Abstract
In a series of three experiments, we recorded ERPs associated with subjects' perception of or attention to the faces or the vase in the ambiguous face-vase figure. In the first experiment, subjects viewed the face-vase figure and reported their perceptions by pressing one of two buttons. Each button-press immediately triggered a probe flash to either the face region, the vase region, or the borders between the two. In the second experiment, the same three probes were presented while subjects selectively attended to the face or vase region in order to detect rare longer-duration probes. In the third experiment, subjects performed the same selective attention task but the probes were flashed to regions with no clear figure-ground configuration (two peripheral regions and one central region were divided by sparsely-dotted lines). ERP recordings showed the well-known face-specific N170/VPP component elicited by probes to the face region to be larger when subjects perceived or attended to the faces. Preceding the N170/VPP, two earlier components were identified. First, when the borders between the face and vase regions were probed, ERPs over parietal-occipital scalp regions differed in amplitude as early as 110ms after probe-onset depending on subjects' perceptual/attentional state. This component most likely reflects boundary detection and border ownership assignment, and its source was estimated to lie in ventral-posterior lateral occipital cortex (LOC). Second, when the face or vase regions were probed, occipital ERPs were more positive (∼150–200ms) when that region was perceived as figure (vs. background) or attended (vs. unattended). This component was considered to reflect figure-ground segregation processes, and source localization suggested ventral-anterior LOC generators. In the third experiment, none of these early components were produced, which rules out the possibility that they reflect a simple enhancement of probe-evoked activity by spatially directed attention. Instead, these early components appear to reflect contour detection and figure-ground segregation processes that can be modulated by attention.