Abstract

We report a series of experiments in which subjects were required to make spatial discriminations about naturally obtained images, as follows. Subjects were shown two natural images on a computer screen, side by side and for a period of 500 ms. Subjects were then shown, on a separate part of the computer screen, a small patch of one of the images selected at random. Subjects were required to decide which of the two full images the patch comes from, and whereabouts in that image it is taken from. They indicated their response by clicking the mouse cursor at the appropriate point on the screen. The proportion of trials on which the subjects selected the correct image and the accuracy of the spatial position of their correct responses were calculated. These two measures independently indicate the general discriminability of the two image sets, and the extent to which the spatial layout of the image has been perceived. By comparing response accuracy for images that are from different sets (such as mountain or town scenes) with response accuracy when the two images are from the same set it is possible to establish the image properties that underlie coarse scene apprehension. By using sets of images that are perhaps best described as being natural textures (such as patterns of foliage or rock surface) it is possible to measure the extent to which texture is processed spatially. Our results are interpreted in the light of statistical differences between the image sets that we have used.