Definition:
The potential for instruments to lose accuracy over time is sometimes called "drift." Drift in thermometers necessitates periodic calibration against standards. In the United States, NIST is responsible for setting the standards by which the accuracy of instruments are checked and reset.
Related Notes and Links:
Electronic thermometers with computer circuitry can sometimes perform very complex calculations, factoring in such things as the effect of ambient temperature on the thermometer's own circuitry to determine a measurement with greater accuracy and reproducibility. But by separating the temperature sensor (the probe) from the temperature calculator and display (the meter) into distinct devices, they also introduce the possibility of additional error.
With mechanical thermometers like liquid thermometers and dial thermometers, the display is directly manipulated by the physical properties of the temperature sensor itself (the expansion of the liquid or bimetal coil). Dial thermometers need frequent (weekly, if not daily) recalibration, but they only need one calibration at a time.
Electronic thermometers, particularly those that take interchangeable probes, may only need to be calibrated once a year (depending upon use) but both the probes and the meter should be calibrated for accuracy. Electronic thermometers and probes that are calibrated together can often mitigate against the potential for composite errors. Such probe/meter calibrations are said to be "system calibrated". Accuracy can also be variable over the full range of temperatures measured by a given thermometer