The Bush "Guard memos" are forgeries! The Hailey Followup

Home Resume Bush Memos

11-Jan-05

Dr. Hailey finally responded to my several critiques of his analysis. Unfortunately, he did not do a very good job of his response. It resembles a random collection of incoherent ramblings rather than a reasoned summary of the issues. He seems to ramble on about bizarre syllogisms which seem to be more concerned with interpreting "who is right" instead of "what fact is true"; frankly, I couldn't follow it very well. At no point does he demonstrate that anything conclusion he has arrived at is actually supported by factual evidence, nor does he explain how he can see features of a font that do not exist. He even attempts to explain scientific method to a scientist. He makes claims about what he said in his report that are actually inconsistent with what is printed in his report!

He seems upset that he is beset by bloggers accusing him of all sorts of nefarious acts. Nobody has seen me whine about those who have accused me of everything from being a member of the Nazi party, to being in the pay of Carl Rove, to simply not existing. This is part of the price you pay for having a Web presence. The difference between the two of us that he actually has committed some serious errors in his analysis. I can prove his work is worthless without even assuming any attempt at fraud; his methodology is so poor that any conclusion arrived at using it is suspect.

Let me start at the beginning.

I use the phrases "prestidigitation" and "red herring" carefully. He also seems to have missed mentioning "legerdemain", which I feel I was justified in using.

Type wear

The type wear argument is nonsensical. First, he does not actually prove the inconsistencies he sees are due to type wear; he ignores other indications of problems, and even in his Figure 12, which I did not cite, he does not actually prove anything. He shows the "e" symbol apparently shows some damage, but he makes several fundamental errors in this analysis. He does not show examples of other characters with curves, such as "c", "d", "o", "p", "q", or "s". Instead he shows characters with very small bottom components, such as "h", "n", "w" and "m". The only good example he has is "f". But that is not enough of a statistical sample to arrive at a conclusion. In addition, it is easily seen that there are other errors in the "e", some of which are larger than the wear he claims, which could be due to ribbon errors. But he does not discuss this. In doing scientific measurement, you cannot place reliability on a metric that has error bounds in excess of the value claimed. He does not relate the apparent wear to type position; the problem may not be in the type ball, but in the mechanism. For example, the layout of a Selectric typeball is

So he starts with a document of unknown provenance, analyzes a few letters (as shown in the above chart), and claims that the "e" shows more wear than other letters. Actually, it is not clear that the artifact we are seeing is type wear. We are supposed to trust his assertion, presented without proof, that the distinctions he sees are representative of type wear. Maybe it is, maybe it isn't. We can't tell from his analysis. He later shows illustrations of a Selectric typeball, but shows no illustrations of what the printout from that typeball looks like. So we have nothing other than his opinion that this illustrates type wear, and this wear would be visible on a typed document. If I were going to make such an assertion, I would find a worn type ball and type something with it, then type something with a brand new type ball, and compare the two. Then I could say something meaningful about type wear. But there's not real evidence that what he is demonstrating is type wear, as opposed to perhaps ribbon wear over time, or mechanism wear.

But the whole analysis of font wear is meaningless in terms of the CBS memos. They were faxed. Faxing reduces the character to 200dpi, black and white. This is a massive amount of information lossage. The type wear he claims to see requires resolution below this limit. But the Nyquist Sampling Theorem (or the Shannon-Nyquist Sampling Theorem, since it was Claude Shannon who provided the formal mathematical proof) says that at 200dpi, the smallest feature you can resolve is at best 1/100 of an inch. The compounded effects of Moiré  effects at the initial scan and Moiré effects of the CBS scan of the faxed memos introduces even larger errors to the images. I can see errors larger than his purported "type wear" just by taking the identical document produced in Microsoft Word and faxing to myself three times! So analyzing the direct output of a Selectric typewriter is meaningless; even if he had proof that type wear shows up, the only way to justify its applicability to the CBS memos would be to take a document which showed wear, a document which showed no wear, fax them both at 200dpi, and compare all the instances of "e" that resulted. I have done this with Word documents, and there is no correlation of any faxed artifacts with the original. The same letter appears distorted in assorted different ways on a single document. The distortions are significantly larger than the "type wear" Hailey claims to see. Given the data is noisy, making assertions based on an ability to analyze data that is both below the Nyquist limit and which is demonstrably impossible to derive from a faxed copy is a meaningless exercise.

I have done this experiment, but the image is quite large, about 233K bytes, so I've put it in a separate page. This is only to save you download time in case you already believe the obvious.

I think his basic approach to type wear is flawed. He shows some nice photomicrographs of an IBM Selectric type ball; this is definitely good. He does not show the corresponding typed characters associated with those images. If I were trying to make this point, I would correlate the images of the damaged type with high-resolution scans of the results of using that type ball. Using Photoshop or Corel Photopaint, I would mirror-image either the type image or the printed image so a direct correlation could be seen. Instead, he takes some documents which he alleges show type wear, extrapolates from those documents (scanned at 4200dpi, he claims) to documents scanned at 200dpi, then rescanned at some other resolution, converted to PDF files, extracted, and then examined. I have given every step of my analysis. I do not see a corresponding sequence of detailed steps in his analysis.

