In an interview with User Interface Engineering, Kim Goodwin of Cooper talks about her company's approach to interaction design using personas and what she calls "scenarios" (what I would call "use cases.") She calls their methodology "Goal-Directed Design" because it focuses on the goals of the user and of the organization providing the product or the service. (She points out rightly that "If design teams only focus on the persona's goals to the exclusion of making profits, the product won't be successful.")
One other point she made caught my eye because it comes up often when creating and discussing personas. When asked if you should base your personas on real people, she said:
Certainly there are some real people who are very similar to a persona the design team may create, but it's a dangerous approach because real humans are idiosyncratic. For example, any individual user might hate the color blue or have some other random opinions that aren't necessarily representative of a larger population.
One of the strength of personas is that they gloss over those little idiosyncratic things and really focus on the essence of what is common to this particular type of person. That's one of the reasons why we rely on personas instead of real users--they are more representative.
People like to make personas as real as they can and impart personality and preferences to them. We give them names, for example, which I think helps to make them real to developers. I've been in discussions of the brand of coffee a software developer persona would prefer, though, and I think that's taking it a bit too far. It's important to talk to enough real users that they begin to blend together (like fine coffee) into a composite that's useful and not overly specific.
This is qualitative research, though, not quantitative, so (as Kim also points out), that doesn't mean you have to take months or spend millions on research. I like to do 10-15 telephone interviews to begin to understand and characterize the personas for a particular project. The clue that you haven't done enough interviews is if you keep referring to a single real life customer by name when talking about project decisions.
The key aspects of a persona are their goals and behaviors. Goals are things like making an airline reservation, getting their printer to work again or meeting a client deadline. Behaviors are things like getting impatient after waiting for a page to load for more than 10 seconds, refusing to use Internet Explorer or doing online research and then buying at the store.
These goals and behaviors are abstracted from individual people so that the individual quirks of those people fall away. Go ahead and add names for these people that help you imagine them and refer to them in documents and discussions, but make sure you don't distract yourself with the names of any actual subjects.