Functional magnetic resonance imaging (fMRI) measures activation-induced signal changes by means of the blood-oxygenation-level-depended (BOLD) contrast. Since signal changes are typically only 1-3%, the method is intrinsically sensitive to all signal instabilities including physiological noise from respiratory and cardiac cycles. Respiration induces a periodic B0-shift [1-3], leading mainly to sub-voxel shifts in phase- encoding direction in single-shot EPI acquisitions. The observed B0-fluctuation scales with the magnetic field and - in particular at fields ≥3T – may severely degrade the accuracy of the fMRI analysis. In this work, we propose a simple pulse sequence adaptation that enables a reliable and continuous monitoring of the respiration cycle sampling at ~10Hz. The respiration-induced B0-shift is traced by monitoring the phase of the FID signal. After introduction of a slice-wise normalization, the method turns out to be thoroughly compatible with multi-slice acquisitions potentially substituting an often used respiration belt. The dynamic frequency information can be used to correct image position errors and to correct for residual respiration-induced fluctuations in subsequent post-processing steps.