Why is water not considered a proper liquid in terms of thermometers? We were just discussing about why liquids in general are used in glass thermometers. I was wondering why water isn't considered a proper liquid. Is it because of the way is expands differently to others? The fact that is boils at 100 Celcius and freezes at 0 Celcius, whereas others will contract in alternate ways?
If water is not a proper liquid, what is it that makes mercury a proper liquid?
I mean I know mercury is a metal, and so has fairly high temperature expansion. But I've never seen liquids referred to in this way before, as "proper" and not so. It must be purely in a thermometer context, because we all know that water is just about the most "proper" liquid there is!
 A: Water is the people's most popular liquid but it is also one that differs from almost all other liquids in some almost qualitative ways.

The most important reason why water is considered an "improper liquid" is that its density is higher than the density of its solid phase, the ice. For "proper liquids", it's the other way around – the solid phase is denser. For water, the maximum density occurs for 4 °C. (That's why a water-based thermometer couldn't distinguish temperatures above from those below 4 °C, even if it were positive.)
Also, its heat capacity per mole is very high – and almost none of the heat is stored in the molecular motion.  The heat capacity is unusually constant between 0 and 100 °C, with a shallow minimum at 35 °C. When pressure goes up, the coefficient of expansion goes up and the viscosity goes down – it is the opposite for other liquids.
In 1892, Röntgen found an amusing "explanation" of these facts. Water isn't just a liquid, a single substance, but "ice dissolved in proper pure liquid water". This is the main microscopic idea behind water an "improper liquid".
A: I only find the expression "proper liquid" (in this context) in works from the 19th century. An obvious reason would be the highly non-linear expansion curve (note that the picture below doesn't show that curve itself, but rather its derivative).
For mercury, the secant volumetric expansion coefficient varies very little, from $0.18165*10^{-3}$ at 0°C to $0.18250*10^{-3}$ at 100°C (at 1 bar pressure, based on equation (8) given in this reference), a change of less than 0.5% over a 100° range. For water, the (tangent) volumetric expansion coefficient is $0.15*10^{-3}$ at 15°C, $0.30*10^{-3}$ at 30°C and about $0.7*10^{-3}$ at 90°C, a change of more than 400% over a 75° range. 

Thermometers use cylindrical tubes with constant diameter, and are calibrated (or used to be in the "early days") by marking two known temperatures, usually 0°C and 100°C, and dividing the distance between those points evenly. This would clearly not work with water since the expansion between 89° and 90° is more than 4 times the expansion between 15° and 16°.
Other reasons why water is a bad choice: water reaches its highest density at 4°C, at lower temperature the water level would rise again, reaching the 8°C mark when the temperature was 0°C. And a water thermometer will break when the temperature drops below 0°C, since water expands when freezing. Mercury or alcohol thermometers may break when exposed to too high temperatures, but too low isn't a problem.     
