Measuring display clutter in different contexts: evidence from performance and attention in websites and maps

Algorithms can be used to measure display clutter, but their validation often relies on comparing to subjective ratings, rather than to performance and attentional decrements. It is unclear which algorithms work best in different contexts. In this study, participants searched for targets in websites and maps. Feature congestion (FC), subband entropy (SE), and edge density (ED; Rosenholtz, Li, and Nakano 2007) were calculated for each image. Results showed that only FC is positively linked to response time in websites and maps, but more so for websites and moderately in both cases. SE was negatively correlated with mean saccade length in websites. In maps, all clutter metrics were positively correlated with mean fixation duration, while FC and ED were negatively correlated to mean saccade length. The findings provide recommendations for which clutter algorithms and eye tracking metrics to use for websites and maps, while suggesting caution about some metrics.