Abstract
Previous studies have shown that higher-order statistics underlie observers’ performance in animal detection task (Banno & Saiki, 2015), and is more efficient than second-order statistics in masking scene gist perception (Loschky, Sethi, & Pydimarri, 2010), indicating that higher-order statistics might play an important role in scene recognition. In the current study, we examined if high-order statistics could evoke perceptual priming effects on rapid scene recognition. For this purpose, it is necessary to extract higher-order statistics from intact scene images without extracting information such as the contour of objects and the overall spatial layout of the scene that could lead to conceptual priming. The algorithm suggested by Portilla and Simoncelli (2000) could meet this need by preserving higher-order statistics of images while disorganizing global configuration at the same time. Thus, we synthesized texture images using Portilla and Simoncelli’s (2000) algorithm to preserve higher-order statistics of scenes while making them unrecognizable for their gist. Two experiments were conducted using the paradigm based on previous studies: participants were first presented with eight primes followed by an intact target scene, and were instructed to answer the basic category of the target scene in a four-alternative forced choice task. Priming level (all primes belonged to the same basic category vs. all primes belonged to the same superordinate category) and prime-target category congruency (congruent vs. incongruent) was manipulated. Experiment 1 used intact images as primes (conceptual priming and perceptual priming), while experiment 2 used texture images synthesized based on the primes used in experiment 1 (perceptual priming). We found facilitation from primes on rapid scene recognition under the prime-target congruent condition of experiment 1, while no robust influence from prime-target congruency was found in experiment 2. The current results suggest that higher-order statistics itself is not sufficient to evoke category-specific perceptual priming effect in rapid scene categorization.