I should point out that his hypothesis of type wear is a good hypothesis. However, there is insufficient data to support it. If he had the original documents, he might have a valid point of discussion; unfortunately, he compares images from documents with images that have gone through several transformations. While there is a great deal of rich information still available from those documents, type wear cannot be among them. Since his basic premise is flawed--that he can see type wear in a faxed, then scanned, then converted to PDF, then converted-to-image document, the details of his analysis of type wear are largely irrelevant. The entire argument can be dismissed as a waste of time and document space. It proves nothing, because there is simply no data capable of supporting it.

I'm not at all sure where his argument about logical failures and prestidigitation fits into the discussion. Hailey himself seems to equate "Typewriter" (the specialized font) with "typewriter" (typewriter fonts in general). He seems to think that because he can find some sort of correspondence between a modern computer font and the memos that it is clear the memos were typed on a typewriter. For example, in his Figure 8 he has a set of incoherent and unrelated sentences, some of which are provably false.

The above examples of variations in Typewriter, digitized.

He does not explain how he scanned them or otherwise got these images.

In some cases, the examples are taken directly from antique typewriters and digitized for use on computers

Unfortunately, he has no basis for this assertion, and the facts are that ITC Typewriter was not created until 1974, and by the evidence presented by ITC,

ITC American Typewriter was never intended for typewriters. It was designed in 1974 as a proportional spaced typeface that mimicked typewriter type.

This is from a message from ITC in response to a query by Geoff Matthews, a private communication quoted here with his permission. Note that the operative parts of this are "designed in 1974" and "mimicked typewriter type". At no time does ITC claim that these fonts were taken from antique typewriters and digitized for use on computers. So the entire basis of Hailey's argument, that ITC Typewriter represents digitized antique type fonts, disintegrates.

The Bush memos are done in a species from this genus.

I seriously doubt that. When Charles Johnson has already demonstrated that there is what is very close to a perfect match with Times New Roman, Hailey's analysis, based on a flawed premise (that the Typewriter font represents typewriter fonts), using an analysis (fine details) impossible to perform on a faxed copy of a document, seems to be untenable. Perhaps I failed to capitalize "typewriter" as often has he thinks I should have, but he's the one that is claiming the equivalence; apparently when I claim it, I'm in error. I guess I don't follow the logic. But trying to analyze font characteristics from individual letters, when you only have 200dpi resolution, is not going to yield meaningful results.

It is worth noting that a certified, highly respected document examiner employed by the Thornburgh panel arrives at the same conclusion Charles Johnson and I do, that the font used is Times New Roman.

Hailey claims that I have faulty logic. My logic at least follows basic rules of scientific evidence.

The scientific method

Hailey misses the point of the scientific method. He doesn't have to take my word for it; he should simply read what experts have to say. For example, he might consult http://teacher.nsrl.rochester.edu/phy_labs/AppendixE/AppendixE.html to see how a scientist explains the scientific method.

1. Observation and description of a phenomenon or group of phenomena.
2. Formulation of an hypothesis to explain the phenomena. In physics, the hypothesis often takes the form of a causal mechanism or a mathematical relation.
3. Use of the hypothesis to predict the existence of other phenomena, or to predict quantitatively the results of new observations.
4. Performance of experimental tests of the predictions by several independent experimenters and properly performed experiments

He claims that all he did was step 1, yet somehow he manages to arrive at conclusions, using the other three steps He even claims, in his incoherent diatribe,

Newcomer sees no validation in my hypothesis because I had none -- I had nothing to prove, only things to discover.

Yet he has lots of things he chooses to prove, including the fact that he can see type wear in the fonts, that the font is not Times New Roman, that it was typed on a typewriter, and so on. His document is riddled with examples of his claims that he has demonstrated that the document had to have been typed. That sure sounds like a set of hypotheses to me. He had not just one hypothesis. He has lots of them! But he apparently doesn't recognize that he has done this!

In fact, his title suggests strong that he has a hypothesis, which he spends most of the report proving: that the font in which the documents are produced can be precisely identified. He even states his hypothesis explicitly on page 5 of his report:. Note that I have pointed out his hypotheses, since he clearly is incapable of seeing them.

