I have been logging the fuel economy and outside air temperature for a month or so; and thought I'd try to make sense of it. There's a nice trend, a linear fit gives mpg = (0.4*temperature) +70. Nice mix of imperial mpg and degrees centigrade - that's just what we use in the UK!
I found some figures for the "dynamic viscosity" of air and so compared figures for 10 degrees C and 20 degrees C.
Air - Absolute and Kinematic Viscosity
The effect on economy is 6% (78.5 and 74mpg (imp) at 20C and 10C respectively) ... and the viscosity changes by 5.3% (3.44 and 3.5).
Convincing - provided I've chosen the right type of viscosity...