A lot of behavioral research is done by asking random people to fill out questionnaires.
However i know that when i do a questionnaire although my answers may not be outright lies the truth is somewhat subverted and the answers i give are not 100% accurate. even if the questionnaire is anonymous i almost sub-consciously give imprecise answers.
Now it could be said that this kind of uncertainty is just a random uncertainty and so by collecting enough data this is cancelled out. However consider this argument.
say we take a 100 people and ask them "would you ever screw over a colleague to get a promotion."
the answers here give a distribution of 25% yes and 75% no.
now if we take into account sub-conscious/conscious lying. Then the results may be innacurate.
Now you could argue that some people who said yes actually meant no, and some who said no actually meant yes. so that a large amount of data means that a balance should be struck.
However it is clear that the people who said yes in this situation have a lower moral compass and therefore their answer is more likely to be truthful as obviously they don't believe there is a problem with screwing someone over.
therefore the "Untruthfulness" of the answer is not balanced from both sides. this means that we have a systematic uncertainty in our experiment, and that actually taking larger amounts of data increases the uncertainty. A systematic uncertainty is caused by a problem with the instrument being used, which in this case is the questionnaire.
So can the questionnaire ever really be trusted?
BTW - how can i get images from my hard disk e.g.JPG,GIF, onto a post like this?
it would make my point easier to explain.