The primary visual cortex (V1) in highly visual mammals is organized into two-dimensional maps of stimulus features including orientation, spatial frequency and visual space. Intrinsic signal optical imaging, which defines maps at a scale of ca 100 um resolution, has revealed an inverse-gradient relationship between certain sets of maps: high gradient regions of orientation, ocular dominance and spatial frequency maps avoid each other in a two dimensional cortical space. This finding is consistent with a model of map self-organization that maximizes continuity and coverage. We have now probed map organization on a cell-by-cell basis, and further examined whether similar relationships exist in visual space, at a scale below the resolution of intrinsic signal measurements. We used bulk loading of fluorescent calcium indicators and two-photon imaging in ferret V1 in vivo to measure multiple tuning properties of individual, spatially identified, adjacent neurons in the same patch of cortex. By combining calcium imaging with intrinsic signal optical imaging, we targeted injections of calcium indicator specifically to orientation pinwheel centers, as well as to specific locations within spatial frequency maps. We find that neurons with the same preferred spatial frequency cluster in a columnar way, and different spatial frequency columns separated by less than 100 um can be clearly distinguished. Retinotopic mapping and receptive field progression is evident across cells spread over 100 um of cortical distance. Receptive field centers have small but significant jumps at one side of the orientation pinwheel centers, yet at locations more than 50 um away from them. Such sites are seen consistently at multiple depths of superficial visual cortex. At orientation pinwheels, where the preferred orientation of neurons changes rapidly, preferred spatial frequencies of the same group of neurons are very similar and show very little scatter. Away from pinwheel centers and on the opposite side of the retinotopic jump, where both preferred orientations and retinotopy change very little, preferred spatial frequencies change rapidly at some regions. Thus, the retinotopic, orientation and spatial frequency maps appear to have a highly precise, fine-scale inter-relationship, so that the rapidly changing high-gradient areas of multiple feature maps are exquisitely positioned to avoid each other in space.