Polygon displays: Optimization for integral effects and comparison with bar graphs in a process control system requiring fault detection

The purpose of the study was to compare polygon and bar graph displays for their effects on detecting fault states in a simulated process control system. Two other factors were also involved in the study: the difficulty of the task and the number of parameters represented in the display. Ninety-six undergraduate students participated in the study. Half the participants viewed polygons and the other half viewed bar graphs. Each of these halves was further divided into three groups of 16 each to view displays containing 8 or 12 or 16 parameters. All participants received two difficulty levels of the task—easy and difficult. The between-subjects independent variables were display format (polygons and bar graphs) and number of parameters (8, 12, and 16). And the within-subjects independent variable was task difficulty (easy and difficult). The dependent variables were reaction time and accuracy. All participants responded to 2-second exposures of the stimuli (polygons or bar graphs with 8 or 12 or 16 parameters in both easy and difficult task conditions representing various process control states) by pressing one of four keys that are mapped to the process control states of the stimuli. The task that the participants performed was the speeded classification task. The experiment recorded the reaction time and the accuracy of the responses.; Results indicated that (1) the polygons were superior to bar graphs in the speeded-classification performance, (2) increases in the number of vertices in the polygon generally did not degrade the speeded-classification performance when the task condition was easy, (3) increases in the number of vertices in the polygon from 12 to 16, however, did degrade the speeded-classification performance when the task condition was difficult, (4) increases in the number of bars in the bar graph display generally did not degrade the speeded classification performance in both the easy and difficult task conditions, and (5) increasing the difficulty level of the speeded classification-task impaired about equally with both polygon and bar graph displays when accuracy was measured.