The IBM Selectric Composer
For a couple of days now we've been talking about whether the CBS memos could have been produced using the technology available in 1972 and 1973. We've talked about two typewriters mainly, both widely used at that time: the IBM Executive series and the IBM Selectric series.
Though the question has hardly been conclusively answered, the consensus of opinion among interested parties seems to be that neither an Executive nor a Selectric could have produced these memos.
My purpose here is not to debate the relative merits of either of those typewriters; that discussion is happening elsewhere. Rather, I want to take a moment to consider the dark horse candidate, the one piece of equipment that is widely believed to have been capable of producing a document similar to these memos, but that has been dismissed as being so improbable an alternative as to hardly bear talking about.
I'm referring to the IBM Selectric Composer. This machine resembles a sophisticated electric typewriter in most respects, but is in fact a full-fledged cold-type typesetting machine. (Cold type as opposed to hot type, machines like the Linotype that would cast entire lines of type in molten lead as the typesetter worked. Ah, those were the days.)
Whenever the topic has turned to the Selectric Composer, it has been dismissed out-of-hand as being far too expensive an item to find in an office on an Air National Guard base: The machine sold for anywhere from $3,600 to $4,400, and fonts were extra and not cheap. Furthermore, the Composer was widely agreed to be far too complicated and slow a machine to use for typing up memoranda, especially ones that were destined to go into a file and not even be distributed.
Update: Many commenters have pointed out — and I'm trying to read 'em all, I promise! — that I'm talking about $3,600 to $4,400 in unadjusted 1973 dollars here. If you use one of the widely available deflation or purchasing-power calculators, you end up with an equivalent in 2004 dollars of between about $16,000 and about $22,000. Bottom line: despite its fairly innocuous appearance, the Selectric Composer was no ordinary office typewriter. It was a pricey little number.
But the nagging question remained: Could an IBM Selectric Composer have been used to produce these documents?
I found my answer the same place everybody finds everything these days: Google. Typing "IBM Selectric Composer" into that search site took me to the aptly named ibmcomposer.org, which describes itself as "the only site on the Internet completely dedicated to the IBM 'Selectric' Composer line of typesetting machines." The site, which is run by Gerry Kaplan, includes information, scanned user manuals, and photographs of the only working IBM Selectric Composer I've been able to find. And, fortunately for me, it also includes an e-mail address.
When I first heard back from Gerry, I felt a little bad for having bothered him. He'd been fielding calls and letters all day, he told me, including an inquiry from CNN. But he was a trouper, willing — enthusiastic even — to help out.
I asked Gerry, in a fit of hubris, if he wouldn't mind trying to reproduce a sample from one of the CBS memos on his Selectric Composer. Just over an hour later, he emailed me back a sample, typed up on his Composer using the 11-point Press Roman type ball and scanned into his computer.
At first glance, the sample Gerry provided looks pretty darned close. The type is proportionally spaced, just like the type in the CBS memos. Gerry was also able to reproduce the now-infamous superscripted "th," though he had a disclaimer about that.
Superscript didn't come out so good because when you change the escapement lever (from the larger spacing to smaller spacing, and visa versa), sometimes the ball actually slips forward by a small amount, so you can see that the superscript looks disjointed.But all in all, I thought it looked pretty close. Was it possible that thirty years ago an Air Force Reserve lieutenant colonel typed up a handful of memos on a state-of-the-art typesetting machine?
I was getting ahead of myself. There's a big difference between looking pretty close and actually being pretty close. I knew I wouldn't be able to tell until I got the samples into Adobe Photoshop and superimposed them. I tinted Gerry's sample red for visibility and then overlaid it on top of the original. Here's the result.
The most obvious discrepancy was that the line-spacing — what typographers call leading (rhymes with "shredding") — was off. I e-mailed Gerry about this, and he replied: "Yes, if I had really tried, I could have matched the spacing (leading). The leading on the composer can be finely adjusted. Don't know if it is down to the single point level, but it probably is since you can set the leading according to the font, and the leading dial goes from something like 6pt up to 14pt."
Rather than asking Gerry to cough me up another sample, I simply split the lines of type apart in Photoshop and slid them down to align with the baselines of the corresponding lines of type in the original. Here's the adjusted version.
Much better … and pretty darned close to the original. But not close enough. The letterforms in the IBM's Press Roman typeface are very close to the letterforms in the CBS memo. Not surprising, since they're both based on the original Times New Roman font commissioned by the Times of London in 1931. But as we've seen already, different versions of the same font always exhibit subtle differences, usually in letterspacing. This case is no different.
