Hands-on Vision Science: "Thatcherizing" real faces with inverting goggles
More #visionscience in the park leads to a neat developmental question about face perception at different ages.
In a previous essay, I talked about heading out to community events armed with an array of goggles and glasses with weird optical properties to try and convince people to take part in our research. In that piece, I spent some time talking about how the Crayola 3D Sidewalk Chalk glasses make it possible to experience chromodepth - a phenomenon that visitors to our community booths have found to be a lot of fun at different events we’ve attended this summer. Here, I want to shift my focus to another pair of interesting goggles I have that are far and away the most popular item that we bring to these outreach visits: Inverting goggles! Besides being a lot of fun for folks to play with, they also led to an interesting observation regarding how kids and adults perceive the faces of people wearing the goggles. This observation has some neat connections to a bunch of different topics including the development (or not!) of holistic face processing in childhood and the Thatcher Illusion, which is a mainstay of my Sensation & Perception courses due to the robustness of the illusory phenomenon.
Inverting prisms: Now available at your Fargo-area Target!
Let’s begin by taking a look at the goggles that got me thinking and reading about all of this stuff. Below you’ll see the inverting prisms that you get with the Upside-Down Challenge game, which is (at the time of writing at least) widely available at retailers in the US for a little under $20 USD. The game itself isn’t terribly interesting - it’s just a deck of cards listing various things that will be difficult/messy/ill-advised to do if everything you see is upside-down. The real fun here is just having these goggles, which are well-made and comfortable to wear. The eye-cups are nice and wide, with a solid rubber ring around the edge to make them both light-tight and easy on your face. The adjustable headstrap also accommodates head sizes from adults all the way down to about 3-year-olds. It’s even easy to disassemble the goggles to remove the prisms, which is useful for the vision scientist who might want to compare upright to inverted vision with the same limited field of view. Overall, this is a surprisingly great little bit of optics gear for a totally reasonable price.
When I first saw these on the shelf at one of our local big-box stores, I snapped them up right away because of how nice the construction looked. I inherited a pair of inverting prisms from a senior colleague at NDSU that worked, but were also a somewhat…well, let’s say bespoke set of goggles. Look, there isn’t a huge market for these things, so if you were a vision scientist who wanted a pair you scrounged up some prisms, spray-painted a set of safety goggles to limit the field of view to what you could see through the acrylic and then figured out some way to bodge the prisms onto or into the goggles so they’d be wearable (see below). While I appreciate the DIY efforts of those that came before me, it’s kind of nice to be able to get a well-made pair of these at relatively low cost instead. I’ve amassed a collection of 4-5 pairs of these now partly for fear that they might be discontinued at some point and partly so that I have bunches of them to use at the various public events we attend. If the Fargo Target keeps track of analytics closely enough, they might be wondering why on Earth this particular game seems to be as popular as it is. Between my lab and the parents I keep mentioning these too after their kids have played with them, I figure there is a pretty weird spike in the data.
The hot-glue, spray paint, and surplus optics shop version of inverting goggles. They do work, but it’s kinda nice to have everything in a slightly sturdier assembly.
How do inverting goggles work (and what are they for)?
First, some very basic optics to show you why the goggles do what they do. If you do disassemble a pair like I mentioned above, you’ll find that in front of each eye there is a trapezoidal prism (a Dove prism to be specific) made of acrylic. This prism bends or refracts the incoming light before it reaches your eye in accordance with a relationship called Snell’s Law. Briefly, this law tells us how the direction light takes after it enters a new medium (in this case the acrylic stuff) depends on the angle of the incoming light relative to the surface of the medium and the material that the new medium is made of. In the case of the Dove prism, the refraction of the incoming light means that light rays parallel to the long axis of the prism are bent towards it’s floor, where the rays then bounce off (or reflect) due to a phenomenon called total internal reflection. This phenomenon depends on the relationship between the two materials we’re working with (in this case, air and acrylic) and the shallowness of the angle a light ray makes relative to a boundary between the two materials. In the Dove prism, the stuff we make the prism out of and the angle made by the sloped sides conspire to make this reflection happen, which leads to the flipped image we experience when we see the light that exits from the other side of the prism (see below). The optics of the prism is also responsible for one of the features of the inverting goggles that nearly all the kids (and some of the adults) wearing them try out eventually: Can I flip the goggles upside-down to see a rightside-up image? Nope! This doesn’t work because the incoming light doesn’t “know” how the prism is oriented, but instead will still refract towards flat surface at a steep enough angle to undergo the critical reflection. Put the goggles on in either orientation and you’ll thus still see a topsy-turvy world.
