We've Been Talkin' 'Bout Jackson, Ever Since the Fire Went Out
That is, until some killjoy with a fractal geometry computer program comes along.
A physicist who is broadly experienced in using computers to identify consistent patterns in the drip paintings of Jackson Pollock has determined that half a dozen small paintings recently discovered and claimed by their owner to be original Pollocks do not exhibit the same patterns.In case you haven't been following this story, Alex Matter has been organizing an exhibition of the paintings, with an eye toward cashing in, of course. So he has quite a bit invested in the art-buying public believing these were indeed painted by Jackson. He used to have images of some of the works on his website, http://www.pollockexhibit.com/, but all you can see on that site now are responses to The New York Times article and counter arguments by Matter's expert, Dr. Ellen G. Landau, who has a fairly public battle raging with her ex-colleague, Eugene V. Thaw (an art dealer who once worked with Dr. Landau on the Pollock foundation's authentication board), who thinks they're fakes. The essence of Matter's rebuttal is that the technology is problematic, or at least not conclusive:
The analysis showed that the techniques in this painting, found in 2003 by man whose parents were friends of Pollock, differ from those in the top work. The finding, by Richard P. Taylor, a physics professor at the University of Oregon, does not prove that Pollock did not paint the works, among a cache of 24 paintings found in 2003 in Wainscott, N.Y., by Alex Matter, whose father, Herbert, and mother, Mercedes, were friends of Pollock. But it casts serious doubt on their authenticity, even as Alex Matter is planning for a major exhibition of the paintings this year. And the finding could deepen a dispute among a once-unified group of Pollock scholars who have disagreed publicly over the works' origins.
Fractal Analysis is still a very new and contested field in art authentication and is but a small part of a much broader range of technical investigations.Now it's easy to suggest "Of course, they're gonna say that...they want to sell these paintings...blah..blah...blah." But they're not the only ones questioning the relevance of the fractal geometry study. The supersmart Quantum of Wantum tribe offer a trifold of concerns with Taylor's study (if you're like me and not familiar with some of these terms, you'll have to read the whole thing on their site, where they kindly link to definitions):
OK, so even Taylor willingly admits that the study is inconclusive:
The first question is the validity of the assumption (I assume it’s an assumption) of the scale-free nature of his measurements. In particular, I’ll mention Cosma Shalizi’s Notebook on Power Laws and All That, which should be required reading around here (and in many parts of the web). Note his note at the end,
If I had, oh, let’s say fifty dollars for every time I’ve seen a slide (or a preprint) where one of us makes a log-log plot of their data, and then reports as the exponent of a new power law the slope they got from doing a least-squares linear fit, I’d at least not grumble. If my colleagues had gone to statistics textbooks and looked up how to estimate the parameters of a Pareto distribution, I’d be a happier man. If any of them had actually tested the hypothesis that they had a power law against alternatives like stretched exponentials, or even log-normals, I’d think the millennium was at hand.
So I don’t know if Taylor has done any of that or not, but there’s no sign in either the NYT or the Nature correspondence. So I’d consider that my First Concern.
I think the second concern here is, even if the assumption of a Power Law is correct, how good a classifier is the power-law exponent? This is more of a statistical question, and it’s one that I don’t see any of the newspaper journalists asking (and Taylor doesn’t appear to volunteer it? I can’t tell; maybe it just didn’t make it into print.) In other words, how well does this statistical process work on all the new Pollocks? On all the old Pollocks? How often can it tell apart different painters? What about different painters with the same style? (That is, different drip-painters.) In fact, even another quick Google search on this brings up evidence that other people are asking the same questions, and that the answers might not always be positive. You could probably find more on this, on your own. So that’s my Second Concern.
And then there’s the Third Concern, which is how I think this stuff is (or appears to be) covered in the press? Taylor’s not the only guy doing this general kind of work; how about getting a comment from another mathematically-astute forensic-art mathematician, when you write an article like this? .... Or even just a professional
statistician. I’m not asking that every journalist have also been a math-major — but if it’s not your area of expertise, at least find someone for whom it is? In general, I think issues like this (even asking questions about the power, specificity, and applicability of these kinds of statistical techniques would be nice) are (a) important, and (b) woefully neglected.
"Certainly my pattern analysis shouldn't be taken in isolation but should be integrated with all the known facts — including provenance, visual inspection and materials analysis," he said.but I think all this fuss was totally unnecessary. All I needed to know the one painting, at least, was made by Pollock was to look at it. It's clearly a self-portrait:
You don't see it?
Look to the leftside of the canvas. Here, I'll highlight it for you.
Why it's the spitting image of the man...and if that's not all the evidence anyone would need, well....
Yes, please consider this an open thread.