Impoverished second-order input to global linking in human vision.

Abstract

Recent evidence points to the importance of global operations across spatial regions larger than individual cortical receptive fields. Studies of contour integration and motion trajectory detection suggest that network operations between local detectors underlie the encoding of extended contours in space and extended trajectories in motion. Here we ask whether such network operations also occur between second-order-detectors known to exist in visual cortex. We compared performance for stimuli composed of either first-order or second-order elements equated for visibility, and we show that unlike the first-order case, there is little or no linking interaction between local second-order detectors. Near chance performance was found for elements defined by second-order attributes when observers had to identify either an elongated spatial contour or an extended motion trajectory embedded in noise elements. This implies that the network operations thought to underlie these two global tasks receive, at best, an impoverished input from local detectors that encode second-order image attributes.