New imaging technologies have increased our capabilities to resolve three-dimensional structures from microscopic samples. Laser-scanning confocal microscopy is particularly amenable to this task because it allows the researcher to optically section biological samples, creating three-dimensional image volumes. However, a number of problems arise when studying neural tissue samples. These include data set size, physical scanning restrictions, volume registration and display. To deal with these issues, we undertook large-scale confocal scanning microscopy in order to visualize neural networks spanning multiple tissue sections. We demonstrate a technique to create and visualize a three-dimensional digital reconstruction of the hypothalamic arginine vasopressin neuroendocrine system in the male mouse. The generated three-dimensional data included a volume of tissue that measures 4.35 mm x 2.6 mm x 1.4mm with a voxel resolution of 1.2 microm. The dataset matrix included 3508 x 2072 x 700 pixels and was a composite of 19,600 optical sections. Once reconstructed into a single volume, the data is suitable for interactive stereoscopic projection. Stereoscopic imaging provides greater insight and understanding of spatial relationships in neural tissues' inherently three-dimensional structure. This technique provides a model approach for the development of data sets that can provide new and informative volume rendered views of brain structures. This study affirms the value of stereoscopic volume-based visualization in neuroscience research and education, and the feasibility of creating large-scale high resolution interactive three-dimensional reconstructions of neural tissue from microscopic imagery.