http://www.mth.kcl.ac.uk/~streater/cannabis.html

"In the study in New Zealand, analysed in [2], 3 out of 29 patients who were heavy users of cannabis suffered long-term psychotic damage. This is more than one in ten. In [3], other studies are included in an analysis, and adjustments made, and the authors conclude that the association overall for heavy users is 3%, compared with 1% for light users and non-users. Thus 97% of heavy users might expect not to suffer ill effects ten years later. This shows that, if heavy use is a cause, it is not effective on its own. One could form the hypothesis that there is a predisposition [present in 3% of the population] for cannabis to cause schizophrenia, and for those without the predisposition, smoking pot is not going to cause long-term schizophrenia. Since we do not know how to identify this condition, each of us has a risk of being among "the chosen few". If cannabis use is a cause, then by using cannabis (fifty times a year for several teenage years) we increase our risk by a factor of about 3.

But if cannabis were not a cause of long term illness, but is merely associated with it, then if a person is among the "chosen few", and he is persuaded not to use cannabis, this decision would not do him any good; he would get ill anyway. This is the implication of the failure of direction."