One is where you know what you want to learn (iOS, say). The question is, how do you efficiently learn it (or at least the bits that you are most concerned about learning). There are several good comments here along the lines of "download the tools, read the basics, start building something, look for help when you get blocked."

The second scenario is when you don't know what you want to learn. Should the next thing you learn be iOS? Android? Ruby? Haskell? And if you're not just going to go by which set of fanboys is loudest, you need to learn a bit about the candidates so that you can decide which is the one to dive into and really learn. The problem is that there are more candidates than there is time (especially if you count not just languages but frameworks and tools).

I don't have a good answer for how to find out enough to decide whether a particular language or framework has enough substance that it's going to pay back learning it.