Magnetic resonance (MR) imaging has been proposed as a method of monitoring the interstitial laser heating of tissue for the clinical treatment of tumors (laser hyperthermia). The treatment causes considerable temperature changes over the time that image data are acquired, and therefore an analysis of the time-dependent effects of heating is required. The problem is expressed mathematically, and computer-simulated images are compared with those obtained from experimental heating and imaging of gel phantoms. Results show that at the rates of heating typical for laser hyperthermia and even with the relatively slow standard imaging techniques used, generation of artifact is not a major concern. It is also shown that a spatial spin signal magnitude distribution, evolving in time, is effectively sampled at the time to when the low-numbered phase-encoding steps are collected at to. It is noted, however, that substantial temperature changes during image data acquisition make accurate temperature determination difficult and place limits on MR imaging for quantitative spatial temperature mapping.