A Dove prism inverting an image via refraction and total internal reflection. Image credit: By Fred the Oyster, CC BY-SA 4.0, https://commons.wikimedia.org/w/index.php?curid=35067681
Besides this being a nice optical trick to baffle and amaze kids and adults alike with an upside-down view, what’s the point of goggles like these? The use of prisms as a tool for studying visual perception has a long history in vision science. I count myself very fortunate to have shared a lab for a few years with Dr. Richard Held, who would sometimes introduce himself to people at the time as “MIT’s oldest post-doc.” That was true, I suppose, but Dick was an absolute giant in the field of vision science, having conducted a number of foundational and creative studies examining visual development. These included the famous “kitten carousel,” which was a means of investigating the role of active movement in developing visual sensitivity to patterns in the visual world, one of the first methods for testing visual acuity in infants, and important results demonstrating that correcting conditions like astigmatism and strabismus optically in early life could prevent neural consequences that would affect later visual function. During the time I spent in the lab with him, we’d hear all kinds of stories and musings about the field, many of which put ideas in my head that I’m still turning over. In particular, Dick and I spent a lot of time talking about various prism studies he had done decades ago in which they had been interested in using observers’ ability to adapt to optical transformations imposed by prisms as a model system for visual development. Babies and children can be tough to work with in the lab for a variety of reasons (many of which are probably apparent to you immediately), but if you want to understand how the visual system changes over time they’re probably the right people to work with, right? The idea behind the series of experiments Dick told me about was to make adult participants deal with some kind of manipulated visual world via prism goggles, forcing their visual system to adjust to the change. That process of adjustment, its timing, the manner in which visual capabilities changed afterwards, and the return to baseline after the prisms were removed, could all be treated as a sort of developmental trajectory. By introducing different optical manipulations, one could ask questions about different mappings between vision and action that might vary in how learnable they were and potentially probe neural indices of spatial vision to see how different types of visual experience led to measurable changes in the visual cortex.
Dr. Richard Held (profile pic from researchgate.net) and one of the best figures in visual neuroscience - the kitten carousel. In this experiment, the visual experience of two kittens was matched by placing them in a circular arena with vertically-striped walls, but the activity of the cats was very different. One kitten may walk around as it likes, actively exploring the visual environment, but the other kitten passively glides around at the same time by virtue of the yoking mechanism.
To give you a simple example of the kind of thing I’m describing here, let me tell you about an easy demo I often do with a different set of prism goggles. By putting a Fresnel prism in front of ordinary goggles, one can impose another transformation of the visual world: A lateral shift of the incoming light so that everything you see is moved some number of degrees of visual angle over from where it would have appeared otherwise. You can change the amount of this shift by using different prisms, but the fun part of this demo happens when you start asking people to do simple things like reaching for objects on a table, giving out high-fives or fist-bumps, or attempting to throw things to hit a target. In all of these case, the (usual horizontal) shift imposed by the prisms leads to people making large errors - you reach off to the side of an object, or throw a beanbag over to the right of the target. If the person keeps doing these activities, however, you’ll find that they rapidly learn to correct for the horizontal offset: It usually only takes people I run this demo with 4-5 rounds of throwing beanbags at a target before they’re back to being fairly accurate, for example. Take the glasses off at this point, and they’ll start making the opposite error: If they whiffed off to the right when they first put the glasses on, they’ll start making mistakes off to the left once the prisms are removed. It’s quick, it’s easy, and it gives a vision scientist a chance to ask questions about the relationship between variables like the length of time wearing the prisms and the rebound after they’re removed, or look for temporary changes in visual abilities once they’ve adapted to the new conditions. Adaptation to prisms has a number of therapeutic applications as well, including the treatment of strabismus and visual neglect. This is all to say that while the Upside-Down Challenge is a fun game, optical devices like these are part of serious basic and applied vision science, too.
