The Digital Dish: Making Headlines, Not Setting Them
Most creative professionals have a hard time explaining the intricacies of their jobs to family and friends. This is especially true for typographers. Details such as the differences between type families, font widths, x-heights, or superscripting are lost on an acquaintance who barely knows the distinction between smart and dumb quotes.
So it was quite fun for most designers and others who deal with typefaces when a significant part of the story about CBS Memos regarding President Bush’s service in the Texas Air National Guard boiled down in part to looking at the fonts in the memos. What kind of typewriter in 1972 could produce the type in the memos? Could a typewriter create proportional spacing? Could a typewriter make superscripts for ordinal numbers such as 111th Group?
Suddenly fonts and typography were front page news.
It was even more exciting for me when I got an e-mail from Thomas Phinney outlining his thoughts on the authenticity of the memos. Thomas has worked with Adobe’s type group for seven years and is currently Program Manager for the Fonts & Core Technologies groups. He’s been an expert witness in court, and called on by all sorts of people. including the US Treasury, to look at document authenticity based on fonts.. If anyone could comment on the fonts used in the memos, it would be Thomas. In fact, he has been quoted in two articles by ‘The Washington Post” (Note: This site requires free registration). ABC News and “Newsweek” also contacted him regarding the memos. However, none of those organizations published all his thoughts, nor did they explain the methods he used to come to his conclusions. Here then, for the first time, is a complete interview regarding his work.
Digital Dish: What got you started examining the memos? Was it just curiosity or did someone contact you?
Thomas Phinney: Well, it started just out of curiosity. I read about the memos on a news list, including accusations of forgery. Of course I was interested, since analyzing documents to figure out if they’re forged is my idea of a good time (laughs). Seriously though, in recent years when people have approached Adobe to ask if we had anybody who could look into the fonts on an allegedly forged document, I’ve been the one to volunteer to take a look at them. I’ve always been surprised I didn’t have to beat my colleagues off with a stick. Doesn’t everybody want to play Sherlock Holmes?
By the way, I need to be very clear that Adobe itself did not ask me to analyze the memos, and I didn’t do this as part of my day job or during work hours. Adobe takes no position on the authenticity of the memos.
Anyway, that night I started looking at these memos to see if I could see if I could figure out whether they could (or could not) have come from a typewriter that existed at that time. Then the very next morning Carol Wahler of the Type Director’s Club in New York put “Newsweek” in touch with me because they were interested in getting my opinion on the memos — which as far as I know they haven’t done anything with. Adobe’s public relations department was also contacted by “The Washington Post,” which they referred to me. So all of a sudden I had a lot more motivation to get serious about looking into it. I did a bunch of research and analysis, and ran a few tests.
DD: What was your method of verification and why did you choose it?
TP: The incredibly bad reproduction of the memos makes it hard to state many things definitively. These things were first faxed, then scanned back into a computer, then exported back at really low resolution (about 120 dpi). At this point, it’s hard to judge subtleties in letterforms. I’ve had debates about what the font was, based on this, and it’s hard. But one thing that is not degraded by the reproduction is the simple question of relative line lengths. Where does each line end, relative to the lines above and below it? What letter lines up at the end of the line, between one and the next? Given proportionally spaced fonts, and a large enough sample (as the full set of four memos are), how each line aligns against its neighbors offers a sort of digital fingerprint of the widths of the font used. The degradation of the copy is no longer important.
Bottom line? The memos precisely match current digital versions of Times (and previous phototype and hot metal typesetting versions).
As shown in Figure 1, Times matches the memo fingerprint.
Figure 1: The text from the CBS memos is compared to the same text set in a computer version of Times (New) Roman (highlighted in yellow). The line widths are a very close match.
DD: Okay. So what does that mean as far as the memos being produced in 1972-73?
TP: First, the proportional spacing used in the documents meant that ordinary typewriters using monospaced fonts could not have created the memos. Fonts that are monospaced have each letter taking up the same amount of space. As shown in Figure 1, the memos use a proportionally spaced font that matches Times Roman. Common monospaced typewriter fonts of the period, such as the Prestige Elite and Courier typefaces, create line widths that are entirely different (see Figure 2).
Figure 2: The text from the CBS memos is compared to the same text set in Prestige Elite and Courier (highlighted in yellow). The proportional text in the memos has a much shorter line length than the monospaced fonts.
DD: Did all typewriters of the period use monospaced fonts? What about the IBM Selectric?
TP: Almost all typewriters used monospaced fonts back then. The Selectric was among those. The only two machines that have been touted as possible sources of the memos, which did not use monospaced fonts, were the IBM Executive typewriter, and the IBM Selectric Composer typesetter.
