[Aims] Neurons in mouse visual cortex encode high-dimensional sensory information and lower-dimensional behavioral variables. Here we asked how behavioral signals are organized across the cortical volume, and whether they follow the organization of visual activity.[Methods] We used Light Beads Microscopy (Demas et al, 2021) to functionally image Ca2+ transients at cellular resolution from large volumes of cortex (3 x 3 x 0.5 mm^3) in awake, headfixed transgenic mice expressing GCaMP6s. To segment individual cells, we developed a novel volumetric cell extraction pipeline by accelerating and extending to three dimensions the Suite2P algorithms (Pachitariu et al, 2017). [Results] We found functional clusters of cells with activity corresponding to different behavioral variables such as running or whisking. While these clusters were distributed across visual cortex, we found that they had a weak spatial organization across short (~10s of um) and long (~1mm) length scales. Next, we compared the spatial structure of correlations to the retinotopic organization of visual activity. Retinotopically matched pairs of neurons in different visual areas did not have higher correlations than equally distant non-retinotopic pairs. Finally, we characterized the dimensionality of behavioral activity. We found that the strongest eigenmodes of activity were global across visual cortex. However, weaker dimensions of activity were spatially localized.[Conclusions] Behavioral signals are widely distributed across the visual cortex, and follow a weak spatial structure. This structure is distinct from the organization of retinotopic activity, and consists of a few globally distributed dominant dimensions and many weaker, spatially localized dimensions of activity.