Most established NMR thermometers rely on temperature-dependent chemical shift differences measured from samples that are either neat or concentrated solutions (e.g. ethylene glycol, methanol). These are unsuitable for modern cryoprobes on account of strong radiation damping resulting from the high Q of the probe. Using perdeuterated methanol, we established a relationship between the chemical shift difference and temperature, and we show that this relationship is well fitted by a quadratic equation. The actual temperature within a sample tube in the probe was verified using a Pt-100 resistor.