September 15, 2004

Typography, truth, and politics

The documents that CBS, Dan Rather, and 60 Minutes presented as 1972 memos from the Texas Air National Guard, with their putative revelations that George W. Bush tried to wriggle out of his obligations, are crude forgeries. The evidence for this claim is basically linguistic. There are weaker points about style (a military officer writing a memo to file with "CYA" as the subject?) and abbreviatory arcana (OETR for OER), but the strong evidence has to do with technical topics often discussed on Language Log and fairly close to the business of many modern linguists: things like character sets, typographical details, and word processing technology. Enough so, anyway, that the story does merit a brief but rather serious discussion here, and a comment at the end.

The forger was too stupid (or careless) to realize that in order to forge a 1972 document it would be useful to get hold of a 1971 typewriter. The evidence from document analysis is discussed in minute detail on numerous blogs. A thorough summary of the bloggery can be found here. Dale Franks attempts a full compendium of the evidence here. There is a highly expert typographical analysis by an expert here. I'm not a primary investigator in this, and I'm not even redoing any of this work (it doesn't need it; it will stand); I'll just discuss a few particularly strong points to give the flavor.

One small but telling observation of typography has to do with two characters I recently discussed in another context here):


The first has the HTML code ' and is known as the apostrophe or tick or pock. The second has the HTML code ’ and also ’ because it is a 9-shaped right single quote, used to match the 6-shaped left single quote. As I remarked, no font distinguishes the functions by consistent uses of these differently shaped glyphs. The Times Roman font standardly uses the character ’-shaped character for both the apostrophe function and the single right quote function, though you can insert the '-shaped glyph if you want to for some special reason. One special reason might be that you wanted to simulate a typewriter: since their invention, typewriters have had only the ' glyph. You were supposed to use it for both left and right single quote functions as well as the apostrophe function. But many people do not seem to notice the difference in shape between these glyphs. And the alleged Bush memos have ’ (see the pictures given as part of the analysis here ), the one never found on typewriters. These memos were not typed in 1972.

A second and even clearer giveaway feature is the appearance of small-font superscripts in words like 117th. In 1972 these could hardly be done at all using office equipment. If you had a fancy typesetter, the IBM Selectric Composer, which would have cost you the 1972 equivalent of about $20,000, then if you knew how you could produce something like this effect, but it was struggle, and involved stopping to adjust the paper position and change the type ball before and after the th (a blog called The Shape of Days gives the full details). But Microsoft Word's AutoCorrect feature and WordPerfect's QuickCorrect feature both automatically change 117th to 117th as you type if you leave them with the default settings the way the programs come from the factory — unless you leave a space to break up the sequence, getting a thoroughly non-standard look (117 th). The alleged Bush memos have a mixture of 117 th (with a space) and 117th (with a small-font superscript). They were typed using a modern word processor, like Word or WordPerfect, using the factory defaults. The forger was not careful enough either to switch off the automatic substitution, or to go back and remove the space in 117 th, or to go back and turn the superscript off in 117th (any of which would have been fairly easy). These memos were not typed in 1972.

A third giveaway is the positioning of the date. It matches perfectly with one of the positions you get if you just tab across the page a bit using the factory tab defaults of Microsoft Word. In fact everything in the document does, as reported here, with screen shots: if you just retype in Word with default margins, default tabs, and default AutoCorrect substitutions, every line break comes in the same place, every line comes out to the same length, even the letter positions are essentially the same down to sub-millimeter levels. These memos were not typed in 1972.

An even clearer piece of evidence lies in something very simple: the centered address at the head of each memo. The memos are not printed on Texas Air National Guard preprinted stationery as you might have expected. The address at the head is typed in the same face as the content of the memo. But the forger made the terrible mistake of using the word processor's centering function, which did the center alignment perfectly. Word processors do such things to an accuracy of something like one twentieth of a point. Typists can only do it at all in a crude way after some careful measurement, and then can only get it to an accuracy of about one character width. The paper can roll into the machine with a few millimeters' difference either way, so it is very unlikely to find the same line typed with matching distances at right and left on two different pieces of paper. Yet the centered addresses at the tops of the alleged memos about Bush match up so perfectly that if you superimpose them you can't see that there's more than one (read a bit further in the reference I gave above to this site for a demonstration). These memos were not typed in 1972.