• I believe the claim that the Bush memos were done in Times New Roman or related typeface is not defensible. (hypothesis)
• I also believe the claim that the memos were done with current word processing is not supported by my data. (hypothesis)
• I believe the documents are printed using an impact printer (typewriter or daisy wheel) and are not digitally produced for the following four reasons: (hypothesis)

1. The font is a common typewriter typeface invented at the beginning of the 20th century and in continuous use until the computer replaced the typewriter. The font's name is "Typewriter". Although the typeface was somewhat modified for civilian communities in the 1960s, it remained commonplace in the military well into the 1970s. In short, the Bush memos were produced in a version of Typewriter commonly used in the military at the time.
2. It may be possible to find worn and damaged characters. The top left of the "t" appears to be worn to the extent that it seldom makes an impression. The "e" shows indicators consistent with physical damage. The "a" and the "s" show similar indicators of wear and/or damage.
3. Seldom used characters such as numbers, capitals, and the lower case "o", "q", and "p" (and the other less used lower case characters) show no sign of damage.
4. Flaws in the characters are not random as one would expect from artifacts from copying but are sufficiently inconsistent to imply inconsistent (mechanical) printing.
5. There are indications of white "blisters" perhaps cause by a character typed on paper that was deformed by the impact of the previously struck character.

OK, I can forgive him the confusion of four with five; I've done it often enough myself, which is why I never use a number in that position, but that's just a detail. But he goes on

• I contend that the memos were probably done in a proprietary IBM typewriter font redesigned specifically for proportional typing. (hypothesis)
• I believe the font used in the memos to be a variant of the font used in Figure 1, below. (hypothesis)

Even to make the four sub-points which are the reasons he believes they are not digitally produced, he has to form a set of hypotheses and test them.  For example,

1. The font is a common typewriter typeface.
1. Observation: it looks like a common typewriter typeface
2. Formulate hypothesis: it is a common typewriter typeface
3. Prediction: artifacts of the characters in the CBS document will resemble artifacts of characters in that typeface
4. Experiment: extract a random character from the document and see if it matches the corresponding character of the hypothesized typeface. Repeat as needed to get a statistically significant correlation.
2. The font's name is "Typewriter"
1. Observation: it looks like a common typewriter typeface
2. Formulate hypothesis: it is a computer font called "Typewriter"
3. Prediction: artifacts of the characters in the CBS document will resemble artifacts of characters in that typeface
4. Experiment: extract a random character from the document and see if it matches the corresponding character of the hypthesized computer typeface called "Typewriter". Repeat as needed to get a statistically significant correlation.
3. In short, the Bush memos were produced in a version of Typewriter commonly used in the military at that time
   1. Observation: it looks like a common typewriter typeface
   2. Formulate hypothesis: it is a computer font called "Typewriter" that was in use in the military at that time
   3. Prediction: artifacts of the document will resemble comparable artifacts of that font as it appears on military typewriters of that time
   4. Experiments:
   1. extract a random letter from the document and see if it matches a letter found in a military document of that time. Repeat as needed to get a statistically significant correlation.
   2. extract entire words from the document that correspond to words found in the Bush Guard memos and demonstrate that the letter placement used by a 1970s typewriter using that font is identical to the spacing of the letters found in the same word in the Bush Guard memos.
4. The documents were typed on a typewriter
1. Observation: characters typed on mechanical typewriters often show signs of wear
2. Formulate hypothesis: if the documents were typed on a mechanical typewriter they would show signs of wear
3. Prediction: we will see characters that exhibit this wear
4. Experiment: examine characters for signs of wear
5. Certain letters show greater wear than other letters
1. Observation: wear will be correlated with letter frequency in the English language
2. Formulate hypothesis: frequently-used letters will show more wear than less-frequently used letters
3. Prediction: letters such as "e" and "s" will show more wear than other letters
4. Experiment: examine the document to locate characters which exhibit wear
6. Inconsistent characters indicate mechanical processing
1. Observation: mechanical processes produce inconsistent character representations
2. Formulate hypothesis: if the documents are typed mechanically, they will exhibit certain inconsistencies which are artifacts of the mechanical production mechanism
3. Prediction: the documents actually exhibit such artifacts
4. Experiment: examine the document and find such artifacts.
7. The letterforms exhibit white blisters which indicate mechanical deformation of the paper
1. Observation: the letterforms exhibit white blisters
2. Formulate hypothesis: white blisters are due to mechanical deformation of the paper
3. Prediction: we will find instances where paper deformation creates white blisters
4. Experiment: Examine the documents for evidence of such white blisters

1. It is a common typewriter typeface.

There is no real evidence to support this. No known typewriter typeface has the letterforms and pseudo-kerning (hence the word spacing) of Times New Roman under Microsoft Word (using, we must re-emphasize, the default settings of Word!). There seems to be no evidence to support his hypothesis.

Of his many methodological errors, the one he fails worst at is taking into account the combination of pixel quantization and Moiré effects in faxing.  Thus at the fine detail level, there is simply insufficient information to reconstruct what the font is when examined on a letter-by-letter basis. Instead of selecting random letters, he instead selects letters that make an "best match". Throw away all the data that disagrees with your hypothesis? This is science?

