reversed even for the licit-to-illicit sequence, in contradiction to the stated first validity criterion of the GH. This order becomes even less consistent beyond involvement with illicit substances, to the degree that the use of illicit drugs other than marijuana, i.e., “hard” drugs, is usually collapsed in one class in GH modeling use sequences (Kandel and Yamaguchi, 1999). The use of “hard” drugs, such as cocaine and heroin, is frequently preceded by “soft” drugs such as marijuana whose legal status depends on the population. In fact, however, when the frequency of marijuana use in the population is taken into account, the true association between marijuana and the “hard” drug use is negligible (Earleywine, 2002). This negates the above quoted second criterion of validity for GH. The high correlations encountered in the literature and establishing the association between marijuana use and other drug use are artifactual, because they are estimated among hard drug users, without taking into account the large population of those who try or even habitually use marijuana but never transition to harder drugs (Earleywine, 2002).
