Abstract
Can working memory (WM) exist outside of conscious awareness? Previous reports of 'unconscious' WM have been challenged on the grounds that above-chance performance could result from a superposition of (i) random guesses in the absence of awareness and (ii) accurate responses with subthreshold awareness. We addressed this possibility using a visuospatial delayed estimation task. In Experiment 1, we presented a colorful dot at a random location around an invisible annulus and instructed subjects to maintain the position of this dot across a brief delay. At the end of the delay, they reported the position of the dot with a mouse click. Experiment 2 was identical, except that we presented a line during the second half of the delay and had subjects mentally mirror the remembered location across this line before reporting this mirrored location. Across trials, we manipulated the visibility of the dot using backward masking: the dot was either masked after 16.7ms (suppressed trials), 50 ms (non-suppressed trials), or absent altogether (absent trials). Subjects rated their perception of the dot on each trial using a 4-point scale. We analyzed suppressed trials that were given the lowest perceptual rating (matching the perceptual rating of absent trials) and estimated subjects' WM precision from their memory reports using a standard mixture model. Even when they were unaware of the dot, subjects were able to report its location with a surprising degree of precision. This result held for Experiment 2, demonstrating that it is possible to manipulate WM representations without awareness. Most critically, we show through simulations that these data could not have resulted from the superposition of trials with no information and trials with subthreshold awareness. These results confirm that WM is possible without awareness and that it shares several properties with conscious WM.
Meeting abstract presented at VSS 2017