Perceptual Discriminability as a Basis for Selecting Graphic Symbols

Abstract

The purpose of this research was to develop a performance-based criterion for selecting among alternative symbols to be used in graphic displays. The specific criterion developed was an index of perceptual discriminability. Through regression analyses of an intersymbol similarity-rating matrix, it was concluded that symbols are judged more or less similar on the basis of the number of shared versus unique configural attributes (an X, a triangle, etc.), as opposed to primitive attributes (number of lines, arcs, etc.). An easy-to-use discriminability-index formula was derived from the regression analysis involving the configural attributes, and this formula was used to predict the results of an experiment involving a search for specific symbols embedded in an array. Indices obtained from a formula such as the one developed here could be used as part of the basis for choosing among alternative candidate symbols for inclusion in an existing symbol domain.

When Beauty is Only Skin Deep: 3-D Realistic Icons are Harder to Identify than Conventional 2-D Military Symbols

“…Jolicouer and Humphrey, 1998). And we know that discriminability among tactical symbols is a strong predictor of performance in operational tasks such as visual search (Geiselman and Christen, 1982).…”

Track Recognition Using Two-Dimensional Symbols or Three-Dimensional Realistic Icons

“…Discriminating between realistic icons makes life unnecessarily difficult for the user because the icons are difficult to distinguish. We know that discriminability among tactical symbols is a strong predictor of performance in operational tasks (Geiselman and Christen, 1982;Remington and Williams, 1986).…”

A rule of thumb of icons' visual distinctiveness

“…Tversky's contrast model [13] stipulated that the perceived similarity between two objects is linearly correlated to the number of common features and negatively correlated to the number of distinctive features of those objects. Geiselman et al [4] proposed a mathematical notation of discriminability as: n n j=l j=L where wu = a weighting factor used for unique feature ws = a weighting factor used for similar feature a~j = the number of features unique to each icon asj = the number of common features…”