Vision science in the park - upside-down version
So what do we actually do with these when we work with the public? Honestly, a lot of silly stuff. We ask them to play Jenga, we have them try to catch beanbags and shake hands with us, and we use this to talk just a little about how the visual brain can adapt to some optical transformations more easily than others which connects to some of the things we study in the lab. We keep it light and fun, but always stand ready to answer some of the really good questions people occasionally ask. A classic example: Some people have heard that long-term use of goggles like these leads the image to subjectively “flip back over.” Is that actually true? The answer, at least according to classic self-experimentation by Dr. George Stratton, is “not really.” Stratton wore inverting goggles for days, eventually finding that after 3 days or so he was able to move around and do things normally, experiencing a long rebound effect of just under 3 days after removing the goggles. The subjective experience of the world appeared to be that it was upside-down the whole time, but his ability to cope with it improved substantially. Even if people don’t ask, I sometimes bring this up because people tend to find the story interesting and it can get kids and their grown-ups talking about their experience of wearing the goggles.
I’m always looking for neat stuff to point out to people while they wear the prisms and one that struck me recently was a simple observation that turns out to make people laugh: For the person wearing the goggles, the prisms invert their whole visual world. For someone looking at the wearer, however, the prisms also invert the image of the eyes within the head! There’s nothing amazing to explain here (it’s all the same optics) but it looks pretty funny (see below). This usually leads to a lot of laughter, people wanting to swap the goggles around so they can look at each other, and much taking of pictures for the Insta, the Finsta, or whatever version of Snaptagrambook people post on these days.
It’s more vivid in person than in this picture, but this is a remarkably hard photo to take! H/t as always to my wife and colleague, Dr. Erin Conwell, who continues to bail me out when I need to do weird stuff in the name of #visionscience.
The Thatcher Effect - Messing up politician’s faces since 1980.
This piecewise manipulation of the way the eyes look is closely related to another vision science effect that has been thoroughly studied: The Thatcher Effect. The full transformation that’s usually used to examine this phenomenon includes inverting the eyes and the mouth within an otherwise upright face (see below). The end result (first implemented by Dr. Peter Thompson on then UK PM Margaret Thatcher’s face) is typically grotesque, which in itself is not so surprising. What is perhaps more surprising (and elevates this manipulation from straightforward defacement of a potrait to a useful tool for vision science) is that turning the entire image upside-down makes the effect of this transformation much, much less apparent.
The original Thatcher Illusion (Photography: Rob Bogaerts - Image manipulation: Phonebox, CC0, via Wikimedia Commons). The two images in the bottom row likely look much different to you, but the two in the top row probably look much more similar to one another.
There are a number of ways to describe what this demonstrates about the visual system. I usually favor pointing out that this is a dramatic example of the Face Inversion Effect, which broadly refers to the much poorer recognition and discrimination abilities most observers have with upside-down faces compared to upright ones. This effect of image inversion is generally much weaker for other types of objects, hinting at the distinct neural mechanisms that support face processing. When I talk to students about the Thatcher Effect, I point out that it shows off just how much less effective/efficient your visual processing of upside-down faces is compared to upright ones: The “Thatcherization” of the face is glaringly apparent when they’re upright, but much subtler in the inverted images despite the low-level differences being the same! If that doesn’t convince you that you have terrible inverted face perception, I’m not sure what would. Other researchers invoke holistic processing when talking about the Thatcher Effect, by which they mean the bias favoring large-scale visual processing of faces rather than small-scale, part-by-part processing. Put more simply, it seems like observers tend to measure face images all at once rather than analyzing the eyes, the mouth, or other micropatterns within the face item-by-item before assembling them into a whole. This is another way in which face processing differs from how other kinds of objects are perceived and recognized.
But listen: Here’s where things got sort of interesting when I started pointing out the inversion of the wearer’s eyes while wearing our goggles. Time after time, I’d point out to Mum or Dad that their child’s eyes looked upside-down with the goggles on, eliciting the laughter and picture-taking I described above. The kids, of course, wanted to know what all the fuss was about and usually asked to see their parent wearing the goggles. What happened next was usually something like this:
PARENT:
KID: Ok. At your eyes?
