What kind of evidence is portraiture?

December 14, 2020 4:09 PM

How Scientists Use and Abuse Portraiture (Hyperallergic). Many scientific studies assume that painted faces are factual representations of flesh-and-blood countenances. Yael Rice and Sonja Drimmer explain why this isn’t just false - it’s preposterous.
posted by adrianhon (29 comments total)

This was great. The "science" discussed in this reminds me of the character in Vonnegut's Mother Night (or possibly the movie--it's been a long time) who concludes that Jesus was Aryan on the basis of inspecting hundreds of portraits of the man.
posted by sy at 4:57 PM on December 14, 2020 [14 favorites]

This is the kind of thing where you should be required to take a semester of art history and maybe a figure drawing class before being allowed to do any kind of “study” like this...and then at least one of your peer reviewers should be from the art department.
posted by rockindata at 5:13 PM on December 14, 2020 [5 favorites]

NONE EVIDENCE
Sorry am art historian. Evidence of conventions of beauty; evidence of social aspiration (of sitter or painter). I’m sure the article covers more and I will read it, this is just my visceral reaction.
posted by Lawn Beaver at 5:31 PM on December 14, 2020 [5 favorites]

None evidence with art beef?
posted by mhoye at 6:48 PM on December 14, 2020 [34 favorites]

It does seem a bit light on evidence of the base claim "Many scientific studies assume that painted faces are factual representations of flesh-and-blood countenances." All that I see cited are two papers and a CNN article.
posted by eruonna at 7:35 PM on December 14, 2020 [2 favorites]

Usually I've heard of people looking at old paintings for fruit:

https://www.atlasobscura.com/articles/fruit-in-renaissance-art


https://www.vox.com/2015/7/28/9050469/watermelon-breeding-paintings




The "trustworthiness" paper, they don't seem to be assuming it's a "factual representation." They are willing to acknowledge the portraits will exaggerate features:

people’s preference for friendly-looking faces leads painters to exaggerate “neotenic” features in their portraits (big eyes or round faces)

Their mistake is in thinking the same features would be consistently associated with "trustworthiness" across 500 years.

We thus checked that the algorithm was susceptible to the same biases as humans, i.e., rating younger, feminine, and happy faces as more trustworthy.

As in they tested it against biases of modern humans that would have been much the same people used to train the algorithm in the first place, no?

I wonder if there's some bit of evo-psych attitude slipping in here, that they're quick to assume any association with "trustworthiness" is innate rather than cultural.
posted by RobotHero at 8:11 PM on December 14, 2020 [4 favorites]

A lot of interesting and valid ideas, but the connections going from facial eugenics to pseudo-scientific studies of portraiture to the biases of facial recognition algorithms are never clearly laid out even if they may be related.
posted by blue shadows at 8:14 PM on December 14, 2020

It makes a lot of sense that the Hapsburg jaw would get exaggerated in the paintings if the painter wants to emphasize that they're a Hapsburg.

My friend and I are now doing impersonations of Hapsburgs getting outraged at painters for painting them with peasant-sized jaws.

Scientific study that babies used to be uglier.
posted by RobotHero at 8:40 PM on December 14, 2020 [8 favorites]

OK but parenting secret: sometimes babies look like old men.
posted by muddgirl at 8:51 PM on December 14, 2020 [14 favorites]

Hot take: the Habsburg jaw is just a series of awkward attempts to apply cutting-edge 1500s science – the Mercator Projection – to portraiture. It’s exaggerated because it’s far from the equator!
posted by oulipian at 9:58 PM on December 14, 2020 [14 favorites]

What better way to “prove” one’s purebred status as a Habsburg than to have a portrait painted with the trademark trait?

"Purebred," LOL
posted by chavenet at 12:50 AM on December 15, 2020

"One way in which an artist could have signaled a sitter’s pedigree would have been to portray him or her with a particular physical feature for which a relative or ancestor was known."

Gosh, they're asking a lot of "could" there. Given that conventional portraiture has a known history of flattering its subjects by showing them as attractive as possible, it would be extraordinary to deliberately emphasise or invent a disfigurement like a jaw so portruding that it makes speech and eating difficult.

Sadly, the authors of the article don't offer any evidence that any portraitists ever did so, either in the case of Habsburgs or anybody else.
posted by vincebowdren at 2:33 AM on December 15, 2020 [2 favorites]

vincebrowdren: That's why it's an article, not a scientific study.
posted by adrianhon at 2:50 AM on December 15, 2020 [3 favorites]

Usually I've heard of people looking at old paintings for fruit...

I guess you have some hope of finding accurate information about elements that the artist had no reason not to paint realistically.

(Although... how many wealthy patrons would have paid an artist for a still life featuring exotic fruit that they couldn't actually afford but wanted everyone to think that they could afford?)

