Obtaining functional magnetic resonance images of the brain is a challenging measurement process having a low characteristic signal-to-noise ratio. Images contain various forms of noise, including those induced by physiologic processes. One of the prevalent disturbances is hypothesized to result from susceptibility fluctuations caused by abdominal volume changes during respiration. To test this hypothesis and characterize the contribution of respiration noise to both magnitude and phase images, a mechanical model of a respiring human was constructed. The model was tested by comparing data from the model with that of a resting human. Power spectrum analyses show that the model induces both phase and magnitude disturbances similar to those in the human. The disturbances are directly related to the frequency of the respiration, with the noise most prevalent in the phase images. Though magnitude image noise is hard to identify in the human, the manikin demonstrates the presence of this disturbance. The construction of the manikin rules out motion as the primary source of the observed fluctuations and variation of the electrical properties of the manikin also indicates that signal fluctuations are not primarily due to eddy currents. Therefore, the changes are most probably induced by bulk susceptibility changes correlating with respiration.