hide
Free keywords:
-
Abstract:
We investigated the use of configural information for the first 30 ms in a scene-perception task, using colour photographs of natural scenes. To remove configural information, a jumbled version of each picture was constructed by dividing the picture into 80 small squares and randomising their positions. A coherent picture (C) was paired with its jumbled version (J), a blank (B), or a random mask (M) to form a display sequence, which was bracketed by a fixation point (500 ms) and a mask (1 s). Two small images were displayed after each sequence. Subjects indicated which of the two was a cut-out from the coherent picture. In experiment 1, three types of display sequences were studied in a within-subjects design: 2B3C (ie 2 frames of blank followed by 3 frames of a coherent picture, at 14 ms per frame), 2J3C, and 2M3C. The sequences beginning with a jumbled image led to significantly higher accuracy (88) than those starting with either a blank (79) or a mask (75). This suggests that a jumbled image carries usable information. In experiment 2, the 2J3C sequences were compared against 3C2J and 5C (5 frames of a coherent picture alone) sequences with new subjects. The sequences that began with a jumbled image produced a lower accuracy (82) than both the 5-frame coherent-picture sequences (93) and the jumbled-last sequences (92). We conclude that during the first 30 ms, visual information other than configural information can be used to facilitate scene perception (experiment 1); however, performance can improve further if configural information is available from the very beginning (experiment 2).