The current research (1) examines empirical evidence to substantiate the relationship between substance choice and chronology of onset of anxiety and substance use disorders, and (2) provides information on the specificity of substance choice among anxiety disorders. A study group of 181 subjects in the Harvard Anxiety Research Project (HARP) who had a history of substance use disorder were the focus of this examination. Subjects whose anxiety disorder had an onset before their substance use disorder (primary anxiety) were compared with those whose substance use preceded onset of an anxiety disorder (secondary anxiety) for differences in distribution of subjects among categories of substance of abuse. Primary and secondary anxiety groups do not have different ages of onset for substance use disorder, nor was there greater likelihood for choosing alcohol for any of the anxiety disorders. However, there is a decreased risk of alcohol use in the small group of generalized anxiety subjects and an increased risk of opioid use in the small group of posttraumatic stress disorder subjects. There was no indirect support for the self-medication hypothesis. Neither age of onset data, specific substance association, nor proximal diagnosis association support a simple interaction. The strongest finding supported an "avoidance" of CNS stimulants.