Understanding Navon: Different designs of local-global tasks capture different bias effects

When processing hierarchical compound figures with local elements in a global shape (often called the "Navon task"), an observer may perceive information from the local or the global level more quickly and/or accurately. This difference is referred to as local-global bias. This basic principle - measuring a difference between the local and global level - is conceptually very broad and can be implemented with many different task designs and figure designs. Importantly, however, there exists a pervasive notion of this bias being a unitary construct or even representative of a stable perceptual or cognitive style. However, few studies have examined the various facets of local-global bias across different task designs or assessed the retest reliability of these measures. Our study aimed to deepen the understanding of local-global bias by using three different hierarchical compound figure tasks: the Navon task and two Kimchi-Palmer task variations (one preference-based, one performance-based). We tested 75 participants across two sessions 1 week apart. Our findings revealed that metrics from the Navon task are not test-retest reliable, whereas two metrics from the Kimchi-Palmer tasks are. Some metrics were correlated between the Kimchi-Palmer tasks, but not between the Kimchi-Palmer task and the Navon task. These results suggest that local-global bias is not a unitary construct, and that different tasks measure different underlying mechanisms. We discuss possible reasons for these differences and suggest that future research should use multiple local-global tasks or otherwise prioritize Kimchi-Palmer tasks for studying individual traits and possibly Navon tasks for temporary and group-level effects.