AK doesn’t think so.
In another of his lengthy and well-researched posts, he argues that the understanding of more or less abstract concepts occurred in primates before a language based on words. This is based on a recent study of the mirror neurons in rhesus macaques. This research seems to indicate that rhesuses divide other rhesuses into two categories when the mirror neurons are activated: those within such a short distance that interaction is immediately possible, and those further away.
The post also includes an interesting reflection on how visual information is encoded in terms of a set of vectors in multidimensional space, suggesting that the same principle applies as a general form for representation in the brain.
In the process, AK also manages to discredit Plato’s idea that concepts are classes of things resembling an “ideal” concept that is by definition beyond our grasp. Instead, we construct concepts “bottom-up”, by grouping together objects and ideas that appear to us to have many similarities.
Implicit to AK’s argument is also the notion of a well-developed spatial modularity in the brain, with different areas encoding different concepts. While there is strong evidence for spatial modularity e.g. from split-brain experiments, showing that the two hemispheres can accurately identify and interact with objects independently of each other, it is very likely that at least some concepts are represented only by the concurrent activation of several areas in synchrony.