In Pricing breadth in skills, T-Bone, of the Games Diner, muses on how to price depth vs breadth in skills:
Games typically address depth in all the detail you’d want – which isn’t much, really. Some systems might offer skills as a binary “you got it or you don’t” switch, which won’t satisfy enthusiastic character designers. Most systems are more accommodating, with the ability to purchase a desired level of skill. Nothing fancy is needed; just by picking a level, players can make the character a laughable novice or an awesome master, or anything in between, where Astronomy or Knife Throwing is concerned. All fine and good.
Breadth, though, isn’t something you’ll see addressed with such a freely-set measure. A system will pre-package the breadth of skills, establishing that combat skills consist of Sword and Knife and Shield and so on, and scientific skills consist of Astronomy and Chemistry and Physics and what not. There may be limited options to tweak breadth, such as ways to learn smaller subsets of those skills, or to lump them into broader Hand Weapon or Science skills. But they’re typically coarse tools.
This neglects one very common and workable approach, perhaps because it shows up mostly in systems on the rules-light end of the spectrum, which is to have everything cost the same but let the player define the breadth of the skill when purchased and assume there’s some GM-guided ad-hoc trade-off between depth and breadth in use. E.g. if you define your skill as Science a successful roll will give you somewhat less information identifying an alien disease than if you had defined it as Xenobiology. Roughly equivalently the target number can be set higher for a really broad skill than a broad skill than a narrow skill than a super-specialized skill. I find this works quite well in practice, because the GM is already making those kind of judgment calls when setting target numbers or deciding what to tell the player when a knowledge skill is employed. Even if you did as T-Bone suggested and had a detailed list with a complicated formula for figuring cost, the GM would still have to make the same sorts of judgment calls as to whether the more or less specific version applied and where to divide the skills.
There are actually two kinds of sets of skills to consider when you think about grouping them: skills that can be arranged in a hierarchy of specialization, and skills that are discrete even if similar. Knowledge skills are often hierarchical. Science -> Biology -> Xenobiology -> Xenoforensics would be an example of increasing specialization in a single category. Performance skills, such as languages or musical instruments, are related but discrete. French and German aren’t really specializations of general training in “European Languages”, nor are cello and violin specializations of the broader skill stringed instruments. Facility in one might make it easier to learn another, and it’s possible to have a general aptitude for being good at learning things in the category (having an “ear for languages”), but it doesn’t really make sense to have a skill rating in Languages without skill at any particular language unless it’s just a placeholder indicating that if you ever do learn a language it will be easier.
Discrete but similar skills probably do require some way of assigning a cost along the lines that T-Bone discusses, where there are increasing returns on investment, or you’ll end up making perfectly logical characters of a sort who actually exist in the real world prohibitively hard to create in the game system. His particular ideas (basing the cost on the square root of the number of discrete skills, or on a sequence where each additional skill costs half what the previous one cost) are, imo, too complex and almost immediately run into problems of fractional costs, but it would be possible to fix that by doing the inverse: instead of saying the first language costs 1, the second 0.5, the third 0.25 and so on say you have to spend 1 at a time, but each 1 gives you an additional number of languages equal to double the prior step: for 1 you get one, another 1 would add two more, another 1 would grant you four on top of that, and so forth. Even simpler would be to do something like: one, two, three, many. That is, make the character pay for the first three, and have the fourth step jump to however many the player likes. These kinds of skills are almost never the focus of the game, and not only does the marginal utility of an additional skill itself drop off quickly, so does the utility of tracking exactly which ones are known. I’d rather the player be able to say “My character knows all the goblin and giant tongues” and move on than have the game stop while everybody looks at their character sheets to see if a particular dialect is listed.