Purchase this article with an account.
Kathryn N. Graves, Brynn E. Sherman, Nicholas B. Turk-Browne; Closer than it appeared: Distorted spatial memory during virtual navigation. Journal of Vision 2020;20(11):1056. doi: https://doi.org/10.1167/jov.20.11.1056.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
As we navigate the world, we tend to visit multiple locations within a local neighborhood (e.g., spots in a parking lot, restaurants on a block, offices in a building). We remember not only these individual locations but also abstract over these locations to represent their underlying spatial distribution. We have previously shown that such patterns guide future navigation when searching for new locations in a familiar environment. However, what is the fate of the individual past locations? Are these representations distorted by the distribution of other locations, for example, pulled toward the mean of the distribution? Or are patterns generated in parallel with high-fidelity memory for individual locations? We could not test this previously because participants in our earlier studies were never instructed to return to an old location. In a new study, participants virtually navigated a circular arena with an outer-space theme. During training, they were exposed to a distribution of five locations across search trials, each appearing as a colored disk. On each trial, participants were prompted with a color word that informed them of the disk they should find. They were then shown the disk in the arena and navigated to it repeatedly. During test, participants were again given a color prompt but no disk was presented in the arena. Instead, they had to rely on spatial memory for the past location of the corresponding disk. Their navigational behavior thus provided a rich source of information about how this location was represented. Participants generally navigated toward the correct location but were biased to stop in a location that was closer to the other experienced locations. These findings suggest that statistical learning helpfully extracts patterns that can guide navigation to new locations probabilistically, but that this comes at the expense of distorted memory for old locations.
This PDF is available to Subscribers Only