On Portraits

This is a theme I revisit occasionally, because I am very, very interested in how pictures of people function. Very long time readers may find large swathes of this familiar, but I believe I have something new to add this time around.

This much is so sufficiently established by wiser heads than mine that we might as well take it, in broad strokes, as fact:

We, humans, are mind-readers. When we are interacting with another human, we maintain a surprisingly accurate and detailed model of the other human's mental state. It goes beyond simple mood, and includes rough approximations of what the other human is thinking, and where, in some detail "their head is at." We guess at how a joke would be taken, whether she would like us to kiss her, and so on. We're wrong some of the time, but right surprisingly often. Our social systems are grounded in these mechanisms.

Two things: First, our mental model is built by continuous observation of body language and facial expression, in concert with how the conversation is going. We use a myriad of thoroughly subtle cues, I don't know the extent to which they're catalogued, but the catalog is deep. Second: we are over-confident in the correctness of our mental model.

How often do we encounter the "hey, but you wanted me to..." which is some combination of wishful thinking and misread cues? At least in those cases in which you did not actually want them to, but are now denying it.

Thus endeth the science lesson. Onwards now to speculation from your host.

Even somewhat less long-term readers will have seen my theory about presence, the idea that a photograph conjures a kind of presence. The idea is that when we look at a photo, any photo, we react, we behave, we feel, in some ways as if we were actually present in the photograph. We reconstruct the world that surrounds the photograph, in space and time, and interpret the photograph in the light of our construct.

So, consider a portrait.

We are, in a sense, there. We react as if we were in the presence of that person. We read their face, their body language, and attempt to construct a mental model of the subject's mind.

At the same time we construct a larger world to contain the portrait. I don't think we literally create a mental filmstrip with 10 seconds of "footage" of the face, our construct is more vague, more generalized than that. We do, though, build the equivalent.

There are, to my way of thinking, two mental models here. One is of the world of the photo, the world we're imagining the photo to be drawn from. The second is our model of the subject's mental state, of their mind. There may be no distinct line between these two things, but I will try to keep them separate in this next few lines, for clarity. We construct a mental model of the subject's mind based on our model of the world the photo comes from.

They appear, say, happy. We extrapolate that to a larger timeframe, add in everything else we can glean from the photo, from ourselves, from what we know of the subject, from accidental resemblances, and so on. We construct our mental model of the subject's mind as-if we were present, as-if we had been present for the last few minutes, or hours, or seconds. We mind conclude that they're happy-tinged-with-sad or whatever. We might conclude that he wants to be kissed, or that she wants to be elsewhere, or that they are distracted by love, or tragedy.

Return now to the two things. First: we're building this model of the subject's mental state on a ground of fine detail. The set of the eyebrows, the shape of the mouth, the crow's feet, the tilt of the head, minor details of the tension of tiny muscles, etc etc. Second: we're quite certain that we've got it right.

This suggests a couple of things.

First, it explains why we imagine we know something intimate, something profound, about the subject when we see a really good portrait. We understand the subject's mind, and we've definitely got it right. We see, we know.

Second, it suggests what makes a "good" portrait of that type. The maze of fine detail must be both visible, and also legible. This doesn't mean that the face need be lit like an operating theater, you can plunge one half into complete darkness since faces are pretty symmetrical. Rendering these details visible in whatever way makes sense, is the job of the photographer, with lighting and direction. Visibility, though, is not enough, the details must be legible, and therein lies the magic of a great portraitist. The sitter need be arranging their face and body into a form which, when seen statically, presents a coherent set of cues.

Coherent need not be consistent, the sitter might well present a mixed emotion, but the mixture must be legible to us as we see the static image and extrapolate it into a larger construct.

Third, this suggests why excessive modern software tends to render portraits terrible. AI driven skin smoothing software is magical, the skin comes out like a baby's ass. The appearance of the sitter is enhanced, the result remains essentially them in appearance. To the photographer who is aiming to get an attractive and flattering representation of a model's appearance, this is marvelous.

At the same time, though, these tools are erasing or muddling the minuscule cues we're reading. Legibility is destroyed. Tools which reshape the face to more pleasing proportions, especially, do this, but I think "skin work" tools are likely mucking up the fine structure of facial expression in ways we can't even quantify. Our mind-reading relies in part on cues too small to consciously note.

All that remains, when these tools are applied with anything but the lightest hand, is a kind of empty representation of someone's face. Of course, if that's all your started with (which is pretty common) nothing is lost. If all you're looking for is a flattering picture of someone's face, you probably didn't have a particularly visible-and-legible gestalt of detail in the first place.

Fourth, this suggests why so many commercially made, or amateur-made, portraits lack what we might call soul. The gestalt of legible detail on your face is inevitably the result of your internal mental state — that's why mind-reading works. If your mental state is nothing more than "ok, smile, don't blink" or "I wonder what's for lunch" or "is this asshole going to hit on me?" then there's not going to be much to read on the face in the resulting photo.

Most portraits are like that. This is the skill of the portraitist, to bring out that legible gestalt of detail, by conjuring a mental state which produces that. This is what all the chit-chat in the studio is for (or if you're Jane Bown, it's what the 10 seconds of observation and a magical invisibility is for.)

And there you have it. That's how portraits work.