Hi David,
I comment on that in the link (third paragraph):
It turns out that refractive index is highly dependent on temperature. When using a refractometer that does not account for this effect, temperature changes can be a large source of errors. Most liquid materials expand slightly when heated and shrink when cooled. For a given material, light can pass through it more easily when it is expanded, so the index of refraction falls when materials are warmed. However, the magnitude of this effect is different for every material, and refractometers must somehow take this into account.
Handheld refractometers account for temperature by employing a bimetal strip inside them. This bimetal strip expands and contracts as the temperature changes. The bimetal strip is attached to the optics inside the refractometer, moving them slightly as the temperature changes. This movement is designed to exactly cancel temperature's effects on refractive index, and generally does a very good job IF the refractometer is designed to cancel out the temperature effects of the specific material being analyzed.
Because many refractometers are designed to use aqueous (water) solutions, the bimetal strip can be designed to account for the change in refractive index of aqueous solutions, although it may not be perfect in some situations because salts and other materials in the water can change temperature's effects on refractive index by a small extent (possibly to a larger extent for very concentrated solutions, like 750% sugar in water, but seawater is not in that category). Other details of this compensation may cause it to be imperfect (for example, the bimetallic strip provides a linear correction while the true temperature effect may be nonlinear), but those issues are beyond the scope of this article, and in general automatic temperature compensation (ATC) is a very useful attribute for aquarists using refractometers.