Electrophysiological studies of the time-course of 3D shape perception and object recognition

This thesis reports the results of three novel studies using event-related potentials (ERPs) to examine (1) how different kinds of shape information across local and global spatial scales are computed, and integrated, during the perception of 3D object shape, (2) the role of stereo information in 3D shape processing and (3) the temporal dynamics of shape information processing. In experiment 1 we examined the time course of information processing at local and global spatial scales during object recognition and the role of stereo information in this processing. ERPs were recorded whilst participants completed a recognition memory task where they distinguished objects learned in training sessions from distracters that were either locally- or globally-similar. Participants completed the training and recognition task in either mono or stereo viewing conditions. The behavioural data showed a stereo advantage in object recognition and enhanced generalisation between trained and untrained views. The ERP data showed that during mono viewing, perceptual sensitivity was greatest for distracters with different local parts to targets during the N1. For stereo viewing, perceptual sensitivity was greatest for distracters with different 3D spatial configuration during the N2/P3 component. The findings show that there is differential ERP sensitivity to shape processing at local and global spatial scales and that stereo information is important in object recognition. The results, therefore, challenge theoretical models of object recognition that do not attribute functional significance to both 2D and 3D shape information. In experiment 2 we investigated the perceptual integration of information from local and global spatial scales, to find a temporal-spatial EEG marker for this integration. A Navon-type paradigm was used, whereby participants’ attention was directed to either the local or global level of the stimulus. The local and global levels could be either congruent or incongruent and the rationale was that congruency effects can be used as a functional marker for integration, 7 as congruency effects presumably arise at the point of global/local integration. ERPs were recorded whilst participants made orientation decisions about hierarchical stimuli made up of Gabor patches oriented to either the left or right. We found that there were interference effects evident in the behavioural and ERP data, particularly, global interference was evident at the N2/P3. We suggest that the global interference is evidence of global/local integration. In experiment 3 we aimed to examine the robustness of the integration of local and global information found in experiment 2 using more complex stimuli and a different task. Our stimuli comprised sets of geometrically coherent (possible) or incoherent (impossible) objects. The impossible objects comprised local parts that were geometrically coherent and global configuration that was possible. The objects’ impossibility becomes apparent when integrating the local and global levels of information. The rationale, therefore, was that the first point that the ERPs differed between possible and impossible object conditions would reflect the integration of information at local and global spatial scales. Using ERPs, we compared the processing of possible and impossible objects in a simple classification task. We found that there were no early processing differences for possible and impossible objects. However there were differences at the N2/P3 (from around 300ms post-stimulus onset) and we were able to verify that these differences did reflect the perceptual integration of local and global shape information. The results provided evidence of the generalisability of the integration effect at the N2/P3 with a more complex stimulus set. The main empirical findings in this thesis show: (1) processing of information from local and global spatial scales first occurs at the N1 and information from these spatial scales are integrated at the N2/P3; (2) local and global processing occur at least partly in parallel; (3) stereo information plays a role in object recognition and ought to be included in models of object recognition; and (4) our findings challenge models of object recognition that do not include independent coding of object parts and their spatial relation