Abstract
In a single eye fixation, people understand the gist of a scene. Scene gist is defined as understanding the general meaning of a scene, which can include objects, actions, and the scene category. Research has shown that scene gist can facilitate object recognition. However, can this facilitative priming effect be transferred to another image? Due to the rapid speed at which scene gist is activated, researchers have theorized that it is the result of a rapid feedforward process, with little, if any, feedback processing. Therefore, each image would generate their own respective feedforward sweep, with little interaction between the two. If so, then understanding scene gist for a picture should not facilitate action categorization of a second image. Conversely, numerous studies have shown that semantically related objects and settings aid identification of those same concepts. This suggests that activating scene gist for one image should prime action categorization in another image. Participants were presented with Prime and Target images that contained semantically congruent or incongruent actions. Their task was to categorize the action in the Target. Various Prime-to-Target SOAs (from 24 to 365 ms) manipulated the strength of the Prime image, while a Target-to-Mask SOA limited target processing time to 24 ms. After the mask, a post-cue appeared that validly cued the target action 50% of the time. The cue remained until participants made a "Yes" vs. "No" decision. A control condition was included where the prime was absent. The data shows that at the earliest prime-to-target SOA, congruent prime images facilitated action categorization. This effect increased until the prime was processed for 100 ms or longer. Conversely, incongruent primes suppressed action categorization. This shows that when various concepts are activated during scene gist, they can prime and speed gist processing for subsequent images.
Meeting abstract presented at VSS 2018