Consider the first line of type. The "14" at the end of the line is almost perfectly aligned in both samples. But the word "to" in "report to commander" is significantly offset. So's "AFB." And, of course, the second line is completely out of whack. The third line is quite close, except for the superscript, the one Gerry said looked disjointed because of a slip in the carrier while he was adjusting the escapement lever.
Hey, what about that superscript? How'd he make it? I asked him via e-mail, and he replied:
To make the superscripted th, I first typed "111", then switched the font to the 8pt font, switched the escapement lever to the smaller escapement (horizontal movement), reverse indexed the paper 1/2 line up, typed the "th", indexed 1/2 line down, switched the escapement lever to the wider escapement, then changed the type ball back to the 11pt font. On other tries, I was able to produce the superscripted th much cleaner (where it looked proper), but on the one I sent you, the carrier slipped forward a little bit when I switched the escapement lever to and from the smaller spacing.Just to be clear, when Gerry says he switched to the 8-point font, he's not talking about pushing a button. He had to remove the 11-point type ball from the machine and replace it with the 8-point type ball, which in a real office would involve digging in the back of a drawer to find the seldom-used thing. Creating that superscript wasn't quick or easy, and when he did it the carrier slipped and the superscript ended up offset. Unlike the perfectly formed and placed superscripts seen in the CBS memos.
So the superscript is slightly off, and the letterspacing is significantly off. What's left? Something I didn't even think to ask about: the centered type.
Another point that is very suspicious is the centered heading. This is a snap to do with fixed spacing (like courier), but the text is centered using proportional spaced text, which means that the typist had to carefully measure the text prior to typing to calculate its exact center point. Typing a superscript, with all its steps, is simple compared to centering text proportionally without digital electronics.This point was so important to Gerry that he went out of his way to mention it to me again later in the day: centering type is hard on the Selectric Composer. Two of the memos, May 4 and August 1, 1972, feature a three-line centered head. Each of those lines of type had to be centered by measuring it carefully, doing some math, then advancing the carrier to just the right point on the page. The margin for error would be pretty wide because type can be off by a few points in either direction and still look pretty well centered. It wouldn't be objectionable unless you went looking for it. So it wasn't necessary for Lt. Col. Killian — or his typist — to be millimeter-precise.
And yet … he was.
Two letterheads typed three months apart can be superimposed on each other so perfectly that no difference at all can be seen. It's the same deal as before: the red in front was superimposed over the black behind it. You just can't see the black copy because the red copy is perfectly aligned with it. These letterheads weren't centered to within a couple of points of each other. They were centered exactly the same. Three months apart.
Remarkable.
Can we draw any conclusions from this? Well, there's always room for doubt, no matter how slim, no matter how slight. But in my opinion … yes. Based on the significant differences in letterspacing between the Composer font and the font used in the memos, the iffy nature of the superscript "th," and the unbelievable coincidence of the precisely centered headlines, I'm ready to say that the IBM Selectric Composer was not used to produce these memos.
Update: Gerry, who I swear is going to have his own blog before the end of this, had a suggestion.
Something that I think would be a good test for your website may be to reproduce the centered heading using MS Word and Times New Roman. If you can produce centered text that matches identically to the letterhead, it is, in my opinion, a true hoax. The reason is, because even if they were able to center text with a typesetting machine such as the composer, a PC (and good word processor), will center the text even more precisely, not at the "point" level, but rather on the twip level (1/1440th of an inch or 1/20th of a point).I live to please. Behold:
This is the composite image from above with the new stuff on top. The bottom layer is the first original memo headers in black. Above that is the second original memo headers in red ink. And on top of that in black is the header I created just now using Microsoft Word's default settings and clicking the "center" button. There's a little slippage because the original scans are not perfectly horizontal while the overlay I put on top is. But beyond that … looks like a dead-on match to me.
What are the odds?
An update, in two parts: First, I would never suggest ballot-stuffing an online poll. Never. However, I'm not above publicizing one that deserves attention. Tonight's CNN/Lou Dobbs "quick poll" asks the question, "Should Dan Rather and CBS News reveal the sources of the Bush memos?" Y'all all know what I think, and now so does CNN. Go give 'em your opinion.
Part the second: if any of y'all felt like dropping a dollar or two in my tip jar, that'd be just fine. I don't want anybody to feel obligated, but just know that donations are both welcome and deeply appreciated.
