story of CBS News (1978) and David Halberstam's The Powers that Be (1979). As I mention in the "On the Media" interview, I contacted Mr. Gates to find out where he got the story from. It turns out he had read it in an early excerpt of The Powers that Be that appeared in the Atlantic Monthly in February 1976.

So Halberstam's piece in the Atlantic would appear to be the birth of the legend in its printed form, though Gates did say that Cronkite had told him he had heard of the Sweden anecdote, as had Cronkite's producers Ernie Leiser and Les Midgley. Nobody could remember where the story had originated from, however — including Halberstam, when Gates asked him about it a few years after their books were published. Since Halberstam passed away a couple of years ago, I'm afraid we'll never know where the anecdote first came from. Like many urban legends, the origins of this factoid will remain murky, even if we can trace how it has subsequently spread, like epidemiologists on the trail of an out-of-control virus.

Mae Sander has passed on this fascinating story from the joint website of the Ghana Institute of Architects and the Architects Registration Council of Ghana. According to the story (attributed to Prof. Ablade Glover of the College of Arts of Kwame Nkrumah University of Science and Technology in Kumasi), every Ghanaian chief

has a linguist. He goes on errands to convey his master's ideas, or appears in public with him.

It is the linguist who puts the chief's whispers into poetic and eloquent language. He is not only a mo[u]th-piece as he is wrongly described today but rather [an] ambassador and a very useful and prominent courtier. Indeed a chief's fame, to a great exten[t], depends upon the wisdom and eloquence of his linguist.

The symbol he carries is the symbol (usually proverbial) of the state he represents. Some depict animal or human forms while others depict just simple abstract shapes. Whatever stands on that staff or stick represents the beliefs and aspirations of the entire state. The staff itself is made of wood wrapped with either silver or gold leaf, or sometimes of solid gold or silver.

When the linguist is about to pronounce judgement he transfers the stick from his right hand for gesticulation. A linguist represents the link between the chief and his people, and the staff is his symbol of authority.

A number of these symbols are illustrated on the site, most of them with accompanying proverbs. For instance, a tsetsefly on a tortoise: "the tsetsefly follows the tortoise in vain" ("unprofitable and therefore useless venture trying to steal from a fortress").

I'm trying to imagine how this might work in the U.S. political system, with senators, representatives, governors, cabinet members, and the like in the role of chiefs. And what symbols to use?

A publication agreement that was just presented to me for signature takes minimization of typographical clutter to a new extreme:

No rights shall revert if it is not possible to reprint or reissue the Work for reasons connected with any war strikes lockouts or circumstances beyond the Publisher's reasonable control.

The underlined part, clearly intended as a 4-member nominal coordination, is not a grammatical phrase at all.

Modern written English allows either (for example) war, strikes, lockouts, or disasters (with the "Oxford comma", which is my preference), or war, strikes, lockouts or disasters (without the Oxford comma); but no current policy permits *war strikes lockouts or disasters (on the assumption that there is no such thing as a "war strikes lockout").

I don't think this is a sign of any new legal policy of total comma elimination, because elsewhere in the agreement there are coordinations punctuated with an Oxford comma policy (as one might have expected, since the publisher is Oxford University Press).

I signed anyway, because what the heck. Technically I am now committed to a condition that is nonsense because it is not even phrased grammatically; but if it is truly nonsense, no one can enforce it. And if it isn't (that is, if common sense is used to interpret the coordination), there is no problem.

I have left comments open below; but please put commas after at least after coordinates 1 thru n -2 in any n-element coordination.