Although the gauge case is earthed, that's for the illumination rather than the reading.

The gauge has a positive feed at +10V from the regulator. The other side of the gauge goes to the sender and then to earth. The sender is probably a Negative Temperature Coefficient (NTC) thermistor. All that means is that its resistance goes down as its temperature rises. As the resistance goes down, the current through the gauge and sender will rise, giving a higher gauge reading.

As Kevin has said, we could quite easily profile a sender, measuring its resistance with temperature, but as I'm about to go on holiday, if I started to heat up a saucepan of water with a sender in it and wires going to a meter I'm likely to be shot at dawn! I'll be an interesting thing to do when we get back though. I'm sure I can find a good original to experiment on! At least that would give us a base line to work from....

The resistance from sender case to pump body is an interesting thought (from Alec was it?). It's not something I'd thought of, but it's only likely to increase the total resistance and give a lower temperature reading I'd have thought.

Cheers,
Mike.