Giessener Elektronische Bibliothek

GEB - Giessener Elektronische Bibliothek

Perceptual qualities and material classes

Fleming, Roland W. ; Wiebel, Christiane ; Gegenfurtner, Karl

Originalveröffentlichung: (2013) Journal of Vision 13(8):9 doi:10.1167/13.8.9
Zum Volltext im pdf-Format: Dokument 1.pdf (2.425 KB)

Bitte beziehen Sie sich beim Zitieren dieses Dokumentes immer auf folgende
URN: urn:nbn:de:hebis:26-opus-104304

Sammlung: Open Access - Publikationsfonds
Universität Justus-Liebig-Universit√§t Gie√üen
Fachgebiet: Psychologie
DDC-Sachgruppe: Psychologie
Dokumentart: Aufsatz
Sprache: Englisch
Erstellungsjahr: 2013
Publikationsdatum: 09.12.2013
Kurzfassung auf Englisch: Under typical viewing conditions, we can easily group materials into distinct classes (e.g., woods, plastics, textiles). Additionally, we can also make many other judgments about material properties (e.g., hardness, rigidity, colorfulness). Although these two types of judgment (classification and inferring material properties) have different requirements, they likely facilitate one another. We conducted two experiments to investigate the interactions between material classification and judgments of material qualities in both the visual and semantic domains. In Experiment 1, nine students viewed 130 images of materials from 10 different classes. For each image, they rated nine subjective properties (glossiness, transparency, colorfulness, roughness, hardness, coldness, fragility, naturalness, prettiness). In Experiment 2, 65 subjects were given the verbal names of six material classes, which they rated in terms of 42 adjectives describing material qualities. In both experiments, there was notable agreement between subjects, and a relatively small number of factors (weighted combinations of different qualities) were substantially independent of one another. Despite the difficulty of classifying materials from images (Liu, Sharan, Adelson, & Rosenholtz, 2010), the different classes were well clustered in the feature space defined by the subjective ratings. K-means clustering could correctly identify class membership for over 90% of the samples, based on the average ratings across subjects. We also found a high degree of consistency between the two tasks, suggesting subjects access similar information about materials whether judging their qualities visually or from memory. Together, these findings show that perceptual qualities are well defined, distinct, and systematically related to material class membership.
Lizenz: Veröffentlichungsvertrag für Publikationen ohne Print on Demand