Associations, expectations and meaning: decoding the neural processes underlying conceptual representations
thesisposted on 28.03.2022, 03:07 by Lina Teichmann
Integrating incoming visual information with prior knowledge is vital when interacting with the world around us. But how do internal concepts and expectations about the environment shape perception? The research presented in this thesis focuses on disentangling neural activity associated with conceptual representations from the visual information that activate them. Using time-resolved multivariate pattern analyses (MVPA), I present four Magnetoencephalography (MEG) studies that aim to increase our understanding of how the neural activity associated with conceptual representations unfolds over time. In the first study, I examine whether there is a shared magnitude representation that can be activated independently of numerical format. The results showed that a shared representation of magnitude can be accessed via different numerical symbols (i.e., digits and dice) but that this representation is accessed slightly earlier via digits than dice. These findings highlight that there is an internally generated magnitude representation that can be separated from incoming visual information. In the second study, I explore whether accessing representations of objects with a strong canonical colour results in typical object colour being activated. When participants viewed greyscale objects that are associated with a specific colour (e.g., a greyscalestrawberry), the typical colour of the object (e.g., red) was decodable suggestingactivation of the object representation includes colour information. Further, I showed thatcolour information accessed via object-colour activation resembles later stages of real colour perception. These findings provide novel insights into the time course of object feature representations that are based on prior knowledge. In the third study, I describe the use of a congruency paradigm to investigate what happens when our prior expectations about object features are violated. The results show that the typicality of feature binding (colour and form) influences the neural response over time and can be differentiated. Whether the conjunction of object and colour was typical had an effect on colour representations but not on object representations. By focusing on the temporal aspect of object processing, this study highlights the effect of typicality of feature binding on object and colour processing. In the fourth study, I discuss data of a special population, grapheme-colour synaesthetes, who experience colours when perceiving achromatic letters and digits. For grapheme-colour synaesthetes, colour information could be decoded from brain activation patterns associated with viewing achromatic symbols in a very brief time window. Colour representations accessed via synaesthetic inducers resembled later stages of colour representations activated via real colour perception. Focusing on the timecourse of the neural signal associated with synaesthetic colours suggests that synaesthesia might be similar to object-colour knowledge. Together, the findings of this show that MVPA applied to MEG data is a suitable method to measure conceptual representations independent from the visual information that activates them. These results give novel insights into the complexities of visual perception and its dependency on prior knowledge.