Our aim was to develop a reliable and valid manual segmentation protocol for tracing the caudate nucleus in MRI for volumetric and, potentially, shape analysis of the caudate. Using the protocol, two inter- and intra-rater reliability studies were conducted using five different raters on two different image analysis platforms (ANALYZE, Mayo Biomedical Imaging Resource, Rochester MN, USA, and HERMES, Nuclear Diagnostics AB, Stockholm, Sweden). Reference images for the detailed protocol are described. Two studies were performed. In study 1, the intra-rater class correlation ICC(1,1) for an experienced rater (JCLL) using this protocol for caudate nucleus volumes was evaluated by repeating right and left caudate measurements on 10 scans (20 comparisons) and was 0.972. The inter-rater class correlation ICC(1,k) with OL was 0.922 on 5 scans (10 comparisons) and with BL was 0.960 on 5 scans (10 comparisons). In study 2, VT obtained an intra-rater class correlation of 0.9 on 5 scans (involving 10 comparisons, e.g. right and left caudate). The inter-rater class correlation ICC(1,k) was 0.988 on 5 scans (again involving 10 comparisons) with EM. We therefore developed a novel, reliable and reference image-based, method of outlining the caudate nucleus on axial MRI scans, usable in two different image analysis laboratories, across two different sets number of tracers reliably, and across software platforms. This method is therefore potentially usable for any image analysis package capable of displaying and measuring outlined voxels from MRI brain scans.