The chemical shift or proton-resonance frequency (phase mapping) can be used to measure temperature changes. As a subtraction technique, it requires scans at exactly the same location, making it prone to respiration-induced artifacts. The accuracy of magnetic resonance (MR) phase mapping for temperature monitoring of interstitial laser coagulation (ILC) was therefore investigated in two ex vivo models with simulated respiration. MR temperatures were calibrated to interstitially measured temperature. Gradual cooling of a homogenous medium (gel) was monitored for four starting temperatures (room temperature, 40 degrees C, 50 degrees C, and 60 degrees C) during 30 min. Temperature increases were measured during ILC in ex vivo porcine liver with Nd:YAG for 6 min with 5 Watt. Experiments were performed at rest and with simulated respiratory motion (both n = 5). In liver, accuracy did not decrease with respiration simulation (P = 0.32), whereas a significant decline was found in the gel model (P = 0.002). In all experiments a small drift over time was observed between temperature determined with MR and thermoprobes. Correction for temperature-independent phase-shift at a reference location did not enhance agreement. Temperatures could be determined correctly by MR in the moving liver within a range of +/-3.5 degrees C after 6 min of laser application (95% confidence interval), justifying further pre-clinical studies.