CLASSIFICATION OF VISUAL SCENES BY OVERALL COLORFULNESS

Classifying objects or events is vital for survival and daily life.Categorization learning varies in its structure, stimulus-response associations, and feedback methods.In typical experiments, observers classify objects and receive feedback after each response, gradually associating stimuli with correct responses through trial and error.Determining which features of objects are relevant for classification can be complex.Color is a common visual characteristic used in this process, though colors are spread among multiple objects in natural scenes.This study explored observers' ability to classify visual scenes based on color dominance when the number of objects varied.The stimuli included 21 red and yellow squares within a 10x10 cm black square, with proportions of red and yellow ranging from 6/15 to 15/6.A total of 16 different stimuli were generated based on these color ratios, whereas the groupings of squares varied from 2 to 18 clusters.The classification was based on the rule that stimuli with more red squares were one category, while those with more yellow squares constituted the other.Each stimulus was shown five times in random order, totaling 80 presentations.Thirty-five observers (23 females and 12 males) aged 20 to 69 (average age 45) participated, learning to classify the stimuli through trial and error with feedback provided via sound signals.All observers were unaware of the classification rule.The cumulative sum of responses was formed and normalized by the number of presentations to reflect the observers' alignment with the classification rule.A robust regression method and a generalized mixed model regression analyzed factors influencing response accuracy and time.Results indicated: an improvement in response accuracy and a reduction in response time among observers; accuracy plateauing at about 75% in the final experimental block; just over half of the observers recognized the classification rule at various times during the experiment, observers who did not identify the classification rule changed their strategies more frequently, the percentage of correct responses increased with the color ratio more for dispersed stimuli; response times decreased with distance from the decision boundary between categories.The challenges in quickly classifying stimuli based on color suggest that the spatial characteristics of the objects may be the dominant feature, and separating the scenes into objects may hinder the ability to perceive individual object colors and the image's overall colorfulness accurately.