Purchase this article with an account.
Timothy Brady; Visual working memory relies on separate viewpoint-specific ensemble and viewpoint-invariant object representations. Journal of Vision 2016;16(12):32. doi: https://doi.org/10.1167/16.12.32.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
When participants are asked to remember objects in working memory, they encode not only information about individual items, but also information about the ensemble properties of the display. For example, when shown a display of fine-lined Chinese characters and heavy 3D cubes, participants remember not only particular items but also global texture, like which regions of the display are darker vs. lighter (e.g. Brady & Alvarez, 2015). These spatial ensemble representations boost memory performance by providing an additional source of information about a visual display above and beyond information about individual items. In the current experiments, I show that participants' spatial ensemble representations are dissociable from their representations of individual items. In particular, spatial ensemble representations are considerably more viewpoint-dependent than item representations. Across a variety of tasks, participants' ability to use ensemble information to detect changes is severely impaired after a 3D rotation of the display. In contrast, their item representations are unaffected by 3D rotations. This effect is specific to 3D, not 2D, transformations: Spatial ensemble representations are unaffected by 2D transformations like translation or scaling. For example, in one experiment, participants were shown displays of 3D cubes and Chinese characters and asked to detect either cross-category changes (cube->character; which disrupts the spatial ensemble) or within-category changes (cube > cube, which requires individual item memories). The ability to detect cross-category changes was significantly impaired by 3D rotation (Dd-prime: 0.37), whereas detection of within-category changes was completely unaffected (Dd-prime: -0.06; interaction: p=0.007). Across a variety of similar experiments, spatial ensemble representations are reliably viewpoint-specific. Thus, spatial ensemble representations may function independently of our object memory system. Spatial ensemble representations may instead be related to the brain's scene processing network, which remains viewpoint-specific further along the visual processing hierarchy than does the object system (Epstein et al. 2003).
Meeting abstract presented at VSS 2016
This PDF is available to Subscribers Only