PARENT: Yes! Don’t they look silly?
KID: No.
PARENT: No, look at my eyes! See how they’re upside-down? Isn’t that weird?
KID:
The first few times, I chalked this up to some potential confusion on the part of the kids as to what they were looking for. Once it started to become a routine, however, I began to think about this more seriously. Some parents even showed their kids the photos they had taken of the child wearing the goggles and even then the kids were hard-pressed to see what everyone was making a big deal about. To be clear, this isn’t an actual experiment - we’re talking about kids giving a totally subjective report of what they see while hopped-up on candy and/or anxiously waiting to take a ride on the horse-drawn cart. Rigorous psychophysics this is not. Still: Adults laugh at loud when they see these upside-down eyes in a lot of case while the kids are at least not reacting that strongly. What gives?
Do kids develop sensitivity to Thatcherization?
One possibility is that this may reflect some aspect of children’s face processing that is still developing during childhood. The tricky thing is that question as to whether or not there is meaningful development of face recognition itself after infancy still has some complex answers. On one hand, some reports examining children’s abilities to carry out various forms of face matching and face discrimination suggest that specific aspects of face perception (possibly those pertaining to measuring differences in the spacing of facial features) may only reach adult-like levels relatively late in childhood - say around age 8-10 years or so (Mondloch et al., 2004; Anes & Short, 2009; Balas, Weigelt & Koldewyn, 2023). On the other hand, some researchers advocate strongly that the effects that are measurable in adults that we often use to define face recognition uniquely (inversion effects, proxies for holistic processing) are also measurable about as early as we can test them, suggesting that face perception may be qualitatively adult-like in terms of any processes that are face-specific (Crookes & McKone, 2009; McKone et al., 2012). All of this debate is complicated by a number of other factors including how carefully image differences between faces used in these experiments were matched across different tasks researchers asked children and adults to do, whether or not the changes we make to faces in such experiments (moving the eyes closer together or further apart) have much to do with the processes we often assume are used to process those images (does the brain bother measuring eye spacing at all?), and even whether or not the tasks that we assume reflect something unique about faces actually do so. The Thatcher Effect turns out to be measurable for other kinds of objects, for example! Wong et al., (2010) Thatcherized various kinds of faces and non-face objects and used a same/different image judgment to find out if the Thatcher Effect (better task performance for upright vs. inverted images) was more pronounced for face images. The answer? Their data sure doesn’t look like it.
Figure 1 from Wong et al., 2010 - The Thatcher Effect was of comparable magnitude across image categories, which may imply that detecting this manipulation of local vs. global orientation may not depend on face-specific processing.
All the same, what do we know about how kids perceive Thatcherized faces? Is there anything in the literature to help us understand what’s going on when kids at the park don’t flinch when you show them inverted eyes via the Upside-Down Challenge goggles? There isn’t a huge literature about this, but what there is suggests that kids look pretty much like adults most of the time. Both newborn babies (Leo & Simion, 2009) and 6-month-olds (Bertin & Bhatt, 2004) respond to a Thatcherization change to a face more robustly when the faces they see are upright compared to inverted. In childhood (at least for kids older than 6 years of age), it also seems like the Thatcher Effect is consistently evident, including asking children to try and discriminate Thatcherized faces from regular ones in the upright and inverted orientations (Riby et al., 2009) or measuring the “grotesqueness” of Thatcherized faces in different orientations (Lewis, 2003). At least one study (Donnely & Hadwin, 2003) comparing kids 6-10 years old to adults suggests a weaker Thatcher Effect in younger kids, but I found these results a little hard to interpret for some technical reasons. Overall, the balance of the evidence of the literature (at least to me) suggests that this aspect of face recognition doesn’t seem to change much for kids.
Well - now what?
