Zodiac's Grammar, Part II
Zodiackiller.com Message Board: Letters: Zodiac's Grammar, Part II
|By Ed N (Ed_N) (ac9b27eb.ipt.aol.com - 18.104.22.168) on Wednesday, August 08, 2001 - 01:31 pm:|
Here's the new thread... the first got way too long too quickly...
|By Ed N (Ed_N) (ac9b27eb.ipt.aol.com - 22.214.171.124) on Wednesday, August 08, 2001 - 01:48 pm:|
As far as Gadsby goes, since Wright consciously chose to exclude the letter
"e" throughout his novel, wouldn't that therefore mean he was forced to choose
words, syntax etc he wouldn't normally use in order to follow through with his idea?
Now, I just counted 22 e's in the preceeding paragraph (I might've missed some, however). How much would I have needed to twist and warp that sentence to exclude all e's? Would it have been recognizable as my own composition?
So I understand, the thing with Shakespeare is that he often changed various parts of his plays to suit the temperament of his audience. So, if they did not react well to one scene, he'd rewrite it for the next performance (we see this today when movies are shown to test audiences. First Blood, for instance, originally ended with Rambo being killed in a shootout with police, but the audience didn't like it, so they refilmed it with Rambo being arrested and led away. I think Stallone liked that too, because of the two sequels it spawned...). Since this happens, is it possible that rewriting to suit an audience would necessitate a different writing style? Or not?
If we think about it in that way, is it possible that Z (probably subconsciously) got himself into a certain frame of mind when he wrote his letters, and while in that frame of mind, wrote as he did, which might not have been as he normally would?
These are just some ideas I had, I'm curious what everyone thinks, and if it's even possible...
|By Sylvie (Sylvie14) (spider-ntc-td083.proxy.aol.com - 126.96.36.199) on Wednesday, August 08, 2001 - 04:05 pm:|
That's true Ed, and as you probably know, there are those experts that fervently
believe that, in fact Sir Francis Bacon wrote the plays attributed to Shakespeare (I'm not
one of them).
I do think it is possible he took on a different persona, but to change one's grammer seems unlikely.
|By Scott Bullock (Scott_Bullock) (spider-mtc-tc041.proxy.aol.com - 188.8.131.52) on Wednesday, August 08, 2001 - 04:10 pm:|
Personally, I feel it is very possible for an author to consciously change his
writing style; perhaps so much so that two separate works may not even come close to being
the same. As I noted earlier, someone attempted to use the stylometry method to determine
if Stephen King was the author of the Richard Bachman books and was unable to do so.
However, King did write the Bachman books.
To use an even more extreme example, compare William S. Burroughs' first novel to his novel, Naked Lunch. The two styles are radically different; more so than King and "Bachman."
I'm not trying to discredit stylometry. Instead, I'm wondering what the boundaries of the method are. If one could show, by using stylometry, that The Exterminator, for example, was written by the same author as Naked Lunch, then I'd be willing to say that it could work with any writer.
Then again, neither King or Burroughs were serial killers. Who knows how much a serial killer's writing style would differ between, say, The Aura phase and the Totem or Depression phase of their murder cycle? A lot? None at all? Interesting dilemma.
|By Ed N (Ed_N) (ac8f0dfe.ipt.aol.com - 184.108.40.206) on Wednesday, August 08, 2001 - 06:54 pm:|
Basically, that's my question: is stylometry foolproof? Or is it accurate only 50%, or 75%, or even 99% of the time? Even DNA is not foolproof; however, the odds of someone else having the same traits are remote indeed (but not within the realm of impossibility), which is why it is considered highly useful.
|By Douglas Oswell (Dowland) (77.philadelphia08rh.16.pa.dial-access.att.net - 220.127.116.11) on Wednesday, August 08, 2001 - 07:20 pm:|
Have a look at this URL and perhaps you can decide for yourself.
|By The Fife (Thefife) (hsa078.pool012.at001.earthlink.net - 18.104.22.168) on Wednesday, August 08, 2001 - 07:45 pm:|
You inviting to discuss this article? That's the only reason for you to throw it up on the board. Right? So where do you want to start?
|By Douglas Oswell (Dowland) (131.philadelphia01rh.15.pa.dial-access.att.net - 22.214.171.124) on Wednesday, August 08, 2001 - 08:29 pm:|
Better yet, why don't you contact some of these people and present them with your findings and your methodology? Any one of them would certainly be a far better judge than such cretins as myself or the others who post here. Since I'm sure one or more of them has mentored you somewhere along the way in your linguistic career it ought not to be too hard to get in touch.
