Abstract
Previous studies on visual statistical learning (VSL) showed that when statistical regularities were constructed by real-world images with semantic information, participants could extract statistical regularities at a basic category level (Brady & Oliva, 2008) as well as at its subordinate object level (Otsuka, Nishiyama, Nakahara, & Kawaguchi, 2013). However, how the statistical regularities at the basic and subordinate levels interacted with one another has not been investigated. The current study examined whether statistical regularities at a basic category level influenced the extent of VSL at a subordinate object level. In a familiarization phase, participants detected a repetition of the same image as a cover task, while different real-world object images appeared sequentially. Within this temporal stream, statistical regularities at a subordinate object level existed among three different images, which constructed sets of triplets and always appeared in the same order (e.g., a parrot, a sports car, and then a gold fish). For half sets of triplets, temporal regularities also existed at a basic category level (e.g., categories of bird, car, and fish in a fixed order), whereas the other half of triplets did not include such basic-level regularities (e.g., categories of dog, flower, and house in a variable order). A total 6 sets of triplets were used. These are more than 4 sets of triplets typically used in the previous studies. In a test phase, participants were instructed to judge whether each triplet was familiar or not (two-alternative forced-choice) while learned and unlearned triplets were presented. Results showed that participants extracted statistical regularities across the six triplets. However, the extent of learning was not different between triplets with and without regularities at a basic category level.
Meeting abstract presented at VSS 2016