From perceptual to semantic representations in the human brain

Humans are capable of recognizing a myriad of objects in everyday life. To do that, they have evolved the ability to detect their commonalities and differences, moving from perceptual details to construct more abstract representations that we call concepts, which span entire categories (such as the one of people) or refer to very specific and individual entities (such as our parents). Organizing our knowledge of the world around concepts, rather than around individual experiences, allows us for more rapid access to behavioural relevant information (for instance, how to behave when we encounter a dangerous animal), and to quickly generalize this information to what we never encountered before. In few words, this is what permeates the world with meaning. The present work is about the neural bases of learning novel object concepts, a process that in our species is vastly supported by symbols and language: for this reason, I talk about semantic representations. The word “semantics” generally refers to the study of meaning (and to what a “meaning” ultimately is) as it is conveyed by a symbol; in the specific case of cognitive neuroscience, it deals with the neural mechanisms that allow symbols to re-present the meanings or concepts they refer to in the brain. For instance, we can easily describe what is the meaning of the word “DOG”, pretty much as we can explain what “DEMOCRACY” means. However, although cognitive neuroscience has focused on the neuro-cognitive bases of semantic representations for decades, the neural mechanisms underlying their acquisition remain elusive. How does the human brain change when learning novel concepts using symbols? How does a symbol acquire its meaning in the brain? Does this learning generate novel neural representations and/or does it modify pre-existing ones? What internal representational format (neural code) supports the representation of newly learnt concepts in the human brain? The contribution of this work is three-fold. First, I show how new semantic representations learned by categorizing novel objects (defined through a combination of multisensory perceptual features) memory systems. Second, I show results converging on the idea that brain regions that evolved in lower-level mammals to represent spatial relationships between objects and locations, such as the hippocampal formation and medial prefrontal cortex, in humans are recruited to encode relationships between words and concepts by means of the same neural codes used to represent and navigate the physical environment. Finally, I present preliminary data on the cognitive effects of using symbols during learning novel object concepts, showing how language supports the construction of generalizable semantic representations.