This thesis introduces several novel methods for the data-driven and ex- ploratory analysis of functional brain images. Functional magnetic resonance imaging (fMRI) has emerged as a safe and non-invasive way to image the hu- man brain in action. In pharmacological MRI (phMRI), a drug's effect on the brain is of interest, rather than the brain's response to a specific task as in fMRI. However, the sometimes prolonged response to a drug necessi- tates different methodologies than those for task related effects, with further methods development needed to deliver robust results so that phMRI may be of practical use during drug development. There are many confounding issues in analysing these data, including under-informed models of response, subject motion, scanner drift, and gross differences in brain volume. In this work, data from a phMRI experiment was analysed to examine the effect of a pharmacological dose of hydrocortisone; a glucocorticoid associ- ated with the body's response to stress, and used in a number of medical conditions. The key findings were that even without using a priori hypothe- ses about the site of action, hydrocortisone significantly reduces a phMRI signal associated with blood oxygenation in the dorsal hippocampi, which is confirmed by decreases in absolute perfusion measured using arterial spin labelling. Methods were developed for the detection and correction of artefacts, includ- ing intra-scan motion and scanner drift. Functional connectivity methods were examined, and methodological issues in comparing groups investigated, revealing that many previously observed differences may have been biased or even artefactual due to gross differences in brain volume. Temporal decom- position techniques were also explored for their use in brain imaging, with wavelet cluster analysis being developed into an interactive and iterative method, while an adaptive analysis method, empirical mode decomposition, is built upon to allow the analysis of many thousands of time courses.