Thursday 6 June 2013

Misperceiving the affordances in anorexia nervosa

Alice had a few problems with doors too
It's always hard to evaluate the impact of your research, although funding bodies are increasingly asking that we do so. I always struggle a bit, because the connection between studying, say, the perception of affordances and impact on policy or clinical practice is rarely a straight forward story. My problem is that I think that, if you ask a question the right way the answer will always come in handy to someone at some point, and that this is a big part of how science has its impact. 

Case in point: in 1987 Bill Warren was investigating the perception of affordances for passing through an aperture, such as a door. Apertures can vary in width, but the question facing the observer isn't 'how wide is it?' but 'can I fit through it?'. Bill was describing affordances using pi numbers, ratios of widths in the world divided by widths of some relevant body scale. An aperture wider than your shoulders, say, would yield a ratio larger than 1. For a given person's shoulder width, different apertures would produce different ratios. Across people, apertures of different sizes will produce the same ratio as shoulder width changes. It's a way of factoring out the effect of the size of the observer, and it's based on the idea that we percieve the world in terms of our own ability to act on it.

Warren & Whang (1987) asked, at what value of the aperture/shoulder (A/S) ratio do people start turning in order to fit through the gap? Large and small people turned for different apertures but the A/S ratio that triggered the turn was fairly constant over the groups and came out at around 1.30. In effect, people give themselves a 30% window for error and turn whenever the aperture is smaller than 1.3 times their shoulder width.

A nice result; the ratio describes the relevant affordance property (a property of the world measured on an action relevant unit) and people's behaviour is generally organised with respect to that affordance. The margin for error reflects how difficult it is to turn while walking (high inertia), the fact of postural sway during locomotion and the significant cost to locomotion of banging into the edge of the gap. 

Impact (and it only took 26 years!)  
Keizer et al (2013) have now taken advantage of Warren's more detailed, basic science analysis and applied this task to people with anorexia nervosa (AN). People with AN are well known to have distorted ideas about their own appearance, but where does this distortion come from? Are they biased in what they pay attention to, or do they actually misperceive their body's size? 

Keizer et al had patients with AN and some healthy controls do a version of the aperture crossing task. Participants walked 6m to an aperture and beyond to a table where they interacted with some haptic stimuli. (The cover story for the study was testing whether different actions, such as walking, influenced haptic perception. People with AN are obsessed with issues of body size and they wanted to distract them from the issue of fitting through a gap.)

The control  participants turned on at least 2/3 trials at an A/S ratio of around 1.25, essentially replicating Warren & Whang (1987). The AN participants, however, turned at a ratio of 1.4, significantly higher.This clearly suggests that the AN patients literally perceive their bodies to be larger than they really are, and they are turning as if their bodies were bigger.

This opens up a whole new potential avenue for treatment. First, it explains why many AN patients persist in body image distortions even after treatment; the treatment has not changed their perception of themselves. Second, it may be possible to recalibrate these people's perception of themselves, Calibration is just the process of perceiving information that specifies the action-relevant unit; you use that information to scale the perception of, say, the size of an aperture. You can recalibrate people almost indefinitely (see some thoughts further down on this post) and perhaps, with perceptual training, rehabilitation for AN can be enhanced.

One caveat
One other interpretation is that the AN participants correctly perceive the affordance but are giving themselves a larger margin for error for some reason. The 30% margin, although pretty consistent, is not compulsory. The authors might think to check if the AN participants sway more from side to side as they walk to rule this out (I posted a comment to this effect on their paper, a nice feature of PLOS One!)

One Note
We often hear that our description of standard cognitive psychology is unfair; that no one really talks about representations the way we suggest, and that we're describing a straw man. This paper does it all, and is a nice demonstration of just how pervasive this thinking is. The problem people with AN have is they haven't updated their body schemas recently. Motor programmes to control locomotion are working fine, but the input they take from perception is flawed. This, in a paper about the perception of affordances, and that word is only mentioned once, in the title of Bill Warren's paper in the Reference section!

