Abstract
Many everyday tasks require the active prioritisation of one feature among competing ones, both during the selection from the rich sensory input and in visual short-term memory (VSTM) when the relevant information has disappeared from view. Here, we address whether we can change priorities in VSTM when, initially, we attended to a different set of features. More generally, does feature-based attention (FBA) independently affect stimulus encoding and maintenance in VSTM? We had observers report from memory the orientation of one of two distributed groups of Gabors (one black, one white), flashed for 150 ms. To manipulate FBA during stimulus encoding, in 60% of all trials, we pre-cued one color 1 s before stimulus onset, indicating the group most likely to be the target (67% validity); in the remaining 40% of trials, the pre-cue was neutral. A retro-cue, displayed 1 s after stimulus offset, then revealed the color of the target orientation (valid, 75%) or did not provide additional information (neutral, 25%). Another second later, we displayed the target group of Gabors (in target color) at a random orientation and observers rotated a knob to adjust their remembered orientation. We fitted each observer's orientation-report distributions using a mixture model comprising target reports, erroneous non-target reports, and random guesses. Valid pre-cues reliably reduced observers' guess rate and increased their report precision (i.e., 1/spread around the target orientation) as compared to neutral ones; invalid pre-cues increased the guess rate and reduced precision. Valid retro-cues did not affect the precision of the report, but further reduced the guess rate, independently of the pre-cue condition. Non-target reports were infrequent (~3%) in all conditions. Thus, FBA had independent effects during stimulus encoding (pre-cuing) and stimulus maintenance in VSTM (retro-cuing). It can change priorities in VSTM, fortifying information that would otherwise be prone to decay.
Meeting abstract presented at VSS 2016