Twenty-two coffee drinkers (three to seven cups per day) underwent repeated double-blind trials to test for caffeine self-administration, withdrawal, and adverse effects. Each trial consisted first of a randomized crossover period of 1 day of decaffeinated coffee and 1 day of caffeinated coffee (100 mg) to assess withdrawal and adverse effects of caffeine. Next, subjects were given 2 days of concurrent access to the two coffees. The relative use of the two coffees was used to assess caffeine self-administration. Reliable caffeine self-administration occurred in three of 10 subjects in study 1 and seven of 12 subjects in study 2. Withdrawal symptoms were headaches, drowsiness, and fatigue. The major adverse effect from self-administration was tremulousness. The occurrence of headaches on substitution of decaffeinated coffee prospectively predicted subsequent self-administration of caffeine. These results indicate that some coffee drinkers exhibit signs of a caffeine dependence, ie, they self-administer coffee for the effects of caffeine, have withdrawal symptoms on cessation, and experience adverse effects.