Beyond brain decoding: methodological and empirical contributions to brain decoding methods and their link to behaviour

Theoretical thesis.Thesis by publication.Includes bibliographical references.Chapter 1. Introduction -- Chapter 2. Decoding dynamic brain patterns from evoked responses -- Chapter 3. Asymmetric compression of representational space -- Chapter 4. Beyond brain decoding -- Chapter 5. DiscussionBrain decoding methods have transformed the field of Cognitive Neuroscience in the last two decades. As this field has matured, researchers are now asking challenging questions about the interpretation of decoding methods, such as whether the information decoded from neuroimaging data is used by the brain for behaviour. The focus of this thesis is to address the challenge of linking information measured with decoding methods to human behaviour. In the first empirical chapter, using magneto-encephalography (MEG), I describe a broad set of options for conducting time series decoding studies and test the effects that different options in the decoding analysis pipeline can have on the experimental results. The results show that decisions made at all stages of the analysis can significantly affect the results and interpretation of decoding studies. In the second empirical chapter, I explore the distance to bound model as a method for linking brain decoding with behaviour. Using MEG decoding, I tested whether this model can account for behavioural changes in reaction time for categorising degraded objects by animacy. I found that the distance to bound model successfully predicted reaction time, accuracy, and decision time parameters derived from a prominent model of decision making. These findings provide evidence for a systematic relationship between decoded brain representations and perceptual decision-making behaviour. In the third empirical chapter, I examine the distinction between decodable information, and information that can be used in behaviour. Using a searchlight approach on functional Magnetic Resonance Imaging (fMRI) data, I first investigated where decodable information existed in the brain. Secondly, I assessed where in the brain the decoded information was suitable for "read out" by the brain for behaviour. I found that behaviour can only be predicted from a subset of the locations that had decodable information. These results highlight the distinction between decodable information, and information that is relevant for behaviour in the brain. In conclusion, this thesis advances current knowledge on brain decoding methods and on approaches to relating brain representations to behaviour, which is a fundamental challenge in cognitive neuroscience. The results show that decodable information has to be interpreted with caution, and emphasize that continuing to develop methods for linking neuroimaging to behaviour is critical for advancing our understanding of the brain.Mode of access: World wide web.1 online resource ( xii, 258 pages) illustration