Liquid-in-glass (lig) thermometers can be used in varying ranges from around - 200°C to +600°C. In calibrating a liquid-in-glass thermometer one must consider the thermometer itself, the environment used for the calibration, and the standards used to provide the reference temperature.

Factors which affect the uncertainty are:

• The range, length and sub-division of the thermometer. The more extreme the range and the cruder the division then the larger the uncertainty.
• The liquid and glass type: ie is it mercury, mercury-thallium or an organic liquid filling?
• The immersion characteristics of the thermometer (partial, total or complete; highest accuracy thermometers are normally total immersion).
• The magnitude of changes in the correction at the reference point (usually 0°C) during a calibration .
• The number of calibration points: (normally more than 5 are required and some high accuracy ligs eg "calorimeters" need an 11 point calibration to ensure linear interpolation between the calibration points).
• The changes in the corrections between successive calibration points (large changes imply larger uncertainties).
• Previous calibration history: (smaller changes in the calibration can mean smaller uncertainties or longer calibration intervals).

The thermometric liquid should have no breaks or air bubbles in the bulb or measurement column and no fluid should be trapped in the top of the thermometer. There should be no flaws in the glass or impurities in the bore and the glass should be of an appropriate type.

Other factors affecting the uncertainty of any calibration are the uncertainty of the reference thermometers and the uniformity of the calibration bath or media, (over time and space) and the level of its control; the local environment (eg unwanted radiative heating from bright lights) and the knowledge of the temperature surrounding the exposed column of liquid; the method of reading the meniscus level. The avoidance of parallax error, and ensuring a properly formed meniscus are essential for good results.

Finally the short-term reproducibility of the measurements affects the statistical (Type A) uncertainty.