Effective functional imaging of the human Superior Colliculus (SC) has often been regarded as difficult because of the small size of the SC and its proximity to sources of pulsatile (cardiac) noise. An optimised approach to functional imaging of the SC with fMRI is presented, based upon the novel finding that visually-induced BOLD responses in the SC are qualitatively different from responses in both cortical (V1) and sub-cortical (LGN) comparison areas. An optimised model with a Haemodynamic Response Function (HRF) which peaks early (4-5 s) and then falls rapidly is shown to be best suited for revealing SC responses, while a model peaking at 6 s and falling more slowly was most sensitive in the two comparison areas. Additionally, a method of correcting for the noise characteristics of fMRI responses proposed recently by de Zwart et al. (de Zwart, J. A., van Gelderen, P., Fukunaga, M., & Duyn, J. H. (2008). Reducing correlated noise in fMRI data. Magn Reson Med, 59, 939-945) is modified for use in the SC, and shown to be highly effective at further improving the statistical detectability of responses by modelling out noise. Together these methods represent a significant advance over previous approaches to functional imaging of the human SC. They permit the routine detection of strong SC activity in single subjects at standard spatial resolutions.