Now, the Executive did not offer switchable fonts, so you literally had to buy a different typewriter to get a different proportional font. None of them is particularly close to Times in design, and the one that I’ve seen which is closest is just much, much wider. Nobody would confuse it for Times. Somebody who claims to have worked for IBM at the time wrote that any font IBM had available was available for the Executive. This is a lie, unless you change the definition of “font” to somehow mean a general appearance similarity, not including the specific and exact widths and design. Any typographer uses it to mean the latter. I’ll get back to this later, regarding widths…
Now, this eliminates all typewriters of that time. What remains are the low end of typesetting machines. First, one has to understand that these were not typewriters. They took more training to use, and were slower and less efficient. They also cost a lot of money, even the cheaper ones cost $4,000, which would be like $15,000 today. But these machines did offer justification, centering, and most of them had proportional spacing.
The primary device proposed by people who’ve done their homework as being capable of doing this kind of proportional spacing, using a Times-like font, was the IBM Selectric Composer — a “Selectric” in name and external appearance only. This is actually one of these low-end typesetting machines, sold mostly to folks doing newsletters and newspapers rather than for general office work. But some offices wanting to do really spiffy stuff could have had one.
DD: Way back when I was young, I worked at my father’s office where I used a Selectric Composer. It was a difficult machine to use. Each letter had a different spacing which meant that if you needed to go back to delete a letter you had to know the proper number of units to move backwards.
TP: Cool. I’ve never actually used one, just researched them and talked to people who are aficionados — like Gerry Kaplan, who’s a sort of historian of the Composer.
Anyway, it’s good that you steered us back to character widths, because that’s what I wanted to get back to. Proportional fonts today on computers can have letters be pretty much any arbitrary width, on a grid that’s usually at least a thousand units. But the number of possible character widths was much more limited for the Composer. The IBM Composer proportional fonts all had the same relative character widths, regardless of font design. All characters were 3 to 9 units wide. Thus there is in essence only one “fingerprint” for any and all Composer fonts.
But what about the fingerprint of Times today? Is it on a thousand unit or higher scale? Oddly, it turns out that it is not. Today’s digital versions of Times have widths that descend from those used in Linotype’s phototypesetting and earlier hot-metal versions. Monotype had previously had an even more original version, with different widths, but when Microsoft licensed Monotype’s version, they wanted it to be compatible with Adobe’s, so the widths were changed to match the Adobe/Linotype versions.
Thus not only do all the main digital versions of Times used today have the same widths, but they are all based on an earlier 18-units-relative-to-height (“to the em” in font-geek-speak) system, with common characters being 5 to 17 units wide. But the IBM Composer has letters that are 3 to 9 units wide, where 9 units is around 3/4 of the point size (the “em square” to typographers). So while current versions of Times have relatively discrete widths for common characters, these widths are at a “finer grain” than early typewriters or low-end typesetters of the ’70s, such as the IBM Executive and Selectric Composer.
So what? Well, there’s just no way you can get consistent matches between the line endings (indicating the relative line widths) of an IBM Composer and the common version of Times, when you look at exactly how things line up from one line to the next, and have a lot of lines of text — like, say, four memos’ worth.
Now, I knew this to be logically true. But my friend Jules Siegel — who runs the news mailing list where I first heard about this (and is covering it on his blog — convinced me that this was far too abstract an argument. People want something they can see.
So I decided to do a little extra work. Since I had the character widths information for the IBM Composer from Gerry Kaplan’s Composer Museum, I made a digital version of a Composer font (see Figure 3). It isn’t trying to match the appearance of any particular font on the Composer, but it exactly matches the character widths of every font on the Composer, since they all have identical widths. The widths are accurate to 1/1000 of the point size (em) for each letter. This allowed me to create “virtual Composer” simulations and prove that the relative line lengths set with the Composer are quite different.
Figure 3: Some text from the CBS memos is compared to the same text set in Phinney’s simulation of the IBM Selectric Composer font (highlighted in yellow). The line widths do not match — some are longer, some are shorter. When setting entire memos, the lines align differently.
Note: You can download a PDF file showing a comparison for the entire memo. Compare it against the matching CBS memo.
DD: So why didn’t you do the same thing for the IBM Executive typewriter?
TP: On the Executive, all characters were 2 to 5 or perhaps 2 to 6 units wide (depending on who you listen to). But I don’t know that different fonts for the Executive had identical widths the way they did on the Composer. Still, with the unit system being even coarser than on the Composer, the closest possible font on the Executive would be even less close than what the Composer could do. The same logic applies, only stronger.
