Learning is typically modelled as a domain-general, data driven, assoc
iative process. Even though the potential influence of top-down knowle
dge is often acknowledged, the typical theoretical approach postulates
two separate modules for knowledge and learning. According to this vi
ew, knowledge may influence the initial defaults and the output of the
learning process but not the structure of the learning mechanism itse
lf. In contrast to this modular approach, this article defends the pos
ition that learning and prior knowledge interact. Theoretical analyses
and empirical studies are presented that indicate that specific and a
bstract domain knowledge influence the structure of the learning proce
sses.