Neural correlates of dynamic object recognition

Abstract

This thesis investigates how colourful dynamic objects are represented in the human brain. The two main theories of object recognition (structural description models and image-based models) make different predictions about which object features (e.g., shape, motion, and colour) are critical for recognition and how individual features can be selectively attended for further processing. To investigate the relevance of different features for object recognition two sets of novel 3D objects which had different combinations of shape, colour, and motion were created. Selective attention paradigms in which participants attended to single object features were used. The effects of changing unattended features on behavioural performance (Experiment 1 to 3) and large-scale brain responses (Experiments 4 and 5) were measured. The experiments in this thesis used psychophysical experiments combined with functional magnetic resonance imaging (fMRI) and diffusion tensor imaging (DTI). In addition, the relevance of structural and functional connections between brain areas involved in object recognition was investigated. The results showed that the representations of shape and motion are closely linked while colour seemed to be processed more independently of other features. At the neural level, shape and motion activated a common network consisting of occipitotemporal, lateral-frontal, and parietal areas whereas colour activated ventral occipital areas. Further support for a common network involved in shape and motion processing was found in structural connections between these areas. For example, the integrity of white matter tracts connecting the occipito-temporal and the lateral frontal areas was correlated with behavioural performance. Although colour activated a distinct set of brain areas compared to shape and motion, activity in colour sensitive brain areas modulated the activity in shape sensitive areas which suggests integration processes of multiple object features. The results of this thesis suggest a synthesis of elements of both of the two main object recognition theories rather than favouring one of them.