DD: Your research focused on the line widths of the memos. But others have made much of the use of superscript characters in some of the CBS memos. Do you have any thoughts about that?
TP: Well, first let’s make a distinction between ordinals, like th and st, and superscript. Some typewriters and low-end typesetting devices of the period had special characters for ordinals that were a little smaller and raised above the baseline. These even appear on some of the other Bush National Guard memos that are known to be authentic. But they weren’t superscripts that rose above the tops of the numbers and capital letters.
The ordinals in the memos turned up by CBS are superscripted so as to extend above the cap height and numbers of the text. Using any of the devices in question, the typist or typesetter would have to manually adjust the platen that holds the paper to create that effect. Then they’d have to move the paper back into place to continue.
However, Microsoft Word will create superscript characters that exactly match that position for superscript characters. Note that the positioning for superscript characters is not shown precisely on screen in Word — it’s higher when the text is printed or converted to a PDF file.
There is some circumstantial evidence that whoever typed the CBS memos may have tried to stop Word’s automatic replacement of ordinal text with superscript characters. There are some instances of a lowercase “el” used instead of the number one. This would have prevented insertion of the ordinals. Also another spot in the same memo inserts a space in between the number and the ordinals. This would also have prevented the automatic insertion of superscripted characters in Microsoft Word.
DD: Interesting. Someone might not have known how to turn off the AutoCorrect feature for ordinals.
TP: Correct! Not that I’m claiming the document was necessarily done in Word. But there are things like this that make one wonder. Plus, to make this document authentic, you need to imagine an incredibly careful typesetter going the extra mile on some superscripts, and making complete beginner errors on other things like adding and omitting spaces.
DD: As I was talking to friends about this interview, some mentioned that because the letters are irregular and the baseline jumps about, the memos must have been done on a typewriter (or a degraded font). How do you respond to that?
TP: It’s goofy. The image degradation we talked about earlier is a natural result of repeated low-res scanning, both from the fax machine scan and the later computer scan. As soon as you fax a hard copy of a document (not faxing directly off a computer), you’ve made all the letters irregular — and not identically irregular, either. Depending on how square the paper goes through (it was a little crooked) and how high a resolution the fax is set to, it gets worse. Make sure your fax is set to the lowest resolution for maximum degradation (usually they’re labeled “regular,” “fine” and maybe a “superfine” resolution). Then, scan the fax back into a computer with a scanner. You’ve just introduced more error and irregularity. If you either scan at low resolution, or reduce the resolution enough after scanning, it will be really bad. This last is what happened to the CBS memos. They may also have been photocopied in between at some stage or another, which would have made the problem even worse — to the degree that’s even possible.
So, every letter is different. The baseline wavers a bit. That’s no surprise. We’re not looking at an original printed directly from a computer. This is an effect of the image degradation from being scanned twice at low resolution. This is why many of the arguments based on letter shape are pretty dubious.
If you want to test it, type a memo in Times New Roman. Fax it to yourself (make sure the fax guides are loose enough that the paper can twist just ever so slightly). Get it scanned back into a computer, and then drop the resolution to 120 dpi on the final image. It will look a heck of a lot like the CBS memo at this point. Absolutely no need to use a typewriter (or a degraded font) up front!
DD: So what’s your bottom-line conclusion?
TP: The preponderance of evidence against the authenticity of the memos is incredibly high (Note: since this interview CBS has acknowledged the memos may not be authentic). The Washington Post created a graphic summarizing all the discrepancies in the CBS memos versus known authentic documents, available here (Again, free registration is required to enter this site).
At the moment, the only devices we know of that were available around 1972 that we know could have done it were high-end typesetting machines that cost a small fortune to own and to operate. Given that, the burden is on the “pro-memo” forces to come up with a specific device and demonstrate its capacity for producing the memos. I don’t believe it will happen.
DD: Anything else you’d like to add?
TP: I’d like to thank all the people who have helped me in my research: Gerry Kaplan, who helped me with information on the IBM Composer; Norm Aleks, who helped me on the IBM Executive and the Composer machines; Frank Romano and Jack Powers, who gave me information on other possible devices that I’m still following up on; a retired Adobe person, who verified the origins of the Times metrics; and Jules Siegel, for pointing me at many blogs and arguing with me enough that I realized how visual I needed to make things.
Finally, although I don’t feel any need to promote my personal political views when talking about typography and forgery; those who know me well will be well aware that my own politics clearly had no influence on this analysis.
Read more Sandee Cohen.