Thursday, October 29, 2015

What does embodied simulation add to understanding?

Observing someone else being touched seems to activate one's own somatosensory cortex (e.g., this report).  It is has been claimed that this effect contributes to action understanding via embodied simulation. Some view this as an example of the "mirror mechanism" by which we understand others by mirroring their experience in our own bodies (or something like that).

First note that this touch-based "mirror mechanism" is quite different from so-called motor mirroring. The motor claim is non-trivial: perceptual understanding is not achieved by perceptual systems alone, but must (or can benefit from) involvement of the motor system.

What about perceptual mirroring?  At the most abstract level, the claim is this: perceptual understanding is based on perceptual processes.  Not so insightful is it?  Perhaps it's even vacuous. But maybe this is too harsh an analysis.  One could presumably understand the concept of someone being touched on the arm without involving an actual somatosensory representation.  So maybe it is non-trivial, insightful even, that we do activate our touch cortex when observing touch.  In fact, for the sake of argument, let's grant that the empirical observation is true and that it does contribute to our understanding.

What might it add to understanding?  Or put differently, how much does that somatosensory "simulation" add to our understanding of an observed touch?  Consider the following narrative scenarios.

Scenario #1: After he expressed his affection during the romantic dinner, the man reached out and touched the girl gently on the arm.

Scenario #2: After subduing his victim during the home invasion, the man reached out and touched the girl gently on the arm.

How much our understanding of the meaning of that touch action is encoded in the somatosensory experience?  Almost none of it.  The "meaning" of the action is determined for the most part by the context as it interacts with the observed action. The touch wouldn't even have to actually happen, or it could occur on a different body part (all very different experiences from a somato standpoint!), and it wouldn't alter our understanding of the event.  Yes, it's true that simulating the actual touch might add something, i.e., having a sense of what the actual gentle touch felt like on the arm, but what drives real understanding is the interpretation of that touch in its context, not the somatopically specific touch sensation itself.

Conceptualized in these terms, to say that somatosensory simulation contributes to understanding of others' touch experiences is like saying that "acoustic simulation" of the voiceless labiodental fricative in the experience of hearing "fuck you" contributes to the understanding of that phrase.  Yes, I suppose the /f/ plays a role, but how it combines with "uck you" and more importantly who said it to whom and under what circumstances is where the meat of the understanding will be found.

It's interesting and worthwhile to understand all the cognitive and neural bits and pieces that contribute to understanding.  Lowish-level embodied "simulation," whether motor or sensory, may have a role to play.  But it is important to understand these effects in the broader context.  Don't for a second think that we've cracked the cognitive code for understanding just because M1 or S1 activates when we see someone do something.


tim faber said...

Somebody in defense of the understanding hypothesis would say that the context allows people to predict and subsequently simulate the specific sensation which results in more refined understanding such as demonstrated in context examples in action understanding (Iacoboni et al., 2005; Kilner, Friston & Frith, 2007). But in that case any understanding happens before subjects perceive any action or touch-action unfolding making it more likely that understanding happens outside of the motor system.

I am pretty convinced that these examples make it hard to maintain the action understanding hypothesis (or sensation understanding for that matter):

Antonia Hamilton said...

This study tested the ideas you discuss, for touch

they find S1 activation changes for touch from opposite-gender compared to same-gender person, and claim S1 has an important role in understanding the meaning of touch. I'm not sure of the mechanisms though.

ikbol said...

How would your system know whether the man is touching her arm gently/caressingly, idly/distractedly, squeezing/irritated, jabbing/angry, doing some pinpointed action which you can't immediately work out, or any of an infinity of other variations on touching... except by embodied simulation?

My guess is you're starting from the assumption that because we have a single concept of "touching", there is a single action being referred to. In fact, like every other concept it embraces an endless, ever-expanding, potentially infinite range of evernew variations on a general action class, in this case touching, just as "smile" is intended to embrace evernew variations on smiling. You can be presented with evernew subtle variations on any action and yet understand each of those variations. You make the most amazing distinctions between thousands upon thousands of smiles, for example.

How on earth do you think we can understand a new variation of any action without simulating that action with our own body? How else do we know what is referred to?

You think we have a general, stock stereotype of "touch" which can automatically be applied to every new example of that action? And when s.o. says "HAND me that knife, rope, cactus, pig et al..." we have a stock stereotype of HANDING?

The reason you need to simulate actions with your body is 1) actions comes in evernew forms and 2) your body is infinitely plastic within constraints and can simulate and thus understand ever new forms of action. It's a "free formulator".

And when you referred to the "meaning/context" of touching, you were referring to the subsequent goal of touching, rather than the act itself. That too can only be understood by simulation. Will that kind of touching you see achieve the desired goal? Again you need to simulate it and compare it with past sequences of action which may or may not have led to similar goals.