It is generally accepted that the temporal resolution of blood oxygenation level dependent functional MRI is limited due to the inherent latency and longevity of the haemodynamic response. However, in this study we introduce a technique for measurement of timing differences from within the same brain region in two (or more) separate tasks that allows accurate determination of cortical timing differences 200 ms. Our technique, based on a novel use of linear regression analysis, is shown to yield accurate results both in simulated and experimental data. We show that cortical timing differences measured using fMRI are consistent with published electrophysiological results. Measurement of timing differences using this technique could prove a useful strategy for identifying neural network components in a wide range of cognitive paradigms.