There is and has been a lot of debate about the magical number of five users for usability testing. Jeff explains the discussion from a statistical perspective.
The five user number comes from the number of users you would need to detect approximately 85% of the problems in an interface, given that the probability a user would encounter a problem is about 31%.
However, when you release a new product the proability that a user encounters a certain problem might be closer to 10%. In that case you would need up to 18 users to detect the problem.
The advice is to test with 5 users, solve problems and test again with 5 users.