|By The Fife (Thefife) (hsa078.pool012.at001.earthlink.net - 126.96.36.199) on Wednesday, August 08, 2001 - 08:37 pm:|
You are asking good questions. The basic problem to answering it is that it all depends on what you have. You walk up to a study and you have to take some passes on the data to know what you have. Are the deviations tight or all over the place? Do you have sample sizes large enough for the tests you really want to do? Are the available tests of any value to this study? It is not a fast answer situation as with any statistical study. For example, for quite a while I just didnt think it was worth trying with Z since I didnt have any candidate texts. I mentioned that and Tom V came up with two pages of Allen. That was great. I ran with it and found out that Z was more stable than I thought AND that Allen matched so very well. Kacsynski was not in the running. This might work out after all, I thought. Once you have done your study then you can tell with what confidence you can state your conclusion, if you have any. I have only put much value in what they call function words (non-contextual words). I look for high-use function words that have tight SDs. These are good. If you have them, then you are home free.
Most of the arguments in the field are over taking the art out of it. There is a desire to just run some text through a program and it will make all the right decisions and choose the right words and voila! the envelope please. It doesnt work that way. You still have to use good judgement and pool your knowledge of statistics and focus it to the task. I find most of the problem is when someone comes up with a new idea and they do a study and it doesnt work so well, so people on the outside take it to be a weakness in the science. Its not. Like I said, I stick to the tried to true ways that have consistently worked.
To quote from the site Doug posted: For discrimination purposes we need context-free or 'function' words and this paper reviews the seminal work of Mosteller and Wallace (1964) on function word frequencies. I agree with this. As I do this: The technique is very much in vogue now as a reliable stylometric procedure and Holmes and Forsyth (1995) have successfully applied it to the classic 'Federalist Papers' problem.
The science is here to stay. It works.
I would love to help answer questions concerning the subject. BTW for you surfers out there, in the general area we are toying with here, it falls under what they call forensic linguistics.
|By The Fife (Thefife) (hsa078.pool012.at001.earthlink.net - 188.8.131.52) on Wednesday, August 08, 2001 - 09:13 pm:|
Scott: "Who knows how much a serial killer's writing style would differ between, say, The Aura phase and the Totem or Depression phase of their murder cycle?"
A study was done on Sir Walter Scott's works before and after a stroke. He kept a recognizable signature through the years. That would be an interesting study to investigate -- serial killers and their ilk over time. I tend to think that it would not be much variance, but it probably should be done.
I also mention earlier about John Fowles writing two different books purposefully in different literary styles. Under tests his signature was seen in each piece.
|By The Fife (Thefife) (hsa078.pool012.at001.earthlink.net - 184.108.40.206) on Wednesday, August 08, 2001 - 09:31 pm:|
"Z ... while in that frame of mind, wrote as he did"
When you use non-contextual words you are using the words that no-one pays much attention to in style. It's the contextual words that they work on changing. The high-frequency function words are wrapped up at a lower level with how you express yourself. It operates at a subconscious level. That is why they are so stable. The way they fail is if they get wrapped up in the context. For example if you are talking about The Bight of Bengal and repeat it all over your texts, then "of" would forfeit use because it has been "contextualized" -- else you would have to make some effort to back the contextual use out of the text.
A study was done where they asked several people to write pieces in different styles. Like pretending you are writing a new book for the Bible. Pretend you have witnessed a murder and you are describing it -- such stuff as that. The peoples' signatures shone through such tests.
|By Bruce (Bruce_D) (pm3-01-12.sle.du.teleport.com - 220.127.116.11) on Wednesday, August 08, 2001 - 10:22 pm:|
As you and others have seen by some of my posts-at times I can be quite combative.
Sir, And I do mean SIR, my hat is off to you. Great job-by what I consider by your posts- A GREAT MIND!!!
|By Scott Bullock (Scott_Bullock) (spider-mtc-tj034.proxy.aol.com - 18.104.22.168) on Thursday, August 09, 2001 - 07:27 am:|
I wonder if we could perform a stylometry test here on the message board?
We could have Tom V post a couple of paragraphs written by somebody who would remain an "unknown" until the test was completed. Then, we could say, "It was written by one of the following 4 or 5 people who are members of the message board." Tom F could use his program, post the results, including the degree of certainty, and we could see for ourselves the science of stylometry in action.
It would be an interesting experiment for those of us who are unfamiliar with this technique. Of course, we'd need Tom V's approval and Tom F's willingness to conduct the experiment.
Tom V, Tom F, members of the board, what do you think?
|By Barb Becker (B_Becker) (px3nr.wp.shawcable.net - 22.214.171.124) on Thursday, August 09, 2001 - 08:47 am:|
A question for you...How does the 'function' of the 'function' word affect your theory? By this I mean do you take into account the difference between 'and' used as a clausal link, and 'and' used to link simple noun groups? The word serves 2 different functions. Is this reflected?
