The borders of human visual areas V1, V2, VP, V3, and V4 were precisely and non-invasively determined. Functional magnetic resonance images were recorded during phase-encoded retinal stimulation. This volume data set was then sampled with a cortical surface reconstruction, making it possible to calculate local visual field sign (mirror vs. non-mirror image representation). This method automatically and objectively outlines area borders because adjacent areas often have the opposite field sign. Cortical magnification factor curves for striate and extrastriate cortical areas were determined, which showed that human visual areas have a greater emphasis on the center of gaze than their counterparts in monkeys. Retinotopically organized visual areas in humans extend anteriorly to overlap several areas previously shown to be activated by written words. Over half of the neocortex in non-human primates is occupied by visual areas. At least 25 visual areas beyond the primary visual cortex (V1) have been identified with a combination of microelectrode mapping, tracer injections, histological stains, and functional studies (1). The analysis of this data has been greatly aided by the use of flattened representations of the cortical surface—made from conventional sections using graphical techniques (2) and flattened wire models (3), or more directly from sections of physically flatmounted cortex (4). A large portion of the neocortex in the human primate is likely to be occupied by visual areas too. It has been difficult, however, to outline unambiguously any human cortical area with non-invasive techniques. Previous studies mapped only a few locations in the visual field, or have relied on stimulus features to activate different areas (5); and the tortuous convolutions of the human neocortex have defied previous attempts to see activity across all of its surface area at once. Many of the cortical visual areas in non-human primates are retinotopically organized to some degree (3, 6). These areas are irregularly shaped and somewhat variable in location; consequently, recordings from many locations (400-600) in single animals have been required in order to define areal borders with confidence (7). Here we demonstrate a technique for generating retinotopic maps of visual cortex in humans with a precision similar to that obtained in the most detailed invasive animal studies. Responses to phase-encoded retinal stimulation (8) were recorded with echo-planar functional magnetic resonance imaging (9) and analyzed with a Fourier-based method. The resulting volume data sets were sampled with a cortical surface reconstruction made from high resolution structural MRI images collected separately for each subject (10). The cortical surface containing the data was then unfolded and analyzed with the visual field sign method to distinguish mirror image from non-mirror image representations (7). By combining these four techniques (multislice functional MRI, stimulus phase-encoding and Fourier analysis, cortical surface reconstruction, and visual field sign calculations) it was possible to reconstruct the retinotopic organization of V1, V2, VP, V3, and V4 in humans in two dimensions, and to accurately trace out the borders between these areas in the living human brain. To map polar angle, 128 asymmetric spin echo MRI images (11) of 8-16 oblique sections perpendicular to the calcarine sulcus (1024-2048 total) were obtained in a 512 sec session (~8.5 min) while subjects (n = 7) viewed a slowly rotating (clockwise or counterclockwise), semicircular checkerboard stimulus. Eccentricity was mapped using a thick ring (dilating or contracting) instead of a semicircle. These four kinds of stimuli elicit periodic excitation at the rotation or dilation/contraction frequency at each point in a cortical retinotopic map (8, 12). The phase of the periodic response at the rotation or dilation/contraction frequency—measured using the (complex-valued) Fourier transform of the response profile over time at each voxel—is closely related to the polar angle or eccentricity