To celebrate the Fourth of July this year, I’d like to share a set of portraits I’ve made of the first six presidents of the United States by applying face-averaging software to groups of contemporaneous paintings of them—portraits that are arguably both new and authentic at the same time, as paradoxical as that might seem. I first tried out the principle behind these portraits a couple years ago with an image of Queen Elizabeth the First made by averaging thirty-one different paintings of her (see here), but I’ve been meaning to do something further with the idea for some time now, and this is what I’ve come up with. For each president, I’ve relied on images created during his own life (as well as I could easily determine), though not necessarily during his presidency.
1. George Washington (born 1732, president 1789-1797, died 1799)
For the “father of our country” and the icon alike of quarter and dollar bill, I started with a list of eighteen non-profile life portraits online at mounvernon.org, augmenting this with another painting by John Trumbull (here) and the well-known Lansdowne portrait (here) to round the total out to twenty. Next, I flipped the source images as needed for George to be facing consistently left, assuming that his face was symmetrical enough for its two sides to be interchangeable for my purposes. Then I loaded the modified images into Abrosoft FaceMixer.
Below you can see some intermediate stages of the averaging process. The image on the left shows the default average which FaceMixer generated after it had detected all twenty faces (with occasional help from me). The middle image came after the “edit dot” stage, where I manually defined individual points on each face—pupils of eyes, tip of nose, and so forth—in order to align the source images more closely with each other. Then I processed that result using “Auto Contrast” in Photoshop to create the image on the right.
Finally, I did some manual retouching to remove distracting artifacts, the most obvious of these being a line across Washington’s forehead that came from him wearing a hat in one of the source portraits. Here’s the end result:
You certainly wouldn’t confuse this with any one of the contemporaneous paintings, but at the same time it’s still recognizably George Washington as we’ve come to expect him to look. Moreover, there’s a source we can use to check the reliability of our results, shown below at left: Jean-Antoine Houdon’s Life Mask of George Washington, taken at Mount Vernon in October 1785, courtesy of the Pierpont Morgan Library:
The features of the life mask were taken directly from Washington’s face by a molding process, so it has a certain claim to accuracy not shared by the paintings. As we read in a piece called “The Real Face of George Washington” at varsitytutors.com: “The life mask shows Washington as he really appeared in life….without the personal and, often, subjective interpretations of the many artists, painters and sculptors who came to sketch, paint and model his countenance.” But our average of twenty paintings resembles it pretty closely, and it too has arguably factored out the “personal and, often, subjective interpretations” of individual painters to a great extent.
But there’s admittedly some room for subjective bias in our composite average, since seven of our twenty source images are by Charles Willson Peale, two are by Gilbert Stuart, and two are by John Trumbull. In some cases, we’re dealing with nearly identical portraits, where one plainly served as the basis for one or more others. If Peale in particular was imposing an idiosyncratic personal touch on all his portraits of George Washington, he might have skewed our results, since his work contributed 35% of the total. With that in mind, I’ve created a second average that incorporates only one source painting each from Peale, Stuart, and Trumbull:
It’s hard to say whether one of these averages can be considered more representative or accurate than the other. The image on the right weights the vision of thirteen different artists equally, regardless of reputation or skill. The image on the left gives the vision of three artists significantly more weight than the others, but they’re the artists whose work was seemingly most influential and well-regarded at the time, so maybe we should give them more weight. Democracy or oligarchy: take your pick, and then be prepared for your choice to affect how you see our Founding Fathers.
We read of one person’s experience meeting Washington in the 1790s, as recounted by his son:
[T]he General entered the room. It was not necessary to announce his name, for his peculiar appearance, his firm forehead, Roman nose, and a projection of the lower jaw, his height and figure, could not be mistaken by any one who had seen a full-length picture of him, and yet no picture accurately resembled him in the minute traits of his person. His features, however, were so marked by prominent characteristics, which appear in all likenesses of him, that a stranger could not be mistaken in the man…. It was observed to me that there was an expression in Washington’s face that no painter had succeeded in taking.
I can’t help but wonder whether either of our averages has distilled that “expression” from the flawed observations of over a dozen painters, even if none of them quite grasped it individually.
2. John Adams (born 1735, president 1797-1801, died 1826)
For John Adams, I started with the National Portrait Gallery’s list of presidential portraits and then added to it whatever else I could scrounge up to reach a total of fourteen source images.