There is no possible way to use any computer font as a basis of proving a document was typed on a typewriter. When I was poking around the Internet, I found the American Society of Questioned Document Examiners has a link ("Products and services") to a page describing the Haas Atlas CD set (6 CDs), which is a digital version of the printed catalog of typewriter examples. If I were trying to compare an unknown document to a possible source typewriter, I would want to use this resource, not some random computer font that contained the word "Typewriter" in it. As a non-member, it would cost me $875. If I joined, I could buy it for $675. Not like it is a big deal to do the job right.

In the measurement business, there are "primary standards", which are usually maintained either by a national organization. In the USA, this is NIST, the National Institute for Standards and Technology. A number of standards are regularly calibrated against these standards. These are known as "secondary standards". People who need to make precise measurements spend a lot of money to buy measurement equipment which has long-term stability, and regularly send it to very expensive laboratories to have it recalibrated against a secondary standard. In the case of the Haas Atlas, this represents a tertiary standard calibrated against the secondary standard (the primary standard being the images typed directly by typewriters themselves; the secondary standard being the scanned and digitized images. A tertiary standard would be the display of these either on paper or on the screen). A computer font, no matter what its name, is an uncalibrated reference, and any scientist would recognize this as a meaningless basis of comparison to determine if an actual typewriter had been used.

Here's another hypothesis:

1. The document is done in Times New Roman under Microsoft Word

1. Observation: the word spacing and letterforms resemble Times New Roman under Microsoft Word
2. Formulate hypothesis: It was produced using Microsoft Word.
3. Prediction: A document produced by Microsoft Word will be a near-perfect match for the document
4. Experiment: Produce a document using Microsoft Word. Compare it to the original document. If the match is near-perfect, the hypothesis is proven; if the match is poor, the hypothesis is false. Charles Johnson did this experiment and discovered that the document is a near-perfect match. I did the experiment with another document and got a bad match.
   1. Sub-observation: My experiment failed to match one of the CBS memos
   2. Sub-hypothesis: The document was distorted by the copying process so that it is smaller than the original
   3. Sub-prediction: scaling the document by a linear scaling factor will produce a better match
   4. Sub-experiment: I reduced the font size to 11.5 point and got a near-perfect match. Conclusion: sub-hypothesis proven

From this I conclude that the document was produced using Microsoft Word. This leads to

1. Observation: the document is produced using Microsoft Word using Times New Roman
2. Formulate hypothesis: Other artifacts of the document will be consistent with Microsoft Word
3. Prediction: there will be pseudo-kerning consistent with Microsoft font technology and specifically contain pseudo-kerning consistent with Times New Roman pseudo-kerning.
4. Experiment: find various pairs of letters that should exhibit pseudo-kerning consistent with Microsoft's Times New Roman font. Compare the documents with a known standard, an original document produced using Microsoft Word. For all combinations I tried, the experiment was successful. I demonstrate this first with simple pseudo-kerning, then a more detailed discussion, Hypothesis proven: pseudo-kerning artifacts of the document show pseudo-kerning artifacts of Microsoft Word.

1. Observation: the document is produced using Microsoft Word using Times New Roman
2. Formulate hypothesis: Other artifacts of the document will be consistent with Microsoft Word
3. Prediction: there will be superscripts consistent with Microsoft font technology
4. Experiment: find instances of superscripts and compare them with a known standard, an original document produced using Microsoft Word. For all instances of superscripts I found, the experiment was successful. I discuss one of those experiments. Hypothesis proven: artifacts of the document show artifacts of Microsoft Word.

1. Observation: there are superscripts in the document
2. Formulate hypothesis: If these superscripts are consistent with the superscripts of Microsoft Word, the document was constructed with Microsoft Word
3. Prediction: the superscripts are consistent with Microsoft font technology
4. Experiment: find instances of superscripts and compare them with a known standard, an original document produced using Microsoft Word. For all instances of superscripts I found, the experiment was successful. I discuss one of those experiments. Hypothesis proven: superscripts in the document are consistent with those of Microsoft Word.

Thus I arrive at the conclusion that the documents are produced using Microsoft Word

So let's examine some of those other issues Hailey discusses. For example

2. The documents were typed on a typewriter

He bases this on issues of type wear. So let's consider the issues of type wear.

1. Observation: mechanical type exhibits wear
2. Formulate hypothesis: documents which exhibit type wear indicate they are typed on a mechanical typewriter
3. Prediction: The CBS documents will show type wear
4. Experiment: examine the documents for type wear

This presumes that there is sufficient resolution to distinguish type wear from the artifacts of faxing. It also suggests that damaged characters will exhibit consistent wear patterns after having been faxed.

Counter-hypothesis: character distortions are due to the set of transformations including possible copies preceding the fax, the faxing operation, the scanning operation, the PDF conversion operation, the conversion to a printed document, and the scanning of the printed document to examine the actual printing effects.