This leaves me in a place that you might see as frustrating, but that I think is pretty interesting. Here is this neat thing that seems to happen quite a lot out in the real world of people playing in parks under the sunshine - why does it happen? It’s connected to a lot of interesting results that I’ve known about for years but it also doesn’t make a whole lot of sense given the data. What to do next? I might or I might not follow up on this funny little effect, but if nothing else it was a nice reminder of a couple of things I’ve been trying to keep in mind more often these days. For one, one of my post-doc mentors (the inimitable Dr. Ruth Rosenholtz) often pushed me to go look for big effects rather than trying to eke out tiny results from tortured stimulus manipulations (my preferred methodology at the time). Paying attention to the kinds of things that kids and adults tell you they see in everyday settings is often a good way of finding those kinds of phenomena if you keep your own eyes out for them. Revisiting some of this literature also reminded me how often there are fiddly and not-so-fiddly details in published studies that are worth picking at. For example, I didn’t find much work at all working with kids between 6 months and 6 years of age in Thatcher Effect tasks, but that 3-5 year age range is almost exactly who I see the most of at these events we attend. Does that age gap in the literature mean we might be missing something cool happening at younger ages? Likewise, do task demands matter more than we think? In the lab, you usually want to use your time to ask a well-defined question of your participants, preferably a few dozen times or so. Does that structure mean that the thing that seems to be happening in the park isn’t so evident in a laboratory setting? I’m not sure how many of these are especially exciting questions to ask, and I may not get to any of them any time soon. Regardless, it’s all at least an indication that a day in the park with weird glasses remains a great way to see how much we understand about why things look the way they do.
References
Anes, M. D., & Short, L. A. (2009). Adult-like competence in perceptual encoding of facial configuration by the right hemisphere emerges after 10 years of age. Perception, 38(3), 333–342. https://doi.org/10.1068/p6092
Balas, B., Weigelt, S., & Koldewyn, K. (2023). Configural properties of face portraits change between childhood and adulthood. International Journal of Behavioral Development, 47(1), 35-46. https://doi.org/10.1177/01650254221111792
Bertin, E., & Bhatt, R. S. (2004). The Thatcher illusion and face processing in infancy. Developmental science, 7(4), 431–436. https://doi.org/10.1111/j.1467-7687.2004.00363.x
Crookes, K., & McKone, E. (2009). Early maturity of face recognition: no childhood development of holistic processing, novel face encoding, or face-space. Cognition, 111(2), 219–247. https://doi.org/10.1016/j.cognition.2009.02.004
Donnelly, N., & Hadwin, J. (2003). Children’s perception of the Thatcher illusion: Evidence for development in configural face processing. Visual Cognition, 10(8), 1001–1017. https://doi.org/10.1080/13506280344000202
Leo, I., & Simion, F. (2009). Face processing at birth: a Thatcher illusion study. Developmental science, 12(3), 492–498. https://doi.org/10.1111/j.1467-7687.2008.00791.x
Lewis, M. B. (2003). Thatcher’s Children: Development and the Thatcher Illusion. Perception, 32(12), 1415-1421. https://doi.org/10.1068/p5089
McKone, E., Crookes, K., Jeffery, L., & Dilks, D. D. (2012). A critical review of the development of face recognition: experience is less important than previously believed. Cognitive neuropsychology, 29(1-2), 174–212. https://doi.org/10.1080/02643294.2012.660138
Mondloch, C. J., Dobson, K. S., Parsons, J., & Maurer, D. (2004). Why 8-year-olds cannot tell the difference between Steve Martin and Paul Newman: factors contributing to the slow development of sensitivity to the spacing of facial features. Journal of experimental child psychology, 89(2), 159–181. https://doi.org/10.1016/j.jecp.2004.07.002
Riby, D. M., Riby, L. M., & Reay, J. L. (2009). Differential Sensitivity to Rotations of Facial Features in the Thatcher Illusion. Psychological Reports, 105(3), 721-726. https://doi.org/10.2466/PR0.105.3.721-726
Thompson P. (1980). Margaret Thatcher: a new illusion. Perception, 9(4), 483–484. https://doi.org/10.1068/p090483
Wong, Y. K., Twedt, E., Sheinberg, D., & Gauthier, I. (2010). Does Thompson's Thatcher Effect reflect a face-specific mechanism?. Perception, 39(8), 1125–1141. https://doi.org/10.1068/p6659