The gateway claim is nonsense - research shows that the correlation between earlier marijuana use and later use of other drugs can be explained by a "common-factor" model, that is, a third factor that causes both results, such as individuals' opportunities and unique propensities to use drugs, or more broadly a social or psychological predisposition towards anti-social behavior. (http://www.rand.org/pubs/research_briefs/RB6010/index1.html, http://rds.homeoffice.gov.uk/rds/pdfs2/hors253.pdf)
Also note that the same sort of correlation between earlier marijuana and later harder drugs also exists between earlier alcohol and tobacco and later illegal drugs - so if marijuana is a "gateway" so are alcohol and tobacco.