1. Observation: the letterforms in the CBS memos are distorted
2. Formulate hypothesis: faxing distorts letterforms
3. Prediction: we will see inconsistent distortions of the characters based on the faxing artifacts
4. Experiment: create a Word document. Fax it several times at 200dpi (I did this once from a Kinko's; twice, several days apart, from another fax machine). If there are distortions seen in the letterforms which were not present in the original, they are artifacts of the faxing process. Such distortions are found, and in fact are inconsistent from faxed copy to faxed copy. Conclusion: the distortions of the characters are predominantly due to the faxing artifacts. Such distortions would mask any evidence of type wear.

1. Observation: faxing distorts letterforms (see previous experiment)
2. Formulate hypothesis: faxing artifacts will mask worn type artifacts
3. Prediction: It will be impossible to distinguish worn type in a faxed document
4. Experiment: (not performed, but suggested)
1. Locate a typewriter (preferably an IBM Selectric) which has a type ball that exhibits significant wear. Type any document at all using this type ball. Buy a brand-new type ball for this typewriter, which uses the same type face. Type a duplicate of the previous document using this new type ball. Fax several copies of each document. Have an independent third party analyst determine which documents were typed by the old type ball and the new type ball.
2. Refinement of experiment: Take the two original documents (one with the worn type, one with the new type) and mark each line in the far left of the margin with an "o" or "n" to indicate old or new. Then cut out the individual lines as strips of text. Paste the text onto a plain page. Lines are selected randomly from the "o" and "n" samples, but in the order of the original document. Lines are pasted with randomly-chosen indentations approximately 1/2 inch misaligned so no information is conveyed by the indentation. The indicator of the source is then covered with a separate strip of paper so it is concealed. Fax the two documents at 200dpi to a third party. Have the third party identify which lines were typed by which type ball. Compare the results with the original assignments. Note that on the average, completely random assignment should work out to 50% correlation. Repeat with several third-party examiners if necessary, or with different arrangements of the lines, if necessary to get statistically significant results.

I have not conducted this experiment, because it requires both a lot of time to prepare, and the problem is that I would be experimenting on myself, so it would not be a classic single-blind experiment. So an independent party has to produce the examples I would need to examine. In addition, I do not have an IBM Selectric typewriter with a known worn type ball that could be used to create a standard example of worn type. However, the very fine details that Hailey demonstrates are considerably smaller than the pixel quantization and Moiré effects of a 200dpi fax.

3. Certain letters show greater wear than other letters.

Given that I believe it is impossible to see wear, due to the numerous artifacts of the copying, faxing, scanning, and conversion processes involved, it would be impossible to determine the relative wear of the letters. This can actually be readily seen just by examining the actual documents. There is not, as Hailey asserts, consistent wear on letters in all memos; I can show completely inconsistent letters within a single document, and in fact, in two cases where I compared "ee" (for example, as in "agrees") I see different letterforms in adjacent letters. It seems unlikely that we would see massive wear changes between two successive keystrokes. Hailey conveniently ignores this data, which means his conclusions are suspect.

4. Inconsistent characters indicate mechanical processing.

1. Observation: faxing distorts character forms (see earlier experiment)
2. Hypothesis: the inconsistent characters are the result of the faxing process
3. Prediction: faxing a document multiple times will produce different inconsistencies, and the inconsistencies within each faxed document will be significant
4. Experiment: See previous experiment. Faxing a Word original document three times produced three different sets of distortions. I have the data to prove it. Conclusion: Hailey is correct in asserting that mechanical processing will cause inconsistent charcters, but he neglects that the faxing is the dominant mechanical process in question. Given the documents were faxed to CBS (we know this), the simpler hypothesis, that the inconsistencies are the result of the faxing and scanning processes, is sufficient to account for the character inconsistencies. My experiment proves this beyond argument.

5. The letterforms exhibit white blisters which indicate mechanical deformation of the paper

1. Observation: there are indications that certain character adjacencies may cause distortions
2. Hypothesis: this is caused by mechanical distortion of the paper
3. Prediction: typing a document in a way that would result in mechanical distortion of the paper will exhibit such artifacts
4. Experiment: type a document on an IBM Selectric typewriter, using six sheets of paper (one on top that is typed on, and five underneath to allow for distortion). Scan the resulting document at high resolution. Compare it with an identical document typed without any paper (directly onto a single sheet). If there is physical distortion of the type suggested by Hailey, we will see significant differences between the two documents, and the six-sheet document will exhibit the artifacts postulated. I did this. No significant differences were found (in exactly one case, there is a slight misplacement of exactly one character, which could also be accounted for by mechanical rotation artifacts of the type ball; I would have needed to do many more experiments to see if the typeball itself registers correctly in every case. One slight error is not statistically significant given no other character misregistered. Examine closely the e in "report" in both of the illustrations below). Given the significant and consistent artifacts postulated by Hailey were not found, the hypothesis is disproved.

