Thursday, 10 April 2014

Does action scaling predict 'the embodiment of culture'? (No.)

I recently reviewed a paper for Frontiers by Arthur Glenberg and colleagues called 'Sensory motor mechanisms unify psychology: the embodiment of culture' (Soliman, Gibson & Glenberg, 2014). This is part of an ongoing research topic on embodied cognition run by Guy Dove. Once the paper had been published, I took the opportunity to write up my main remaining problem with the paper as a commentary piece (Wilson, 2014) and I'd like to review that here (and also please see my comment at the end about my role here as reviewer). Go download the paper, though, I worked hard to produce a focused critique in the 1000 word limit and I think it went well.

Glenberg and colleagues are trying to develop a broad embodied framework that can encompass both traditional hunting grounds like perception and action but also 'higher order' cognition like social cognition and culture. I admire the effort but to my mind this is grounded cognition, not embodied and so this is not an approach I'd endorse. I like to review papers like this though because it's healthy to have an adversarial reviewer who tries not to be evil (I promise!), and because I also like to keep myself informed about the work I'm rejecting so that I don't start fighting straw men.

Soliman et al ground their approach in Proffitt's theory about the action-scaling of distance perception. I reviewed a recent point-counterpoint on this topic (Firestone, 2013; Proffitt, 2013) that summarises the key facts about Proffitt's theory. In particular, one key finding is that people judge things to be farther away when it requires more effort to cross that distance (because they are tired, for example, or carrying a heavy load).

Soliman et al go looking for this kind of change in distance perception as a function of social group membership. If they can find it, they argue that this would license them to discuss perceptual and cultural/social factors in the same theoretical framework, unifying psychology. They therefore ran a study in which they had people judge the distance to ingroup and outgroup members. The logic is this (bear with me, it's a long walk):
  1. Interacting with someone requires you to (unconsciously) mentally simulate things they do
  2. We do this simulation ahead of time, in anticipation of an encounter
  3. This simulation will require more effort with out-group members, because they do things differently to you
  4. This increased simulation effort will affect your judgment of distance, as per Proffitt
  5. Therefore out-group members should look further away
They also measured whether people doing the judging has a independent or interdependant self-construal. Independent people will have more experience with out-group members than interdependent people, and hence the required simulation should be easier for this group. 

The results seemed to support this set up. In two experiments, there was an interaction between self-construal and judged distance to in- and out-group members. Interdependent people rated the distance to ingroup members as smaller compared to independent people (suggesting their increased practice interacting with ingroup members resulted in less effortful simulations). In the second experiment, the result flipped for outgroup members. This interaction suggests the two self-construal groups are using different metrics to judge the distance, and Soliman at al claim that these metrics must be effort based because of Steps 1-3 above.
Figure 1. a) Results of Experiment 1. b) Results of Experiment 2. See Soliman et al for details
My commentary
I wrote a commentary that summarised my primary concern from the review process. I didn't even get into the problems with the idea that we mentally simulate upcoming social interactions, because to be honest I barely know where to start. I think this is a ridiculous idea but I haven't had time to get into it deeply enough to justify a critique on this basis. I instead focused on the (mis)use of Proffitt's action scaling theory. The discussion between Firestone (2013) and Proffitt (2013) was extremely useful here, as was the fact I'd reviewed it for the blog.

In our embodied cognition paper (Wilson & Golonka, 2013) we laid out a 4 step research strategy that embodied cognition research should follow. Those steps are
  1. What is the task to be solved?
  2. What resources are available to solve the task?
  3. What are the viable ways you could combine your resources in order to solve the task?
  4. What does the organism actually do?
Step 3 is crucial. We specify this step in order to to require that the hypothesised mechanism is built out of resources that definitely exist and that definitely can be hooked together that way. Psychologists often waver here, and although Soliman et al try to tie their pieces together, they can't because they have misunderstood Proffitt's result.

Proffitt's theory (derived from his data) is that effort affects distance perception, but in a very task specific way. If I make your legs tired and ask you to judge a distance to cover by walking, the effort manipulation will make the distance seem greater. But if I ask you to judge a distance to cover by throwing (i.e. in a way that doesn't depend on tired legs) then the distance does not seem bigger. In Proffitt's words, "An important finding across our studies is that the influence of an action unit – such as graspability – is evident only within its action boundary’ (Proffitt, 2013, pg. 477 and see in general his reply to Firestone's Argument 2). 

