Numerosity underestimation in sets with illusory contours. (May 2016)

• Record Type: Journal Article
• Authors: Kirjakovski, Atanas; Matsumoto, Eriko
• Is Part Of: Vision research. Volume 122(2016)
• Page Start: 34
• Page End: 42
• Publication Date: 2016-05
• Subjects: Numerosity estimation -- Segmentation -- Object perception -- Configural processing -- Illusory contours; Vision -- Periodicals; 573.88
• DOI: 10.1016/j.visres.2016.03.005
• Languages: English
• ISSNs: 0042-6989

Abstract

Highlights: Collections of line-connected dots exhibit underestimates of numerosity. Similar to line-connected collections, illusory contours led to an underestimation. Less easily perceived illusory contours or larger numerosities eliminated the effect. Estimation depends on segmented inputs, but is limited by the collection size. Abstract: People underestimate the numerosity of collections in which a few dots are connected in pairs by task-irrelevant lines. Such configural processing suggests that visual numerosity depends on the perceived scene segments, rather than on the perceived total area occupied by a collection. However, a methodology that uses irrelevant line connections may also introduce unnecessary distraction and variety, or obscure the perception of task-relevant items, given the saliency of the lines. To avoid such potentially confounding variables, we conducted four experiments where the line-connected dots were replaced with collinear inducers of Kanizsa-type illusory contours. Our participants had to compare two simultaneously presented collections and choose the more numerous one. Displays comprised c-shaped inducers and disks (Experiment 1), c-shaped inducers only (Experiments 2 and 4), or closed inducers (Experiment 3). One display always showed a 12- (Experiments 1–3) or 48-item reference pattern (Experiment 4); the other was a test pattern with numerosity varying between 9 and 15 (Experiments 1–3) or 36–60 items (Experiment 4). By manipulating the number of illusory contours in the test patterns, the level of connectedness increased or decreased respectively. The fitted psychometric functions revealed an underestimation that increased with the number of illusory contours in Experiments 1 and 2, but was absent in Experiments 3 and 4, where illusory contours were more difficult to perceive or larger numerosities were used. Results corroborate claims that visual numerosity estimation depends on segmented inputs, but only within moderate numerical ranges.