01whole.pdf (37.13 MB)
Download file

The temporal dynamics of visual object recognition

Download (37.13 MB)
thesis
posted on 29.03.2022, 00:25 by Erika Wilson Contini
Visual object recognition is a complex problem, with much still to be discovered about how the visual system achieves this task. Several studies have examined the emergence of object category structure, focusing particularly on animacy as an overarching principle of the neural organisation of object representations. Results from fMRI studies have highlighted additional organisational principles for category structure, such as real-world size, and biological class, however the temporal dynamics of these category organisations are yet to be established. The aim of this thesis is to build upon our understanding of visual object recognition, with a specific focus on evaluating the temporal dynamics of object category structure as measured with MEG. Using representational similarity analysis applied to MEG data, the first empirical chapter compares the temporal dynamics of animacy and real-world size dimensions of object representations. The results replicate previous findings for the animacy time-course, however there was no evidence for a distinct time-course associated with real-world size. The second empirical chapter examines alternatives to the animacy category organisation of object representations, using a novel stimulus set that includes objects which do not clearly belong to the typically evaluated 'animate' or 'inanimate' categories (e.g., robots and human-/animal-like toys). This study evaluates a range of models based on current theories of object categorisation including animacy, and the biological classes based 'animacy continuum', as well as novel behaviourally-generated models related to human-similarity and experience. Results show that the model of human-similarity is the best predictor of object representations late in the time-course of visual object processing. The aim of the third empirical chapter is to link these human-similarity results from the MEG data to behaviour. This study shows that object categorisation reaction times predict representational distance not only for object animacy (as shown in previous studies), but also when objects are grouped according to human-similarity. In contrast, other plausible object category organisations for the same stimulus set (i.e., living/non-living; has movement/no movement) do not show the same relationship between brain activation patterns and behaviour. To conclude, the findings from these three studies are discussed within the broader context of the current literature related to object representations in the human brain. This thesis highlights the efficacy of a new human-similarity model of object category representations and critically evaluates what aspects of decodable neural representations are informative for understanding the link between brain and behaviour.

History

Table of Contents

Chapter one: General introduction -- Chapter two: Evaluating the temporal dynamics of object representations as a function of animacy and real-world size -- Chapter three: Neural coding of visual objects: New insights into categorical representations -- Chapter four: Reaction times predict dynamic brain representations measured with MEG for only some object categorisation tasks -- Chapter five: General discussion

Notes

Includes bibliographical references Thesis by publication.

Awarding Institution

Macquarie University

Degree Type

Thesis PhD

Degree

PhD, Macquarie University, Faculty of Human Sciences, Department of Psychology

Department, Centre or School

Department of Psychology

Year of Award

2018

Principal Supervisor

Mark Williams

Rights

Copyright Erika W. Contini Copyright disclaimer: http://mq.edu.au/library/copyright

Language

English

Extent

1 online resource (xii, 176 pages) illustrations

Former Identifiers

mq:72175 http://hdl.handle.net/1959.14/1282147