I've done some studies of my own counting the use of 'and' in each of the Zodiac letter, and 5 of TK's known texts. I was developing a real pattern, until I took that one step too many. I decided to do a forum study. Three members of this board were selected. They all fell well within the range of being highly probable Zodiac suspects. I then realized that to be valid, the study would have to look at function as well as numerical occurrence. Possibly register... letter vs paper, and level of formality. So, my question is what role does 'function', or register, or level of formality play in your studies?
Having said all that, I would have to say that my study showed problems with Bates, Mikado follow-up, Paul Avery, the Exorcist, and the Concerned Citizen. The occurrences of 'and' in the above was way out of wack with the other letters, all of which scored under 2%. These letters were all over 3%. Looking at 'Mikado follow-up' which scored 5.2% which was 10/191 occurrences, and comparing it to 7/31/69 which scored 1.1% and was 2/174 occurrences, the two are not a good match. They are roughly the same length of document, but the 'and' use is considerably higher in the 70 letter. 69 is an authenticated letter, whereas 70 is questionable. Bates and 8/7/69 are close to the same number of words as well. Bates is 3.3%, 14/420, and 8/7/69 is 1.4%, 6/416. Again, not a good match for texts of similar length. I have problems with lumping all these letters together and comparing them to the manifesto. They don't even compare well to each other.
Further analysis of the function of 'and' shows the following:
Bates... 36% joining noun groups and 64% clausal links.
8/7/69... 50-50% split
Mikado follow-up... 90% clausal links and 10% joining noun groups.
7/31/69... 100% clausal links
If there is a pattern here, I don't see it.
I'm also curious...if you're using computer software, who designed it and what controls were put on the system.
|By The Fife (Thefife) (host020.bro.capgroup.com - 126.96.36.199) on Thursday, August 09, 2001 - 12:09 pm:|
I do not know your level of expertise with statistics. I will tell you in general that people who are not comfortable and intimate with statistics have a hard time getting stylometry. Keep in mind this is not arithmetic, but statistics.
Lets say you have Allen standing in front of you. You cut him in half. Can you recognize him? Into quarters, into eighths. Eventually you get to the point where youre staring at hamburger and trying to make sense of it. Stylometry is more like taking a DNA sample and working it through the lab. I just have the feeling youre looking at hamburger.
That said, to make a broad-brush statement, anything is grist for the mill if it can be used to distinguish authors. If it doesnt distinguish, you throw it out. Within the halls of statisticians this area of study is often called discriminate analysis. You use it to discriminate.
The different uses of a word can sometimes be used to help in the study. An approach that is fruitful for one text might not work for another for any of several reasons. Another statistic that is often of value is the ratio of some words like the and/but ratio. I like to keep it as simple as possible. I tell novices, Look for high-frequency words with small standard deviations in all texts involved and you are on your way to a good indicator.
Interesting note: in studying the Federalist Papers Mosteller and Wallace noted that when a paper was down to being attributed to either Hamilton or Madison they could tell by just looking for one word. The first exclusively used while and the other exclusively used whilst.
One reason that the Federalist Papers is a well-loved subject for tests is that they had been mauled over for years by classic stylists. They had been studied and attributed to Hamilton, Madison or Jay, but there remained some heavily disputed papers. Mosteller and Wallace were in agreement with the bulk of the classic attributions and settled the dispute on the rest.
Now, concerning the letters you looked at, I am not sure what you used. What is Bates? What is Mikado follow-up? What is Paul Avery? What are the 70 letter and the 69 letter? Remember, I excluded the Little List portion since thats basically Gilberts.
Concerning software, I wrote it in C++, verified it and validated it. I initially wrote the routines in the early 80s in FORTRAN to perform some disputed authorship studies and have refined them over the years. They are not generic statistic programs but programs designed specifically for the task at hand. BTW, what do you mean by controls?
|By Scott Bullock (Scott_Bullock) (spider-wq062.proxy.aol.com - 188.8.131.52) on Thursday, August 09, 2001 - 02:57 pm:|
When did this turn into a private discussion?
|By Bruce (Bruce_D) (pm3-01-16.sle.du.teleport.com - 184.108.40.206) on Thursday, August 09, 2001 - 04:33 pm:|
It's not private. If they email each other and don't post,we'll all be in the dark.
From now on I'll start all my posts with,"Hi,Yawl."
Seriously, if people want to address questions that are posed to them, it is only courtesy to acknowledge their existence(The message is for all of us though) If they just email each other we'll all be deprived of a learning experience)
TO SCOTT BULLOCK, YOU"RE ONE OF MY FAVORITE PEOPLE ON THIS BOARD
|By Scott Bullock (Scott_Bullock) (spider-mtc-tj062.proxy.aol.com - 220.127.116.11) on Thursday, August 09, 2001 - 06:09 pm:|
I'm still wondering if anybody is interested in doing a stylometry experiment on the message board? If we personalize this method, I feel it would be a great learning experience.
Members of the board, what does everyone think?
|By Barb Becker (B_Becker) (px3nr.wp.shawcable.net - 18.104.22.168) on Thursday, August 09, 2001 - 09:47 pm:|
Fife, dare I say Fife...and y'all...
I am aware of one computer program where undergrads were used to input the data. They were basically unsupervised...What's that famous maxim...garbage in...garbage out. That's what I mean by controls.
Sorry about the letters, but I don't know how people commonly refer to them. The 69 and 70 letters are just short for the full dates that are their antecedents. 'Bates' would be Nov 1966 RPD. I don't bother with 'Gilbert' either. The follow-up is the letter which arrived shortly after the Mikado letter and had just a brief reference to the Mikado in it. Paul Avery is my error. I meant the Oct 27/70 card to PA.
I don't understand what you mean by Statistics...in what context is this relevant?
|By Tom Voigt (Tom_Voigt) (acb6b8d7.ipt.aol.com - 22.214.171.124) on Friday, August 10, 2001 - 01:07 am:|
I'm not interested in participating in a stylometry experiment, mainly because the results don't matter. There are already a million circumstantial points to connect Allen with the Zodiac crimes...I can't get too excited about another.
|By Linda (Linda) (207-172-73-76.s76.tnt1.fdk.md.dialup.rcn.com - 126.96.36.199) on Friday, August 10, 2001 - 04:26 am:|
Tom (and all) I agree that an experiment would be fun but, like you, I won't hold any results in stylometry in excitement one way or the other since there are already unlimited, extremely incriminating and undeniable circumstantial pieces of evidence (especially where writing, writing styles, grammar and codes are concerned) linking KACYZNSKI to being a VERY STRONG, VIABLE SUSPECT to the crimes...
|By Bruce (Bruce_D) (pm3-01-48.sle.du.teleport.com - 188.8.131.52) on Friday, August 10, 2001 - 04:45 am:|
WHEN YOU DON'T LIKE THE MESSAGE, SHOOT THE MESSENGER!!!!!!
|By Scott Bullock (Scott_Bullock) (spider-wq014.proxy.aol.com - 184.108.40.206) on Friday, August 10, 2001 - 07:53 am:|
So, are we saying that any results obtained through the use of stylometry would be
nothing more than circumstantial? Is the method so inconsequential that a prosecutor
wouldn't be willing to try a case based upon stylometry results alone? I ask because I
honestly don't know.
|By Sylvie (Sylvie14) (spider-ntc-ta083.proxy.aol.com - 220.127.116.11) on Friday, August 10, 2001 - 09:15 am:|
I don't think any prosecuter would ever try a case on stylometry alone, as for every "expert" that says "this is a match", you can bet your bottom dollar that the defense can find another "expert" that will say "no way".
|By The Fife (Thefife) (host020.bro.capgroup.com - 18.104.22.168) on Friday, August 10, 2001 - 09:55 am:|
Stylometry has been used in court cases more widely in other countries than in the US. As you can imagine, the defense tends to try to knock it down so courts have not tried to use it as their strong leg in a suit. I can tell you one thing, though. Outside of the court itself, it is held very highly and is used to help narrow down lines of suspects by investigators. As far as I am concerned, that's what this has done here. It's eliminated Kacynski as a suspect.
Stylometry suffers the same criticism as DNA. In the end its statistical. Even if you say the odds are 25 million to one that it is Mr. Brown, theres always the guy that questions the integrity of the sample and raises his finger in punctuation to the remark, But there still is that possibility of error, isnt there?
Stylometry, BTW, was used in the case against Kacynski as the Unabomber. Re: Author Unknown by Don Foster.
|By Douglas Oswell (Dowland) (131.philadelphia01rh.15.pa.dial-access.att.net - 22.214.171.124) on Friday, August 10, 2001 - 10:16 am:|
Fife, stylometry was never used in the case against Kaczynski because the case never
got that far. He pled guilty before his case came to trial and got life imprisonment. The
only way in which it could have been used against Kaczynski is in the affidavit used to
secure his arrest and search of his cabin. But the only stylistic comparisons mentioned in
that document are conventional. See http://www.unabombertrial.com/documents/turchie_affidavit.html
You can see these comparisons at http://www.weblogz.org/ehaugsjaa/unabom/docs/compare/ .
|By Sylvie (Sylvie14) (spider-ntc-ta021.proxy.aol.com - 126.96.36.199) on Friday, August 10, 2001 - 10:55 am:|
I do agree that other countries place more emphasis on, not just stylometry, but other areas of psycolinguistics as well, and a great deal of creedence is given to the actual handwriting. As far as usage of the auxillary verbs and as we already know handwriting -- Allen as Z fails.