Mostly it is conservative bloggers who are making these points. A few liberal blogs are resisting the conclusions and some hair-splitting is going on about micro-details of line spacing and superscript heights. It's beside the point as far as I can see: I would say that the forgeries were subjected to repeated faxing and/or graphical scanning to make them look fuzzy and sort of old. Faxing a fax and then making a PDF from the faxed fax will play minor havoc with letter definition and apparent position. Still the stunningly stupid enormity of the forgery is perfectly clear. You really have to be pretty ignorant about word processing (as plenty of journalists and even some bloggers may be) to doubt this evidence.

Where does that put things for the current Presidential campaign? This is Language Log, and we don't get into politics much. If the textual and typographical evidence of the Texas memo forgeries were a mainly political topic it would not be discussed here. But I am going to allow myself to say one thing about political discourse. I do have a modest proposal about the present battle of competing allegations of wartime mendacity and neglect of duty that is afflicting the Presidential campaign. But before I present it, I must stray from linguistics into a neighbouring discipline for a moment, psychology, to bring the Swiftboats Veterans for Truth story into this.

Human memories from a time over thirty years ago (even if people had not been recently talking in prejudicial terms about the events) are worth nothing. Show students a film of a fender bender accident and then ask them for a guess at the speed of the cars "smashing into each other" and some of them will report seeing broken glass in the video when there was none. Call it "bumping into each other" and the speed estimates are lower and they don't have false memories of broken glass. Stage a brief struggle between a black man and a knife-wielding white man in front of a psychology class and get them to write reports and quite a few will report that the negro had the knife. And this is the state of memory reports from only minutes ago using observers who have nothing to gain or hide. The prospects for getting from a committed Republican veteran in 2004 a totally uncolored memory of a couple of minutes on the Mekong river in 1968 when the crux concerns what a certain fleeing Vietnamese man was wearing and the outcome of the Presidential election might hang on what happened? Zero. Nul. Nada. No chance. Not even given the very best intentions, which we probably do not have in this case. It's possible that even contemporary reports might get things wrong. Forget remembered reports by highly interested parties over three decades later.

The Swifties' stories about Kerry therefore align with the story of dereliction of duty told in the forged memos about Bush: none of this nonsense is worth a serious person's time. There is exactly one thing we can do about those stories that is rational: accept military records and actions as definitively settling the question, for both sides. Was Bush in effect a deserter? The records of the Texas Air National Guard say he was not; they gave him an honorable discharge. End of story. Was Kerry a minor hero? The records of the Navy say he was; they gave him a Silver Star Medal for gallantry in action, a Bronze Star Medal with Combat V device for heroic achievement, three Purple Hearts for shrapnel wounds to his arms, legs and buttocks, and an honorable discharge. End of story.

I'm not saying end of true story. I'm saying end of story. We can go no further than this with any hope of arriving at truth. Mark Liberman served in the Army in Vietnam and was discharged. George W. Bush served in the Texas Air National Guard and was discharged. John Kerry served in the Navy in Vietnam and was discharged. After all these years, we must just let the accepted official permanent records of such bygone military service stand, put a bilateral stop to this inexpert fiddling with Vietnam-era history, and turn to more pressing contemporary matters.

We'd better. Because there are political issues (they will not be discussed here) about which I need to hear some answers. Not just the stupid contentless political blather of which I wrote light-heartedly a few days ago, but actual answers to compelling economic and law enforcement and governmental and military policy issues on which the fortunes of my country are going to turn.

I get no answers though, because the conduct and content of the two main Presidential campaigns is dominated and driven by lying, forging, conniving, slandering, mendacious, frothing, snarling assholes who seem to think that spreading innuendo and forgery and calumny and fraud and rumor across the landscape will help to turn voters toward their favored guy even if he completely avoids substantive discussion of anything that could be of relevance. Well, they have profoundly misjudged at least one very angry voter.

We have free speech in this country, and access to a magnificently flexible and expressive language. This power of linguistic expression, granted to our species alone it would seem, is strong magic. We must be very careful what we do with it. This descent into slander and false memory recovery and document forgery and history denial and mutual accusations of cowardice and treachery is not the free discussion of political matters that the authors of the First Amendment envisaged for us. It is not political activity at all; it is the destruction of political activity.

Posted by Geoffrey K. Pullum at September 15, 2004 11:18 AM