Abstract
What visual information is used by observers in detection of a form: contour or surface? We reported previously (Devinck, Delord & Knoblauch, 2002, IOVS, 43) that the classification image (Ahumada, 2002, JoV, 2) for detection of a weak, square luminance increment (d′=1) of 3 deg in a 10 deg background shows a central, bright blob about the size of the test stimulus indicating that summation of luminance in the stimulus center contributes importantly to the detection process. These results were found to be insensitive to the size (1 to 6 deg) and duration of the stimulus (40 to 400 ms). We argued that since surface information is used to perform this task, filling-in was not necessary. Here we test two additional hypotheses. We presented the signal monocularly over the blind-spot, where filling-in is generally acknowledged to occur. In spite of this, there was a gap in the classification image over the region of the blind-spot, indicating that the classification image reveals no direct information about filling-in. In a second experiment, we instructed observers to report the presence of the signal only when they could detect its edges. Under these conditions, edge information dominanted in the classification image and a central gap occurred, as if the information was transmitted over a band-pass channel. These results taken with those previously reported demonstrate that observers can use edge information but do not do so automatically. They also indicate that the instructions suffice to dissociate the classification image from the signal. Overall, these results lend support to our previous conclusions that surface information is transmitted through early visual pathways, and, thus, does not need to be reconstructed at a higher level.