Classifying ambiguous everyday objects: the effects of stimulus duration, visual field, and background contextTools Sankar, Harini (2022) Classifying ambiguous everyday objects: the effects of stimulus duration, visual field, and background context. MRes thesis, University of Nottingham.
AbstractVisual perception in humans is optimised to function in its habitat. When sensory information is ambiguous, our prior knowledge and expectations about the habitat biases perception. By measuring perceptual biases for ambiguous stimuli, we are able to infer what expectations we have about our environment. Recently, Hussain Ismail et al. (2019) found a perceptual bias in people living in cities, where ambiguous objects (made of one man-made and one natural component images) were more likely to be classified as a man-made object (i.e., manufactured objects such as a vehicle or a house) rather than a naturally occurring object (e.g., flower or animal). They speculated that this bias may be a result of our expectations to see man-made objects more often in our living environment. The aim of this thesis is to examine whether the “man-made bias” is susceptible to factors that are known to alter perceptual biases in vision. The first experiment examined whether the man-made bias is influenced by presentation duration when ambiguous objects (“hybrids”) are directly fixated at. We found that a shorter presentation duration (50 ms) that increased the uncertainty of the stimulus increased the magnitude of the man-made bias compared to a longer duration (150 ms), in line with predictions of Bayesian perception. Experiment 1 also demonstrated that the man-made bias is replicable when object classification is measured with an online experiment, using a new collection of object stimuli on a new set of participants. Experiment 2 measured the same perceptual bias by presenting hybrids in participants’ peripheral vision. Although we found a significant, albeit small, perceptual bias, we did not find any effect of stimulus duration on the man-made bias in the periphery. In Experiment 3, participants viewed hybrids at fixation, and we measured the effect of background scenes on which hybrids were superimposed on. The superordinate semantic category of the background scene (man-made or natural) did not affect the magnitude or the direction of the man-made bias, indicating that background scenes did not alter the detectability of semantically congruent component images in the hybrids. Overall, we show that the man-made bias is resistant to changes in presentation duration, visual field in which objects are viewed and the semantic congruency of the spatial context.
Actions (Archive Staff Only)
|
Tools
Tools
