One of the fundamental questions in cognitive science is how information is stored in the brain and in the mind. There are innumerable different models, each with its own strengths and weaknesses, but the one that I will be addressing here is known as embodiment.

From a neurolinguistic perspective, embodiment is the idea that the semantic content belonging to words is linked to sensorimotor representations in the brain. So if you talk about doing something, your brain behaves (to some extent) as if you are actually doing it.

In a view known as “weak embodiment”, linguistic and conceptual representations overlap in certain sensorimotor areas. In the “strong embodiment” view, linguistic and conceptual representations are much more strongly linked, with some researchers suggesting that action words actually trigger the neuronal assemblies that are associated with those actions.

A recent study sought to discover if either of these two views could be supported by neuroimaging. The researchers played sentences describing dynamic (“the mechanic is walking toward the airplane”) or static (“the mechanic is looking at the airplane”) situations and measured the brain response. The researchers looked in particular both in the visual cortex area V5 — which is known to respond to visual motion perception — as well as other temporal areas of the brain.

They suggested that, if the strong embodiment view holds, dynamic descriptions should trigger activity in areas that are very strongly linked with visual motion perception, like V5. If the weak embodiment holds, they said, it was more likely that these descriptions would not trigger V5, but would activate other areas related to motion perception (presumably ones that are less strongly linked to visual cues).

After using localizer stimuli to determine the location of each participant’s V5 area, the authors analyzed the effect of static and dynamic language stimuli on this area. They found activation in the left posterior middle and superior temporal gyri, near V5, but crucially, no overlap between language-activated areas and V5 proper, supporting the weak embodiment theory of language.

So what does this all mean? The authors point out that the sensorimotor representations that are triggered by language are not as specific as the strong embodiment view suggests. In this case, language triggered activation in areas that are more generally linked, amodally, to motion. These parts of the cortex seem to be activated by amodal information related to motion, such as animacy or intention. This has significant implications for theories of embodiment, as strong claims of embodiment must now come up with a way to explain the finding that “once-removed”, schematic areas are activated by language, instead of modality specific representations  (as far as I am aware, there hasn’t been a response to this article from proponents of the strong embodiment view, but I’m sure it’s coming).

Finally, it’s worth pointing out that this may have some relevance for theories on the relationship between language and non-linguistic cognition. Visual sensorimotor representations, being non-linguistic, seem to be triggered, though somewhat indirectly, by language. Exactly how this works could also be important in understanding the other ways in which language and memory interact.

References

Humphreys GF, Newling K, Jennings C, & Gennari SP (2013). Motion and actions in language: semantic representations in occipito-temporal cortex. Brain and language, 125 (1), 94-105 PMID: 23454619

Image via benchart / Shutterstock.