I've heard from a historical re-enactor that if you want to use paintings as sources for costuming you should look at the people on the peripheries -- because the religious figures in the middle of the triptych are going to be dressed in a completely fantastic "olden times" style, but the people in the crowd scene may actually be the very wealthy burghers who paid for the triptych and wanted everyone to remember them blinged out in their very real and very fancy fashionable clothes.
posted by confluency at 3:28 AM on December 15, 2020 [4 favorites]

I take the notion that commissioned portraits were painted less to represent reality and more to please the patron as established scientific fact because as a sometime professional portrait photographer jesus fucking christ.
posted by seanmpuckett at 4:47 AM on December 15, 2020 [14 favorites]

Whoever is doing these "studies" suffers from a bad case of engineer's disease. You see it in medicine, too, where Breughel's Yawning Man is trotted out to illustrate a fairly rare dystonic syndrome rather than ... a guy yawning in the tradition of apparently a lot of Dutch yawning guys (maybe there was an epidemic of yawning in 17th c Holland? all that tulip collecting must have been rough)

I will say, though, when I went to Italy on vacation, Google Photos kept asking me to tag statues as people. Who is this? It's Bernini's David, man.
posted by basalganglia at 5:15 AM on December 15, 2020 [2 favorites]

Although the authors of the article may be right that the scientists aren't looking at these works with an artist's eye, I find it a bit frustrating that they commit the same mistake they're claiming to correct for. They're saying that scientists are neglecting the value of an artistic perspective to fully understand the material they're studying, but the authors of this article similarly shun the scientific perspective.

The scientists who performed the original studies did discover trends in art, and interpreted them. The authors of this article condemn the interpretations, but often don't provide an alternate explanation or hypothesis for the observed trends:

For instance, a recent paper that argues that “trustworthiness displays” in portraiture increased in accordance with a “rise of democratic values observed in Western Europe.” The authors quantify trustworthiness by measuring the curve of lips and width of eyes. But what, exactly, is “trustworthiness” to begin with? This is a term that the authors of the article leave undefined, and, as those who work in the field of intellectual history would argue, the concept of trust and how it is expressed are by no means universal.

And yet - there _were_ changes to the curve of lips and width of eyes over time! And even if that doesn't reflect trustworthiness, discovering a gradual change over time that might not be obvious to the naked eye is interesting and the authors of this piece ought to at least make the case that it's not interesting rather than just treating that as a given.

This isn't an article intended to be a part of a dialogue (the authors give as much away in the last two paragraphs), it's meant to condemn scientists without offering corrective suggestions or expressing the slightest interest in the use of the scientific toolset.

That said, I did find the examples the article gave with regard to the Habsburg jaw interesting and thought-provoking. I hope that there's room for collaboration between art scholars and AI researchers to discover historical phenomena through depictions of people over time!
posted by LSK at 5:40 AM on December 15, 2020 [3 favorites]

idk why anybody wouldn’t think paintings are 100% accurate
posted by Huffy Puffy at 6:16 AM on December 15, 2020 [4 favorites]

In the case of the trustworthiness paper, the scientists did not start by noticing interesting changes in the curve of lips and width of eyes in portraits over time and wondering what might explain that. Instead, they began with the hypothesis that social trust increased in Europe in recent centuries and wondered if this might be reflected in portraiture.

The authors of the Hyperallergic article are saying that using portraiture is a bad method – they aren't interested in the changes in the curve of lips over time, and nor should they be expected to be. If they read as angry or frustrated, it is because scientists frequently rely overly heavy on novel data and tools without considering the full context of that data, not to mention the sordid history of using visual and physical characteristics to determine intelligence and psychology.
posted by adrianhon at 6:53 AM on December 15, 2020 [3 favorites]

Using software to seek out signs of a valued personality trait in paintings may not exactly be phrenology, but its intellectual forebears are.

Say what you want about the tenets of phrenology, dude, but at least they measured real skulls.
posted by bumpkin at 6:56 AM on December 15, 2020 [8 favorites]

The paper of "trustworthiness" linked in the article is just flat out idiotic. Big surprise, the closer the portraits are to our time the better fit to currently alleged ideals of "trustworthiness" they find, with the biggest jump happening in the middle of the twentieth century, almost as if the artistic conventions were driving the response and not the other way around. The more like "now" the artistic style, the more it fits current values! How astonishing!

It's not as if people are actually good at judging trustworthiness from images, I mean look at the some of the people heading major elected governments around the world and how divisive those people are to those who might invest their trust. What people are pretty good at is recognizing the conventions of their day, even if they can't articulate them, so artistic conventions change over time becoming what we understand as the current ideals, and we hold those concepts as defining certain values because of how we engage with the world through media and emulate that as part of our shared culture.

