The Effect of Relative Size Consistency on Object Recognition

Thumbnail Image
Journal Title
Journal ISSN
Volume Title
Objects that frequently co-occur in the real-world appear to have canonical size relations. For example, a milk carton should be twice the size of the glass next to it. Similarly, a chair next to a table needs to be appropriately sized to form a functional group. Although the real-world size of objects has been identified as an integral property of object processing and search, little is known about how canonical size relations between objects influence recognition for semantically congruent object groups. Here, we used electroencephalography (EEG) and behavioural measures to test whether real-world size consistency between familiar object pairs facilitates grouping and recognition. We constructed silhouette object pairs containing semantically associated objects drawn from two possible real-world scale categories (eg., large objects – desk and chair, small objects – a bottle and a glass). We perturbed size consistency by rescaling one of the two items by 1:2 ratio. In Experiment 1 we validated the stimulus set through behavioural testing. In Experiment 2 brain activity was recorded using EEG while participants performed a one-back task during which size consistent and inconsistent targets were viewed. We tested whether event related potential (ERP) magnitudes differed as a function of size consistency. We found differences between mean amplitudes for size consistent and inconsistent trials specific to scale category. That is, for targets that appear large in the realworld responses significantly differ as a function of consistency earlier (P200) as opposed to small targets (P300, P600). We further hypothesized that scene understanding, measured by scale decodability, should be better for consistent as opposed to inconsistent pairs. However, multivariate classification analysis found no evidence that those underlying representations differ significantly. Keywords: relative-size, size constancy, object recognition, visual cognition, EEG decoding
Faculteit der Sociale Wetenschappen