Fig 1.
A target appeared in either one of the circular placeholders and was preceded or followed by a pre or retro cue in the form of a brief dimming of one of the placeholders. Subjects reported the orientation of the target using the central Gabor patch. On the response screen, a report cue (thickening of one side of the fixation circle) indicated where the target had been presented so that there was no location uncertainty at the time of the response. Note: stimuli are not to scale on this representation.
Fig 2.
According to a first hypothesis (a) retro-cued attention prevents the decay of an existing conscious percept. In this proposition, even when the target is conscious, in the absence of retro-cued attention the precision of this conscious representation decays slightly with time (right column). When retro-cued attention is focused on the target’s location (left column), it would prevent this slight decay of the conscious representation. This hypothesis thus predicts that the precision of the response on target’s orientation should be increased for valid retro-cues (blue curve) compared to no cues or invalid retro-cues (red curve) (b). The alternative hypothesis (c) is that retro-cued attention triggers conscious perception on trials where the target would otherwise have been missed. In this proposition, the target is not always consciously accessed, and thus not always consciously seen following its presentation (right column), but it always leaves a sensory trace in the visual cortex (left column). On trials were the target initially failed to reach conscious access, retro-cued attention at the target’s location could still promote the remaining sensory trace in visual cortex at this location to be consciously accessed (middle column). This hypothesis predicts that valid retro-cues (blue curve) should decrease the number of guesses compared to no cues or invalid retro-cues (red curve) (d). It also predicts a decrease in the precision of the accessed information: indeed valid retro-cues trigger conscious access to a degraded sensory trace on trials that otherwise would have counted as guess. Thus, less precise representations get included in the standard deviation estimate.
Fig 3.
Observed response distributions.
Distributions of the angular response errors around the target’s true orientation (kernel density estimation) for valid and invalid cues at the three different SOAs: pre-cues (-100 ms, left panel) or retro-cues (100 ms and 400 ms, middle and right panels). A reduction in the uniform component (Pguess) is apparent across SOAs as a difference between the valid and invalid distributions in the extrema of the curve.
Fig 4.
Angular error and parameter estimates.
Effect of cue validity and SOA on mean absolute response error (a), on percentage of guesses (parameter Pguess of the model) (b), and on standard deviation (SD) (c). Error bars represent standard error of the mean effect size. For “no cue” trials, error bars represent the standard error of the mean.