Abstract
Previous studies (Hwang et al., 2011) indicate that our strategies for memorizing objects in naturalistic scenes can be predicted by the semantic relationships between objects in that scene. That is, we tend to make saccades to objects that are most semantically related to the object in the current fixation. A possible explanation is that consecutive inspection of semantically similar objects facilitates object memorization. Previous work in our lab has shown that indeed, characteristic objects from a specific context have a close semantic relationship that facilitates recognition. The observed high recognition rate was not induced by gist-based errors, as performance did not decline even when false recognition rates were significantly reduced. When and how does this ‘semantic advantage’ arise during processing? We investigated this question using a rapid serial visual presentation task, which mimics sequential eye movements. A series of eight grayscale object images on a white background were shown for 250 ms each. Subsequently, participants saw another image and indicated whether it had been in the series (same/different judgment). The object sets were randomly chosen or taken from specific contexts such as airport, park or bedroom. In three experiments, we measured recognition accuracy as we varied target position (excluding first and last image in the series) and primed participants with context labels either before or after a trial. The context labels provided for random trials were randomly chosen and misleading. Results show that the semantic advantage starts after viewing approximately three objects in the context trials when compared to random trials. Priming with context labels (either before or after the trial) diminishes this advantage and in general hurts performance. These results reveal that object memory benefits from the semantic structure established by an efficient, unconscious mechanism that is impeded by conscious category processing.
Meeting abstract presented at VSS 2015