The research was done by Roman Landes, a German knifemaker. He cited a reference where surface temperatures were measured by embedded thermocouples for a sample slid dry, by hand, across a piece of fine abrasive. Peak temperatures were measured well above the austenizing temperature, but only for a few microns deep and for brief periods of time. At the very edge, beyond detectable by fingers or temper colors seen visually, the edge heats and cools very rapidly.
He also points out that this is but one part of the equation. For a knife with a botched heat treatment, it won't even be noticable. For a knife with a superlative heat treatment, it can make a difference.
I sharpen my knives occasionally with a 1x30 belt sander and finish with a Surgisharp leather belt loaded with compound. For the sake of time, I'm just willing to accept that A) I'm getting somewhat lower performance than I otherwise would, and/or B) there are other factors that make up for any decrease that I would otherwise detect.
Other makers I've communicated with have noticed differences between wet and dry sharpening, and switched to either wet belts or wet hand abrasives for the final sharpening. The original place I heard about it was from a maker who attended the Ashokan seminar/show where Roman presented his findings/research.
It's worth noting that Roman's standards are extremely high, and he uses steels, heat treatments, and hardness levels that many makers would just roll their eyes at, like 64-68 HRc on very thin bevels, 0.25 mm or so.