The concept of a "gateway drug" is a logical fallacy.
Now, here's why: suppose you didn't like fox hunting, a peculiar British sport where country-dwellers dress in red and gallop across fields on horseback with a pack of dogs after a fox. It's been banned in this country, but many similar activites are still legal such as chasing a pre-laid trail of fox musk, or chasing a toy doped in fox musk and dragged by another horse. All very well and good, economic substitution at work, however a small amount of illegal hunting continues.
It comes to your attention that everyone who participates in illegal fox hunting started off by learning to ride horses, so you ban horse riding -- it's clearly a gateway to illegal fox hunting.
"But that's stupid!" you cry, "And nothing to do with drugs!"
But wait: Here are the facts -- nearly everyone who currently does heroin started off with a "soft" drug such as ecstasy or cannabis.
You can argue about this two ways -- that smoking dope makes you more likely to shoot up smack, or that the kind of people who shoot up smack are the kind of people who would start off smoking dope.
Both result in the same correlative evidence, but one is a causal relationship, and the other isn't. How would you tell the difference?
Back to fox hunting. There are large numbers of people who participate in riding-related activities that aren't fox hunting. If riding did really "cause" fox-hunting, then we would expect a far greater number of people who do riding also doing fox-hunting.
Now consider drugs; consider drugs in Portugal, where both cannabis and heroin have similarly declassified legal statuses. In 2006, heroin use occurence amongst 16-18 year olds was 1.8% and cannabis occurence was 15.1%, using Bayes law, even assuming that every heroin user also used cannabis, and in the knowledge that heroin and cannabis use do not equate to usage, the probability that any given cannabis-user also uses heroin is 0.12. Interestingly, when drugs were decriminalised in 2001, heroin usage dropped, and cannabis usage rose, strongly implying a fact that seems obvious when applied to horse riding:
If horse riding were illegal, then less people would ride horses, but those who did ride, as they were already breaking the law, would be more inclined to participate in fox hunting.
The evidence from statistics in favour of the "gateway drug" concept is weak indeed, but argument from anecdote is weaker still. I saw a program the other week where a panel discussed the potential legalisation of drugs, and despite having some excellent rationalists on both sides (Tim Carpenter of LPUK is the only one I can remember), the debate devolved into an argument between a medical-marijuana user and a former drug addict.
Come on gentlemen.