I thought it funny in France that every single job interview must come with a sample of one's handwriting to be psychoanalyzed. They, however, could not understand why I found this a bit bizarre.
At any rate, to put DNA up against stylometry is not a good comparison. The courts now have standards as to the "legal accuracy" of DNA. Take Paternity tests -- one can be legally determined
to be the biological father of a child (I believe it is at 98%). And no defense attorney in the world can touch that. Remember doubt is ever present ("aliens could have done it"), but the standard is REASONABLE doubt.
Stylometry has not yet reached a level of legal
acceptability. As of yet there are no legal standards on stylometry. For the moment it just falls to the jury to decide in which "expert" they choose to believe.
|By The Fife (Thefife) (host020.bro.capgroup.com - 188.8.131.52) on Friday, August 10, 2001 - 11:36 am:|
I am sorry, but your research of the investigation of the writings of the Unabomber/Kacynski is incomplete. They used stylometry in the process and there was no disagreements in its conclusions.
It seems your links are broken under "Documents referenced in the comparison*"
|By The Fife (Thefife) (host020.bro.capgroup.com - 184.108.40.206) on Friday, August 10, 2001 - 12:06 pm:|
I don't want to argue with you on this matter. We are close on a lot of it except the conclusion. To be very frank, however, I look upon classic stylistic studies as being very subjective and somewhat of a voodoo science. I'll take the conclusion from a stylometry study any day over a stylist.
Allen is the front-runner and Kacynski's out. The statistics are out of the ballpark.
I understand Tom V's coolness toward this whole thing. If you think Allen did it, then he's already hung on the standing circumstantial evidence. I do believe that stylometry goes beyond the purely circumstantial, but you have a hard time convincing non-techno-professionals. So that's just a bunch of empty arguing. The only people left concerned over the whole thing are the Kacynski-ites. Their focus seems to be to throw stones at the science whose results they don't agree with. But that doesn't matter either, because they won't be convinced if we found a confession from Allen wrapped in Stine's bloody shirt.
I think some fence sitters have moved and the rest are polarized. That's where I think this stands.
|By Bruce (Bruce_D) (pm3-01-46.sle.du.teleport.com - 220.127.116.11) on Friday, August 10, 2001 - 01:17 pm:|
Fife, I reiterate your thoughts on your post of 12:06 p.m. and my post of 4:45 a.m.The
position of the non-scientific minds on this issue is- If YOU DON'T LIKE THE MESSAGE
-SHOOT THE MESSENGER
|By Douglas Oswell (Dowland) (28.philadelphia01rh.15.pa.dial-access.att.net - 18.104.22.168) on Friday, August 10, 2001 - 02:48 pm:|
Bruce, no one is "shooting the messenger." That implies that you're taking your frustrations out on the person whose words or ideas upset you. Go back over these posts, or indeed, over the entire message board from about July 4 on, and find the one who taunts and ridicules (on a personal level) those who disagree with him.
|By Sylvie (Sylvie14) (spider-ntc-ta024.proxy.aol.com - 22.214.171.124) on Friday, August 10, 2001 - 03:36 pm:|
I was just saying that so far there are no legal difinitions of it's legal accuracy (such as with DNA), that's all.
I think it is very interesting and I am impressed at your prowess in this field.
I'd love to see some more comparisons, but the owner of the 7-11 has spoken and apparently he does not like the flavor of this slurpee.
|By The Fife (Thefife) (host020.bro.capgroup.com - 126.96.36.199) on Friday, August 10, 2001 - 04:06 pm:|
You are right. There are no sanctioned standards for court use. I think the problem is the technique can be very damning and at the same time easily attacked by the naive. It also lacks a high-profile champion. So much of the criticism that flies about it is nit-picking infighting among people with turf to protect. It looks bad to outsiders, and I wish they would stop it and settle down. In the old days people would study each word and only include words that were very good indicators. But they could be accused of picking variables that swayed the results. They didn't, but they could be accused of it. So the tendency now is to make studies that use as many variables as possible. By being very inclusive you can be criticized less.
If you are really interested in some good (classic) discussion of the science you should try getting a hold of Applied Bayesian and Classical Inference : The Case of the Federalist Papers by Mosteller and Wallace. Its heavy-duty math, but its the nature of the beast. This is really an excellent coverage of what they did.
|By Barb Becker (B_Becker) (px3nr.wp.shawcable.net - 188.8.131.52) on Friday, August 10, 2001 - 10:44 pm:|
a few more words...you can slant stats to reflect just about anything you want as you very well know. Why are you so vague about the formula you are using?
