01 October 2004 Blog Home : October 2004 : Permalink
Wizbang and various links to/from (and update) are all over an analysis by a Dr David Hailey, Associate Professor at Utah State University. Wizbang and co. are all over his retroactive fixing of language etc. to make clear that he never tried to actually locate a typewriter that could reproduce the doc just dug out a suitable digitized font and played with it. All this is well and good but I have two rather more fundamental criticisms that I throw out to see if the true experts can agree with.
My hypothesis is that the good professor is missing the wood for the trees. Now I have a couple of assumptions here that could be wrong (and if so bang goes this ankle-biting into the same grave as yesterday's), but if true really help. The first is that Dr Hailey is working from similar quality digital images to the rest of us (i.e. he doesn't actually have CBS' "originals" either) and secondly he has deliberately sought to explain things away in the best possible light he can. Dr Hailey has three or four good hits in his typeface analysis where he makes some cogent points but I believe that some of them are susceptable to alternative reasoning.
It is worth noting that none of Dr Hailey's actual genuinely typed examples (as opposed to his photoshop recreations) show proportional spacing and no documents in the Bush AWOL archive (including at least one typed at Ellington) show proportional typing. I was merely starting my career as an inky schoolby in 1973 but I later (early/mid 1980s) had access to a core memory computer from that era with numerous teletypes and letterprinting printers. None of the old printers (golfballs and daisywheels IIRC) could do proportional spacing and I recall being immensely impressed when I forced a 1980s era electronic typewriter to do proportional spacing instead of its standard fixed width elite font. We know there were certain models of IBM typewrite could do proportional typing and the Shape of Days blog did quite a bit of research to find out just how much of a high end model they were.
Secondly it is interesting that the real smoking gun memo - the 18 August 1973 one - is the one he makes no attempt at using as an example. This is the memo that LGF (and I and many others) replicated in word and discovered the utterly coincidental tab spacing and line breaking. The first three lines of note 1. of that memo (ending "running", "regarding" and "rating" respectively) only line up as they do in the CBS document if you print from MS Word on American defaults incliding a font of Times New Roman. As I noted on Sept 15 with all other fonts I tried these words do not line up the way they do on the CBS document and others have also failed to get the same alignment using real typewriters etc etc.
Dr Hailey describes the number 1 as having a horizontal "flag" serif in typewriter fonts as follows:
One of the important characteristics of early Typewriter is the nature of the “one.” As I
mentioned earlier, the character "1" in early Typewriter is unique. It consists of a broad,
thick, sometimes slightly curved base and a horizontal top serif that is often as long as
left side of the base. Another characteristic is the uniform width of the stroke used to
create the character.
If you look at an enlarged MS Word 12pt Times New Roman "1" such as that over on the right, you see that it doesn't do badly when it comes to matching the description above. It most certainly has a horizontal top serif as long as the left side of the base whereas, as the good Doctor himself notes, many other computer fonts (such as Palatino Linotype) have drooping flags of a different length. The issue is the perceived different stroke width between the horizontal serifs and the vertical main stroke. My rejoinder is that this is an element of either the copying or scanning process where at a certain combination of contrast brightness etc the thin horizontals become slightly fatter. Certainly as far as can we can tell Dr Hailey has not had access to CBS's documents since his images are as blurry as ours therefore I suggest he cannot tell if this particular copy/scan error has occured.
Dr Hailey also descibes the letter t as worn in the following manner:
By copying a “t” typed with an IBM typewriter and adding “wear” to the cross stroke and
ascender consistent with my hypothesis in figure 11, I am able to replicate the problems
seen in the “t’s” in the Bush memos. The character appears to be worn in a pattern of
fading toward the left and top of the cross stroke and ascender (Figure 12).
This is jolly interesting because the enlarged MS Word 12pt Times New Roman "t" on the right has a rather similar pattern of "wear". I.e. its top left hand bits are cut out. That is an artifact of the font and not surprisingly matches his note that "[t]he top left of the “t” is clearly worn to the extent that it seldom makes an impression.". One other way of saying this is that if any "t" could be found which clearly failed to exhibit this behaviour it would be evidence that the font was not Times New Roman. Dr Hailey kindly confirms my visual inspection that this is not the case.
This is in my non-expert opinion the big elephant in living room of Dr Hailey's analysis. Dr Newcomer has stated that (pseudo-)kerning was impossible in the era these documents where purportedly created and that they show indications of precisely the (pseudo-)kerning inherent in Microsoft's Times New Roman font. For example consider the fragment below (enlarged from his PDF file):
In this fragment we see the classic "fl" ligature that pseudo-kerning produces. We also see another one in the "ay" pair earlier. All in all this looks remarkably similar to the following enlarged MS Word 12pt Times New Roman screen capture of the same text:
I admit it isn't 100% identical, I suspect this is artifect of my being lazy and doing a screen capture rather than printing and scanning. I have noticed that there screen layout does tend to vary very slightly from what is printed and indeed it may vary between HP printers and printer drivers. I also consider this fragment to be decent evidence of my hypothesis that repeated copying thickens things. The words in the two images line up quite well in length as does the over all fragment but each letter looks a lot blockier. I found it curious that Dr Hailey used a bolded version of TNR when he did his compare of the same text. This was blockier which is at first sight reasonable, but it also nicely camoflages the fatc that the overall word widths are the same. If you wanted to hide an embarassing fact this could be one way to do it.
One piece of embarassing counter evidence for Dr Hailey is the letter "h". If you examine the two images in the previous paragraph it is clear that the CBS document's letter "h" has a top serif that descends slightly, it does so at precisely the same angle as the Times New Roman "h". On the other hand none of the various examples of typewriter fonts produced by Dr Hailey and copied at the start of this paragraph exhibit this top serif except just possibly the one in the middle and that one has otherwise no resemblence to the font used in the memos. Likewise the font used in the PDF's fig 5 purported memo reconstruction fails to show this feature as far as can be told (but it is hard to confirm this because the image is rather low resolution).
I was about to conclude this post that the above point when it occured to me that printing this text out on an INKJET printer rather than a LASERJET could 100% explain the noted thickness of the font cross-strokes and the overall thickness of the characters versus the enlarged screen versions. Inkjets are not as high resolution and tend to spatter slightly which could easily account for the observed differences between the CBS memos and the beautiful clean screen captures or laserjet copies that we bloggers have generally analysed.
(Note that where I quote Dr Hailey's document it is the PDF created at 14:38:50 on 29 Sepember 2004 and last modified at 14:39:33 that was stored by wizbang)