Purchase this article with an account.
Benjamin Balas, Jamie Schmidt; Children's use of visual summary-statistics for material recognition. Journal of Vision 2017;17(10):1080. doi: 10.1167/17.10.1080.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
A growing body of work describes adult material recognition and perception,but the development of material perception during childhood has been largely ignored. Presently, our goal was to investigate how material perception changes during middle childhood, both with regard to overall abilities as a function of material category and also with regard to the use of summary-statistics for material recognition relative to higher-order/global appearance descriptors. We carried out two experiments: A material categorization task (Experiment 1) and a material matching task (Experiment 2). For both tasks, we recruited children aged 5-10 years old (N=40) and compared their performance to adults (N=20). In Experiment 1, participants performed a 4AFC material categorization judgment using natural images of metal, stone, water, and wood, and synthetic images created from these using the Portilla-Simoncelli texture synthesis algorithm. In Experiment 2, we used these stimuli to implement a material matching task. Here, participants were presented with a sample image on each trial and indicated which of two test images depicted the same material. In both experiments, the synthetic appearance condition allowed us to examine how performance was affected by representing material appearance using only the Portilla-Simoncelli summary-statistics. In Experiment 1, we observed a significant interaction between age, material category, and appearance (p = 0.004): Observers of all ages were less accurate with synthetic textures, but young children (5-7 years old) were disproportionately worse, especially when categorizing wood and water. In Experiment 2 (which did not require labeling textures), we observed a slightly different outcome: Water was again disproportionately impacted by texture synthesis, but this did not interact with age. We conclude that the ability to use summary-statistics for material recognition develops during middle childhood, and that different material categories depend differentially on such descriptors, leading to asynchronous development of material perception across distinct material classes.
Meeting abstract presented at VSS 2017
This PDF is available to Subscribers Only