I pointed out several problems with your method, that you have yet to address; Register, formality, and 'function' of the function word. If you factor this in I would be curious as to how that works. If you sweep it under the carpet, I say...not good enough.
Linguistic analysis has been used in Canadian courts and interestingly enough later validated by DNA testing. How about your method? Has it ever been validated by DNA testing? Where is it used, and how successful has it been, forensically that is?
Playing around with the literary style of dead authors may be interesting, but when all is said and done, there's no one left standing to argue. Kaczinsky is still standing...but fails to argue. Odd that, don't you think?
|By Linda (Linda) (207-172-73-251.s251.tnt1.fdk.md.dialup.rcn.com - 184.108.40.206) on Saturday, August 11, 2001 - 08:34 am:|
I'm certainly not one to try to shoot the messenger nor ruffle anyone's feathers
relating to the hard work put into any analysis of the case. However, if I am to be drawn
to considering acceptance of the results of stylometry in the comparison of the Zodiac
suspects to the Zodiac letters (and more particularly Kaczynski), I would very much like
to see the following comparisons.
1. Each Zodiac letter to one another to indicate which of the Z letters more closely matches the other, thus ruling out of the experiment, through stylometry, possible non-Z letters or "unconfirmed" letters such as the Citizen letter, Bates letter, etc.
2. Comparison of Kaczynski's Manifesto to other formal works of Kaczynski (i.e. the 1971 Essay and other educational papers). I would think that the stylometry results in these instances should match perfectly.
3. Comparison of all Unabomber letters to victims, editors, etc. to obtain results in comparison with each other. I would suspect that the stylometry figures would come out closely, if not perfectly. (Since we know that Ted wrote all of the items we would be comparing in Nos. 2 & 3, and the results do come out predictably accurate, through stylometry,
THEN, go to the next step and perform
4. Comparison of all Unabomber letters to victims, editors, etc. to all "known" Zodiac letters. (By comparing similar type correspondences of Unabomber and Zodiac after received controlled and acceptable results as indicated in 2 & 3, I would feel comfortable in considering the accuracy of stylometry testing in this instance).
I can't help but feel that stylometry results would tend to be more accurate when comparing similar type material to one another. In other words, when stylometry is used to identify authorship of novels or formal written works, I would think that comparison would be done through use of other "known" formal and/or similar works of the author or authors attempting to be identified.
To compare a formally written piece of work with other less formal or casual pieces (i.e. letters, poems, notes) does not seem practical nor, to me, capable of obtaining accurate results. What you have initially compared between Allen, Zodiac and Kacyznski are totally different type of written data. Allen's sample was a short, informal letter to a friend; Kaczynski's was a formally written document of persuasion; whereas, it appears that you took all Zodiac correspondences (not just the "known" ones) and lumped them into one creating a document large enough for testing purposes.
I would venture to say that Ted Kaczynski himself would scoff at results of any experiment (be it stylometry or otherwise) unless control factors are met and comparisons are made between similar and known material. As himself, Ted would always require proof and/or provide "proof" of results in any of his undertakings (i.e. mathematical works, etc.). Even his professors indicated he would always provide "more proof than necessary). Additionally, the Unabomber granted "proof" in his correspondences that he was the author. Further, in several letters, Zodiac insisted on "proving" that his missives were truly from him.
Sorry for the lengthy post; however, I felt it important for you to understand exactly where I was coming from as far as stylometry goes.
|By Eduard (Eduard) (1cust4.tnt55.rtm1.nl.uu.net - 220.127.116.11) on Thursday, August 16, 2001 - 04:18 am:|
Hi Fife (or is it High five?),
I think you should try to contact a user called "Glen". He has the same thoughts on this subject as you. Like some sort of a evil twin,hahaha!
|By Douglas Oswell (Dowland) (191.philadelphia01rh.15.pa.dial-access.att.net - 18.104.22.168) on Wednesday, August 22, 2001 - 05:51 pm:|
Just to clear up the confusion as to whether stylometry was used in the Unabom investigation, I recently e-mailed James Fitzgerald, the textual analyst who did the Unabom/Kaczynski comparisons. Fitzgerald informed me bluntly that stylometry was not used in the document comparison, nor was it used by the defense in its failed attempt to refute the prosecution's analysis. According to Fitzgerald "[T]he project was geared toward content analysis, focusing on similar topics, phraseology, semantics, and syntax."
|By Linda (Linda) (207-172-74-26.s26.tnt2.fdk.md.dialup.rcn.com - 22.214.171.124) on Wednesday, August 22, 2001 - 08:03 pm:|
Thanks for having this verified, Doug. The link you presented on August 11th, above,
showing the comparisons used in identifying Kaczynski as the Unabomber is clear in
identifying content analysis, similar topics, phraseology, etc. Concurrently, you have
done the same in comparing known documents of both Zodiac and Kaczynski, with striking
similarity, I might add:
This is the same type comparison that should be performed on all known correspondence and/or writings of EVERY suspect... Again, I feel that the key to solving the Zodiac mystery will eventually be found in the writings and WHOEVER the Zodiac turns out to be, we can almost assuredly bet that he not only wrote as the Zodiac, but was, in his own right, an individual who had a need, desire and determination to put his feelings down in writing whether it was in personal correspondence, letters-to-the-editor, diaries, etc.
