Place in perception: The role of structure and individual differences in scene processing

Our ability to rapidly process scenes despite the complex and varied visual information they contain represents a fundamental contradiction for cognitive science to explain. However, despite their great variation, natural scenes contain visual regularities that are highly diagnostic of their identity and function. Past research has found the visual system takes advantage of these regularities in order to facilitate this efficient processing through the use of internal models, which contain representations of typical scene information and act as a referential template for incoming visual information. In chapter 2, we use a jumbling paradigm to investigate how global scene structure is extracted from scenes, and under which conditions structure impacts categorising accuracy. We demonstrate that whilst disruptions to coherent global structure impacts processing, potential vertical biases observed in previous studies may be better explained by regularities in low and mid-level visual features. In chapter 3, we developed a novel drawing paradigm to describe the contents of internal models of the visual world, in order to investigate how individual differences in conceptions of typicality may drive efficient scene processing. Here, we found that drawings could be used to produce approximations of internal scene models, and that the strength of the match to these internal models was predictive of behavioural categorisation performance. In chapter 4 we conducted 2 further experiments to explore the contents of these internal models, by manipulating the structure and content of renders based on these drawings. However, we failed to find clear evidence for object identity and location being key features of internal models. Taken together, this thesis demonstrates how the visual system utilises structural regularities to facilitate scene processing and provides evidence for a possible influence of individual differences. We further show the ability of drawing paradigms to successfully represent internal models and investigate their application.