Any questions asked correctly produces an answer that will be of use to someone, some day. Warren's theoretically motivated exploration of affordance perception laid out some methods and analyses for measuring how people perceive their own bodies, and this question, it turns out, is very interesting with respect to anorexia. It turns out getting the right answers to the right questions is reason enough to study things like affordances.
Keizer, A., Smeets, M., Dijkerman, H., Uzunbajakau, S., van Elburg, A., & Postma, A. (2013). Too Fat to Fit through the Door: First Evidence for Disturbed Body-Scaled Action in Anorexia Nervosa during Locomotion PLoS ONE, 8 (5) DOI: 10.1371/journal.pone.0064602

Warren, W., & Whang, S. (1987). Visual guidance of walking through apertures: Body-scaled information for affordances. Journal of Experimental Psychology: Human Perception and Performance, 13 (3), 371-383 DOI: 10.1037//0096-1523.13.3.371 Download


  1. some thoughts:

    1. Isn't it equally possible that AN individuals just misperceived the size of the doorway, rather than their own body? As far as I can tell, they never actually asked them to estimate the door width.

    2. WRT the perceptual re-calibration, would this be something akin to repeated exposure to fattening to thinning mirrors (like at a circus funhouse)? Presumably AN have opportunities to recalibrate with mirrors all the time, or would you feel that they are not really salient way to calibrate stuff (no real action outcomes/errors, such at bumping into things, occur with mirrors perhaps)

    3. Could you say more on sway? I'm not sure I understand why it's important here.

    1. 1. There is apparently evidence that AN people don't generally misperceive the size of other things, even bodies if not their own, although none of that is affordance work. You could describe what they're doing as misperceiving the door size, too, although given that they are misperceiving the door affordances it seems more likely they are poorly calibrated.

      2. I don't know what you'd do to recalibrate them. It's not clear that mirrors would work, given that they see themselves in mirrors and it doesn't seem to have helped. I think some kind of distorted feedback might be necessary.

      3. Sway: Warren & Whang note that people sway from side to side while walking and that this is probably the main driver for the 25-30% error margin. You turn for gaps that you can fit through given your shoulder width and your sway. If AN people sway more for any reason, their 40% error margin might be correct.

  2. Cool!

    Re point 1:
    From an affordance-entrenched perspective, when Andrew says,

    "AN patients literally perceive their bodies to be larger than they really are, and they are turning as if their bodies were bigger."

    It should be read as

    "AN patients literally perceive their bodies to be larger than they really are, relative to objects and events in their environment, and they are turning as if their bodies were bigger relative to the doorway."

    Because all perception is scaled to "what I can do" units, there is no difference between thinking yourself bigger and thinking the door smaller; both thoughts entail each other.

    P.S. This seems like an easy-to-sell grant. "We will get AN patients and calibrate them to several body-size-relevant tasks. The two groups of participants will either have this treatment alone, or in conjunction with some other normal therapy. Weight before and after will be measured, and compared with the huge amount of normative data on the effectiveness of normal therapy."

    Obviously you would need a list of tasks to recalibrate the people to, not just door walking, but such a list shouldn't be too hard to make. There might even be a first grant in determining the tasks.

    1. The problem is that their current calibration is being actively maintained by something (or else they'd sort it out themselves using their real body size); so temporarily recalibrating them won't work without addressing the underlying problem. Why the hell would they be poorly calibrated to begin with?

  3. Fascinating study! I haven't read the actual paper yet, but 2 things come to mind;
    1) I think we need to establish that in AN their perception of other affordances is intact, eg. graspability of an object (hand aperture / object size), so we can rule out a general perceptual problem.
    2) The study reminds me of Proffitt's work on embodied perception, showing that judgment of geometric properties of the environment can be modified by action system variables such as fatigue.
    Here is a fascinating fragment broadcast on Dutch television, where young AN patients judged the circumference of body parts (belly / ankle) using a piece of rope. (click on play / fragment 9:30 - 10:30).


    1. The video clip is amazing. They are very, very wrong!

      I like the idea of the control. Bingham's done some work recently showing that prehension affordances do vary detectably with hand size though. I wonder what the differences are for a fat vs large hand?

      Proffitt's heart is in the right place, but his stuff bugs me because it's so lacking in precision and detail, and the whole point of affordances is precision! For example, he thinks the affordances for crossing spaces by walking and throwing are scaled relative to 'effort' - well, ok, but what the hell does that mean? We've been quantifying affordances for throwing in some detail and this is where it's at, I think. So I like Proffitt's thinking, but he and Jessica Witt need to stop running tiny little demo studies about how golf holes really do look bigger, and get into doing this properly.

  4. Ruling out a general perceptual scaling problem would be a good place to start - I suppose one possible problem with asking questions just about the size of objects is that this misses the 'action' aspect. Perhaps comparing performance on an experiment such as that by Linkenauger et al (2011) where they asked participants about the graspability of an object before asking them to estimate the size, thereby engaging the critical 'what could I do with this?' thought.