These portraits differ more from each other than George Washington’s did. They don’t even look much like pictures of the same person (except where one is clearly a copy of the other), and it’s hard to get any sense of what John Adams might actually have looked like by examining them side by side in this way. But here’s the result we get from averaging the fourteen portraits:
Some of the source portraits look a bit stylized, or perhaps inclined to flatter rather than reveal, which could be skewing our averaged result. It’s true that John Adams’s appearance seems to have changed dramatically over the course of his life, due in part to the loss of his teeth, which (unlike Washington) he opted not to replace with dentures, eventually giving him a muffled, whistling lisp when he spoke. On the other hand, if we choose to average only the six least flattering portraits—which I take to be 1, 5, 6, 8, 11, and 14 in the grid above, numbered from left to right and then top to bottom—we end up with an image of someone who’s a bit older-looking, but who still shares the same general facial features seen in our first composite, as though we’re seeing another picture of the same person at advanced age.
Compared with the averages, the John Adams life mask shows a face even more deeply wrinkled and bearing a distinct frown, with furrowed brow and downturned corners of the mouth. After all, having a life mask taken by Browere wasn’t much fun—more about that in a moment. But as with George Washington, we do get the sense that these are all images of the same recognizable individual. I’m pretty sure we’d be able to pick John Adams out of a police lineup based on them.
3. Thomas Jefferson (born 1743, president 1801-1809, died 1826)
It was surprisingly hard to round up a decent number of different color, non-profile portraits of Thomas Jefferson created during his lifetime, and I ended up including one in the mix (by Bass Otis) that I could only find in grayscale, as well as a couple near duplicates, in order to raise my total to fourteen source images.
There’s really not much difference between the two. But let’s proceed to our usual check on accuracy: a comparison with the life mask taken by Browere on October 15, 1825 (image of unknown provenance, borrowed from Michael Beschloss’s Twitter feed):
Once again, I’d say these all look recognizably like the same person. Granted, the life mask displays a pronounced frown that’s absent from our averages, much as we saw before with John Adams, but that’s only to be expected, given Jefferson’s account of the ordeal of having the mask made:
I was taken in by Mr. Browere. He said his operation would be of about twenty minutes and less unpleasant than Houdon’s method [Houdon had made the life mask of George Washington which we examined earlier]. I submitted therefore without enquiry but it was a bold experiment on his part on the health of an Octogenary, worn down by sickness as well as age. Successive coats of thin grout plastered on the naked head, and kept there an hour, would have been a severe trial of a young and hale person. He suffered the plaster also to get so dry that separation became difficult and even dangerous. He was obliged to use freely the mallet and chissel [sic] to break it into pieces and cut off a piece at a time. These thumps of the mallet would have been sensible almost to a loggerhead. The family became alarmed, and he confused, till I was quite exhausted, and there became a real danger that the ears would separate from the head sooner than from the plaster.
Our averages might also help in evaluating portraits where the identification of Thomas Jefferson as the subject is tentative and uncertain. Back in 2013, a team of researchers proposed that a portrait of an unidentified subject painted by Benjamin Nicolas Delapierre in 1785 is in fact the earliest known Jefferson portrait, and there’s a website here dedicated to making that case. One point the researchers consider, naturally enough, is whether the subject of the Delapierre painting looks like Jefferson or not, and as they point out, this isn’t a simple thing to establish:
A factor further complicating comparative analysis of Jefferson’s physical appearance using known life portraits of him is the very different way he was depicted by various artists. Some portraits painted within months of each other appear to be of altogether different people. (“Thomas Jefferson’s Physical Appearance”)
They bring a number of different sources to bear on the question of what Jefferson really looked like, but I’d like to think our composite average might also shed some light on the matter. Below is a comparison of the Delapierre painting (left), our average of ten paintings of Thomas Jefferson by ten different artists (right), and a fifty-fifty mix of these (middle):
4. James Madison (born 1751, president 1809-1817, died 1836)
You know the drill by now. I gathered fourteen portraits of James Madison created during his lifetime, prepared them as before—
Here’s the obligatory life-mask comparison, based on another of Browere’s works:
5. James Monroe (born 1758, president 1817-1825, died 1831)
I found a lot of duplication among the available portraits of James Monroe—that is, paintings based closely on other paintings—and poor documentation of some others which I left out due to uncertainty over date (who the heck was Wanderayn?). After some deliberation, I wound up with these twelve source images:
After going through the trouble of generating and retouching this average, I remained uncomfortable with how much it favored a few original source paintings, especially one by Gilbert Stuart, that had been repeatedly copied; and then I realized to my chagrin that two of the sources I’d used were actually digital files of the same exact painting—I’d been misled because they’re credited differently online (#5 to John Vanderlyn, #10 to William James Hubbard). Oops. After that discovery, I went back and generated a new average based only on the six clearly distinct portraits in the group. Here’s the result, which doesn’t actually look all that much different from the first:
Monroe escaped the ordeal of having a life mask taken by Browere, fortunate guy. However, Browere did take a death mask of Monroe’s face, so we at least have that available to compare with our composite averages:
In terms of facial expression, Monroe’s Browere mask comes closer to my averages than it did in any of the other cases we’ve examined. Since he was dead when it was made, I suppose he was more at ease about the process.
