Muscle functional MRI (mfMRI) has been proposed as a tool for non-invasively measuring the metabolic and hemodynamic responses to muscle activation, but its theoretical basis remains unclear. One challenge is that it is difficult to isolate individually those variables affecting the magnitude and temporal pattern of the mfMRI response. Therefore the purpose of this study was to develop a computer model of how physiological factors altered during exercise affect the mfMRI signal intensity (SI) time course, and then predict the contributions made by individual factors. A model muscle containing 39,204 fibers was defined. The fiber type composition and neural activation strategies were designed to represent isometric contractions of the human anterior tibialis muscle, for which published mfMRI data exist. Sustained isometric contractions at 25 and 40% MVC were modeled, as were the vascular (capillary recruitment, blood oxygen extraction) and metabolic (lactate accumulation, phosphocreatine hydrolysis, pH) responses. The effects on the transverse relaxation of MRI signal were estimated, and the mfMRI SI time course was measured from simulated images. The model data agreed well qualitatively with published experimental data, and at long exercise durations the quantitative agreement was also good. The model was then used to predict that NMR relaxation effects secondary to blood volume and oxygenation changes plus the creatine kinase reaction dominate the mfMRI time course at short exercise durations (up to ~45 s) and that effects secondary to glycolysis are the main contributors at later times.