Avery Caulfield, Justin Halberda¹, E. J. Green¹; ¹Johns Hopkins University

Abstract: An object file is a perceptual representation that sustains reference to an object over time while retaining information about the object’s currently and recently perceived features (Kahneman et al. 1992). Here we investigate whether object files incorporate a distinction between individuating and non-individuating features of objects, and whether this distinction governs how object files interface with the rest of the mind. In one series of experiments, subjects detected a pop-out target among homogenous distractors and then made a memory judgment about that object’s features (with the precision of each feature staircased across trials: the pop-out individuating feature and the ancillary non-individuating feature). Findings suggest that, surprisingly, participants demonstrate less precision for the individuating feature than for the non-individuating feature. We hypothesize that individuating and non-individuating features play distinct computational roles within object files. The representation of the pop-out feature functions to pick out the object to be attended, while the representation of the non-individuating feature is prioritized for higher-precision processing in visual working memory. This asymmetry between individuating and non-individuating features was mirrored in another series of experiments where subjects were given a pop-out display and were asked which English sentence best described the array. Subjects preferred descriptions placing the pop-out feature within the scope of the demonstrative and the ancillary feature in the predicative clause. For example, when shown a red tilted bar among green tilted bars, subjects preferred the description “That red bar is tilted” over “That tilted bar is red.” This pattern suggests a correspondence between genuinely discursive representations (i.e., language) and a visual language of thought.