It is widely agreed that formal learning can influence people's social attitudes and can encourage them to take part in civic activities. Less investigated, however, has been the question of whether these effects flow from any kind of education, or whether particular subjects of study are more relevant to civic values than others. The 1958 and 1970 British birth cohort studies are used to investigate this question. Longitudinal data are essential here in three respects. They allow the accumulation of information about many episodes of learning, and so allow attention to be given to learning in adulthood, vocational courses, and informal learning as well as to courses that are part of formal initial education. They allow numerous confounding factors to be controlled for. And, with such controls, they allow the question of whether there might be long-term influences of learning. The conclusion is that courses in the social sciences and humanities are more strongly associated with socially liberal values and with participation than other courses, but that (by comparing cohorts) this distinction might be declining over time. (Contains 5 tables and 3 notes.)