Boundaries of Efficient Object Recognition: Animacy, Ecological Expectation, and Connectomics
Object recognition is subject to constraints on its performance. These constraints are necessary, given the limited capacity of the brain to process the abundance of visual information we receive at any time. Consequently, the visual system is tuned to process objects optimally under some conditions. Optimal conditions can be described as the boundaries within which object recognition performance is most efficient. My thesis focuses on object recognition, with a particular emphasis on the boundary conditions which influence the efficiency of object recognition. In the first empirical chapter, I examine how object animacy affects visual memory, attention, and visual search performance. The results showed that search for memorised real-world objects was significantly more efficient when an animate / inanimate distinction was present between the objects held in memory and those being searched through. These findings imply that perceptual information about object animacy were used in the visual search strategy, which drastically reduced the cost of the search process. In the second empirical chapter, I consider how visual object real-world size information interacts with stereoscopic distance, doing so through manipulations of object visual size, apparent physical size, and egocentric distance in virtual reality. The results showed that object recognition and the perception of spatial relations were both more efficient when object size and distance information were congruent with real-world size expectations. These findings point to the integral role real-world size plays in object perception. In the third empirical chapter, I used publicly available MEG and resting-state fMRI data to investigate if the efficiency of category-level object processing (as inferred via machine learning classification of MEG data) differs within a population. The results showed considerable individual differences in the efficiency of object processing. Further analysis revealed significant covariation between the efficiency of object processing and specific networks of functional connectivity. ii Collectively, the work presented in my thesis contributes to our knowledge of object recognition by describing factors which influence the efficiency of object recognition. These findings also open avenues to future research on object recognition.