Soliman et al, however, require Proffitt's mechanism to work across action boundaries. In particular, they need the effort involved in mentally simulating an upcoming social interaction to influence your judgment of a distance you must cross by walking. There is exactly nothing in Proffitt's data or theory that would allow this, and a lot that explicitly rules it out as an option. Distance perception is changed because effort, etc recalibrates the effectors involved in the upcoming action. This simulation effort (even if it exists) does not recalibrate anything to do with walking, and so Proffitt's account simply does not predict or explain 'the embodiment of culture'. 

Why pick on this point? Because it matters, and because Soliman et al very explicitly hung their hat on this peg. They want to unify psychology, but they fail because the pieces they use simply cannot be assembled the way they claim, and it's this fairly common problem in psychology that bugs us the most. If we ever want to do great science, we need to get serious about our mechanisms.

Commentary at Frontiers
I enjoyed the ability to write this commentary based on my reviews. I was reviewed by Guy Dove and Arthur Glenberg, both of who were fair and constructive. My critique is now part of Guy's special topic and a permanent companion to the Glenberg article, and so I hope people reading the main paper will at least be made to think. I think this format and connection is a nice feature of Frontiers and helps treat published work as part of an ongoing discussion, rather than a static document. 

Reviewing at Frontiers
Why did I let the paper through? I had the option to withdraw as a reviewer if I thought the paper was so flawed, and I nearly did. 

My original review raised many problems with the data analysis as well as with the theory. Glenberg and colleagues actually did a good job of fixing those problems and I thought the resulting manuscript was a better piece. Once these problems were fixed, I felt uncomfortable ruling out publication on this theoretical grounds, even though I think they are making a critical error. The data are what they are, and they require explanation; I just think their current explanation is flawed. That doesn't mean there isn't something here to be explained and the study itself was adequately run (not especially sophisticated, but within the range of this kind of work). Given that it reached this bar, allowing the paper to be published then meant publishing my critique and getting the conversation into the literature, which I thought was not a bad idea.

All that said, I still wonder if the right move was for me to withdraw from the review on this issue. This is what you do at Frontiers when you reach an impasse with the authors; it's then up to the editor to replace me or agree with my rejection of the paper, as I understand it. Should I have taken that stand on this critique? I'd be interested in people's thoughts on this!

Firestone C. (2013). How "Paternalistic" Is Spatial Perception? Why Wearing a Heavy Backpack Doesn't--and Couldn't--Make Hills Look Steeper, Perspectives on Psychological Science, 8 (4) 455-473. DOI: 

Proffitt D.R. (2013). An Embodied Approach to Perception: By What Units Are Visual Perceptions Scaled?, Perspectives on Psychological Science, 8 (4) 474-483. DOI: 

Soliman T, Gibson A and Glenberg AM (2013) Sensory motor mechanisms unify psychology: the embodiment of culture. Front. Psychol. 4:885. doi: 10.3389/fpsyg.2013.00885 (Download)

Wilson AD (2014). Action scaling of distance perception is task specific and does not predict ‘the embodiment of culture’: A comment on Soliman, Gibson & Glenberg (2013). Front. Psychol. 5:302. doi: 10.3389/fpsyg.2014.00302 (Download)

Wilson AD and Golonka S (2013) Embodied cognition is not what you think it is. Front. Psychology 4:58. doi: 10.3389/fpsyg.2013.00058 (Download)


  1. Just to add one wrinkle: Frank Durgin has some evidence that the backpack effect might actually be a demand effect.

    1. Yes, I've seen Frank's stuff. I guess I should also say I'm not that into Proffitt either. I think he's right in spirit but wrong or incomplete on details. This commentary is really just saying that if you're going to hang your hat on a theory, you should probably know whether that theory has a hat hook :)

  2. For whatever it might be worth I think staying as reviewer and responding to the paper publicly was the right move.

    I've always thought that as a reviewer there was an onus on me to apply my own critical judgement (perforce informed by my own thinking), but the role also involves acting on behalf of the community of researchers. To that end, while I might disagree personally with a particular piece of work, if it abided by existing standards in the community then it should go forward.

    Where I disagree with the community the place to try to change things is in public, through published research and argument, not by silently throttling other points of view. (Though an editor might make such decisions, I don't think the reviewer should.)