The further back you go the farther from current ideals you start to drift as conventions change over time and slowly become more familiar. The paper does the brilliant work of showing how "Now" is more like now, then "Then" was, in terms of how we see things. What people might have deemed "trustworthy" then was either somehow opposed by the portraits artists made and people sat for (evidently paying to be rendered as "untrustworthy" must have been a big fad or something for a few centuries), or the nature of the question and perceptions thereof have changed along with all the other parts of the culture.

I mean this isn't hard, you can watch a movie from the 1930s or even up to the 1980s and one from today to see all sorts of changes in presentation around looks and trust and all sorts of other details that have occurred in the space of decades, where we'll of course be more comfortable with things the closer it gets to our own time. Any algorithm based on sampling current tastes will, unsurprisingly, reflect better on the conventions of more current eras. That's kinda how conventions work, but evo-art/science types keep trying to pretend that they're "discovering" something new about people by stripping art history of its context and selling it piecemeal as measurements fit to their just so stories. Utter garbage.
posted by gusottertrout at 8:45 AM on December 15, 2020 [5 favorites]

I wonder if there's some bit of evo-psych attitude slipping in here, that they're quick to assume any association with "trustworthiness" is innate rather than cultural.

I remember when the "trustworthiness" paper hit my corner of Biologist Twitter, and uh. Yep. That was an explicit argument, as I recall. Pity that the "trustworthniness" metric is, er, not racially neutral, huh?

Although the authors of the article may be right that the scientists aren't looking at these works with an artist's eye, I find it a bit frustrating that they commit the same mistake they're claiming to correct for. They're saying that scientists are neglecting the value of an artistic perspective to fully understand the material they're studying, but the authors of this article similarly shun the scientific perspective.

The scientists who performed the original studies did discover trends in art, and interpreted them. The authors of this article condemn the interpretations, but often don't provide an alternate explanation or hypothesis for the observed trends:

That would be because the scientific perspective does not require you to propose an alternate explanation when falsifying a wrong hypothesis. I do not need to fully account for all existing natural phenomena to say that the balance of evidence suggests that the world isn't flat. No, the scientists are not looking at these works with an artists' eye--but neither are they looking with scientists' eyes because they fail to consider things like "what if artistic standards of beauty or power change over time OR artists subtly change their portraiture to appeal to subjects OR what if desirable facial expressions for portraiture change over time" as alternate hypotheses to be ruled out. In point of fact, they don't even look at that notion, let alone consider it.

Just because these assholes are using computer algorithms and Big Data, it does not fucking follow that they are taking a "scientist's perspective" acceptably or accurately to this dataset. Fuck's sake, man.
posted by sciatrix at 8:53 AM on December 15, 2020 [5 favorites]

idk why anybody wouldn’t think paintings are 100% accurate

I really expected your link to lead us to this docu-series.
posted by nickmark at 10:11 AM on December 15, 2020 [2 favorites]

So aren't there any Habsburg skulls hanging around for people to measure?

Anyway, it's interesting when a powerful king like, say, Henry VIII says to his painter, "bring me a picture of her and make it accurate or I'll kill you." You get this instead of this. Still not something you'd want to use as data but it's definitely a comment on the fact that art -- even photography -- is almost never about the representation of spatial relationships alone, since a) it's impossible to do that accurately anyway; and b) if an artwork contains less information than the object it represents, what's the point in making it?
posted by klanawa at 2:09 PM on December 15, 2020 [2 favorites]

The 'trustworthiness' paper is like the Sokal hoax in reverse. Any historian can see at once that the claims it makes are ludicrous. Yet it passed peer review in a scientific journal.
posted by verstegan at 2:20 PM on December 15, 2020 [1 favorite]

Wait, you’re telling me Mrs. Matisse didn’t actually have a big green face tattoo?
posted by kevinbelt at 5:07 PM on December 15, 2020

Honestly I think it's widely underappreciated how difficult it was for Picasso to find so many brightly multicolored, angular people to sit in contortions for hours while he painted their portraits. I mean, sure, they're commonplace enough today with the advent of computer graphics, but at the time finding people made entirely out of exaggerated geometric shapes to sit for portraits was quite a burden on the artist.
posted by mhoye at 6:21 PM on December 15, 2020 [4 favorites]

Having looked at hundreds of old black and white photos of relatives from the early 20th century, I conclude that back then everyone scowled and was angry all the time and wore full suits or fancy dresses every day.
posted by freecellwizard at 9:29 AM on December 16, 2020 [1 favorite]

@freecellwizard also, a lot more people were translucent or had multiple heads, on account of the much higher rate of transporter accidents.
posted by confluency at 3:42 PM on December 16, 2020 [1 favorite]