Here is the result I got in an actual experiment. Of course, to get statistical significance I should perform this experiment with 1, 2, 3, 4, 5 and 6 sheets of paper, on a variety of typewriters, and perform them several times, to collect enough information to state a confidence interval of my results. Consider this an informal experiment. However, it someone wishes to prove that multiple sheets of paper do result in the phenomena Hailey hypothesizes, they are free to do so. Full details of the experiment must be disclosed.

The upper line was typed on the topmost of six sheets of ordinary 20# typing paper; the bottom line was typed on a single sheet, using an IBM Selectric typewriter. The image was scanned at 2400dpi.

Always note that you, too, can perform any of these experiments. All you need is a typewriter and access to a fax machine. The experiment with worn type is harder to perform, but could be. I have not performed it. Access to a scanner is handy, but you can do the same thing with a camera with a good macro lens and a calibrated reticule.

Then there are his absolutely hilarious statements presented as if they were facts, with nothing to substantiate them at all. Not even an attempt is made to justify the following absurd statement, which appears on page 12 of his report:

The ability of the military to produce the proportional text with a superscript "th" with a typewriter is beyond question. If I remember correctly, IBM is on record as saying they have made machines capable of that since 1944.

This has so many ridiculous claims in one place it just boggles the imagination. Let's look at them independently (by the way, he's using a cheap lawyer trick here, of building a compound sentence for which there is no simple yes/no answer. As an expert witness, I have to watch out for statements like this and corresponding questions framed in the same way). Let's break this down a bit into its component parts:

• The ability of the military to produce proportional text is beyond question.

Sure, lots of military offices had IBM Executive typewriters. Too bad TANG was not among them! So what relevance does this sentence have to analyzing the documents in question? None whatsoever, since there were no Executive typewriters available. But let's assume he couldn't know that (for example, assume he was writing his report before the existence of the IBM Executive typewriter was categorically denied by Mary Knox, Killian's secretary, instead of weeks after she made this statement).

To make this assertion, he presumes that these typewriters could produce absolutely identical output to Microsoft Times New Roman. The actual evidence, from output produced by an IBM Executive typewriter, shows this is not so. IBM Executive typewriters did not do pseudo-kerning identical to Times New Roman, for example. The letterforms were different, and the word layout was different. So even in the absence of first-person testimony that there were no such devices available, his statement is incomplete. To make it meaningful, it should say "The ability of the military to produce proportional text, in 1972, that is precisely identical in all respects to Microsoft Times New Roman, is beyond question". But if he had said that, it would be so obviously absurd that it could not hold up. (I know, I know, he doesn't believe it is Times New Roman. But no other font yet postulated can reproduce the Johnson experiment as well. And besides, who cares if another computer font is a better match? It still means it was produced on a computer).

• The ability of the military to produce a superscript "th" with a typewriter is beyond question.

Too bad he can't actually reason about this point. It is clear that an Olympia typewriter could do this; in fact, according to an actual typewriter expert, it was the only typewriter in the 1970s that had this as a standard option (ah, I hear the cries but what about nonstandard options? OK, find an example. Don't just wish there might have been an example. Find one, and find one that produces superscripts, letterforms, and layout identical to Microsoft Word using Times New Roman with the default settings). And the incredible difference in appearance between the ligature "th" of the Olympia typewriter could never be confused with the actual superscripting done by Microsoft Word. Yet the superscripts we see in the CBS documents do not at all resemble the Olympia typewriter, and precisely resemble the automated superscripting of Microsoft Word.

Now, somehow, he takes these two independent statements, one of which is true in general but not true for TANG, and the other of which is simply nonsensical because he cannot tell a ligature from a superscript, and combines them into a single sentence that we are supposed to accept as "beyond question". And this is supposed to prove the CBS documents were typed on a typewriter which did not exist at the office of origin, using a feature that did not exist on that typewriter, and in both cases the output would be visually incompatible with the documents CBS claims is authentic!

• If I remember correctly, IBM is on record as saying they have made machines capable of that since 1944.

Sure, IBM has been manufacturing the Executive typewriter since, well, I guess I could take the additional 45 seconds to look it up, rather than trust my memory: google, "IBM Executive typewriter". Click. That's not it. Scroll. "IBM Archives: 1941". Click. Oh yes, 1941, with the IBM Electromatic Model 04. When doing a scholarly work, it does pay to spend a little bit of time to actually do the research. But I digress.

Nonetheless, he uses "that" to suggest that both proportional spacing and superscripts were available on the same typewriter. There is no evidence this is so. In addition, we always have to return to the key point: none of these devices were capable of producing output precisely like Times New Roman. And no such typewriter existed at TANG anyway! So ascribing to a product a set of nonexistent features, and the ability to produce text that looks precisely like Microsoft's font, somehow proves that the CBS documents could have been typed on this typewriter. Handwave, handwave. This is a true piece of legerdemain.

