The Roles Of Scene Consistency In Object Categorization

Object categorization can take place at different levels of abstraction.Some studies found that the visual system deals with information at the basic level(e.g.,dog)faster than at the superordinate(e.g.,animals)levels or subordinage(e.g.,pug-dog)levels.But other studies suggested that the surperordinate category information can be processed more quickly than the basic level information.These studies uses the pictures of the real world so that they have high ecological validity,but there are more extraneous variables.This study focus the roles of the relationship between the scenes and objects in object categorization.we adopted a forced-choice saccadic task in experiments.In this task,the target image and the distractor image were displayed simultaneously,the observers must saccade as fast as possible toward the target image on the screen.We explored the advantage of the object categorization levels and the roles of scene consistency in object categorization through four experiments.We found the advantage of the superordinate levels in Experiment 1 when we controlled the factors of the scenes.We investigated the object categorization by manipulating the context of the scene in Experiment 2,3,4.We explored the role of the semantic consistency in object categorization,and found that the semantic inconsistency can improve the accuracy in basic level.We found that syntactic relations between the scene and the object has no influence on object categorization.We explored the role of the stimulus attribute(nature,artificial)in object categorization,and found interesting thing: on the superordinate level,the stimulus attribute consistent objects can be categorized better,but on the basic level,the stimulus attribute inconsistent objects have high scores.The result of this study supports the superordinate level advantage,and the semantic consistency and stimulus attribute consistency influence the object categorization process.These findings place powerful constraints on theories of object recognition.