Like most factors which change a person's state, the effects of caffeine are selective and influenced by contextual factors such as the nature of the activity being performed, the personality of the subject, and so on (e.g., Clubley et aI., 1979; File et at., 1982; Leathwood and Pollet, 1983; Lieberman et aI., 1987a,b; Lieberman, 1992; Blittig, 1994). The two major issues considered in this paper are the dose of caffeine and state of arousal of the person. The major question under consideration is whether there are positive effects of caffeine or does it just restore function. This has recently been re-phrased to consider whether there are positive effects of caffeine or whether the claimed improvements in perfoonance reflect withdrawal effects in decaffeinated conditions. This issue will be considered at the start of the paper as it has important methodological implications. The present paper will be mainly concerned with discussion of performance of attention tasks, for example, cognitive vigilance tasks, sustained self-paced responding, or a variable fore period simple reaction time task where the subject does not know exactly when to respond). This is not because the effects of caffeine are restricted to such tasks.