I don't know that the definitive study has been done on this. As the usability people know, opinion often contradicts direct observation of behavior. In other words users may say they like one choice even when most will use another. However, there may be other issues.
The basic idea of "tell a friend" is that you like the content of the site. This suggests location at the bottom -- after you've read enough to form some kind of opinion.
User testing into questions suggets more specificity creates a better response. If you prompted the reader not what to say, but which specific about the site to say it about, you might get higher response. And quite frankly, the success of "tell a friend" may have more to do with desirability factors rather than usability. Friends find talking about money problems taboo. You may find a "How to talk to talk to friends about money management issues without losing them" report gets a higher response.
To sum up, this is something to test based on your layout rather than adhere to a "magic usability formula."
|