Research On The Level Priority Of Scene Gist Extraction

The scene gist extraction is defined operationally to classify scenes,which can be performed at three levels:superordinate level,basic level and subordinate level Observers have different processing times for classification at different levels.The study found that the subordinate level classification processing time is the longest,but there is a level priority dispute between the superordinate level and the basic level processing time.In the historical research,there are two viewpoints:basic level superority and superordinate level superority.It is difficult to unify the two viewpoints because there is a lack of systematic discussion about the influence of global attributes and object information on the level priority,so that it's hard to solve the dispute at the level of the scene gist extraction mechanism.This study aims to reveal the theoretical mechanism of level priority by exploring the influence of the processing of global attributes and object information on level priority.In Experiment 1,we manipulated the stimulus onset asynchrony between the scene and the mask to see the difference of performance between basic level classification and superordinate level classification in different processing time.In Experiment 2,we explored the impact of global attributes on level priority using category discrimination paradigm.In Experiment 3,we explored the impact of object information on level priority using dual task paradigm.These are the results:1.In the early stage(when ISI is less than 30ms),the performance of basic level classification is better than superordinate level.With the increase of processing time,the performance gap between superordinate level and basic level gradually disappears.2.When the processing of global attributes in basic level classification is limited,the priority of basic level is reversed in the early stage,making the performance of superordinate level discrimination task was better than basic level.3.When the processing of object information in superordinate level classification is limited,the performance of superordinate level classification is still lower than basic level in the late stage(when ISI is more than 70ms).The above results show that the processing of the global attributes in the early stage helps to obtain the basic level classification,which makes the classification performance of the basic level better than the superordinate level,and the processing of the object information helps to obtain the superordinate level classification,so that there is no classification performance difference between superordinate level and basic level in the late stage.