Over the past decades, a wealth of findings has led to a substantial c
hange in the assumed complexity of classical conditioning. The combine
d evidence indicates that temporal pairing is neither necessary nor su
fficient for the formation of an associative connection. At the same t
ime, studies of model invertebrate nervous systems have allowed us to
ask a series of questions about the molecular basis of associative con
ditioning, The discovery of a pairing-sensitive mechanism in the gill-
withdrawal circuitry of Aplysia is regarded as the hallmark of the red
uctionist approach. This review outlines the insights gathered from be
havioral and neurobiological studies. Furthermore, the conceptual fram
eworks guiding research at the 'what' and 'how' levels of analysis are
compared and contrasted. I argue that a rich cognitive view of condit
ioning has emerged at the 'what' level, whereas the traditional notion
of temporal pairing still drives research at the 'how' level. A compl
ete account of classical conditioning has to await the resolving of th
is discordance. (C) 1997 Elsevier Science B.V.