VISUAL GENERALIZATION : THE EFFECTS OF GLOBAL/LOCAL PROCESSING AND SAMPLING ASSUMPTIONS

SOPHIA SONG XIU LING MUHAMMAD DANIAL

Publication

VISUAL GENERALIZATION : THE EFFECTS OF GLOBAL/LOCAL PROCESSING AND SAMPLING ASSUMPTIONS

SOPHIA SONG XIU LING MUHAMMAD DANIAL

Abstract

The Bayesian framework of Tenenbaum and Griffiths (2001) describes generalisation as the product of the individual’s prior learning and their assumptions on how the exemplars were sampled. Through a model of visual object recognition proposed by Gerlach (2006), the present study links Bayesian priors to the individual’s subjective perception of stimulus difference and similarity. We compared generalisation between participants primed with either a global or local processing-bias and exposed to cover stories indicating that exemplars were either sampled purposefully or randomly. We found that participants primed with global processing-bias tightened generalisation under a strong sampling assumption, whereas those primed with a local processing-bias failed to generalise. Additionally, the interaction between the two variables depended on the extent of similarity the novel stimulus had to the exemplar stimuli. Our results suggest that how individuals utilise category traits during generalisation is subject to their perceptual processing style and inference from cover stories.