Before Kacyznski was even heard of, what writing comparisons were done by investigators on the KEY suspects? What writings/documents did they compare the Z writings to for each of the suspects? What were each suspects NEEDS as it related to writing and/or corresponding with oneself or others? I don't think this is an unreasonable question to ask of the investigators.
|By Linda (Linda) (207-172-74-26.s26.tnt2.fdk.md.dialup.rcn.com - 126.96.36.199) on Wednesday, August 22, 2001 - 08:14 pm:|
Fife, on August 11, I asked if you could perform comparisons of documents of Z against
one another and documents of Unabomber against one another to see how they ranked in
stylometry (see above for details). I'm not sure if you missed that post or was unable to
perform the comparisons for one reason or another. I'd be curious what the outcome would
be especially in the comparison of one Z document to another to see how closely they
match... The same with all known Kaczynski documents. I'd like to see how they matched
against each other.
Thanks for any information you can provide.
|By The Fife (Thefife) (host020.bro.capgroup.com - 188.8.131.52) on Thursday, August 23, 2001 - 10:17 am:|
Just for the record, to clear up the confusion as to whether stylometry was used in
the Unabomer investigation: The text studies that were performed and used in the
investigation were contextual studies. These are readily accepted in courts for one very
big reason. They are just a step away from tearing apart a persons oral testimony.
Its been done for centuries and people accept this approach. You look at the
contextual words and their use and come to relationships based on this. What this lacks is
a quantitative authority, but that is not what they are after. They are following enough
threads between texts to display a preponderance of evidence, etc.
Now stylometry was performed on the texts late in the process as a verifying pass. It was not formally used. It was not entered into evidence because they just wanted it done and not allowed to be used as a red herring. It was performed and checked against the conclusions drawn by the mainstream analysis. I was told the two agreed. I was also told it took much less time than the conventional study. It was my understanding that the study was performed to verify in both directions partially to help show that stylometry worked and that it could provide a quantifying value to the study.
I can imagine very easily that Mr. Fitzgerald never even knew the stylometry study took place.
|By The Fife (Thefife) (host020.bro.capgroup.com - 184.108.40.206) on Friday, August 24, 2001 - 11:51 am:|
FYI, I have been deeply under the weather lately with a particularly nasty flu. I have completely avoided even trying to do anything that took concentration and real thought. I'm coming out of it now and will address myself to some of these Stylo questions soon.
|By Bookworm (Bookworm) (ro02-24-29-217-79.ce.mediaone.net - 220.127.116.11) on Saturday, August 25, 2001 - 06:49 am:|
Will there be any way to tell the authentic Zodiac letters from copycat ones by using
stylometry or contextual comparisons?
I know the Unabomber letters and manifesto being more complex would seem to be easier to authenticate their authorship.
Hi, I'm new and don't know a lot about the Zodiac except what I've seen on TV. Currently, I'm reading Zodiac by Graysmith. Are there any other good Zodiac books out there anyone can recommend?
|By Ryan Olesin (Ryan) (d141-193-74.home.cgocable.net - 18.104.22.168) on Saturday, August 25, 2001 - 01:52 pm:|
You'll find a number of books that mention Zodiac in one chapter, like the Serial
Killer Encyclopedia. I think John Douglas has a new book that briefly mentions Zodiac.
Graysmith's new book Zodiac Unmasked will be out in February.
Check the links section for the Zodiac-Manson connection, there is a book about that. Also check for the link to 'This is the Zodiac Speaking', on that page there is a link for Mike Kelleher's Zodiac book due November 30, 2001 I believe.
Tom's Zodiac video is also good with plenty of TV shows of Zodiac.
I should get a royalty check for all this advertising.
|By Tom Voigt (Tom_Voigt) (acb70413.ipt.aol.com - 22.214.171.124) on Saturday, August 25, 2001 - 03:34 pm:|
I've e-mailed Bookworm about straying off topic...
Back to "Zodiac's Grammar, Part II".
|By Bookworm (Bookworm) (126.96.36.199) on Thursday, September 27, 2001 - 03:02 pm:|
Happy Christmas could be Spanish. Feliz Navidad.
Feliz translates as "happy" and the Zodiac may well have liked the "z" at the end of Feliz.
Clothes and clothesline have that Z sound.
