Tuesday, September 14, 2004

Forgery and the Morelli Method

It seems just about everyone and their dog is talking about the Killian memos and the allegations that they are forged. A useful selection of links on the subject can be found at "blogicus" here.

In any case, I've been intending to put up a post about the Morelli method for ages; one with links that I can use (this weblog, after all, is in great measure intended to be a collection resources for my own use; naturally, if any of you know of any errors in anything below, or have further resources, I'd love to learn them). So this provides a great opportunity.

The Morelli method, invented by Giovanni Morelli, was a revolution in the detection of art forgery. Prior to Morelli, determination of whether a work of art was a forgery or not was decided by the general impression of someone thoroughly familiar with the artist in question about whether this work of art was the sort of thing that the artist could have, and would have, painted. Morelli's revolution was to recognize that the obvious characteristics were precisely the ones forgers could most easily forge; the ones the expert forgers slip on are not the things we notice on general impression, even on expert and careful general impression, but the trifles. A competent forger can paint like Fra Angelico; but it would require another level of competence entirely to paint so like Fra Angelico that you paint even the eyes, fingers, and ears like Fra Angelico. It's the little slips that mark the forger. God is in the details. Some resources:

An extract from Morelli's work on the subject.

A very readable introduction to how the Morelli method would actually be used.

A general introduction to the method.

The Morelli method is easily adapted to other fields. One of the instances I find most interesting is A. Q. Morton's work on literary authorship. By analyzing occurrences of minor words - things like 'the', 'of', 'an' - vowels, sentence length, etc., Morton is able to identify differences in authorship to a reasonable degree of probability (assuming there's enough text to analyze). This involves text analysis (CUSUM), although other things need to be added to it (expert knowledge about the writing material, etc.). There are limits to this, of course; the method only detects differences in immediate authorship. For instance, Morton's work showing that Romans, 1 & 2 Corinthians, and (perhaps) Philemon are the only Pauline Epistles that match Galatians doesn't mean (assuming, as Morton does, that Galatians is certainly Paul's) that the other Pauline Epistles are not Pauline, but only that whoever is the immediate author of the others would probably have to have been different (for a good discussion of this, see here). Beyond this additional evidence would be needed. It would be entirely possible, for instance, for the difference in authorship to be merely a difference in the scribe dictated to, if the scribe had considerable freedom to put it all into good literary form. This is important to keep in mind because in the ancient world scribes often did have considerable freedom; they didn't take dictation, they provided a product given the information they were given. Thus, in Morton's work on the gospels this social fact, that scribes standardly compressed and expanded what they were told to write in order to fit as exactly as possible the 'page limit' (e.g., the length of the scroll) the client had paid for, turns out to be key to a number of the arguments. It's important to recognize this: some have treated the CUSUM technique as, in and of itself, sufficient to determine authorship; but it's clear that other knowledge must be added to the mix. It has limits and complexities. But it's clear that combined with other evidences, this sort of analysis has immense potential. See here, for example, with regard to some of Francis Bacon's writings and Thomas Hobbes; here, with regard to Shakespeare and Fletcher.

And, of course, word frequencies and the like are not the only sorts of trifles useful in detecting forgeries, frauds, and differences of authorship - depending on what exactly you have in front of you, there are things like handwriting, typography, ink type, etc. (One of the problems in the Killian memos controversy is that CBS doesn't appear to have ever had the originals, nor even close copies, but only copies several places removed. This degrades the evidence and sharply reduces the trifles one can use to help decide the issue.)

I first became interested in the Morelli method while reading Eco & Sebeok's The Sign of Three, which is on precisely the semiotic and evidential importance of trifles.