The study of brain temperature is important for a number of clinical conditions such as stroke, traumatic brain injury, schizophrenia and birth asphyxia (for neonates). A direct method to estimate brain temperature non-invasively will allow assessment of brain thermoregulation and its variation in clinical conditions. Magnetic resonance imaging is a powerful technique widely used for diagnosis of a range of neurological conditions. All magnetic resonance procedures involve manipulation of the hydrogen nuclei in the water molecules of the human body. The resonance frequency of the water molecules is temperature dependent, thus MR thermometry is a powerful tool for non-invasive temperature measurement. Using internal reference MR spectroscopic imaging (MRSI), absolute brain temperature maps can be estimated. However a number of temperature independent factors influence MRSI data acquisition, thus a thorough validation is necessary and is the focus of this PhD study. In this PhD study using phantom (test object) studies it was shown that optimization of the MRSI pulse sequence is necessary to reduce systematic error in temperature maps and extensive in-vitro validation of MRSI temperature mapping was performed. A custom made temperature-controlled phantom was designed for this purpose and is presented in this thesis. MRSI data acquired from healthy (young and elderly) volunteers was employed to assess regional brain temperature variations and repeatability. Finally, the feasibility of employing fast echo planar spectroscopic imaging for volumetric MRSI temperature mapping will be presented in this thesis.