Because water is commonly available in fairly pure form, it has historically been used as a reproducible standard for defining physical quantities. Most of those old standards using water have been superseded by more precise standards. However, it is still interesting and instructive to trace the ways in which water has been used as a measurement standard.
Probably the most familiar such use of water is in connection to the temperature scale. The Celsius (sometimes called Centigrade, though use of that term is no longer considered correct) temperature scale was originally defined so that the freezing point and boiling point of pure water, both at one atmosphere pressure, were 0 and 100 degrees, respectively. This definition ceased to be valid with the adoption of a new International Temperature Scale in 1990.
The thermodynamic definition of temperature is based solely on the behavior of an ideal gas; also one fixed point is needed to set the size of the degree. The fixed point used is the "triple point" of water, which is the pressure/temperature condition where solid, liquid, and vapor all coexist. This is used because the triple point is a unique condition that can be precisely reproduced; water's triple point is specifically chosen because it is relatively convenient to realize in the laboratory. The temperature of the triple point of water is defined to be exactly 273.16 kelvins (where 0 K is the absolute zero of temperature). While this completely determines the thermodynamic temperature scale, temperature measurements require approximating the thermodynamic temperature by a "practical" scale that contains other fixed points at which instruments can be calibrated. Temperatures are assigned to these points based on the best scientific estimate of their true thermodynamic temperatures, and procedures are specified for interpolating between the fixed points. While previous temperature scales used the atmospheric boiling point of water as a fixed point (assigning it 373.15 K, which is 100 degrees Celsius), the reproducibility of that point is not as good as other choices. The new International Temperature Scale adopted in 1990 (known as ITS-90) covers this region with the solid/liquid equilibrium (melting/freezing) points of gallium (302.9146 K) and indium (429.7485 K). On ITS-90, the atmospheric boiling temperature of water turns out to be approximately 373.124 K (99.974 degrees Celsius).
So, have the properties of water changed? Of course not. What has changed is our ability to precisely determine temperatures in closer approximation to the true thermodynamic temperature. It turns out that the true temperature of water's boiling point is not quite what people thought it was when the Celsius scale was first defined long ago.
It is sometimes asked why one could not redefine the temperature scale so that the familiar 0 and 100 degrees Celsius would still hold for the freezing and boiling points of water. This could be done, but it would require changing the size of the degree; this would distort another familiar relationship because the difference between absolute temperature in kelvins and the Celsius scale would have to become approximately 273.22 rather than the familiar 273.15. Also, such a definition would require changing the whole scale if more precise measurements were ever made for water's boiling point. It is better to base temperature on fundamental physics (in this case, the laws of thermodynamics applied to an ideal gas) and use one precisely reproducible point (such as water's triple point) to define the scale. In this way, water is still an important part of defining the temperature scale, but it is the triple point, rather than the freezing and boiling points, that is used. Of course for most practical uses, it is an adequate approximation to think of water as boiling at 100 degrees Celsius rather than 99.974.
The other important historical use of water as a measurement standard has been in the definition of mass. The gram was originally defined as the mass of one cubic centimeter of water at some standard condition. However, mass is now referenced to the standard kilogram, which is a platinum/iridium cylinder kept in Paris. This is advantageous because it is independent of the standard of length and because a solid is easier to weigh precisely than a liquid. Careful measurements have shown that liquid water at its density maximum has a density slightly less than 1 g/cm3; the currently accepted number is 0.999975 g/cm3.
For more on the fundamental definitions of SI units, see the NIST Reference on the International System of Units (SI).
Updated June 15, 2000