6. John Quincy Adams (born 1767, president 1825-1829, died 1848)
We’re on a roll now. I was able to pull together sixteen paintings of John Quincy Adams made at various stages of his life from the 1790s through the 1840s.
John Quincy Adams’s painted portraits don’t differ from each other quite as much as those of his father, John Adams, but they still vary enough that it’s hard to get an overall sense of his appearance from looking at them side by side. At one point, he himself wrote: “I gave a sitting of an hour this morning to Mr. John Cranch and Mr. Bingham, neither of whom is likely to make out either a strong likeness or a fine picture.” Things don’t look promising here as far as the accuracy of individual paintings goes. But here’s a composite average of the sixteen paintings:
If we compare this with a life mask taken by our old friend Browere in 1825, we might easily persuade ourselves that we’re looking at the same person:
On the other hand, John Quincy Adams was also the earliest-serving president to have his image captured photographically later in life, and some doubt about the accuracy of our average might arise once we factor in four daguerreotypes taken in the 1840s (see here, here, here, and here):
If we take these photographically captured images as unimpeachable, objective data about what John Quincy Adams’s face really looked like—at least during the last decade of his life—the distance from the base of his nose to his mouth appears consistently greater than it does in the average of the sixteen paintings, as it also does in the life mask if we look more carefully.
Of course, we can combine these different sources of evidence if we like. Here’s a fifty-fifty mix of (1) the sixteen paintings and (2) the daguerreotypes and life mask, with the area around the bright white collar considerably retouched:
Below I’ve placed side by side for comparison (1) the average of the paintings only; (2) the composite based 50% on the paintings and 50% on the daguerreotypes and life mask; and (3) an average of only the daguerreotypes and life mask.
The differences we see here are presumably due in part to the subject’s aging. The daguerreotypes were all taken during the last decade of John Quincy Adams’s life, whereas the paintings extend as much as half a century in time further back than that. With this in mind, we might interpret the image on the left as showing a younger John Quincy Adams than the images in the middle and on the right. But I’m not sure that could account for all the differences we see in the appearance of the mouth, which I think must owe something to the aesthetic conventions of painted portraiture as well. And if there’s one consistent difference between the masks and paintings we’ve compared of earlier presidents, it does seem to lie in the distances between tip of nose and mouth.
In any case, it’s clear that once we enter further into the era of photography, we’ll need to strike a well-considered balance between data from paintings and data from photographs. But that’s a puzzle for another day.
Here, then—just in time for the Fourth of July—are averages of contemporaneous portraits of the first six presidents of the United States (now with further processing by the “Auto Tone” filter in Photoshop):
If I decide to continue on down the list of U. S. presidents to the present, that would make six down, thirty-seven to go (soon to be thirty-eight). But let’s stop for a moment and reflect on what exactly we have here.
These are all new portraits, unlike any that have existed before, but they’re also authentic, insofar as they’re based on contemporaneous images averaged together more or less objectively. Indeed, I believe they could give us a better sense for what each president looked like in general than any one “real” portrait does. They also offer a fresh alternative to hackneyed canonical portraits that have been recycled again and again and again until they’ve lost all their impact.
Need an attractive cover for your biography of (say) James Madison? Well, what could be better than a new portrait that draws faithfully and impartially on a variety of primary sources, just as a good biography does?
And there are plenty of scenarios where there’s value in lining up images of all the presidents in order, as for instance with this placemat available from the White House Gift Shop—
—or the Presidential $1 Coin Program—
—or this mug offered by Cafe Press—
—or any one of the countless other similar examples I could mention. But the portraits used in these cases aren’t ordinarily very consistent: the presidents are shown facing in different directions, or as represented in different media or styles, so that they end up being as much about the superficial quirks of portraiture as they are about the presidents themselves. By contrast, my averaged presidential portraits are nicely consistent and compatible with one another, and ideal for projects and products along these lines. Isn’t it easy to imagine a set of postage stamps, for instance?
In the meantime, I hereby issue the averaged portraits included in this blog post under a Creative Commons Attribution-NonCommercial-ShareAlike license in the vain hope that they’ll go viral. If you like them, knock yourself out!