Year: 2025
Type: article
Abstract: Semantic information is used to guide attention during scene-viewing, but how exactly does this occur? We tested the hypothesis that "anchor objects" are a special class of semantic associate essential to efficient target search in naturalistic scenes. Adult participants completed a RSVP search task for a target object preceded by a prime. In Experiment 1a (N = 210), the prime was either a semantic anchor or an unrelated object followed by a target in a congruent or incongruent spatial location relative to the prime. Experiment 1b (N = 157) served as a direct replication of Experiment 1a with a more selective stimulus set. In Experiment 2 (N = 158) the prime and target objects were reversed to assess the directionality of the effects. Results showed that semantically related objects facilitated target identification no matter which object was the prime. However, only anchor objects acted as predictive spatial cues, relative to control object pairs. The results indicate that anchor objects are a unique type of semantic associate, providing information about "what" objects are nearby and "where" they might be. The role of anchor objects in predicting the location of associated objects goes beyond both general scene regularities (e.g., "small things go on top of big things") and semantic associations (e.g., pots and stoves).
Authors: Makayla Souza-Wiggins, Joy J. Geng
Institution: University of California, Davis
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $3,090
Award: G6058395490
OpenAlex ID: W4416920645