Several other hypotheses were suggested, either directly to me via email, or on the blog sites. These included:

A. The documents were typed on an IBM Executive typewriter

1. Observation: The IBM Executive typewriter uses a variable-pitch font
2. Hypothesis: The documents were typed on an IBM Executive typewriter
3. Prediction: The letterforms and word spacing will match those of an IBM Executive typewriter
4. Experiment: Compare the output of an IBM Executive typewriter to output seen in the CBS memos. If the letterforms and words match, the document may well have been typed on the IBM Executive typewriter. I performed this experiment. I have actual output from a 1970 IBM Executive typewriter. I compared the letterforms and word spacing. The letterforms and word spacing of the IBM Executive typewriter are inconsistent with the letterforms and word spacing found in the CBS memos. Hypothesis disproved.

B. The documents were typed on an IBM Selectric Composer

1. Observation: The IBM Selectric Composer was capable of variable-pitch typing, multiple font sizes, ability to superscript, and had a Times-like font.
2. Hypothesis: The documents were typed on an IBM Selectric Composer
3. Prediction: The letterforms and word spacing will match those of an IBM Selectric Composer
4. Experiment: Compare the font-relative letter spacing of Microsoft's Times New Roman to the font-relative letter spacing of the IBM Selectric Composer (as documented in the IBM Selectric Composer manuals, available online). Under the metric of font-relative letter spacing, divide the font into equivalence classes. Examine the equivalence classes of the two technologies. If there is a perfect match, there is a distinct possibility that the documents could have been typed on a Selectric Composer. I did this experiment. Twice. The two devices show non-compatible equivalence classes. It is extremely unlikely that the document could have been typed on an IBM Selectric Composer. Hypothesis disproved.

or an alternate approach

1. Observation: The IBM Selectric Composer was capable of variable-pitch typing, multiple font sizes, ability to superscript, and had a Times-like font.
2. Hypothesis: The documents were typed on an IBM Selectric Composer
3. Prediction: The letterforms and word spacing will match those of an IBM Selectric Composer
4. Experiment: Attempt to type one of the documents using the IBM Selectric Composer. Compare the image produced by the IBM Selectric Composer to the image of the CBS documents. If they are compatible, there is an excellent chance that the document could have been typed using the IBM Selectric Composer. This is the experiment done by Jeff Harrell. The result is incompatible with the CBS memos. Hypothesis disproved.

C. Typewriters of the day could produce superscripts

1. Observation: A "th" resembling a superscript is seen in a document released by the White House detailing Bush's National Guard record.
2. Hypothesis: The CBS documents were typed using this typewriter
3. Prediction: The letterforms and word spacing, and the superscript, of the CBS memos will match those of this typewriter.
4. Experiment: Compare the "th" of the National Guard memo, and the font, to the fonts used in the CBS documents. If the forms are compatible, the CBS documents could have been typed using that typewriter. If the shapes are incompatible, the documents could not have been typed using that typewriter. I performed this experiment. The National Guard memos are typed using a fixed-pitch font; the "th" is not a superscript, but a ligature, the same width as the characters. The superscript of the CBS documents is in a different position, has a different shape, and is in a document of variable-pitch font and is itself variable-pitch. Hypothesis disproved.

I could do more, but you get the idea. While these latter hypotheses were not addressed directly by Hailey as such, they are consistent with the sort of questions he asked. Hailey, in spite of his protestations, actually attempted to formulate hypotheses and prove them. He used faulty logic, ignored actual facts before him, selected his examples carefully instead of using all the data available (see his caption to Figure 2, in which he states "By extracting a large number of examples from the memos, I can hope to find a few good enough to act as templates" [emphasis added-jmn]. Choose the characters you like, and ignore the rest, and you can probably match a whole lot of different fonts. But this presumes that character-by-character matching is sufficient; my approach was to look at interletter spacing and consequently at word presentation, which is a lot more meaningful), and in general failed to follow the scientific method. Just about every example he has chosen is contradicted by the data available, which did not prevent him from ignoring the actual data and arriving at erroneous conclusions.

I hope I have explained this in sufficiently simple language that both David Hailey and Corey Pein will understand it.

Copyright issues

Dr. Hailey apparently understands copyright law about as well as he understands scientific method. He even sent me a note, on 9-Dec-04, which said,

Dear Dr. Newcomer,

I have read your discussions of my work several times and find it interesting. I also read your own report.
I worry that you cannot examine the issues you examine, but you already know that. 

In the end, physical evidence will win out. But then I'm prepared to accept that; it's simply part of the process. 

You may have noticed that although I discussed your work obliquely, I was not so rude as to use your images. 

As a patent holder, you are, I assume, familiar with the various ownership issues involved here. 
Unless you get my permission first, you may not display any art I have produced -- even the press 
is not allowed to do that. Nor may you adapt my art to your purposes. You may point to it on my site, 
but you have no permission to serve it from your site. 

