Tuesday, 19 February 2013

Embodied vs. Disembodied Approaches to Language and Conceptual Processing: Finding a Middle Ground

In the last two decades, Embodied cognition approaches have been gaining increasing popularity among researchers concerned with the study of the neural organization of Language. The version of embodiment that has attracted the most attention is the so called ‘strong embodiment’ or ‘full simulation’ approach (For a useful review of the different flavors of embodiment that deal with language and semantic processing I recommend this paper). In a nutshell, strong embodiment views conceptual content as being “reductively constituted by information that is represented within the sensory and motor systems. According to the embodied cognition hypothesis, ‘understanding’ is sensory and motor simulation.” (Meteyard et al., 2010). This view of embodiment was met with even greater enthusiasm with the discovery of mirror neurons.

Researchers who work within the ‘strong’ or ‘full simulation’ embodied cognition framework argue that language comprehension is mostly dependent on the motor system. The most prominent of these views is the Motor Theory of Action Semantics which provides the foundations for Glenberg and Gallese’s ABL or Action Based Language. It’s main proposal is that the comprehension of action verbs is dependent on the activity of the motor system, which allow us to simulate the action referred to by the verbs and thus provides the basis for our understanding. This idea is supported by what G&G term the ACE. The ACE or Action-Sentence Compatibility Effect, has become somewhat of a poster child for the embodied camp. To take a concrete example, the ACE is thought to occur whenever a sentence implying motion in one direction (eg. He pushed the box away) facilitates responses (in the form of pushes on a button) that are directed away from the subject and interferes with responses that are toward the subject. A number of experiments seem to support the ACE and its influence on Language comprehension (for those interested, see Vega (2008); Zwaan and Taylor (2006)).

However, as mainstream as these approaches appear to be (based on the number of publications that adhere to the strong form of embodiment), not everyone is buying into them. Neuroscientist Greg Hickok is one of the most ardent (and my opinion, humorous) critiques of ABL and the ACE. Here is an amusing extract from his blog:

“They (G&G) want to argue that this means that the meaning of say push is grounded in actual pushing movements that must be reactivated to accomplish understanding.  The ACE is interesting but not surprising nor conclusive.  Just because two things are correlated (the meaning of the word push and the motor program for pushing) doesn't mean one is dependent on the other... If I blew a puff of air in your eye every time I said the phrase "there is not a giraffe standing next to me", before long I could elicit an eye blink simply by uttering the phrase.  Furthermore, I could probably measure a There-Is-Not-A-Giraffe-Standing-Next-To-Me-Eyeblink Compatibility Effect (the TINAGSNTMECE) by asking subjects to respond either by opening their eyes wider or by closing them to indicate their decision. This does not mean that the eye blink embodies the meaning of the phrase.  It just means that there is an association between the phrase and the action.”

In this paper, Hickok also bemoans the Motor Theory of Action Semantics’ lack of explanatory power. Indeed, while this version of embodiment can be said to work for verbs depicting actions that the human body can perform (eg. kick, run), how does it cope with action concepts that humans beings cannot execute (e.g. fly, hover, tail-wag, locomote-via-jet-propulsion, that some humans cannot perform (e.g., juggle, swim, play-the- sousaphone), that don’t involve the motor system (e.g., think, grow, bleed, sleep, compare), and for any number of non-motoric verbs (e.g., digitise, melt, erupt). Hickok rightly wonders “If we have a mechanism for representing these concepts such that we can understand fly or juggle or hover even if we cannot perform such actions, then why do we have to invoke the motor system as a necessary component of concepts such as kick or kiss?”

Taken to its extreme, the strong version of embodiment’s reliance on inconclusive neuropsychological evidence can lead to some bizarre conclusions. Indeed, one of the main neuropsychological arguments used by the embodied camp is the results of a study showing that people suffering from ALS (amyotrophic lateral sclerosis or motor neuron disease) have specific difficulties in processing verbs. However, it turned out that in all 6 cases used in the study, the pathology spread beyond the motor cortex. Moreover, those cases were from a subset of patients who presented a dementia/aphasia variant of the disease further calling on the validity of the findings. Hickok suggested that we should let Glenberg & Gallese tell Steven Hawking that he no longer can understand verbs.

A growing number of researchers now agree that neither a strongly embodied (ie-calling on ‘full simulation’) nor a fully disembodied theory can be supported. So where do we go from here?

In his aptly named article Disembodying Cognition, Chatterjee argues that the issue at hand is not discussions of embodied vs. disembodied but rather reconsidering to what extent cognition really is embodied. Chatterjee favors a more graded approach. According to him, “a focus on disembodying cognition, or on graded grounding, opens the way to think about how humans abstract”. For those interested in a rather technical neuroscientific description of his approach in which he proposes that three functional anatomic axes could help frame questions about the graded nature of grounded cognition, here is a link to the full article.

Mahon & Caramazza (2008) also offer an alternative (and in Hickok’s view, better) way of thinking about the problem by suggesting that motor information merely forms a part belonging to a broader representation whose role it is to augment action concepts (as opposed to being the basis for our understanding of those concepts). Caramazza & Mahon’s hypothesis seeks to combine the view that, at some level at least, “concepts are ‘abstract’ and ‘symbolic’ with the idea that sensory and motor information may ‘instantiate’ online conceptual processing”.


  1. While I personally don't agree with this hypothesis, especially the "strong version," I think it makes a relevant point. One could argue that there exists a component of the verb "to fly" that cannot be completely understood unless one can fly, just as one cannot understand what it is like to be a bat unless one is a bat. At the same time though, because no humans can fly, we as a species are simply deprived of this extra component of meaning.

    This line of thinking though becomes complicated when one considers people with acquired disabilities; ex. someone who was healthy but suffered an accident and was paralyzed. Can this person, who experienced things like running and walking in the past, no longer fully understand the meaning of the words because he cannot walk or run in the present?

  2. In relation to embodied semantics, Hauk & Tschentscher (2013) agree that both neuroscience (neuroimaging data) and behavioural data have failed to crack the fundamental connection between process and representation. They therefore call for future research to place more of an emphasis on 'the effects of task and context on semantic processing'.

    They state that "Strong conclusions can only be drawn from a combination of methods that provide time-course information, determine the connectivity among poly- or amodal and sensory-motor areas, link behavioral with neuroimaging measures, and allow causal inferences"

    For those interested, here is the link to this article called 'The Body of Evidence: What Can Neuroscience Tell Us about Embodied Semantics?'