A handful of years ago when I first started ripping mp3s it was commonly held that 128kpbs was as high as you needed to go, because after that the improvement in sound quality was unnoticeable.
I've stuck to that, and a recent online experiment at Cognitive Daily supports it. Read the full story here, and as usual the comments are worth reading too.
Cognitive Daily home: http://scienceblogs.com/cognitivedaily
Cognitive Daily feed: http://scienceblogs.com/cognitivedaily/index.xml