Caveat emport 2: It is also based on how I use a certain product feature, and my usage patterns. For all I know, I may be in the minority.
I've been favoriting a lot of videos lately, but I realized I'm probably not doing what the designer of the feature intended me to do.
Let me start at the beginning. If I designed a feature to "Favorite": essentially a list of videos that I allowed the user to mark as ones she liked, I'd assume that the primary intention of this list was to allow me to save videos that I can come to visit later. Other benefits would be sharing this with others on YouTube, as well as implicitly ranking this video so that it can be considered for YouTube honors and in lists.
However, I find that when I favorite something, my real intention is that it gets up by my Friendfeed and Facebook feeds and is shared with others in my social networks. If I want to watch the video again, I typically end up still just searching for it. It still seems quicker, and something I'm more conditioned to do than visiting my Favorites list.
So I'm using the Favorites button/tab to do what the button/tab to its right might be used to do: "Share". The reason: it was something that I've got used to that and now it's also much more convenient for me.
This is pretty common with designing any feature on absolutely any product. Users end up doing unexpected things with the feature. Changing the feature is tricky because its going to break users learned habits, and in the Internet world where everything has an API, other products as well.
The take-away is simply to learn the user-behavior and move from there when designing another features. For example, the feature of having a "Previously viewed" tag on the video search results while having many benefits, also solves this problem of finding a previously-enjoyed "favorite" video.