The entries of the LSJ lexicon are structured using capital letters, minuscule letters, roman numbers and arabic numbers. For example under the heading λέγω (B) we have
A.
2.
II.
b.
2.
III.
and so on. I know they represent some sort of derivation and variation of meaning, but what exactly are their meanings?
Just a way of grouping different meanings of the word and their derivations. So in λέγω the first (and let us say “original”) meaning is “to pick up”. From there you have “to count” and “to say”, each with their sub-senses. So “to count” will have “to recount”, and “to say” will have “to call by name”. The digital version gets rid of the numerals altogether and just indents each meaning.
Also, generally speaking, the lower in the rank the more uncommon the use. So, to stay with λέγω, the sense of “to say” will be far more common than the one of “to maintain a thesis”. This can also be guessed by the number of examples quoted.
What kind of derivation of sense is this? Historical? λέγω in the most ancient authors meant “pick up” and from that use there came two later senses “to count” and “to say” that appeared roughly at the same time, and so on?
Well I wouldn’t say that definitions at the same level appeared necessarily at the same time, but they are all dependent on the previous sense. So in the most ancient authors (or better said, according to its original concept) it meant to pick up, and from there you had other senses.
All the senses derived from “to speak” must necessarily be dependent on the sense “to speak” being already available (so that is chronological), but from there they aren’t arranged chronologically but from rough order of frequency.
The two senses of λέγω have different principal parts, though, and the 1985 printing of the middle Liddell lists them separately. Maybe the one was spun off so much earlier that it evolved as if it were another word entirely? On the one hand that clarifies συλλέγω but on the other I had always assumed that λέγω and the Latin “lex” were related: now that seems spurious but (the Latin) “lego” makes much more sense. But I’m no etymologist.
I think we’re talking about two different things. Sometimes a word has a double entry, like you correctly pointed for λέγω, but that is indeed because it has a different etymology and to prevent confusion and assimilation of sense. When that happens, each sense will have its own subsets, but we are not to make any assumptions concerning their interrelation.
As I understand it, Greek ὅτι is from ὅ+τι, like the Latin quod that’s lost its markedly pronominal sense.
The “say” sense of the word, however, is listed as subordinate to the “pick up” sense in the picture, with their divergent principal parts listed, but in my copy the two senses are listed seprately, with their separate parts. Surely we’re to assume they’re related, and that “say” evolved from “pick up”?
I’m not trying to start a debate and I’ll concede my ignorance but I’m curious.