“…Accordingly, these spatial relationships of the content contained in real-world imagery! have been extensively examined and quantified by numerous techniques, some of which include optical and digital Fourier analysis, higher-order autocorrelation analyses, principal component analysis, and convolution with various types of gradient kernels (Switkes, Mayer, & Sloan, 1978;Burton & Moorhead, 1987;Tolhurst, Tadmor, & Chao, 1992;Van der Schaaf & van Hateren, 1996;Keil & Cristobal, 2000;Oliva & Torralba, 2001;Thomson, 2001a;Baddeley & Hancock, 1991;Craven, 1993;Coppola, Purves, McCoy, & Purves, 1998;Howe & Purves, 2002;Hansen & Essock, 2004a;2004b). Regardless of the analysis technique utilized, or the scene specific content (e.g., trees/grasslands or buildings/roadways) in the types natural scene imagery investigated, two general regularities of spatial content are found: (1) The amount of physical energy (or intensity) of any real-world image peaks at the larger spatial scales (Le., lower spatial frequencies) and decreases with increasing spatial scale (i.e., higher spatial frequencies) of the image content, and (2) for typical scenes, an anisotropy with a bias in image content favoring the horizontal and vertical axes (cardinal axes) relative to the oblique axes (45 0 and 135 0 oblique).…”