Tuesday, 22 July 2014

Embodying Culture: My ongoing conversation with Soliman & Glenberg

Would a formal reply make me this guy?
I've been exchanging views with Art Glenberg and his colleagues about a paper he published recently in Frontiers. I reviewed it, had reservations but eventually let it through, then published my concerns as a commentary on the original paper. Soliman and Glenberg (2014; S&G) then replied to my reply which I didn't know about until I noticed the commentary had a citation in my Google Scholar profile

I could simply publish a reply to their reply, but to be honest I'm not sure it's worth it; it feels a little too much like arguing on the internet. I'll link to this in the comments section of the Frontiers page, however, and if people think it's worth the DOI then I'll write this reply up as a formal submission. I'd be interested to hear from you all on this.

The short version of my reply is that in the process of dodging my criticism they concede it applies to them, and they swerve into a literature that doesn't help. I do think they've applied some serious and valuable consideration to the details of their proposal, though, so I think this has been a useful process.

The Basic Claim (Soliman et al, 2013)
Our perception of our upcoming interactions with the environment are scaled with respect to our current ability to implement that interaction. For example, fatigue makes hills look steeper (see this post reviewing a head to head on the topic between Chaz Firestone and Dennis Proffitt. Side note: writing those papers up for the blog tooled me up enough to review and sensibly critique Glenberg's paper, so score 1 for blogging!).

Soliman et al (2013) claimed that the anticipated effort of an upcoming social interaction with an out-group member should be larger than for an in-group member. Applying Proffitt's mechanism, they predicted that this anticipated effort should make the distance required to go engage in the interaction look farther, and this is basically what they found.

The Critique (Wilson, 2014)

My problem relates to a central part of Proffitt's theory: that this perceptual scaling is task-specific. Effort in one domain (e.g. making your legs tired) changes judged distances when you intend to walk that distance but not when you intend to throw that distance. As the head-to-head revealed, Proffitt considers this to be a non-negotiable feature of the theory. The problem for Soliman et al is that this means their proposed chain of events is neither predicted nor explained by Proffitt's theory. I wouldn't have minded so much but it was the central framing of their paper, so this issue matters.

The Reply (Soliman & Glenberg, 2014)

The reply has two parts; I'll handle them in the reverse order S&G present them.

1. Proffitt's stuff is entirely different from ours

S&G walk through the logic of Proffitt's work and compare it to theirs. They note that for Proffitt,
the manipulation phase targets one motor system and then tests the effect of the manipulation on perceived distance as the participant intends to perform another task. For example...participants are adapted while temporarily turned into throwing phenotypes, and then tested while in the walking phenotype. Typically, it was found that the visio-motor scale developed while in one phenotype did not transfer to the other...
pg 2
and then 
In our experiments, however, no behavioral phenotype was turned on, manipulated, switched off, replaced by another, and then examined. Instead, our participants were walker-then-interactor phenotypes throughout....We believe that these subtle design differences render our original results and theoretical arguments immune to Wilson's critiques.
pg 2
It does, but at the cost of entirely rejecting their original framing, and my point was always that Proffitt's work could not possibly motivate or explain their work. As far as I can tell, this reply just admits I was right.

Given this, do they have a replacement motivating theory? Yes, but it's not going to save them.

2. End state comfort and motor contagion 
Their new description for their task implies that instead of having two task specific systems that don't interact, they actually have people engaged in a composite task; not walking and then interacting, but walking-so-as-to-interact. 

There are well known examples of these kinds of tasks in which the later parts affect planning and execution of the earlier parts. The main example they cite is Rosenbaum's end state comfort effect. The classic example is reaching-to-grasp-and-turn an upside down wine glass; people adopt an initially uncomfortable upside down hand posture because this means that when they turn the glass, they end up with the hand the right way round in a comfortable posture. People are therefore not reaching-to-grasp, then turning; if they were, they would execute the first part more efficiently. Instead, they suck up the initial cost so that the whole movement is efficient and the final result stable. (I've done a little work on this in the past; Kent et al, 2009; van Swieten et al, 2010).

On the surface it sounds like this work applies nicely to S&G's study; but, as with Proffitt's work, as soon as you consider the actual mechanism at work the match fades away. These cases of 'motor contagion' all show that the earlier movements are reshaped in functional ways to serve the higher order task requirements. In the wine glass case the initially uncomfortable reach-to-grasp makes the overall reach-to-grasp-and-turn action more efficient and stable. S&G describe another study where grasping requirements affected locomotion over to the object to be grasped; but again here, the effect was that people adjusted their final steps to facilitate grasping the object. 

