We used multidimensional scaling to model the semantic network of alcohol and marijuana expectancies (N = 897). Preference mapping was used to estimate vectors representing patterns of activation through the network as a function of levels of behavioral inhibition (BIS) and behavioral activation (BAS). Individuals with low BIS combined with high BAS levels exhibited patterns of activation emphasizing behavioral activation similar to heavier drug users in previous research. High BIS, low BAS individuals exhibited activation patterns with greater emphasis on inhibitory expectancies similar to low-level users. Differences in expectancy activation patterns were maintained after controlling for substance use and gender. Individual differences in BIS/BAS are associated with the organization of semantic networks and patterns of activation of expectancies contributing to differences in substance use behavior.