Skip to comments.The "Kerning" Issue (no doubt - CBS documents were kerned)
Posted on 09/11/2004 10:10:33 PM PDT by Catphish
I don't usually talk about politics in my journal, but I just can't help it in this case. This CBS forged memos thing is just absolutely hilarious and sickening at the same time. I've been glued to the computer over the past few days reading everything there is to read.
I've been very interested in all of the technical explanations and the many examples of comparisons between the memos, including the exact match between one of the memos and the exact same thing typed in Microsoft Word and the inexact match between one of the memos and the exact same thing typed on an IBM Selectric Composer.
Through all of this, there's been much talk about kerning, which is where pairs of letters in a certain order are closer together than they are when they are in the opposite order. However, no one has actually done an analysis of the document with pictures to prove that there is kerning, so I decided I'd try it in Photoshop.
I downloaded the documents from CBS, and opened up the one dated August 18, 1973. Then I picked a letter pair that kerns in Microsoft word, but doesn't kern (or at least not as much) in the opposite order. I found several instances of each letter pair in the document and pasted them into a new document so that they would line up. Here are the results:
Both of the ta sets were lined up by the leftmost line on the t and the right most line on the a. This was somewhat difficult because the document has been copied so many times that the all of the letters look slightly different. However, I'm certain that it is exact enough to prove my point.
Likewise the three at sets were lined up by the rightmost curve of the a and the rightmost line of the t.
Then the leftmost characteristic, the slash of the t on the ta pairs and the curve of the a on the at pairs, of the two different sets were lined up with one another. Then a line was lined up with the rightmost characteristic of the ta pairs to see if it lined up with the rightmost characteristic of the at pairs.
Obviously it does not. The line actually intersects with the next letter after the at pairs. The at pair is thinner than the ta pair. This is kerning.
As an example, I did the same thing using 12 pt Times New Roman font in Word.
Here again, it is obvious that there is kerning. Since this image has not been copied over and over, we can see the source of the kerning: the tail of the a is under the slash of the t. If you look at the at pair in "coat" from the memos, you can also see that the tail of the a is under the slash of the t.
NO TYPEWRITER CAN KERN because the typewriter does not know what letters are next to the letter you are typing when you type it. A word processor on a computer can do that, because it makes adjustments to what you have typed previously as you go. A typesetter can also do this because they can have sets of letters on the same block.
No typewriter wrote this memo. The only other options are that the document is a forgery, or that Killian sent out his personal memos to file to a typesetter.
Yet another nail in the coffin.
By the way, I also wanted to do an analysis of some typewritten text from an IBM Selectric Composer to prove (for absolute morons) that typewriters can not do kerning, but I can't find an example in good enough quality that has both the at and ta pairs. If anyone comes across anything, email me at raina at rainabear dot org.
No, since typewriters could have characters assymetrically positioned. What few typewriters could do is move the carriage by variable amounts after each character, and what no typewriter could do, at least without ridiculous amounts of pointless finagling, would be to move the carriage just the right amounts to match Microsoft Word's default spacing.
Perhaps "kerning" would have been the correct term when dealing with lead type; certainly it does not match modern usage.
The documents cited to not exhibit what would be called "kerning" in the modern usage. There are some slight overhangs, but such things can exist even with typewriters (note the Courier New examples above). The giveaway is not that the document is proportionally spaced, but that it happens to match Word's default behaviors so perfectly.
If you want to go even more technical, Times New Roman does support kerning, but as an optional feature, which is by default off in MS Word. Again, technical definition of "kerning" is adjustment of glyph positions for specific pairs of glyphs in addition to normal positioning (which might have overhang) Here is a demo: open Word, change zoom setting to 500% and type "f". Observe where the caret blinks - it is in the midst of "f" glyph, shifted by the "advance width" of "f" from the beginning of the line, which is less than the full width of the glyph (which inludes the overhang). This is the normal position where the next character will be placed. Kerning would adjust this position further, but only for specific glyphs. For "f" kerning is actually not needed, but it would be useful for "W" followed by "o", for example (you can use Format Font dialog to turn on and play with kerning)
To everything, kern kern kern....
negative "right offsets", meaning that the next glyph is normally placed closer than the
advancefull width of the previous glyph.
I've just addressed this on the updated page:
The upshot of it is that yes, f does have an overhang. However, a does not (and neither does t) and those are the letters that are used in the example. So an overhang can not account for the difference in widths of the at ta pairs.
Also, I typed the same sentence twice in word and turned kerning on for one of them. They look exactly the same. My conclusion is that Times New Roman must be kerned automatically.
Therefore everything I said in the original post is completely accurate, unless you want to call what it's doing something other than kerning. But that doesn't change the fact that the f and the a are closer together in af than they are in fa, and that Word and the memo do both these things, and that this isn't something you'd find on a typewriter.
Make that:But that doesn't change the fact that the t and the a are closer together in at than they are in ta.
The assertion that Times New Roman is "kerned automatically" is incorrect. See my explanation in the posts above
Look, this is a huge deal and let us be extra careful.
F-overlap and kerning BUMPmark.
And look, the top pixel of the a tail is clearly underneath the cross bar of the t.
The a and t letters don't any have overhang like f does.
So if it's not technically kerning, it's still something, whatever the technical name of it is, that's making those letter pairs different sizes.
The technical name for this is "character metrics" and they can be the same for an IBM Composer font. My point is that this is not a clincher...
Kerning algorithms were added to MS Word in 1981.....
In order for the type to overlap, the center to center spacings between type blocks must be less than 1/2 the width of the two blocks. First this requires proportional spacing, and second, unless your typewriter has memory the designer cannot allow this, because you must have kerning or you cannot guarantee that you will not get overlap of typed characters (i.e. you must have knowledge of kerned pairs).
Character metrics is computer terminology. It just refers to parameters describing a character, its width, bearings, etc. In particular different characters have different bearings from left and right. They are responsible for the effects you observe, there is no evidence of kerning (in precise, computer technology sense of the word) in the 'docs'.
Perhaps your Courier font is different from mine, but on my computer the two "M"'s in the Courier New sample definitely touch each other, and there is definite (albeit slight) overhang on the characters in the second column.
To confirm this latter point, select characters individually and note that each extends just past the edge of its selection box. Given that the selection boxes are directly adjoining, this implies that they overlap (albeit by only a pixel or two).
You are correct in your statement that kerning is necessary to allow characters to reach beyond their escapement boxes without any possibility of overlap, but because characters CAN overlap this is generally not a major issue. Unlike lead type which can be damaged if character glyphs overlap, computer type has no such limitations.
Bingo. This is the point which needs to be hammered home. If someone had a photo of Einstein standing in front of a blackboard on which were written the numbers "9911 2198 0725 2194", the photo would have to be a fake produced after 9-10-04, and almost certainly after 9-11-04. While in one sense, that set of figures should be no less likely to have been written than any other set of random figures, the likelihood that Einstein would happen to write out the numbers for four consecutive future Pick-4 drawings is basically nil.
The question is not whether someone in 1972 could have produced a document with the particular font metrics used, but whether it's plausible that anyone (or, more specifically, Mr. Killian) would have done so.
Those were the days, my friend. All of this has reminded me of how much I love Word.
Includes the technical definition of "kerning" which I was surprised to learn is not a default in Microsoft Word. However he then goes on to say that the infamous "CBS memos" were made using TrueType fonts, even more damning than "kerning."
I did a similar exercise with the letter "f" which has a setting as a computer font that causes it to overhang the next letter. See: