In one chapter, they gave a "historical introduction" on the "sensitometry of photographic plates." It seems the standard method of establishing speed was to find a minimum exposure where any effect on the plate could be seen. It was later set to some specific density, at least for one system.
Here's a small excerpt where H&D began to rock the boat:
Sheppard and Mees note that "there has been considerable doubt thrown on the fundamental experiments upon which it has been founded" and that they intend to largely repeat those experiments, with better accuracy.In 1890 there was published by Messrs. Ferdinand Hurter and Vero C. Driffield, in the Journal of the Society of Chemical Industry, a paper entitled "Photo-chemical Investigations," in the course of which they gave for the first time a method of determining the sensitiveness of plates which depends on the measurement of a series of densities instead of a single reading.
Curiously, the modern speed systems have settled back on a single fixed density point being used, albeit with a specified (sort of) development contrast. If you've read Steven Benskins posts, he seems to be a fan of the "fractional-gradient speed method," which is more complicated. Stephen doesn't get much support for his viewpoint either. Just an interesting parallel.