Abstract
Recent behavioral evidence has shown that certain images are consistently remembered and others forgotten irrespective of subjective experiences. Even though this intrinsic property of visual stimuli, termed memorability, is highly replicable across individuals, the neural circuitry subserving this behavior is currently unknown. To determine a lower boundary in the visual hierarchy for memorability-related neural signals, we investigated whether memorability information is represented at early stages of the visual system even for unconsciously perceived stimuli. We collected MEG data in a rapid serial visual presentation (RSVP) experiment. Participants (N=12) viewed streams of 11 images at 34ms/pic, a rate extremely effective in masking stimuli. The middle image (target) was randomly sampled from a set of 30 faces or 30 objects (15 memorable and 15 forgettable each), all controlled for low level features. The distractor images comprised 150 scene images with mid-level memorability scores. Participants performed a two-alternative forced choice task reporting whether a face image was present in each stream. Following the MEG experiment, a subsequent memory test evaluated memory encoding. MEG data were analyzed with time-resolved multivariate pattern analysis to decode the attribute of memorability. Behavioral results revealed that while the attended face targets were remembered, with a significant advantage to the memorable images (p = 0.002), the unattended object targets were at chance level, suggesting unconscious perception with no memory encoding. MEG decoding results for these object images revealed memorability information with significance onset at 131ms (95% CI: 69-232ms) and peak at 156ms (86-261ms). Taken together, our results indicate the RSVP paradigm was effective in suppressing conscious perception of object stimuli and revealed memorability-related neural signals as early as 131ms, offering corroborating evidence that the fate of a stimulus is decided early in the visual stream.
Meeting abstract presented at VSS 2017