It depends on the extent of the research, i.e. a 4 week 2nd yr project versus a PHD. Also it has to do with the length of the questionnaire, rule of thumb being the longer it is the more it should be tested. I've found that time spent thinking of what you're trying to get out of the questions, how leading they questions are, how you'll interpret the data etc is probably more useful than running a trial.
Always test the questionnaire with a friend of a friend, someone you know will give more bias. Once done go through it with them and see how truthful/accurate the answers were that they gave. I think once you're above 80-85% you're in a good place to go.