Proportsioonide eristamise täpsuse sõltuvus diferentseeriva tunnuse tüübist ja eristatavate elementide arvust

Abstract

The aim of this study was to examine observers’ ability to discriminate numerical proportions depending on the number and attributes of the elements. Observers’ task was to decide which one of the two overlapping sets of elements, distinguished by colour or orientation, was more numerous. Bernoulli response model (based on hypergeometrical distribution) with a single free parameter K (the supposed number of inspected elements on which the choice was made) was applied to describe the choice probabilities. According to the Bernoulli model, these K elements are chosen randomly, their number limited by perceptual capacity. With the growth of the total number of displayed elements N = 9, 13, 33, 65, the number of accounted elements increased disproportionately, with colour being a stronger discriminating feature than orientation. It is concluded that the Bernoulli model is an alternative for the habitual Thurstonian-type model for examining observers’ ability to discriminate proportions.