Visual Search Efficiency for Features in Chernoff Faces

Click on the DOI link below to access the article (may not be free).Chernoff faces (Chernoff, 1973) graphically depict multidimensional data by correlating features of a cartoon face with values from a data set. Chernoff's original motivation for their use was that people perceptually prioritize faces and are acutely aware of variations in facial expression, so mapping data onto faces would support better understanding of complex data sets. Chernoff faces are usually depicted in groups where each face represents one particular case or unit and are compared to represent an entire dataset and its relationships. In this study, the authors sought to identify: i) the efficiency of visual search through sets of Chernoff faces relative to extant visual search efficiency benchmarks (e.g., spatial configuration search efficiency), ii) if face recognition processes do indeed provide advantages in visual search for such data, and iii) whether some facial features support more efficient search than others. The study consisted of four conditions in which two sets of four Chernoff faces were compared, each differing from a neutral face by one feature, for both upright and inverted versions. Inverting faces is known to interfere with face processing (e.g., Yin, 1969), so if face processing provides an advantage in searching through Chernoff faces, we reasoned that inverting the stimuli should slow search. The task was an oddball search with unlimited display durations and response time as the major dependent variable. Participants reported whether they saw a target face that differed from the neutral faces on each trial. Set sizes 5, 10 and 15 were tested, with target trials occurring 50 percent of the time. Results indicate differences in search efficiency for the various features, with particularly slow search for eyebrow position. Additionally, there was no observable effect of face inversion on search efficiency, for either set of facial features