Finally, I know this is reaching, but Navidad looks like "Navy dad." It is said there is a Navy connection to the Zodiac case.
|By Ray N (Ray_N) (188.8.131.52) on Friday, October 05, 2001 - 12:06 am:|
I have searched the message board and haven't found where anyone has addressed this small and probably inconsequential point, so I thought I'd offer it. We know Z made several "nasty" references in his letters, e.g. "nasty buttons" and "nasty enough" [to trace the photos back to me]. I find this fairly odd language, and I don't remember it being at all common in the 70's when I was growing up. Does anyone else have any recollection of this word usage back then? The reason I ask is that in reading the police reports of VPD regarding Allen's child molestation arrest, there were at least two instances in the report where Allen reportedly told the boy, "I know you don't like this, but I'm just a nasty man." I know this is just more "titwillow" psycholinguistics, just thought it was quite noticeable, and for me to notice something like a clue is really saying something!
|By Bookworm (Bookworm) (184.108.40.206) on Friday, October 05, 2001 - 04:50 am:|
Ray, Some points seem inconsequential, but small details are what identify one serial
killer or suspect from another. I may be wrong, but when you get a lot of small details
pointing to a certain suspect, that's evidence.
I think you've made a legitimate observation. What I remember from the late sixties is the expression "Man!" like "Man, look at that car!"
Nasty may have been any generation. I hear kids today, say, "That's nasty!" speaking of food or anything else in general.
|By Sylvie (Sylvie14) (220.127.116.11) on Friday, October 05, 2001 - 09:22 am:|
Good Point! I remember as a child and living in SF, that "nasty all had to do with something sexual, as in the way Allen (the molester) used it. In other words, nasty kind of meant "horny", but in a negative way. Now Z gives it an altogether different kind of usage almost like mean or the way it is used today which is essentially "gross".
Does anyone know if that form of "nasty" was used in such a way regionally anywhere around the U.S.?
I know that in England it is used as "mean hearted",
so there again someone with some sort of English connection.
|By Ray N (Ray_N) (18.104.22.168) on Friday, October 05, 2001 - 10:07 am:|
Yeah, in fact I still say "Man" all the time myself. I'm getting old enough that people think that is strange language! But it seems to me that Z used "nasty" in some ways that might have been nearly unique. For instance, the way he said "I would have sent you some photos, but you would have been nasty enough....trace them to me." I find this extremely uncommon word usage. Not sexual at all and almost like he has his own context for it. You would be "mean hearted" enough... something Z knew the police would do as part of a normal investigation, but still he used "nasty" instead of "smart enough" or "clever enough" or "sneaky enough". Or with the buttons, "some nice Zodiac buttons, not nasty ones." He seems to use "nasty" in place of "insulting" or "unkind". I think Z may have just liked this word and found many opportunities to work it into his speech, thereby establishing his own quirky usage of the word. Does anyone know of any other instances where Z used "nasty"? What about Allen? For some reason, this word has really sparked my interest. Perhaps it's because I realize this is how Unabomber was caught.
|By Ray N (Ray_N) (22.214.171.124) on Friday, October 05, 2001 - 10:45 am:|
Bookworm - "I may be wrong, but when you get a lot of small details pointing to a
certain suspect, that's evidence."
You are NOT wrong. Case in point, the Unabomber. It took only a few details in his manifesto to ferret him out. Phrases like "sooner rather than later", "no big deal", and "they want to have their cake and eat it to". Anyone who says this kind of thing isn't evidence is just whistling dixie, 'cause Ted is in jail. (which is my whole reason for bringing up nasty.)
|By Douglas Oswell (Dowland) (249.philadelphia01rh.15.pa.dial-access.att.net - 126.96.36.199) on Friday, October 05, 2001 - 11:49 pm:|
Many English translations of Dostoyesvsky novels use the word "nasty" quite frequently, indicating a Russian word that has roughly the same equivalent.
|By Bookworm (Bookworm) (cb23775-b.rmvll1.il.home.com - 188.8.131.52) on Saturday, October 06, 2001 - 05:16 am:|
Maybe Dostoyesvsky novels should be checked into. If the Zodiac is reading/watching opera, then he could be reading Dostoyesvsky.
|By Douglas Oswell (Dowland) (199.philadelphia01rh.16.pa.dial-access.att.net - 184.108.40.206) on Saturday, October 06, 2001 - 12:19 pm:|
Kaczynski read both opera and Dostoyevsky. How about some of the other suspects?
|By Tom Voigt (Tom_Voigt) (acb41bd5.ipt.aol.com - 220.127.116.11) on Saturday, October 06, 2001 - 12:28 pm:|
One of Allen's favorite albums in the 1960s was the soundtrack to "Porgy and
We should start a new thread for this.