Although it's an interesting question, it seems like it'd be difficult to quantify due to confounding factors. Namely:

1. I would guess that how common the word is would play a larger part than similarity with other words. For instance, "understand" shares a root with "stand"; but it's far more common than "grok", and therefore more understandable.

2. In a similar vein, some words have more or broader meanings than synonyms.

We also need to distinguish words which would cause confusion and words which wouldn't. For instance, mixing up "to" and "too" is so common because they can never occur in the same place, and as such we don't actually need to distinguish them by spelling; we can distinguish them by sentence structure. The same is the case for many of your other examples, "then" (conj) vs "than" (prep), "lie" (intransitive) vs "lay" (transitive), "lose" (verb) vs "loose" (adjective), and "through" (prep) "thorough" (adj) "though" (conj).

This poses another confounding factor as well: context. Not just syntactic context like i described above, but also sematic context. Take for example, "Their stock's value decreased"" vs "The company lost millions as their stock's value decreased". A reader might need to be a bit careful to read the first properly, but the latter gives enough context that "increased" is simply eliminated from the possible word candidates before even reading "--creased". Any study to test differences in word processing these pairs would either need to be done with only the words in a vacuum (unnatural) or take into account how context would affect the result (hard)