My document specifically reserves all such rights except by permission. Please remove any images you may 
have taken from my site from any documents (digital or otherwise) you have produced until you receive my 
permission to use them.


David E. Hailey, Jr, Ph.D.
Associate Professor -- Technical Communication
Department of English
Utah State University

I received exactly one of these. He did not send me "repeated" messages, as he claims in his incoherent ramble. Just exactly one. During the time he sent this, I was out of town. I returned on 20-Dec, and frankly had little time before the holidays to do much of anything. I was weeks behind on my email, and had to prioritize some issues. I was gone the first week of January, 2005. I did not return to my computer for serious work until 11-Jan-05. Perhaps in his world people have nothing to do all day but sit around and answer email; I, alas, have real work to do and it takes precedence over everything else.

I have both his original Web site, and the PDF file from which I extracted the image. Neither one of these contains any claim that "specifically reserves all rights except by permission". I read them very carefully. In any case, such a phrase is not necessary under the new copyright law; all publications are copyrighted at creation. And it wouldn't matter even if this phrase appeared, since Fair Use comes into play (see below).

He is also confused about the difference between patent law and copyright law, but only someone who actually knows these are different issues would know that he is confused. I was working at the Software Engineering Institute during the time that Pam Samuelson was there, and I can say that I understand a lot about patent and copyright law. You can't attend seminars by someone as good as she is and not learn something significant.

However, I did check copyright law. I actually am familiar with copyright law, having been an expert witness in a copyright dispute over code. I commend the following section for perusal by all (you can visit the Copyright Office at http://www.copyright.gov/title17/ and download the appropriate files of the U.S. Copyright Law. The excerpt below is from page 18 of the complete document you can download from that site (1.4MB PDF file, if you click that previous link). I felt that I should actually find the verbatim copy of the law to demonstrate that I in fact know what my rights are.

It took me about five minutes to locate this paragraph. Perhaps Dr. Hailey is not as familiar with copyright law as he claims to be. Perhaps he never bothered to look it up at all, but instead relied on something he once heard someone say. I have no idea. It is worth noting that, in opposition to his claim, a newspaper may actually be allowed to use his material for purposes such as news reporting. My use is criticism, comment, quite possibly even news reporting, and I believe I could make a case that teaching is another purpose of my analysis, certainly scholarship, and most certainly research. But I only need one of those to justify the use. I properly cited his work; I did not claim it for my own. I even gave Web links to his work. Therefore, I fulfilled the requirements for proper scholarly citation, and met the requirements of copyright law.

In addition, if we look in detail at the points to consider

1. The purpose of the use is to expose a poor piece of pseudo-scientific work. My work was not commercial in nature, and is clearly for nonprofit educational purposes.
2. The nature of the copyrighted work is one of discussing a controversial work, producing a specific conclusion, and I am providing a counter to that discussion, for which I need to quote certain parts of the aforesaid work.
3. I used very few of his illustrations, and they formed a very small part of his total work.
4. Since there is no market for his work, and he published it as a freely available document without any plan of financial recompense, I could not possibly have any impact on the "market" for his work.

Finally, his work was most definitely "published"; that is implicit on its being on a Web page.

So my initial reaction on seeing his request was to say "fair use, I have crises to deal with". Now that I have time to deal with it, it is clear that I have met the criteria for "fair use". I see no limitation here that a copyrighted work is limited to text only; in principle, if he had published a musical work as a MIDI file and I wanted to critique some aspect of it, or use some snippet of it for educational purposes, I am within my rights. Therefore, I am not obligated to remove any of his images from my site.

In addition, he says he is "not so rude as to use my images". Please note the permission given below, which appears on each of my Web pages on this topic. He has my explicit permission to use my images, and it would not be rude of him to do so. I consider my use of his images to be covered by §107 Fair Use, and therefore there is no rudeness involved either. I am exercising my rights under law as a researcher.

All told, the work of Dr. David Hailey Jr. is very poor. It uses an undocumented methodology, he ignores the actual data which contradicts his results, reports conclusions which are inconsistent with the actual factual data which anyone can examine, and consequently its conclusions are erroneous. As I have stated, a 12-year old with access to Microsoft Word can readily produce an experiment which contradicts his result. I said so. He doesn't like the fact I say so. That's the nature of research: adopt a bogus approach, you get bogus results, and you get poor reviews. And as I stated in my original posting, he made claims that his critics did not have the qualifications to denigrate his work. I have the qualifications, and under peer review, I would reject his work for publication based simply on its poor methodology. Given that I, and you, too, can examine the same data and demonstrate that his results are inconsistent with the actual data, this leaves very little credibility to his work.

Please feel free to quote this material, use any of my images, etc. if you are reposting. I do ask that you provide a link to this page.