S&G just don't have this set up. The increased effort of interacting with an out-group member will not be reduced by perceiving the distance to that person as farther away. Returning to Proffitt, people rescale distance perception when tired, for example, because that the distance really will be harder to cover and making it look that way facilitates effective planning. The upcoming social effort has no consequences for how hard it will be to walk over the distance (the root of my initial objection) and so making that distance seem farther is not a functional adjustment (the root of this objection). 

S&G also mention example of tasks contaminating other tasks in non-functional ways (e.g. people unintentionally opening their mouths wider when simultaneously reaching for wider objects). This also doesn't help; this is about simultaneous actual actions affecting each other, not sequential upcoming actions. 

I get a kick out of the fact that Glenberg takes me seriously enough that he thought it was worth replying to me. He reviewed my reply and was engaged and fair throughout. I disagree with him about embodied cognition but he's still a major name in the field and I've enjoyed having this out with him and his colleagues.  I also like the fact that my reply really made them examine their task in detail. I wrote my commentary in the spirit of constructive criticism and I'm thrilled it's been taken as such. It's been a refreshing change, really.

All that said, I still don't buy it. All the similarity between these various tasks and Glenberg's experiment is superficial, at the level of the broad strokes verbal description of the effects. The actual mechanisms in play, either in Proffitt's work or the end state comfort effect, simply do not apply to Glenberg's study and so they still have failed to embody culture.

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

Soliman TM and Glenberg AM (2014) How intent to interact can affect action scaling of distance: reply to Wilson. Front. Psychol. 5:513. doi: 10.3389/fpsyg.2014.00513

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


  1. The domain specificity you mention is crucial but I suspect there's another issue overlooked by G&S and this concerns the difference between declarative and procedural knowledge of a task. The contrast is nicely exemplified actually by Proffitt's distinction between the cases of walking and throwing. When you ask someone to evaluate a distance they have to perform a linguistically enabled judgement whereas when they throw something no judgement need be involved at all for a successful result — hence the difference in the walking case and the throwing case. The same kind of contrast is involved in optical illusions like the Müller-Lyer diagram. If optical illusions were just down to judgement then there would never be a conflict between two ways of responding to the diagram.

  2. Thanks, for the link. I look forward to reading it. I was also thinking of the many grip aperture studies that have been carried out with optical illusions and the consistently reported results that these illusions have no measurable influence on the accuracy of sensorimotor action.
    I think it's reasonable to conclude that perceiving something often involves more than one way of efficaciously responding to the thing perceived.
    Taking the declarative account as the definitive version of what an individual is capable of successfully doing with a thing, is to miss this important fact and to privilege language skills over skills that have either been much longer in the making or, as is the case with many optical illusions, skills that have been discovered only relatively recently, i.e. linear perspective.
    So, when we say that fatigue makes a hill look steeper, we don't actually mean that the hill appears to have changed shape. What we are actually doing is inflecting our report with an expression of the anticipated difficulty involved. What someone is inclined (excuse the pun) to say or think about an out-group member is often quite different from what one is actually disposed to do.

    1. Agreed. I'm with Geoff in advocating action measures as the gold standard for measuring perception for this reason. Judgment studies can be done well, but it's always harder and noisier.

    2. This reminds me that I should do a second "myth of knowledge" post. The hypothesis of "knowledge" was (among other things) that there should be a very strong match between the declarative and performative. Thus the idea that those are two different types of knowledge kind of misses the point. (It is the same error inherent in the term "multiple intelligences".)

      Also, for what its worth, I think they might be able to pull off the contagion argument, but it would need to be MUCH broader. Faced with engaging in a task you do not want to do, most people become depressed - their motivation drops (he says as a behaviorist), their metabolism shuts down, etc. It would be really cool to know if you can find the undesired social interaction causing such systemic shut downs, which could then in turn be connected with a perceived increase in distance. If you connected to metabolic changes, you might well be able to create a metric that was something like "effort needed to accomplish the walking task". I think you are spot on though that this is far from what Proffitt and Rosenbaum were getting at.

      That type of possibility, incidentally, is the type of thing I am waving at when I argue that ecological psychologists and developmental psychobiologists should really talk more.

    3. Sabrina's getting more into the physiology end of things as she develops her taxonomy for these kinds of reasons.

      Re the contagion argument, I just think that they are mis-understanding what, say, the end state comfort effect is about. I have an affordance based analysis in mind which would make this clearer but I'm cutting my teeth on the throwing papers first.

  3. I'm not sure I understand you Eric. Are you saying that the declarative/procedural distinction is as faulty as the multiple intelligences theory?